text
stringlengths 4
2.52M
| meta
dict |
|---|---|
1. Introduction {#sec0005}
===============
Testicular trauma can be classified aetiologically as blunt, penetrating or degloving. Blunt testicular trauma accounts for the majority of cases, typically affecting males aged 15--40 years of age [@bib0005]. Interpersonal violence, sporting injuries and road traffic accidents are the three most common causes. Road traffic accidents involving bicycles and motor bikes account for between 9 and 17% of all blunt trauma. A study conducted in 1988 revealed that 98.5% of blunt testicular trauma resulted in unilateral testicular injury [@bib0010].
This case sets out to outline the assessment and management of blunt testicular trauma in secondary care.
2. Case report {#sec0010}
==============
A 48-year-old male motorcyclist presented to the accident and emergency department with an acute scrotum following a road traffic accident. The collision occurred as an oncoming car attempted to make a right turn just as the motorcyclist had begun to set off at a set of traffic lights just as it had turned green. The impact of the motorcycle crashing into the left tyre of the car resulted in the motorcyclist being thrown into the air. The patient was wearing full protective gear including a helmet when his head struck the car windscreen before landing in a supine position on the road. No loss of consciousness was reported. The patient immediately experienced severe pain in his scrotum and on moving his right wrist. There was no nausea, vomiting or abdominal pain. On arrival, he was noted to be fully conscious. According to his Body Mass Index (BMI), the patient was defined as being obese class I and had been diagnosed with hypertension in the past. He was a non-smoker and consumed approximately 14 units of alcohol per week. The patient had a penicillin allergy, was not on any treatment for his hypertension and had no regular medications.
On examination, his GCS was 15 and his cervical spine assessed and cleared. His pulse was regular at 130 beats per minute with a blood pressure of 157/65. His abdomen was soft but tender in the suprapubic region. Examination of his genitalia revealed ecchymosis of the right hemi-scrotum and perineal bruising. The right hemi-scrotum was grossly swollen but the left testis was normal. No blood was visible from the urethral meatus. He was also noted to have swelling, tenderness and restricted range of movement of both wrists.
Blood tests confirmed a neutrophilia, raised urea and alanine aminotransferase (ALT). A Focussed Assessment with Sonography for Trauma (FAST) scan did not reveal any free intraperitoneal fluid. Ultrasonography of the urinary tract revealed approximately 100 ml of urine in the bladder, a normal left testicle and gross haematoma and ruptured capsule of the right testicle. A plain film of his right wrist revealed a comminuted intra-articular fracture of the palmar aspect of the distal radius with anterior displacement and associated anterior displacement of the carpus. There was also a transverse fracture of the ulnar styloid with slight retraction of the fractured fragment. In addition to this, he sustained a displaced volar Barton fracture of the left wrist. A pelvic X-ray did not reveal any further injury and he was subsequently catheterised Shortly after being examined, the patient experienced a vasovagal episode. He was laid supine and oxygen was administered via a facemask. He recovered over the next five minutes and was transferred to the Resuscitation Area. The patient was consented for emergency exploration of his right testicle, manipulation under anaesthesia (MUA) of the right wrist fracture and application of plaster of Paris (POP). He was informed that there was a possibility that intra-operatively, his right testis may be discovered to be so damaged that an orchidectomy may be required.
Intraoperatively, emergency exploration of the right hemiscrotum, manipulation of the fractured left wrist and application of plaster of Paris (POP) was performed concomitantly under general anaesthetic. A midline raphe incision was made and the right testis was exposed by laying open the tunica vaginalis. The presence of a haematocele confirmed intratesticular haemorrhage had occurred. Subsequently, approximately 200--300 ml of clots were evacuated. Closer inspection of the testicle revealed a tear in the tunica albuginea as well as lower pole rupture. Debridement of necrotic testicular tissue was subsequently performed followed by closure of the tunica albuginea with vicryl sutures. Good haemostasis was achieved as the tunica vaginalis was partially closed around the testis. A corrugated drain was left in situ.
During the same admission, the patient was under the joint care of the Orthopaedic team for management of a left volar Barton fracture. The patient recovered well; his drain and catheter were removed two days post-operatively. He was discharged three days after the procedure on a weeks' course of cephalexin with district nurse follow up to review the groin wound. In addition to this, outpatient follow up was arranged by Urology and Orthopaedics within three months and one week respectively.
3. Discussion {#sec0015}
=============
The majority of all testicular ruptures are diagnosed secondary to sport-related injuries [@bib0015], most commonly in the context of being struck directly in the groin. The second most common aetiology is motor vehicle or motor bike accidents, which account for between 9% and 17% [@bib0020]. Other causes include falls and straddle injuries. However, analysis of the literature has revealed a total of five cases of rupture which have been linked to testicular tumours, the most recent of which was reported in 2014 [@bib0025]. In two out of these five cases, trivial trauma preceded the diagnosis. It raises the question whether the presence of malignancy decreases the threshold of suffering a blunt testicular injury hence increasing the likelihood of testicular rupture. Seeing as both testicular malignancies as well as blunt testicular injuries commonly affect those between 16 and 40 years of age [@bib0030], it is something that should be actively considered in those presenting with these types of injuries. In patients who are managed conservatively, should we be at least testing for serum tumour marker levels such as a-fetoprotein (AFP), human chorionic gonadotrophin (hCG) and lactase dehydrogenase (LDH)?
4. Conclusion {#sec0020}
=============
Emergency assessment and diagnosis as well as scrotal exploration are important components of the management of acute testicular rupture. Analysis of the literature proves that timely surgical intervention is crucial; early intervention results in higher rates of preservation and avoids the need for an orchidectomy. In this case, we did not look at whether the patient was able to preserve the function of his right testicle by assessing his fertility following this event.
Conflicts of interest {#sec0025}
=====================
The author declares that there is no conflict of interests regarding the publication of this paper.
Funding {#sec0030}
=======
None.
Ethical approval {#sec0035}
================
Did not require ethical approval.
Consent {#sec0040}
=======
I have obtained written confirmation from this patient for publication of this case report.
Author contribution {#sec0045}
===================
Main author (sole author): Dr. Natasha J. Bauer---solely involved throughout whole process of producing this case report.
Guarantor {#sec0050}
=========
Dr. Natasha J. Bauer.
None.
| {
"pile_set_name": "PubMed Central"
} |
All relevant data are within the paper and its Supporting Information files.
Introduction {#sec001}
============
As a major public health concern worldwide, cancer is responsible for one in four deaths in the USA and Canada \[[@pone.0144890.ref001],[@pone.0144890.ref002]\]. Ovarian and pancreatic cancers are two aggressive cancers that share the characteristics of spreading insidiously while displaying atypical symptoms, and readily shifting to a drug resistance phenotype. Ovarian cancer is a heterogeneous disease that afflicts yearly 225,000 women worldwide \[[@pone.0144890.ref003]--[@pone.0144890.ref005]\]. It is the most lethal among the gynecologic malignancies, due to its asymptomatic nature in its early etiology, and the lack of efficient diagnostic tools \[[@pone.0144890.ref002],[@pone.0144890.ref005]\]. As a result, 75% of women already present themselves with advanced stages of ovarian cancer (stages III- IV of the International Federation of Gynecology and Obstetrics classification) when first diagnosed \[[@pone.0144890.ref003],[@pone.0144890.ref006]--[@pone.0144890.ref009]\]. For these women, the 5-year survival rate varies from 25% to 35% \[[@pone.0144890.ref003],[@pone.0144890.ref010]\] and the gold standard treatment is surgical debulking and chemotherapy based on a combination of paclitaxel- and platinum-based regimens \[[@pone.0144890.ref006],[@pone.0144890.ref011]--[@pone.0144890.ref013]\]. Although initial response to ovarian cancer treatment is favorable, the majority of patients become resistant to currently used treatments and more than 90% are subject to relapses after 18 months \[[@pone.0144890.ref014],[@pone.0144890.ref015]\]. On the other hand, pancreatic cancer, which is the fourth most lethal cancer, is an aggressive neoplasm presenting with a very poor prognosis \[[@pone.0144890.ref001],[@pone.0144890.ref002],[@pone.0144890.ref016]\]. Its 5-year survival rate is of only 6% \[[@pone.0144890.ref001]\] due to late initial diagnosis, rapid progression of disease, and resistance to chemotherapy regimens in current use \[[@pone.0144890.ref017]\]. As pancreatic tumors are characterized by extensive local invasion and early lymphatic as well as hematogenous metastases \[[@pone.0144890.ref018]\], only few patients are candidates to resection, and therefore, systemic gemcitabine-based chemotherapy is the most currently used form of treatment \[[@pone.0144890.ref019]\]. Despite recent improvements in drug development, the length and quality of life of pancreatic cancer patients have not improved \[[@pone.0144890.ref020]\]. Thus, there is an urgent need to develop new approaches for the management of these diseases, which are among most aggressive and lethal cancers and have limited therapeutic options, especially for patients with unresectable pancreatic cancer, whose general condition rapidly deteriorates \[[@pone.0144890.ref016],[@pone.0144890.ref020],[@pone.0144890.ref021]\].
The above considerations indicate that the development of novel therapeutic agents acting via different mechanisms of action is critically needed to overcome the major problem of drug resistance encountered with these two cancer types, and therefore, to improve the overall survival of patients. Compound RM-133 ([Fig 1](#pone.0144890.g001){ref-type="fig"}) is an aminosteroid that has been developed in our laboratory \[[@pone.0144890.ref022]\]. It has shown antiproliferative activity on several cancer cell lines (HL60 human promyelocytic leukemia cells, T-47D human breast carcinoma cells, WEHI-3 mouse myelomonocytic leukemia cells, and LNCaP human prostate cancer cells) with IC~50~ values ranging from 0.1 to 2 μM \[[@pone.0144890.ref023]\]. RM-133 also showed a low to moderate risk of drug-drug interactions based on its weak inhibition of the liver enzymes CYP3A4 and CYP2D6 \[[@pone.0144890.ref023]\]. In a foregoing *in vivo* assay, RM-133 blocked by 57% the growth of HL60 tumor xenografted in nude mice \[[@pone.0144890.ref022]\]. In a preliminary study of the mechanism of action of this family of aminosteroids, an analog of RM-133 blocked HL60 cells in G0/G1 phase and induced apoptosis \[[@pone.0144890.ref024]\].
{#pone.0144890.g001}
Herein, we report the promising anticancer activity of RM-133 on human adenocarcinoma OVCAR-3 cells, which is a model for drug resistance investigation in ovarian cancer \[[@pone.0144890.ref025]\], and on human pancreatic carcinoma (PANC-1) cells. We also report pharmacokinetic studies on RM-133 involving various vehicles, as well as its antitumor activity in models of ovarian and pancreatic cancers, namely the OVCAR-3 and PANC-1 tumors xenografted into nude mice.
Materials and Methods {#sec002}
=====================
Cell lines and cell culture {#sec003}
---------------------------
Human ovarian (OVCAR-3) and pancreatic (PANC-1) cancer cells were purchased from the American Type Culture Collection (ATCC, Rockville, MD) and maintained in exponential growth in a 5% CO~2~ humidified atmosphere at 37°C. OVCAR-3 cells were routinely grown in RPMI-1640 medium (Sigma, St. Louis, MO) supplemented with 20% FBS, L-glutamine (2 mM), antibiotics (100 IU penicillin/mL and 100 μg streptomycin/mL), insulin (50 ng/mL), and estradiol (1 nM). For OVCAR-3 cell proliferation assays, the medium used was identical, except for the omission of estradiol. PANC-1 cells were maintained in DMEM-high glucose (Invitrogen, Burlington, ON, Canada) containing L-glutamine (2 mM) and antibiotics (100 IU penicillin/mL and 100 μg streptomycin/mL), and supplemented with 10% FBS. For PANC-1 cell proliferation assays, the latter maintenance medium was replaced with DMEM/F12 containing the same supplements.
Cell viability assays {#sec004}
---------------------
The cell proliferation assay was performed with a colorimetric method using 3-(4,5-dimethylthiazol-2-yl)-5-(3-carboxymethoxyphenyl)2-(4-sulfophenyl)-2H-tetrazolium (MTS) (Cell Titer 96 Aqueous, Promega, Madison, WI). Microtiter 96-well plates (Becton--Dickinson Company, Lincoln Park, NJ) were seeded with 1 × 10^4^ cells/well suspended in 100 μL of medium and pre-incubated for 24 h at 37°C in a 5% CO~2~ atmosphere. A stock solution of RM-133 was prepared in EtOH (1×10^−2^ M) and was diluted into experimental medium, which was added to each well at time zero. After a 72-h incubation, MTS (20 μL of the solution provided by the manufacturer) was added to each well and the plate incubated for 4 h. MTS is converted to water-soluble colored formazan by dehydrogenases present in metabolically active cells, and therefore, the MTS assay allows the immediate determination of absorbance of the soluble formazan directly in the cell medium and thus the measurement of viable cells. The *A* ~490~ of the medium was determined using a 96-well microplate reader INFINITE 200 PRO series (TECAN, Männedorf, Switzerland). The IC~50~ value of RM-133 was determined for each of the two cell lines using the GraphPad Prism 7 software (GraphPad Software, Inc., San Diego, CA).
Animals {#sec005}
-------
All *in vivo* experiments were approved by our Institutional Animal Care and Use Committee (Comités de protection des animaux de l'Université Laval) and carried out according to the guidelines of the Canadian Council on Animal Care. For xenografts, homozygous ♀ *nu/nu* nude mice (24--42 days old) were purchased from Charles River Inc. (Saint-Constant, QC, Canada) and housed (four to five) in vinyl micro-isolated ventilated cages, equipped with air lids, which were kept in laminar airflow hoods and maintained under pathogen-limiting conditions. During the acclimatization and study period, the animals were housed under a controlled environment at 22 ± 3°C, with 50 ± 20% relative humidity and light set at 12 h/day (light on at 07:15). Rodent food (Rodent diet \#T.2018.15, Harlan Teklad, Madison, WI) and water were provided *ad libitum*. For *nu/nu* nude mice, food and water were sterilized prior to dispensing to the animals. Standard Balb/c mice were used for pharmacokinetic studies. Animals were anesthetized with isoflurane and killed by cervical dislocation.
Effect of RM-133 on OVCAR-3 xenografts and drug plasma concentration in nude mice {#sec006}
---------------------------------------------------------------------------------
### First part {#sec007}
Twenty-two ♀ *nu/nu* nude mice (22--24 g) were inoculated s.c. with 5×10^6^ OVCAR-3 cells (in 0.1 mL of growth medium containing 30% Matrigel (BD Biosciences, Bedford, MA)) into both flanks in each mouse via a 2.5-cm long 22-gauge needle. After 11 days, tumor-bearing mice were randomly assigned to two groups of 11 mice each according to tumor size, i.e. a control group (*n* = 12 tumors) and a treated group (*n* = 15 tumors). RM-133 was administered s.c. daily at 60 mg/kg in 0.1 mL of propylene glycol:EtOH (92:8). Animals in the control group received 0.1 mL of vehicle alone. Tumor size was measured twice weekly using a caliper. Two perpendicular diameters (*L* and *W*) were measured, and the tumor area (in mm^2^) was calculated using the formula (*L*/2) x (*W*/2) x π. Animals were weighed at various intervals during the experiment.
### Second part {#sec008}
At the end of the xenograft experiment, and in order to obtain preliminary pharmacological data, mice in treated groups were separated into 4 subgroups of 2--3 mice each, injected s.c. with RM-133 (60 mg/kg) and sacrificed after 3, 7, 12, or 24 h. In parallel, mice in the control group were also separated into 4 subgroups, treated with RM-133 (60 mg/kg) as for treated mice, and sacrificed at the same intervals. Blood was collected from mice by cardiac puncture, and the plasma concentration of RM-133 determined by liquid chromatography--tandem mass spectrometry (LC-MS/MS) as previously reported \[[@pone.0144890.ref023]\]. Tumors of the mice in the treated group were collected at necropsy, pooled, homogenized according to a known procedure \[[@pone.0144890.ref026]\] and the quantity of RM-133 was determined by LC-MS/MS.
Plasma concentration of RM-133 in mice at different injection doses {#sec009}
-------------------------------------------------------------------
Eight ♀ Balb/c mice weighing approximately 20 g were separated in 4 groups. Mice in each group received a single s.c. injection of RM-133 (30, 60, 120, or 480 mg/kg). The vehicle used was propylene glycol:EtOH (92:8) and injection volume was 0.1 mL for all concentrations except at 480 mg/kg, where 0.2 mL was used. Blood was collected by cardiac puncture after 12 h, and the plasma concentration of RM-133 determined by LC-MS/MS as above.
Analysis of different vehicles for RM-133 administration {#sec010}
--------------------------------------------------------
Eleven different vehicles were investigated by s.c. injection \[[@pone.0144890.ref027]--[@pone.0144890.ref035]\]. These vehicles were formulated as follows: vehicle \#1 \[propylene glycol:EtOH (92:8)\]; vehicle \#2 \[aqueous 0.4% methylcellulose:EtOH (92:8)\]; vehicle \#3 \[castor oil:EtOH:benzyl alcohol:benzyl benzoate (65:10:10:15)\]; vehicle \#4 \[sunflower oil:EtOH (92:8)\]; vehicle \#5 \[sunflower oil:tetrahydrofuran (92:8)\]; vehicle \#6 \[aqueous 25% β-cyclodextrin:EtOH (92:8)\]; vehicle \#7 \[sesame oil:EtOH:benzyl benzoate:benzyl alcohol:Tween 80 (89.8:7.8:1:1:0.5)\]; vehicle \#8 \[soya oil:EtOH (92:8)\]; vehicle \#9 \[RPMI medium:EtOH (92:8)\]; vehicle \#10 \[saline:dimethyl sulfoxide (DMSO):Tween 80 (89.5:10:0.5)\]; and vehicle \#11 \[saline:benzyl alcohol:carboxymethylcellulose:Tween 80 (98.2:0.9:0.5:0.4)\]. The eight vehicles producing the best RM-133 solubility (\#1--8) were injected s.c. (0.1 mL) without the test compound (RM-133) for tolerance and behavioral evaluation studies, by observing mice at 0.25, 1, 3, 7, 23, 26, 28, 30, and 96 h post-injection. Finally, RM-133 was injected s.c. once at 120 mg/kg/0.1 mL of the best seven tolerated vehicles (\#1,2,4--7) into ♀ Balb/c mice (4 per group), and mice were next monitored with blood collection by cardiac puncture after 3 and 12 h. The plasma concentration of RM-133 was then determined by LC-MS/MS as above.
Effect of repeated s.c. injections of RM-133 using 3 vehicles {#sec011}
-------------------------------------------------------------
Ten ♀ Balb/c mice weighing approximately 23 g were separated in 3 groups. Group 1 (3 mice) was treated s.c. with RM-133 (240 mg/kg/0.2 mL, twice a day (AM and PM), every other day) in aqueous 0.4% methylcellulose:EtOH (92:8). Group 2 (3 mice) was treated s.c. with RM-133 (120 mg/kg/0.1 mL, twice a day (AM and PM), every other day) in sunflower oil:EtOH (92:8). Group 3 (4 mice) received only one injection of RM-133 (120 mg/kg/0.1 mL) in aqueous 25% β-cyclodextrin:EtOH (92:8) and were sacrificed 24 h later. Mice in groups 1 and 2 received a total of 8 injections over 4 days and were sacrificed on day 8. Mice behavior was monitored during the whole duration of the experiment, and macroscopic observation of organs was performed upon necropsy.
Effect of RM-133 on OVCAR-3 xenografts using the two selected vehicles {#sec012}
----------------------------------------------------------------------
Fourty ♀ *nu/nu* nude mice (22--24 g) were inoculated with 5×10^6^ OVCAR-3 cells per flank, in 30% matrigel-containing culture medium (0.1 mL). After 23 days, tumor-bearing mice were randomized in 4 groups of 8 to 9 mice each. Group 1 (12 tumors) was treated s.c. with RM-133 (240 mg/kg/0.2 mL, once (AM) every other day) in sunflower oil:EtOH (92:8), and group 2 (13 tumors) was treated s.c. with RM-133 (240 mg/kg/0.2 mL, twice a day (AM and PM), every other day) in aqueous 0.4% methylcellulose:EtOH (92:8). Two control mouse groups, each representing 12 tumors, received only the vehicles. As mentioned above, the tumor area was measured twice weekly and body weight monitored.
Effect of RM-133 on PANC-1 xenografts {#sec013}
-------------------------------------
Twenty female *nu/nu* nude mice (22--24 g) were inoculated with 5×10^6^ PANC-1 cells per flank, in 30% matrigel-containing culture medium (0.1 mL). Ten days after inoculation, tumor-bearing mice were randomized in 2 groups of 10 mice each. Mice in the first group (*n* = 16 tumors) were treated s.c. with RM-133 (240 mg/kg, twice a day (AM and PM), every other day) in 0.2 mL of aqueous 0.4% methylcellulose:EtOH (92:8). Mice in the second (control) group (*n* = 18 tumors) received the vehicle only. As with previous xenografts, the tumor area was measured and mouse body weight monitored.
Synthesis and preparation of RM-133 {#sec014}
-----------------------------------
The aminosteroid RM-133 (2β-\[1-(quinoline-2-carbonyl)-pyrrolidine-2-carbonyl\] *N*-piperazine-5α-androstane-3α,17β-diol) was synthesized, characterized, and purified as previously described \[[@pone.0144890.ref022]\]. The purity of RM-133 was found to be 99.5% as determined by high performance liquid chromatography (Apparatus: Shimadzu (Kyoto, Japan); Column: Altima, HP, C18-AQ, 4.6 x 250 mm, 5 μm; Solvents: A gradient from methanol/water (70:30) to 100% methanol; UV detection at 190 nm). For *in vivo* treatments, RM-133 was suspended in the vehicle one day prior to its injection into mice and stored at 4°C under constant agitation until use.
Statistics {#sec015}
----------
The Duncan-Kramer test was used to analyze data, and statistical significance accepted at *P* \< 0.05 \[[@pone.0144890.ref036]\].
Results and Discussion {#sec016}
======================
RM-133 is cytotoxic towards human OVCAR-3 and PANC-1 cells {#sec017}
----------------------------------------------------------
To evaluate the effect of RM-133 on the proliferation of OVCAR-3 and PANC-1 cancer cells, the latter were incubated with increasing concentrations of RM-133 for 72 h, and the IC~50~ of the drug then measured. RM-133 clearly had an antiproliferative activity on the tested cell lines, with IC~50~ values of 0.8 and 0.3 μM for OVCAR-3 and PANC-1, respectively ([Fig 2](#pone.0144890.g002){ref-type="fig"}). The IC~50~ values in the 0.1--1 μM range obtained here were similar to those previously measured in a panel of human cancer cell lines with this family of aminosteroids \[[@pone.0144890.ref023]\].
{#pone.0144890.g002}
The OVCAR-3 cell line was derived from a patient refractory to cytotoxic chemotherapy, and is therefore an interesting model to investigate drug resistance \[[@pone.0144890.ref025]\]. The fact that RM-133 exhibited potent antiproliferative activity on a model (OVCAR-3) known for its chemoresistance, as well as on PANC-1 cells, prompted us to further investigate the *in vivo* activity of RM-133 in two tumor xenograft models using these two cancer cell lines.
RM-133 inhibits growth of OVCAR-3 tumor xenografts {#sec018}
--------------------------------------------------
Female nude mice were inoculated in both flanks with OVCAR-3 cells. Mice bearing tumors were randomized in 2 groups: one group was treated with RM-133 (60 mg/kg) and the other (control) received the vehicle only (propylene glycol:EtOH (92:8)). Tumors in the control group grew clearly more rapidly than those treated with the aminosteroid ([Fig 3A](#pone.0144890.g003){ref-type="fig"}). Starting from day 15 of drug therapy and until the end of the experimental period, RM-133-treated tumors became significantly smaller than in the control group, with a 60% difference between the two groups measured at the end of the treatment. Since the toxic potential of a given compound depends on its concentration and exposure time \[[@pone.0144890.ref037]\], and as mortality, clinical symptoms, and body weight changes are among the main indicators of its toxicity \[[@pone.0144890.ref038]\], it is noteworthy that over a 21-day treatment period with RM-133, there was no apparent effect on body weight ([Fig 3B](#pone.0144890.g003){ref-type="fig"}) nor apparent toxicity of the drug (e.g. abnormal behavior, death).
{#pone.0144890.g003}
At the end of the OVCAR-3 xenograft experiment, both control and RM-133-treated groups separately received a single s.c. dose of RM-133 (60 mg/kg). Blood was collected after 3, 7, 12, and 24 h, and the drug plasma concentration measured by LC-MS/MS. The time course of RM-133 plasma concentration was similar in groups A and B. The drug exhibited a classic drug-releasing profile \[[@pone.0144890.ref039]\], with a bolus delivery of RM-133 followed by a sustained decrease in its plasma concentration ([Fig 4](#pone.0144890.g004){ref-type="fig"}). The maximum average plasma concentration (460 ng/mL) was observed 3 h following the bolus injection, and had decreased to 81 ng/mL after 24 h. These results demonstrate that RM-133 does not accumulate in blood during the 21 days-xenograft experiment. Tumors of the mice treated with RM-133 were also collected at necropsy in order to measure the quantity of RM-133. This later was present inside the tumor at a concentration of 1.2 μM (773 ng/g), which corresponds roughly to the concentration of RM-133 that inhibits 50% of OVCAR-3 cell proliferation (IC~50~ = 0.8 μM). This result is in accordance with the 60% reduction of tumor size progression observed in the first OVCAR-3 xenograft experiment. Thus, RM-133 is not sufficiently concentrated inside the tumor to fully inhibit its growth.
{#pone.0144890.g004}
RM-133 plasma concentration exhibits a logarithmic dependence on the amount injected {#sec019}
------------------------------------------------------------------------------------
In previous experiments, RM-133 had demonstrated a maximal antitumor activity of 57% against HL60 xenografts \[[@pone.0144890.ref032]\], which is comparable to the 60% inhibition value measured in this report towards OVCAR-3 xenografts. At the end of the OVCAR-3 xenograft experiment, RM-133 plasma concentration measured 12 h after a single bolus injection of 60 mg/kg (in 0.1 mL) in groups A and B ([Fig 4](#pone.0144890.g004){ref-type="fig"}) was 178 and 91 ng/mL, respectively. This result prompted us to examine the dose-plasma level relationship of a single s.c. injection of RM-133 using a range of different doses of the compound. As shown in [Fig 5](#pone.0144890.g005){ref-type="fig"}, the plasma concentration (167 ng/mL) reached after the injection of 30 mg/kg represents only 17% of the injected amount, and this percentage decreased further with higher doses of RM-133. Thus, the trend of the dose-level relationship suggests that increasing the RM-133 dose leads to a plateau, likely as a result of low plasma solubility or of the poor bioavailability of RM-133 in the vehicle used (propylene glycol:EtOH (92:8)). However, the dose of 480 mg/kg in 0.2 mL of vehicle was not tolerated by mice ([Fig 5](#pone.0144890.g005){ref-type="fig"}), which manifested with motion discomfort. Following these observations, the vehicle alone (0.2 mL of propylene glycol:EtOH (92:8)) was administered to the mice. As suspected, reactions were the same in the animals, indicating that the vehicle itself (propylene glycol:EtOH (92:8), at 0.2 mL) was the culprit. The maximum volume tolerated was next determined as 0.1 mL. These observations led us to investigate alternative vehicles that would allow to yield higher RM-133 plasma levels with minimal adverse effects.
{#pone.0144890.g005}
Vehicle optimization: screening of alternative vehicles for optimal RM-133 administration {#sec020}
-----------------------------------------------------------------------------------------
Non-aqueous solvents such as dimethyl sulfoxide and polyethylene glycol, detergents (e.g. Tween, vegetable oils, etc.) and solubilizers (e.g. β-cyclodextrin, methylcellulose) offer strategic options for enhancing compound solubility in the drug discovery process \[[@pone.0144890.ref030], [@pone.0144890.ref039]\]. For this reason, 11 different vehicles previously used with steroidal derivatives were investigated. Firstly, based on the solubility of RM-133 in each vehicle (Table A in [S1 File](#pone.0144890.s001){ref-type="supplementary-material"}), eight vehicles were selected for a second round of investigation. In the latter experiments, mice received a single s.c. injection of each drug-free vehicle for tolerance and behavioral evaluation (Table B in [S1 File](#pone.0144890.s001){ref-type="supplementary-material"}). Based on our observations at different post-injection times (0.25--96 h), we eliminated the vehicle \#3 (castor oil-based). In fact, following the administration of vehicle \#3 alone, mice displayed many discomfort symptoms (itching, jumping, and microphtalmia), and during the following four days, mice presented wounds averaging \~24 mm^2^ at the site of injection.
In a third experiment, the seven best tolerated vehicles containing RM-133 were injected s.c. in mice (120 mg/kg) and blood samples collected after 3 and 12 h for RM-133 plasma concentration measurement ([Table 1](#pone.0144890.t001){ref-type="table"}). The highest RM-133 plasma concentration reached after 3 h was obtained with vehicle \#6, followed (in that order) by vehicles \#5, \#2, \#1, \#7, \#4, and \#8. As expected, RM-133 plasma concentration had decreased after 12 h, but vehicle \#6 still yielded the highest levels after the latter interval. In contrast, vehicle \#8 yielded the lowest plasma concentration at both time intervals. Based on the results obtained from the three experiments described above (Tables A and B in [S1 File](#pone.0144890.s001){ref-type="supplementary-material"}, [Table 1](#pone.0144890.t001){ref-type="table"}), we concluded that (i) RM-133 was poorly soluble in vehicle \#5 (small lumps), (ii) that vehicles \#1 and \#7 led to skin lesions and vehicle \#8 yielded the lowest plasma concentrations of the analyzed set. On the other hand, vehicle \#6 led to the highest plasma concentrations of the set, whereas vehicles \#2 and \#4 did not show any adverse effect on mice. We thus selected vehicles \#2, \#4, and \#6 to perform an additional tolerance test.
10.1371/journal.pone.0144890.t001
###### Time course of RM-133 plasma concentrations with various vehicles.
{#pone.0144890.t001g}
\# Vehicle Plasma concentration
----------------------- ------------------------------------ ------------------------------------------ -----------
1 Propylene glycol (92%) 314 ± 2 189 ± 5
EtOH (8%)
2 Aqueous 0.4% methylcellulose (92%) 449 ± 5 171 ± 10
EtOH (8%)
3 Castor oil (65%) NT[^b^](#t001fn002){ref-type="table-fn"}
EtOH (10%)
Benzyl alcohol (10%)
Benzyl benzoate (15%)
4 Sunflower oil (92%) 285 ± 5 173 ± 6
EtOH (8%)
5 Sunflower oil (92%) 642 ± 20 374 ± 9
Tetrahydrofuran (8%)
6 Aqueous 25% β-cyclodextrin (92%) 2963 ± 121 1783 ± 47
EtOH (8%)
7 Sesame oil (89.7%) 302 ± 7 214 ± 8
EtOH (7.8%)
Benzyl benzoate (1%)
Benzyl alcohol (1%)
Tween 80 (0.5%)
8 Soya oil (92%) 137 ± 2 52 ± 1
EtOH (8%)
^a^Single injection of RM-133 (120 mg/kg) in 0.1 mL of vehicle.
^b^NT: this vehicle was not tested in this experiment based on tolerance and behavioral evaluation in mice.
Effect of repeated subcutaneous injections of RM-133 in 3 selected vehicles (tolerance test) {#sec021}
--------------------------------------------------------------------------------------------
RM-133 was administered to mice every other day using vehicles \#2, \#4, and \#6. Animals were treated for one week with vehicles \#2 and \#4, but only once with vehicle \#6. Mice behavior was monitored during the experiment, and macroscopic observation of organs was performed at necropsy (Table C in [S1 File](#pone.0144890.s001){ref-type="supplementary-material"}). Ten hours following the s.c. injection of the first dose of RM-133 (120 mg/kg/0.1 mL) in aqueous 25% β-cyclodextrin:EtOH (92:8) (vehicle \#6), mouse mobility was considerably reduced. Twenty-four hours later, their general condition had further deteriorated, as seen with seclusion, dehydration, microphthalmia, bristling, accelerated breathing rate, and abdomen distension. Based on these observations, we abruptly terminated experiment with vehicle \#6, 24 h post-injection. At necropsy, all mice in that group showed hemorrhagic lungs, loaded stomachs along with empty intestines, and 75% exhibited slightly enlarged kidneys. When correlated with the mouse deaths previously noted when administering vehicle \#6 alone, these observations clearly indicated that the cyclodextrin-based vehicle was not tolerated by mice, and vehicle \#6 was therefore eliminated from our screening studies.
Interestingly, and in accordance with the observations from the experiment performed with vehicles only, no adverse effect was noted when mice were treated with RM-133 using vehicles \#2 and \#4. Because the oily vehicle \#4 (sunflower oil:EtOH (92:8) generated blisters at the injection site, we limited treatment to a single injection of RM-133 (240 mg/kg/0.2 mL /every other day) for the xenograft experiments. However, the aqueous vehicle \#2 (aqueous 0.4% methylcellulose:EtOH (92:8)) allowed to perform multiple injections (240 mg/kg/0.2 mL, BID/every other day) without adverse effect on mice. At necropsy, organs did not show any detectable abnormality. Thus, both vehicles \#2 and \#4 were included for comparison for the *in vivo* studies on RM-133 antitumor activity.
Antitumor activity of RM-133 on OVCAR-3 xenografts using two different vehicles {#sec022}
-------------------------------------------------------------------------------
With the sunflower-based vehicle ([Fig 6A](#pone.0144890.g006){ref-type="fig"}), RM-133 started to inhibit the tumor growth after 20 days, but the effect became significant only at day 28. By the end of treatment (40 days), RM-133 had completely inhibited (i.e. by 100%) OVCAR-3 tumor growth. However, the oily nature of vehicle \#4 generated big blisters, which also hampered tumor measurements. When RM-133 treatment was stopped at day 40 due to vehicle-induced effects, tumor growth resumed and no significant difference subsisted between the two groups seven days after treatment termination ([Fig 6A](#pone.0144890.g006){ref-type="fig"}). The latter results show that the aminosteroid RM-133 was responsible for tumor growth inhibition. No significant difference in body weight was observed between the untreated (control) and treated (RM-133) group ([Fig 6B](#pone.0144890.g006){ref-type="fig"}). Furthermore, no toxicity symptom was either observed during the study or at necropsy.
{#pone.0144890.g006}
With the methylcellulose-based vehicle, tumor growth became significantly different between the treated and control groups as early as at day 15 ([Fig 7A](#pone.0144890.g007){ref-type="fig"}), and this difference persisted until the end of the experimental period (day 35). In fact, RM-133 treatment efficiently abrogated the rapid growth of OVCAR-3 tumors, whose surface area was decreased by the aminosteroid. After 28 days, RM-133 had completely inhibited (by 100%) tumor growth, and had further decreased tumor size to 78% of its initial value by the end of treatment (35 days). After treatment interruption, RM-133 still maintained its complete blockade of tumor progression until at least 12 days after the end of treatment ([Fig 7A](#pone.0144890.g007){ref-type="fig"}). Interestingly, neither abnormal behavior nor death was recorded over the 35-day treatment period with RM-133. Moreover, vehicle \#2 did not affect body weight ([Fig 7B](#pone.0144890.g007){ref-type="fig"}) or lead to any overt toxicity symptom over the whole experimental period as well as at necropsy.
{#pone.0144890.g007}
Antitumor activity of RM-133 on PANC-1 xenografts {#sec023}
-------------------------------------------------
As with ovarian cancer, pancreatic cancer typically shows resistance to currently available therapies and presents a very poor prognosis. Based on the results obtained with the OVCAR-3 xenografts, we used vehicle \#2 (aqueous 0.4% methylcellulose:EtOH (92:8) to assess the effect of RM-133 on the PANC-1 pancreatic cancer tumor xenograft model. RM-133 progressively inhibited PANC-1 tumor growth until day 22, where a 75% reduction of tumor size progression was observed ([Fig 8A](#pone.0144890.g008){ref-type="fig"}). At the end of the treatment period (40 days), PANC-1 tumor progression was reduced by 63%, indicating that the aminosteroid possessed significant, albeit only partial antitumor activity towards the PANC-1 pancreatic tumor model. As already shown with the OVCAR-3 tumor xenograft experiments, body weight was not affected by treatment ([Fig 8B](#pone.0144890.g008){ref-type="fig"}), and no apparent toxicity could be detected either during the experimental period study or at necropsy.
{#pone.0144890.g008}
Conclusion {#sec024}
==========
The cytotoxic activity of the aminosteroid RM-133 has been first evaluated *in vitro* in two human cancer cell lines, and the IC~50~ values in the low micromolar range that were obtained (0.8 μM for OVCAR-3 (ovary) and 0.3 μM for PANC-1 (pancreas)) prompted us to assess the potential of RM-133 in tumor xenograft models. RM-133 (60 mg/kg/day) reduced OVCAR-3 tumor xenograft growth by 60% in nude mice, using propylene glycol:EtOH (92:8) as vehicle and s.c. injections. The failure to achieve a stronger tumor growth inhibition pointed towards a poor bioavailability of RM-133 with the vehicle used. Upon investigating the solubility, plasma concentration, and tolerance of RM-133 delivered with 11 different vehicles, two alternative vehicles (aqueous 0.4% methylcellulose:EtOH (92:8) and sunflower oil:EtOH (92:8)) were selected for further xenograft experiments. Thus, long-term treatment with RM-133 reduced PANC-1 pancreatic tumor growth by 63% in methylcellulose-based vehicle. Interestingly, RM-133 completely inhibited OVCAR-3 tumor growth using either sunflower-based and methylcellulose vehicles; and using the latter vehicle only, RM-133 further reduced tumor size down to 78% of its initial size. However, the antitumor response observed using the oily, sunflower-based vehicle was delayed and less efficient compared to the methylcellulose-based vehicle. Furthermore, as observed with the methylcellulose-based vehicle, RM-133 maintained its antitumor action against the OVCAR-3 model, even one week after cessation of treatment. RM-133 was well tolerated by mice during the whole experimental period, and no weight loss was recorded. These promising results that have been obtained with ovarian and pancreatic cancer models, which are known as highly refractory to current therapies and as having a very poor prognosis, urge us to pursue our studies with the aminosteroid RM-133 and optimize its dosage, schedule, and mode of delivery, especially regarding its effect towards pancreatic cancer.
Supporting Information {#sec025}
======================
###### Table A. RM-133 solubility in 11 injection vehicles; Table B. Effect of a single s.c. injection of 8 vehicles in mice behavior; Table C. Effect of repeated s.c. injections of RM-133 using 3 preselected vehicles.
(PDF)
######
Click here for additional data file.
The authors thank Patrick Caron (LC-MS/MS analyses) and Sonia Francoeur (animal house) for their expert technical assistance. The careful and critical reading and editing of this manuscript by Dr. Richard Poulin are also greatly appreciated.
[^1]: **Competing Interests:**Authors JR, RM, and DP of this manuscript have the following competing interests: 2-(N-Substituted piperazinyl) steroid derivatives: Patent 13/130,621 (US); Patent Applications 2,744,369 (Canada) and 09828506.7 (Europe). This does not alter the authors' adherence to PLOS ONE policies on sharing data and materials.
[^2]: Conceived and designed the experiments: LCK DA JR RM DP. Performed the experiments: LCK DA JR. Analyzed the data: LCK DA JR RM DP. Wrote the paper: LCK DA JR RM DP.
| {
"pile_set_name": "PubMed Central"
} |
INTRODUCTION {#s1}
============
The Wnt pathway is a highly conserved signaling pathway throughout evolution. It controls numerous stages of embryo development and adult tissue homeostasis through atiming- and location-specific activation \[[@R1], [@R2]\]. The pathway a timing is closely controlled by an overabundance of potential ligands, receptors, inhibitors, and second messengers acting as transcriptional regulators or post-translational modifiers. Thus, alterations of Wnt activity are often associated with developmental disorders and diseases, including cancer, neuronal diseases, and skeletal disorders \[[@R3]--[@R10]\]. Over the last decade, the role of Wnt signaling has been recognized also in cardiac pathophysiology and its dysregulation can result in processes leading to cardiac dysfunction, such as cardiac hypertrophy, fibrosis, arrhythmias, and infarction \[[@R11]--[@R13]\]. More recently, increasing evidence has indicated that the canonical Wnt/β-catenin pathway is implicated in the molecular pathogenesis of arrhythmogenic cardiomyopathy (AC) \[[@R14]--[@R15]\].
In this review, we aim to introduce the Wnt signaling pathway and to discuss its involvement in AC, together with the development of more selective therapeutic approaches based on Wnt targeting therapies.
The Wnt signaling cascade {#s1_1}
-------------------------
The Wnt pathway is coupled to a highly conserved family of secreted protein growth factors (cysteine-rich glycoproteins) known as Wnt proteins, that have been identified in animals from Hydra to Human \[[@R2], [@R16], [@R17]\]. The name Wnt, used in mammals, arises from the combination and the relationship between the *Drosophila melanogaster* segment polarity gene Wingless and its vertebrate orthologous, Int-1, a proto-oncogene discovered in 1984 in mouse \[[@R2], [@R18], [@R19]\]. At present, about 100 Wnt genes have been identified in various species, and 19 diverse Wnt proteins have been isolated in humans \[[@R16], [@R20]--[@R22]\]. Although Wnt family members show a high degree of sequence similarity, the expression of specific Wnt growth factors can lead to largely different signaling cascades \[[@R23]\].
At the cell membrane, Wnt ligands bind the Frizzled receptors (Fzd) as well as low-density Lipoprotein Receptor-related Protein-5 or -6 (LRP5/6) co-receptors. Fzd receptor family includes eleven members in Humans, which show seven transmembrane-spanning regions, an extracellular N-terminus Wnt-binding domain, and an intercellular C-terminal tail, crucial in the signal transduction \[[@R12], [@R24], [@R25]\].
Different Wnt ligands activate different intracellular signal transduction pathways, including the canonical β-catenin-dependent Wnt pathway, and the non-canonical β-catenin-independent Wnt/Planar cell polarity and Wnt-calcium/protein kinase C pathways \[[@R4], [@R26]--[@R28]\]. Wnt/Planar cell polarity (Wnt/PCP) and the Wnt/β-catenin pathway primarily regulates cell cycle-related proteins that are cell-type and context dependent \[[@R29]\], where as the Wnt/PCP is responsible for sensory cell orientation, cytoskeleton re-organization and directed migration \[[@R30]\]. The biological function of the Wnt/Calcium pathway, resulting in increased cytosolic Ca^2+^ concentration and activating Ca^2+^-sensitive proteins, is unclear, but it seems to be also involved in the controlling of cell fate and cell migration \[[@R26], [@R31]\].
In this review, we focus on the canonical Wnt pathway.
Canonical Wnt pathway {#s1_2}
---------------------
The β-catenin protein is the main signal transducer in the canonical Wnt pathway (Figure [1](#F1){ref-type="fig"}). β-catenin is an Armadillo multifunctional protein that can play different roles in the cell, combining one as the crucial transcriptional activator mediating Wnt signal transduction, and another as a structural protein at cell-cell adhesion junctions \[[@R32]\].
{#F1}
In the canonical Wnt pathway, the absence of Wnt ligands determines the rapid phosphorylation of the cytosolic β-catenin protein by a cytoplasmic destruction complex. This complex consists of the scaffolding protein axin, adenomatous polyposis coli (APC) tumour suppressor protein, glycogen synthase kinase 3β (GSK3β), directly responsible for β-catenin phosphorylation at key amino-terminal Ser33, Ser37 and Thr41 residues, and casein kinase 1 (CK1), which phosphorylates β-catenin at the amino acid Ser45 \[[@R22], [@R33]\]. The post-translational phosphorylation of the protein leads to its ubiquitination and subsequent proteosomal degradation thus preventing its translocation in the nucleus, where, as a consequence, the binding of Groucho to T cell factor/lymphoid enhancer-binding factor (Tcf/Lef) blocks the transcription of Wnt responsive genes \[[@R34], [@R35]\] (Figure [1](#F1){ref-type="fig"}).
When Wnt ligands bind to Fzd receptor and LRP5/6 co-receptors, the Wnt signaling is activated. The binding of Wnt recruits the cytoplasmic protein dishevelled (Dvl) to the cell membrane, thus sequestering the rate-limiting components axin and GSK3β from the degradation complex and determining the destruction of the complex \[[@R36], [@R37]\]. The hypo-phosphorylated/stabilized form of cytosolic β-catenin accumulates and translocates to the nucleus, where Wnt responsive genes transcription is activated \[[@R22], [@R34]\] (Figure [1](#F1){ref-type="fig"}). Multiple target genes include WNT1-inducible signaling-pathway protein 1 (WISP1), cell cycle-related proteins (c-Myc, cyclin D1) and the negative regulator Axin2 \[[@R12]\].
Wnt/β-catenin in the heart {#s1_3}
--------------------------
### Wnt signaling in heart development {#s1_3_1}
Heart development and morphogenesis are meticulously controlled by the relationship of several signaling pathways that cooperate to provide cell fate specification. The rigorous expression of Wnt growth factors, together with several families of secreted factors including TGFbs (Transforming Growth Factor-Betas), FGFs (Fibroblast Growth Factors), BMPs (bone morphogenetic proteins), Notch and Hippo proteins, is essentially responsible for heart processes, including myocardial specification and proliferation, cardiac septation, chamber and cardiac valve formation, as well as endothelial and vascular smooth muscle cell proliferation \[[@R1], [@R11], [@R38]--[@R40]\].
Knocking out β-catenin results in mouse embryonic defects at gastrulation, not allowing the investigation of a possible Wnt function during cardiogenesis \[[@R41]\]. Additional developmental studies using inducible promoters to activate or repress Wnt/β-catenin signaling in zebrafish embryos at different times of development and studies in mouse and human embryonic stem cells demonstrate opposite and temporally distinct effects for Wnt/β-catenin signaling during early cardiogenesis. Wnt/β-catenin signaling at pregastrula stages appears to be required for subsequent heart precursor specification, whereas this pathway inhibits heart formation during gastrulation. This suggests that cells destined to become cardiomyocytes are redirected by Wnt signaling towards an alternate mesoderm fate \[[@R42]--[@R44]\]. On the other hand, premature suppression of the Wnt signaling prevents the proper formation of endocardial cushions and heart valves, and can also result in malformation of the outflow tract, including lack of septation of the large arterial trunks, transposition of the great arteries or double outlet right ventricle \[[@R45], [@R46]\]. An upregulation of specific Wnt ligands has been observed during differentiation of embryonic cardiomyocytes into Purkinje fibres, suggesting a regulative role of the Wnt signaling in the proper development of the cardiac conduction system \[[@R11]\].
A delicate balance between canonical and noncanonical Wnt pathways has been implicated in cardiac precursor differentiation, as well as in epicardium formation and differentiation. The canonical Wnt signaling appears to retain the cardiac precursors in a proliferative and precursor state \[[@R43]\], whereas the noncanonical signaling promotes their differentiation \[[@R47], [@R48]\]. In the epicardium organization, the Wnt/β-catenin signaling is essential in pro-epicardial and epicardial cell expansion and Wnt/PCP signaling in epicardial-derived cell differentiation \[[@R49]\].
The Wnt/β-catenin signaling also controls cardiomyocyte proliferation \[[@R50]\]. *In vitro* treatment of neonatal and adult rat cardiomyocytes with BIO, a pharmacological agent inhibiting GSK3, results in increased β-catenin activity and promotes cell proliferation and mitosis. Thus, the activation of canonical Wnt signaling participates in controlling the proliferative capability of the differentiated cardiomyocytes by promoting maintenance of stem cell properties \[[@R51]\]. In embryonic mouse hearts, the inactivation of Hippo, a cardiomyocyte growth and organ-size control pathway, leads to overgrown hearts due to increased stimulation of cardiomyocyte proliferation. In such hearts β-catenin and Wnt target genes are overexpressed, indicating that Wnt/β-catenin signaling might regulate pro-growth genes transcription, under an accurate control of Hippo pathway \[[@R52]\].
Wnt/β-catenin reporter mice unveiled the spatial and temporal distribution of the endogenous status of β-catenin activation in several tissues. Transgenic animals exhibit an evident reporter gene activity in the cardiac region, where it persists up to E10.5, whereas the Wnt/β-catenin pathway appears not active in the adult heart \[[@R53]\]. Therefore, while the key role of Wnt/β-catenin signaling in several processes of early stages of cardiac development has been well documented, its activation in the adult heart is mainly associated with pathological conditions.
Wnt/β-catenin signaling in cardiac disease and repair {#s1_4}
-----------------------------------------------------
The adult mammalian heart includes interacting populations of cells, predominantly myocytes, fibroblasts, smooth muscle cells, endothelial and epicardial cells, all arranged in a precise three-dimensional network. Under pathological conditions, they can interact with each other directly or indirectly, through the expression of growth factors and cytokines, to affect a physiological cardiac response.
In terminally differentiated cardiomyocytes, the canonical Wnt signaling is normally quiescent \[[@R53]\], but it becomes reactivated upon pathological stress, including hypertrophy. Cardiac hypertrophy is characterized by an increase in cell size and upregulation of fetal-gene expression \[[@R54]\]. It can arise from physiological myocyte growth or, on the other hand, from pathological cardiac remodelling induced by cardiovascular diseases, often resulting in fibrosis substitution, diastolic dysfunction, and, eventually, heart failure \[[@R55]--[@R57]\]. Suppression of the signaling by overexpressing GSK3β, the key molecule regulating canonical Wnt transduction, attenuates the stress-induced hypertrophic response of ventricular cardiomyocytes in transgenic mice and leads to the development of smaller heart with markedly depressed contractility \[[@R58]\]. Both myocyte-specific deletion of β-catenin and lack of the Dvl-1 protein gene, responsible for the interruption of the Wnt pathway, also result in a significantly decreased hypertrophic response to pressure overload in transgenic mice, when compared with *wild-type* littermates \[[@R59], [@R60]\]. On the contrary, inhibition of GSK3β or overexpression of Dvl-1leads to augmented hypertrophy, severe systolic and diastolic dysfunction and progressive heart failure, in a pressure overload rat model and in transgenic mice, respectively \[[@R56], [@R61]\].
Many studies have emphasized an important role for the canonical Wnt signaling in cardiac fibrogenesis \[[@R49], [@R62], [@R63]\]. Fibrosis deposition can result from both cardiac repair post myocardial infarction and cardiac diseases, and it generally impedes electrical wave propagation, potentially causing arrhythmias \[[@R64], [@R65]\]. The aberrant activation of Wnt signaling by overexpression of Wnt ligands or nuclear accumulation of β-catenin in different experimental models, suggests that the canonical Wnt signaling alone is sufficient to promote the expression of a fibrogenic program in fibroblasts of several tissues \[[@R66], [@R67]\]. In epicardial cells of post myocardial infarcted heart, Wnt signaling is enhanced and promotes epithelial--mesenchymal transition into fibroblasts \[[@R49]\]. Accordingly, in paediatric heart allographs with diastolic dysfunction and severe epicardial fibrosis involving either epicardial surface or underlying adipose tissue or both, immunohistochemical investigation revealed nuclear accumulation of β-catenin and TCF4 (T-cell factor 4, the main Wnt pathway effector) in fibroblasts, consistent with the activation of Wnt signaling in these cells \[[@R63]\].
Moreover, Wnt signaling has been suggested to inhibit the apoptotic process involved in the transition from hypertrophy to heart failure. Indeed, in heart from patients who have developed heart failure, Akt activation and consequent GSK3β inhibition seem to protect cardiac cells from apoptosis \[[@R68]\].
Wnt1, a major element of the early Wnt/β-catenin signaling, is a specific and potent inducer of connexin-43 expression in cardiomyocytes. This effect results in enhanced accumulation of connexin-43 protein and formation of functional gap junction channels, supporting the hypothesis that dysregulated signaling contributes to altered impulse propagation and arrhythmia in the myopathic heart \[[@R69]\].
In conclusion, the Wnt/β-catenin signaling is a very complex system involved in a variety of cardiac pathologies with quite varied effects, according with the cell system and timing of intervention. For this reason, it can offer possibilities for new insights into disease pathogenesis and identification of new therapeutic targets.
Arrhythmogenic cardiomyopathy {#s1_5}
-----------------------------
Arrhythmogenic cardiomyopathy (AC) is a rare, inherited cardiac disease characterized by ventricular arrhythmias, right ventricular dysfunction and sudden cardiac death \[[@R70], [@R71]\]. It has a prevalence between 1:2000 and 1:5000 and accounts for up to 10% of deaths from undiagnosed cardiac disease in the \< 65 age group \[[@R72], [@R73]\]. Clinical diagnosis of AC is often difficult because of the non-specific nature of the disease features and the broad spectrum of phenotypic expressions, ranging from concealed to severe forms.
The early "concealed" phase is characterized by propensity toward ventricular tachyarrhythmia in the setting of well-preserved structural morphology and ventricular function. As the disease progresses, however, myocyte loss, inflammation, and accumulation of fibrofatty scar tissue become evident \[[@R74]\]. The typical pathological profile consists of the progressive replacement of the myocardium by fatty or fibrofatty tissue, starting from the epicardium or midmyocardium and then extending to became transmural. The infiltration may be diffuse or regional and is located at the angles of the so-called "triangle of dysplasia", including the inferior, apical and infundibular walls \[[@R75]\]. Recent observations in genotyped AC patients revealed a characteristic pattern of cardiac disease that involves the epicardial surface of the basal right ventricle and lateral left ventricle, displacing the right ventricular apex \[[@R76]\].
A scoring system to establish the diagnosis of AC has been developed on the basis of the fulfilment of major and minor criteria encompassing structural, histological, electrocardiographic, arrhythmic and genetic features of the disease \[[@R77], [@R78]\].
AC is an inherited cardiomyopathy characterized by incomplete penetrance and variable expressivity and it is usually transmitted as an autosomal dominant trait, even though recessive forms have been reported \[[@R15]\]. Since 1994, when the first locus was mapped \[[@R79]\], 13 disease genes have been identified \[[@R80]\] (Table [1](#T1){ref-type="table"}). One or more disease-causing mutations are detected in about 60% of the patients \[[@R81]\] and most of them affect genes encoding for mechanical junction proteins, including desmoplakin (DSP) \[[@R82]\], plakophilin-2 (PKP2) \[[@R83]\], desmoglein-2 (DSG2) \[[@R84]\], desmocollin-2 (DSC2) \[[@R85]\], plakoglobin (JUP) \[[@R86]\] and alpha-T catenin (CTNNA3) \[[@R87]\]. In cardiomyocytes, mechanical and electrochemical coupling between adjacent cells relies on complex entities called intercalated discs (IDs), which in turn comprise different kinds of junctional apparatuses, namely desmosomes, gap junctions and the recently described *area composita* (this latter found only in cardiac muscle) \[[@R88]\] (Figure [2](#F2){ref-type="fig"}). The main components of cardiac desmosomes are two types of cadherins, desmocollin-2 and desmoglein-2, two armadillo proteins, plakoglobin and plakophilin-2, and desmoplakin. Function-wise, desmosomal cadherins are responsible for the cell-cell adhesion *via* their extracellular domains, while their cytoplasmic domains interact with the armadillo proteins. These latter interact with desmoplakin, which in turn binds desmin intermediate filaments, thereby creating a continuous link between the cytoskeletons of adjacent cardiomyocytes. The *area composita* comprises some of the desmosomal proteins described above as well as proteins normally found in *adherens* junctions. Amongst these, N-cadherin interacts with the adjacent cell *via* its extracellular domain and, on the cytoplasmic side, is indirectly linked to α-catenin *via* β-catenin or plakoglobin. In turn, α-catenin binds the F-actin cytoskeleton through vinculin \[[@R88]\].
###### Human genes associated with AC
GENE CHOMOSOME LOCUS ENCODED PROTEIN MUTATION FREQUENCE REFERENCE(S)
------------------------- ----------------- --------------------------------- ---------------------------- ------------------------------ --------------------------------------
INTERCALATED DISC GENES **DSP** 6q24 Desmoplakin 1--16% \[[@R82], [@R84], [@R117]--[@R119]\]
**PKP2** 12p11 Plakophilin-2 10--45% \[[@R83], [@R118]--[@R121]\]
**DSG2** 18q12 Desmoglein-2 7--12% \[[@R84], [@R118], [@R122]\]
**DSC2** 18q12 Desmocollin-2 1--5% \[[@R85], [@R121]--[@R123]\]
**JUP** 17q21 Plakoglobin 1% \[[@R86], [@R118]\]
**TMEM43** 3p25 Luma rare \[[@R124]\]
**DES** 2q35 Desmin rare \[[@R125], [@R126]\]
**CTNNA3** 10q21 αT-catenin 2% \[[@R87]\]
OTHER GENES **RYR2** 1q42-q43 Cardiac ryanodine receptor rare \[[@R127]\]
**TGFb3** 14q23-q24 Transcription growth factor β-3 rare \[[@R128]\]
**TTN** 2q31 Titin rare \[[@R129]\]
**LMNA** 1q21.2-q21.3 Lamin A/C rare \[[@R130]\]
**PLN** 6q22.1 Phospholamban rare \[[@R131]\]
![Models of AC pathogenesis\
AC gene mutations involve *area composita* and desmosome proteins (asterisks) and lead to abnormalities in intercalated disc (ID) consisting of cell adhesion defects, gap junction and ion channel remodelling. (**A**) Mutations in ID proteins can result also in the perturbation of the Wnt/β-catenin signaling. (**B**) Wnt/β-catenin signaling is suppressed in concomitance with the activation of Hippo pathway \[[@R93]\]. The impaired localization of phosphorylated protein kinase C-α to the perturbed IDs is associated with activation of neurofibromin 2 (NF2) and results in cascade phosphorylation leading to increased phospho-Yes-associated protein (pYAP) and its retention in the cytoplasm. pYAP interacts with β-catenin (β-cat), which is not able to translocate into the nucleus and, as a consequence, the expression of the effectors of both Hippo and canonical Wnt pathways (TEAD and Tcf/Lef, respectively) is suppressed and adipogenesis is enhanced (B1). Other experimental data reveal that loss of plakoglobin leads to the activation of AKT and the subsequent inhibition of glycogen synthase kinase 3β (GSK3β) resulting in the stabilization of β-cat and its translocation in the nucleus. Here, β-cat interacts with Tcf/Lef causing the increase of the expression of c-myc, c-fos, and cyclin D1, as well as cardiac hypertrophy \[[@R98]\] (B2). The disruption of junction integrity can result in increased presence of β-catenin at IDs without the involvement of Wnt/β-catenin signaling. However, increased expression of transforming growth factor β-1 (TGFβ1), phospho-SMAD2 (pSMAD2), and Pai1 is consistent with the activation of TGFβ pathway responsible for the progressive fibrosis in AC hearts \[[@R99]\] (B3) (DSC2, desmocollin-2; DSG2, desmoglein-2; DSP, desmoplakin; PKP2, plakophilin-2; DES, desmin; N-cad, N-cadherin; αT-cat, αT-catenin; αE-cat, αE-catenin; Tcf/Lef, T cell factor/lymphoid enhancer-binding factor; TEAD, SV40 transcriptional enhancer factor domain).](oncotarget-08-60640-g002){#F2}
As mentioned above, AC is often caused by mutations in genes involved in cell-cell adhesion. Indeed, a recent report has demonstrated that siRNA-mediated silencing of the PKP2 gene in neonatal rat ventricular myocytes subjected to mechanical stress led to a much greater destabilization of the monolayer structure, thereby indicating weakened cell-cell junctions \[[@R89]\]. However, other recent evidences suggest that the pathogenetic role of ID genes\' mutations in AC does not originate exclusively from the mechanical alteration of the junctions but also from alterations in the signal transduction processes. For example, mutations in several of the ID genes have been associated to perturbation of the Wnt/β-catenin signaling (see below for specific references).
Wnt signaling in AC: conflicting evidences {#s1_6}
------------------------------------------
In the last years, studies on zebrafish and mouse models and on cultured cell lines, such as induced pluripotent stem cells (iPSCs) from affected patients, revealed the involvement of the canonical Wnt signaling in the molecular pathogenesis of AC (Table [2](#T2){ref-type="table"}).
###### Studies of the Wnt/β-catenin signaling involvement in AC
--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
HUMAN AND ANIMAL EXPERIMENTAL MODEL(S) SIGNALING COMPONENTS PHYSIOLOGICAL EFFECT CONCLUSIONS REFERENCE(S)
------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- ------------------------------------------------------------------------------- --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- ------------------------------------------------------------------------------------------------------------------------------------- ----------------------
DSP-deficient HL-1 cellsTg mice with cardiac-restricted DSP deletion (DSP^−/−^; DSP^+/−^) Wnt: ↓cyclinD1, c-MycAdipogenesis: ↑PPARγ, CEBPα, adiponectin Fat droplet accumulation in DSP-deficient cellsHigh lethality in DSP^−/−^ embryosVentricular dilatation and disfunction, fibrofatty replacement of myocytes in DSP^−/−^and DSP^+/−^ adult mice Nuclear PG translocation**Suppression of canonical Wnt/β-catenin signaling**Transcriptional switch to adipogenesis and fibrogenesis \[[@R14]\]
Tg mice overexpressing cardiac truncated PG In cardiac progenitor cells\ Increased fibroadiposis, cardiac dysfunction, and premature deathEnhanced adipogenesis in cardiac progenitor cells **Suppression of canonical Wnt signaling**Induction of pro-adipogenic gene expression \[[@R92]\]
Wnt: ↓CTGFAdipogenesis: ↑KLF15, IGFBP5
Conditional Tg mice with cardiac-restricted JUP deletion Wnt: ↑active β-catenin, c-Myc, c-Fos; ↓GSK3β, ↑AKT Progressive loss of cardiac myocytes, extensive inflammatory infiltration, fibrous tissue replacement, and cardiac dysfunction **Activation of canonical Wnt signaling** \[[@R98]\]
Tg mice with cardiac-restricted JUP deletion Wnt: ↑β-catenin, but unchanged cyclinD1, c-MycTGFβ: ↑p-Smad2, TGFβ1, BNP, ANP Cardiac fibrosis and dysfunction, ventricular arrhythmiasAbsence of desmosomes to the intercalated discs of cardiomyocytes **Wnt signaling not altered**Increase of TGFβ signaling \[[@R99]\]
iPS-CMs from patient carryingthe heterozygous PKP2 p.K672Rfs\*12mutation;iPS-CMs from patient carryingthe homozygous PKP2 p.G828Gmutation Wnt: ↓β-cateninUpon lipogenic stimulation:Adipogenesis: ↑PPARα, PPARγ, FABP4 Intracellular lipid accumulationAccelerated pathogenesis upon lipogenic stimulation Nuclear PG translocation**Reduction of Wnt signaling**Induction of lipogenesis and apoptosis \[[@R102], [@R104]\]
iPS-CMs from patient carryingthe homozygous PKP2 p.A324fs335\*mutation;iPS-CMs from patient carrying the homozygous PKP2 p.T50Sfs60\* mutation Wnt: ↓β-cateninAdipogenesis: ↑PPARγ Intracellular lipid accumulationDesmosomal distortionAccelerated pathogenesis upon adipogenic stimulationReversion of intracellular lipid accumulation upon treatment with a GSK3β inhibitor (BIO) **Reduction of Wnt signaling**Pro-adipogenic potentialActivation of Wnt pathway rescues the lipid accumulation \[[@R103]\]
Tg mice with cardiac restricted DSP deletion (DSP^+/−^)\[[@R14]\];Tg mice overexpressing cardiac truncated PG \[[@R92]\];PKP2 knockdown HL-1 cellsAC patient myocardial samples Hippo: ↓pPKC-α, ↑NF2, ↑pYAPWnt: ↑pβ-catenin Molecular remodeling of IDsNo discernible localization of pPKC-α to IDsmembrane localization of pYAPCytoplasmatic binding of pYAP, pβ-catenin and JUP Activation of Hippo pathway**Suppression of canonical Wnt signaling**Enhanced adipogenesis \[[@R93]\]
myocardial samples from Boxer dogs affected with AC Wnt: ↑β-catenin β-catenin retention to the endoplasmic reticulum of cardiomyocytes **Perturbation of canonical Wnt signaling** \[[@R100]\]
Tg zebrafish expressing cardiac truncated PG (p.W680Gfs\*11 mutation)Neonatal rat ventricular myocytes expressing cardiac truncated PG (p.W680Gfs\*11 mutation)iPS-CMs from patient carryingthe heterozygous PKP2 p.K672Rfs\*12 mutationiPS-CMs from patient carryingthe heterozygous PKP2 p.Q617\* mutationAC patient myocardial samples / Reduction in I~Na~and I~K1~ current densities, interruptions in cell boundaries, structural disarray in fishes and in neonatal rat cardiomyocytes;Abnormal subcellular distribution of PG, connexin-43, Nav1.5 and SAP97 in cultured cells linesReversion of AC features upon treatment with SB216763 (a GSK3β inhibitor) Activation of Wnt pathway (by SB216763) rescues the AC pathobiological features \[[@R96]\]
Tg mice expressing truncated Dsg2Tg mice expressing cardiac truncated PG (JUP^2157del2^)AC patient myocardial samples Neonatal rat ventricular myocytes expressing cardiac truncated PG or PKP2PKP2-knockdown HL-1 cells Wnt: ↑GSK3βUpon SB216763 treatment:Wnt: ↓GSK3β, but unchanged pGSK3β GSK3β ID redistribution in AC cardiomyocytes and in AC human myocardium, but not in normal cardiomyocytes Activation of Wnt pathway by pharmacologic GSK3β inhibition (using SB216763) improves cardiomyopathy \[[@R97]\]
Conditional Tg mice with heterozygous DSP deletion in fibroadipocyte progenitor cells (FAPs) In FAPs:\ Mild cardiac dysfunction and increased cardiac fibroadipogenesis **Suppression of canonical Wnt signaling** in FAPsEnhanced adipogenesis \[[@R94]\]
Wnt: ↓cyclinD1, CTGF, Serpine1Adipogenesis: ↑FABP4, CEBPα, PPARγ
Tg mice overexpressing cardiac mutant DSP^R2834H^ exposed to endurance exercise Wnt: ↓β-catenin, pGSK3β, pAKT1 Progression of AC phenotype (right ventricle dilatation and wall thinning, myocyte disarray and fibrofatty infiltration) Exercise-induced **suppression of the Wnt signaling** \[[@R95]\]
--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
The progressive replacement of the cardiomyocytes by adipocytes and fibrosis is the pathological hallmark of AC, which contributes to both cardiac dysfunction and arrhythmias. In this regard, in addition to the inactivation of junctional mechanical functions, which is supposed to lead to myocyte death under physical stress, the alteration of canonical Wnt/β-catenin signaling could act as a major transcriptional switch regulator of myogenesis towards adipogenesis \[[@R90]\].
Several studies from Marian\'s group demonstrated the suppression of Wnt/β-catenin/Tcf/Lef pathway in AC due to nuclear translocation of plakoglobin \[[@R14], [@R91], [@R92]\]. Plakoglobin, also known as γ-catenin, is highly homologous to β-catenin, the effector of the canonical Wnt signaling, and shares with β-catenin common protein partners thus fulfilling some of the same functions \[[@R32]\]. Suppression of desmoplakin expression in cultured atrial myocytes and in mouse hearts causes the impairment of junction assembly and frees plakoglobin from the cell membrane. Consequently, plakoglobin translocates into the nucleus where it interferes with β-catenin/Tcf transcriptional activity leading to an adipogenic switch and explaining enhanced adipogenesis, fibrogenesis, and myocyte apoptosis observed in human AC \[[@R14]\]. Thereafter, through a series of genetic-fate mapping experiments, the same group indicated that cardiac progenitor cells of the embryonic second heart field represent a cell source for excess adipocytes observed in the heart of Dsp^+/−^ mouse model for AC \[[@R91]\]. Direct evidence for the essential function of nuclear plakoglobin in repressing Wnt signaling in cardiac progenitor cells and inducing differentiation to adipocytes was demonstrated in hearts from transgenic mice overexpressing cardiac truncated plakoglobin \[[@R92]\]. Furthermore, the activation of Hippo pathway, another regulator of cellular differentiation and proliferation, was identified as the mechanism for suppression of the canonical Wnt signaling in AC. Pathogenic activation of the Hippo kinase cascade, resulting from perturbations in ID attributable to AC gene mutations, leads to phosphorylation and cytoplasmic retention of the transcriptional coactivator Yes-associated protein (YAP), that can sequester β-catenin and plakoglobin, thus resulting in a reduction of both β-catenin/TCF and YAP/TEAD transcriptional activities and enhanced heart adipogenesis \[[@R93]\] (Figure [2](#F2){ref-type="fig"}).
Very recently, a conditional mouse with cardiac deletion of Dsp was crossed to the PDRFRA (platelet-derived growth factor receptor-α)-Cre deleter mice in order to obtain a novel mouse model with traceable cardiac fibroadipocyte progenitors (FAPs) lacking of Dsp. These mice showed increased fibroadipogenesis in the heart and mild cardiac dysfunction, if compared with control mice. FAPs isolated from the same mouse hearts showed enhanced differentiation to adipocytes through a Wnt-dependent mechanism \[[@R94]\].
Finally, Wnt pathway perturbation, due to reduced AKT and GSK3β activation, has been recently associated to exercise-induced acceleration of AC phenotype in mice overexpressing a mutant form of desmoplakin and exposed to a daily running \[[@R95]\]. The identification of a Wnt pathway activator as the molecule that can rescue the features of the AC phenotype both in a zebrafish and in additional mouse models \[[@R96], [@R97]\] strongly supports the association between AC and suppression of this signaling pathway.
On the other hand, conflicting evidences have been reported about the canonical Wnt signaling involvement in AC pathogenesis. In a conditional mouse model with cardiac-restricted JUP deletion showing cardiac alterations and dysfunctions similar to those of AC patients, the remodeling of the IDs causes an increased β-catenin stabilization associated with activated AKT and consequent inhibition of GSK3β. In these animals, a Wnt/β-catenin signaling activation was suggested to contribute to the AC cardiac phenotype \[[@R98]\] (Figure [2](#F2){ref-type="fig"}). On the contrary, in a different mouse model with stable cardiac-restricted JUP deletion, the increase of β-catenin levels at IDs of cardiomyocytes does not affect the canonical Wnt signaling; the TGFβ pathway, mainly involved in the regulation of myocyte cell death, including both apoptosis and necrosis, and influencing cardiac fibrosis and hypertrophy, results upregulated during the early stages of the disease in mutant hearts \[[@R99]\] (Figure [2](#F2){ref-type="fig"}). A 50% increase in levels of β-catenin protein, that resulted mislocalized to the endoplasmic reticulum, is also observed in heart samples from Boxer dogs affected with AC, a natural animal model for the disease, compared to non-AC dogs, suggestive of a perturbation of the canonical Wnt pathway \[[@R100]\].
Up to now, several *in vitro* studies examined the effect of point mutations in the PKP2 gene using iPSC-derived cardiomyocytes (iPS-CMs) from patient affected with AC \[[@R101]--[@R104]\]. In all of the cases, mutant iPS-CMs appear unable to reproduce the pathologic phenotype in standard cardiogenic conditions, but they are inclined to lipid accumulation following different treatment with adipogenic stimuli, displaying a functional pro-adipogenic state. Electron microscopy analysis showed larger AC iPS-CMs containing lipid droplets compared with controls, thus suggesting an increased adipogenic potential in these cells. The adipogenic stimuli on the mutant cardiomyocytes seemed to be prevented by GSK3β inhibitor, thus supporting the possible role of the canonical Wnt pathway in AC pathogenesis \[[@R103]\].
Altogether, these findings suggest a central role of the Wnt/β-catenin signaling in the disease pathogenesis and appear to expand the cellular spectrum of AC, commonly recognized as a disease of cardiac myocytes, towards nonmyocyte cells in the heart. However, the direct causal relationship between mutant junctional proteins and perturbed Wnt pathway remains poorly understood.
Development of Wnt activators as therapeutics for AC {#s1_7}
----------------------------------------------------
The intricate involvement of Wnt signaling in so many biological processes and disease conditions made this pathway an obvious but yet difficult target for therapeutic intervention. Numerous fields of study have investigated both natural and synthetic compounds able to induce or inhibit Wnt at multiple stages within the pathway. However, despite the evidence of successful methods to rapidly screen thousands of compounds to identify Wnt pathway modulators, unfortunately these molecules cannot be translated yet into the clinical practice \[[@R105]\].
To discover potential chemical modifiers in AC, a zebrafish model with cardiac-specific expression of a truncated form of plakoglobin and manifesting a fully penetrant cardiomyopathy, has been generated and used to screen a library of bioactive compounds for disease modifiers \[[@R96]\] (Table [2](#T2){ref-type="table"}). SB216763 molecule has been identified as a compound able to prevent heart failure and to reduce mortality in the fish model. It also rapidly reverses the highly abnormal action potential and corrected the marked decreases in I~Na~ and I~K1~ exhibited by mutant fish myocytes. Interestingly, SB216763 is annotated as an inhibitor of GSK3β thereby increasing canonical Wnt signaling \[[@R106]\]. Additional experiments performed in neonatal rat ventricular myocytes expressing the same mutant plakoglobin and in iPS-CMs from two AC probands with PKP2 mutations supported the SB216763 efficacy in the reversion or prevention of AC pathobiological features. In particular, the distribution of connexin-43 appears to be restored to normal in truncated plakoglobin--expressing cells exposed to SB216763, consistent with previous studies showing the connexin-43 up-regulation by activation of the canonical Wnt pathway \[[@R96]\].
More recently, the SB216763 molecule has been reported to prevent myocyte injury and cardiac dysfunction *in vivo* in two murine models of AC at baseline and in response to exercise (Table [2](#T2){ref-type="table"}). GSK3β inhibition improves left ventricle function and survival in sedentary and exercised mice and normalizes ID protein distribution in mutant mouse cardiomyocytes. Redistribution of GSK3β to cardiac IDs appeared to be specific for AC, as it was not evident in myocardia from patients with dilated, hypertrophic or ischemic cardiomyopathy, cardiac sarcoidosis, or giant cell myocarditis. Cytoplasmic distribution of GSK3β was restored in all *in vivo* and *in vitro* models treated with SB216763 Collectively, these data suggest a key role for aberrant GSK3β signaling in the final common disease pathway in AC \[[@R97]\]. However, how localization and activity of GSK3β relate to each other and how the pharmacologic inhibition of this kinase could affect its localization remain to be clarified.
The identification of a molecular activator of the canonical Wnt signaling as a promising pharmacological agent for AC open the door for new mechanism-based therapeutic strategies for patients affected with this disease. Such advances would provide a welcome alternative to current reliance on traditional therapeutic tools (life-style modification, devices, anti-arrhythmic drugs, endocardial and epicardial catheter ablation, implantable cardioverter defibrillator and transplantation) targeting life-threatening ventricular arrhythmias and congestive heart once they occur.
MicroRNAs: potential linkers between Wnt/β-catenin signaling and AC? {#s1_8}
--------------------------------------------------------------------
Tissue development, physiology, and functions are strictly regulated by multiple factors. Among them, microRNAs (miRNAs) have been recently described as 22--25 nucleotides long noncoding RNAs which regulate gene expression at post-transcriptional level. This is achieved through imperfect base-pairing with complementary sequences in the 3′UTR of their target mRNA, leading to translational repression or transcript degradation \[[@R107]\]. A single miRNA can regulate multiple genes and, in Human, among the 1,800 detected miRNAs, 800 have been found expressed in the heart \[[@R108]\].
Increasing evidence indicates that miRNAs participate in signaling networks, such as the canonical Wnt/β-catenin signaling, and in promoting or inhibiting the progression of many human diseases, including cardiomyopathies. In the last years, several studies described the correlation between miRNA expression level and the regulation of the Wnt/β-catenin signaling, mostly during cardiac differentiation. Overexpression of miR-499 in rat bone marrow-derived mesenchymal stem cells is sufficient to induce cardiac differentiation, as indicated by increased expression of cardiac specific markers Nkx2.5, GATA4, MEF2C, and cTnl, through Wnt/β-catenin signaling pathway \[[@R109]\]. In particular, miR-499 drives the activation of β-catenin, as shown by the decreased ratio of phosphorylated/dephosphorylated (Ser33/37/Thr41) β-catenin, but the detailed mechanisms involving the Wnt pathway haven\'t been identified yet \[[@R109]\].
Among the most expressed miRNAs in the heart, miR-1 is known to promote cardiomyocyte differentiation. Lu and colleagues showed that this miRNA suppresses the canonical Wnt/β-catenin signaling in human iPSCs by targeting the receptor Fzd7 and antagonizing the ligand Wnt3A. As a consequence, cardiomyocyte differentiation is promoted and endothelial cell commitment from multipotent cardiovascular progenitors is suppressed \[[@R110]\].
Moreover, several papers reported miR-19b as a negative regulator of the Wnt/β-catenin signaling. Overexpression of miR-19b has been shown to cause aberrant heart tube and chamber development and to regulate the contractile function of the heart in zebrafish embryos by inhibiting the Wnt signaling pathway directly targeting Ctnnb1β \[[@R111]\]. Furthermore, inhibition of GSK3β leads to β-catenin accumulation, together with a partial rescue of the cardiac phenotype \[[@R111]\]. In multipotent murine P19 cells, miR-19b overexpression is responsible for cellular proliferation and differentiation into myocardial cells, and apoptosis inhibition with the parallel decrease of Wnt/β-catenin signaling activation \[[@R112]\]. In this study, luciferase assays show the direct link between the overexpression of miR-19b and the reduction of the expression of WNT1 \[[@R112]\]. Accordingly, miR-19b knock-down results in the inhibition of proliferation and apoptosis of C19 cells, together with the increase of Wnt1 protein levels/β-catenin signaling expression levels \[[@R113]\].
Less frequently, miRNAs have been described to modulate Wnt/β-catenin signaling also during cardiac diseases. Gain and loss of function studies show that miR-145 overexpression leads to an increase in Wnt/β-catenin signaling activity and the downregulation of Dab2, a TGFβ receptor adaptor protein required for cardiac protein expression in mesenchymal stem cells, in response to TGFβ1 \[[@R114]\]. Interestingly, miR-145 is found downregulated in acute myocardial infarction, in association with an increase in Dab2 expression level in cardiomyocytes in the infarct border zone. These results suggest that the Wnt/β-catenin signaling is downregulated in cardiomyocytes after acute myocardial infarction, but further studies are required to clarify the underlying associated mechanism \[[@R114]\].
Interestingly, the expression profile of miRNAs recently performed by qRT-PCR in 24 heart samples from AC patients revealed that miR-21-5p and miR-135b, which are correlated with Wnt and Hippo pathway in cancer, are significantly associated with both the myocardium adiposis and fibrosis, and might to be useful as therapeutic targets \[[@R115]\]. On the other hand, the pathogenic downregulation of miR-184 has been identified in PKP2-deficient cell lines and mouse models of AC \[[@R116]\]. Overexpression of miR-184 had no discernible effects on the transcript levels of selected canonical Wnt signaling targets such as CTNNB1 and TCF7L2 in HL-1 cells and, *vice versa*, the activation of the canonical Wnt signaling does not affect miR-184 levels in the *in vitro* model. The authors show the relationship between the downregulation of miR-184 and perturbations in E2F1 pathway, suggesting a novel mechanism concurring to AC development \[[@R116]\].
Taken together, all these data underline the crucial function of miRNAs in the regulation of theWnt/β-catenin signaling and are suggestive of a potential role of these small RNAs in the pathogenesis and the phenotypic variability of cardiac diseases such as AC. Further efforts will be needed to find out whether miRNAs represent the link between the perturbations at the IDs and the alteration of the Wnt/β-catenin signaling activity observed in AC.
Concluding remarks {#s1_9}
------------------
Increasing evidence from experimental studies defines the canonical Wnt/β-catenin signaling as a critical player in the regulation of cardiac function and dysfunctions. To date, the function of this pathway in cardiac hypertrophy, fibrosis and arrhythmias have been well established, however further investigations are required to unravel its role in other pathological conditions. Controversial results have been reported about the involvement of Wnt/β-catenin pathway in the development of AC, an inherited cardiomyopathy determined by mutations mostly in genes encoding for ID proteins and typically characterised by the fibrofatty replacement of the myocardium. Current investigations in zebrafish and mammalian models of AC mainly indicate that suppression of the canonical Wnt/β-catenin signaling promotes cardiac remodelling in affected hearts. The dual function of β-catenin and plakoglobin both in cadherin-mediated adhesion and in gene transcription determines a highly regulated and complex protein network seemingly able to integrate disease signals from the membrane with cellular programs of target gene transcription. However, there are still some questions to be answered, particularly, which signal molecule(s) can play a dominant role in Wnt/β-catenin downregulation upon molecular destabilization of ID junctional structures. A more comprehensive understanding of the dynamics of this signaling in AC cardiomyocytes is therefore required, also for the design of safe and effective drug molecules. Recently, a small molecule activating the Wnt pathway has been experimented and it appears to be remarkably effective in preventing or reversing selected features of the molecular AC pathology. Based on these promising results, Wnt signaling components could represent interesting therapeutic targets for novel treatment strategy in AC, hopefully in the near future. Furthermore, a deeper investigation on the cross-talk between Wnt and other cellular pathways, such as Hippo or TGFβ, may explain other molecular mechanisms leading to AC and assist the design of efficient combinatorial therapeutic protocols. In addition, the relationship between miRNAs-mediated regulation of Wnt signaling, and *vice versa*, needs to be further investigated to contribute in the understanding not only of AC pathogenesis, but also of other, more common forms of heart disease, as well as to provide the basis to develop mechanism-based therapies to treat or prevent diseases at high risk of sudden death.
**FUNDING**
This work was supported by TRANSAC, University of Padua Strategic Grant CPDA133979/13, Padua, Italy; Veneto Region Target Research, Venice; H2020-MSCA-IF-2014.
**CONFLICTS OF INTEREST**
Authors disclose any potential conflicts of interest.
| {
"pile_set_name": "PubMed Central"
} |
Introduction {#s1}
============
Plastids of higher plants, mosses, and algae contain lipoprotein particles known as plastoglobules (PGs) ([@CIT0033]). PGs are surrounded by a lipid monolayer membrane, are found in both non-photosynthetic and photosynthetic tissues, and are associated with the thylakoids in chloroplasts. The particles are filled with hydrophobic molecules, and a number of different proteins have been identified which are associated with the lipid monolayer ([@CIT0057]). Strikingly, PGs are of a highly dynamic nature, and their numbers and sizes have been shown to increase considerably in different developmental stages. For instance, they were found to accumulate in etiolated tissues, decrease upon illumination and greening, and they appear to accumulate again during senescence where they play an important role in thylakoid breakdown ([@CIT0055]; [@CIT0056]). Moreover, exposure to nitrogen starvation, drought, or light stress stimulates formation and growth of PGs ([@CIT0012]; [@CIT0016]; [@CIT0065]; [@CIT0032]).
In contrast to the thylakoid membrane, PGs are composed of neutral lipids (triacylglycerol, phytol esters, and free fatty acids), prenylquinones (α-tocopherol, plastoquinone, plastochromanol, and phylloquinone), as well as carotenoids---the last category being especially prevalent in the PGs of chromoplasts ([@CIT0057]). PGs have been purified from maize and Arabidopsis, and subjected to MS ([@CIT0037]; [@CIT0021]). All of the \~30 proteins identified to date seem to be very specific and are almost exclusively localized to PGs; only a few are thought to be recruited to PGs under particular conditions, for example enzymes involved in jasmonate biosynthesis or chlorophyll degradation ([@CIT0034]; [@CIT0038]). The most abundant PG proteins belong to the family of plastid-specific fibrillins (FBNs), which represent \~53% of the total protein content and are suggested to function as structural components ([@CIT0018]). Another large portion (\~19%) of the PG proteome is composed of members of the 'activity of BC1 complex kinase' (ABC1K) family, which may phosphorylate other PG proteins and thus regulate their activity ([@CIT0036]). Apart from these abundant protein classes, only a few other PG proteins have been investigated in detail. One of these, the enzyme tocopherol cyclase \[vitamin E deficient 1 (VTE1)\], functions in the formation of tocopherol and plastochromanol-8 ([@CIT0058]). Moreover, the phytol ester synthases 1 and 2 (PES 1 and 2), which play a role in the chlorophyll degradation pathway, have been characterized from PGs ([@CIT0034]). Another metabolic enzyme is the NADP(H) dehydrogenase C1 (NDC1), which acts as a reductase in the phylloquinone biosynthetic pathway ([@CIT0011]; [@CIT0014]). Most of the remaining PG core proteins have not yet been studied at the functional level. Interestingly, none of these proteins possesses typical transmembrane domains, as would be required for insertion into a lipid bilayer. It is therefore assumed that PG proteins insert short hydrophobic domains into the lipid monolayer or associate with the membrane via protein--protein interactions.
However, PGs may not be exclusive to higher plants. In fact, the *Chlamydomonas reinhardtii* eyespot and PGs resemble each other in appearance and protein composition ([@CIT0028]). A number of common homologous proteins have been identified, such as proteins with plastoglobulin domains and ABC1 kinases ([@CIT0046]; [@CIT0036]; [@CIT0008]). Moreover, lipid-rich droplets whose metabolite and small molecule composition is similar to that of PGs have also been observed in several cyanobacteria ([@CIT0054]; [@CIT0044]). Furthermore, homologs of plant FBNs have been found in *Synechocystis* sp. PCC 6803 (hereafter *Synechocystis*), and loss-of-function mutants are characterized by light sensitivity and accumulation of photoprotective pigments ([@CIT0005]). This suggests that PGs in Arabidopsis and lipid droplets in cyanobacteria may be evolutionarily related and could share conserved roles.
In this study, we investigated the role of a plastoglobular protein, which we termed pastoglobular protein 18 (PG18; AT4G13200), due to its molecular weight. Initially identified as UNKNOWN1 in [@CIT0037], PG18 is an as yet uncharacterized member of the core proteome of PGs. PG18 has no predicted functional domains and is one of the less abundant PG proteins ([@CIT0037]). Nevertheless, the loss-of-function mutant in Arabidopsis shows a distinctly pale green phenotype and a defect in the formation of thylakoid membranes, as well as in the accumulation of thylakoid membrane protein complexes. Additionally, functional loss of the homolog of PG18 in cyanobacteria has comparable effects, supporting the idea that its role in thylakoid biogenesis has been conserved in higher plants.
Materials and methods {#s2}
=====================
Plant material and growth conditions {#s3}
------------------------------------
The T-DNA insertion line GK-439D01 (*PG18* accession number: At4g13200) was obtained from the GABI-Kat collection, and homozygous mutants were screened with oligonucleotides given in [Supplementary Table S1](#sup1){ref-type="supplementary-material"} at *JXB* online. For complementation the coding sequence of PG18 was cloned into pK7FWG2 under the control of the 35S promoter ([@CIT0026]). The construct was introduced into *Agrobacterium tumefaciens* strain GV3101, and *pg18* mutants were transformed by floral dip ([@CIT0006]).
Unless indicated otherwise, Arabidopsis (*Arabidopsis thaliana*) wild-type (WT) Columbia ecotype and the mutants were grown on soil under normal light conditions (NL, 16/8 h light/dark, 21 °C and 120 µmol photons m^−2^ s^−1^). For *in vivo* labeling experiments, plants were grown on half-strength Murashige and Skoog (MS) medium under the same light regimes as above for 21 d. Pea plants (*Pisum sativum* L., cv. 'Arvica', Prague, Czech Republic) were grown on sand in a climate chamber under a 16 h light (220 μmol photons m^−2^ s^−1^) and 8 h dark regime at 21 °C.
*Synechocystis* strains, growth conditions, and analysis {#s4}
--------------------------------------------------------
WT and mutant *Synechocystis* cells were grown on solid or in liquid BG 11 medium ([@CIT0047]) supplemented with 5 mM glucose (unless indicated otherwise) at 30 °C under continuous illumination at a photon irradiance of 30 μmol photons m^−2^ s^−1^ of white light. Doubling times were determined after 2 d and 4 d of photoheterotrophic and photoautotrophic growth, respectively.
To generate the insertion mutant *synpg18*^−^, fragments flanking the *synpg18* gene (*sll1769*) were amplified from WT genomic DNA and the resulting fragments synpg18-up and synpg18-down were cloned into the pJET1.2 vector (Thermo Scientific). The synpg18-down fragment was inserted into pJET-synpg18-up via *Sal*I and *Eco*RI, and a kanamycin resistance cassette was subsequently inserted between the upstream and downstream fragments via *Bam*HI. Oligonucleotides are given in [Supplementary Table S1](#sup1){ref-type="supplementary-material"}. *Synechocystis* WT cells were transformed as described, and segregation was achieved by growth on increasing concentrations of kanamycin and confirmed by PCR ([@CIT0009]). Isolation of whole-cell proteins was performed as described ([@CIT0046]). Chlorophyll contents of *Synechocystis* WT and mutant cells were measured according to [@CIT0062], and *F*~v~/*F*~m~ values were determined with a FluorCam 800 MF (Photon System Instruments, Drasov, Czech Republic).
Analysis of chloroplast and cyanobacteria ultrastructure {#s5}
--------------------------------------------------------
Plant leaves were cut into small pieces (≤1×1×1 mm) in 75 mM cacodylate buffer containing 2 mM MgCl~2~ and 2.5% glutaraldehyde. After over-/normal pressure infiltration with fixation buffer and storage of the samples overnight at 4 °C, the samples were post-fixed with 1% osmium tetroxide for 1 h. This step was followed by dehydration in a graded acetone series: samples were successively incubated in 10% acetone for 15 min, 20% acetone supplemented with 1% uranyl acetate for 30 min, and in 40, 60, and 80% acetone for 20 min each. Finally, the samples were put into 100% acetone at least twice (for 5 min, then overnight). Afterwards, the plant tissue was infiltrated with Spurr's resin and polymerized at 63 °C for at least 16 h. Prior to fixation, *Synechocystis* cells were concentrated via gentle centrifugation. Chemical fixation was also carried out by resuspending the resulting pellet with 2.5% glutaraldehyde in 75 mM cacodylate buffer containing 2 mM MgCl~2~ and keeping the samples at 4 °C overnight. The post-fixation with 1% osmium tetroxide and the following dehydration, infiltration, and embedding were carried out as mentioned above for Arabidopsis.
After thin sectioning, the material was examined on a Zeiss EM 912 with an integrated OMEGA filter for TEM. The acceleration voltage was set to 80 kV and the microscope was operated in the zero-loss mode. Images were acquired using a 2k×2k slow-scan CCD camera (TRS Tröndle Restlichtverstärkersysteme, Moorenweis, Germany).
SDS--PAGE and immunoblotting {#s6}
----------------------------
Total protein extraction as well as thylakoid isolation were performed as described ([@CIT0051]; [@CIT0042]). For PG18 extraction, thylakoids were resuspended in lysis buffer (5 mM HEPES-KOH, pH 8.5 mM EDTA) and incubated on ice for 30 min. Sonicated thylakoids were mixed with salt solutions \[final concentration of 1 M NaCl, 100 mM Na~2~CO~3~, 3 mM urea, or 1% lithium dodecylsulfate (LDS), respectively\] and incubated for 30 min on ice. The samples were centrifuged for 5 min at 6000 *g* and 4 °C to separate the soluble and membrane fraction of the thylakoids. SDS--PAGE and immunoblotting were performed as described ([@CIT0051]). For PG18 antisera production, full-length purified PG18 was injected into rabbits (Biogenes). Antisera against Arabidopsis proteins were purchased as indicated (see [Supplementary Table S2](#sup1){ref-type="supplementary-material"}). ATP synthase and D1 antisera were provided by S. Greiner, TIC110 and FBPase antisera were from B. Bölter, and PsaG and OE33 antisera were from Jörg Meurer.
Analysis of protein extracts by Blue Native (BN)--PAGE {#s7}
------------------------------------------------------
BN--PAGE was performed as described previously ([@CIT0051]). Thylakoid membranes equivalent to 100 µg of protein were separated on a 6--15% acrylamide gradient. Lanes were excised, denatured, and subjected to SDS--PAGE in the second dimension. Gels were silver stained according to [@CIT0004]. *In vivo* labeling was performed according to [@CIT0040]. Twenty-one-day-old plants of the WT and mutant were incubated for 20 min and isolated proteins were subjected to BN--PAGE as described above. Proteins were detected by autoradiography.
*Agrobacterium*-mediated transient expression of fluorescent proteins in tobacco {#s8}
--------------------------------------------------------------------------------
The *Agrobacterium tumefaciens* strain AGL1 was transformed with PG18 (At4g13200) in the vector pK7FWG2 ([@CIT0026]) and used to infiltrate 4- to 6-week-old *Nicotiana benthamiana* leaves as described by [@CIT0050]. Fluorescence was observed with a confocal laser scanning microscope at 20 °C (Leica TCS SP5).
Isolation of PGs {#s9}
----------------
Intact chloroplasts from 3-week-old peas were isolated according to Waegemann *et al.* (1992). The chloroplasts were then separated in a discontinuous sucrose gradient according to [@CIT0058]. The gradient was fractionated by taking 1 ml samples, and the proteins in each fraction were precipitated with trichloroacetic acid.
Chl *a* fluorescence measurements and PSI activity {#s10}
--------------------------------------------------
The kinetics of induction of Chl *a* fluorescence in WT and mutant leaves were measured using a pulse-modulated fluorometer (Imaging PAM and DUAL-PAM100; Walz). Leaves, dark adapted for at least 15 min, were used to analyze minimal (*F*~0~) and maximal (*F*~m~) fluorescence yields, the latter being determined by application of a saturating light pulse (1 s duration, 1000 μmol photons m^−2^ s^−1^, 4 min illumination between each pulse). The potential maximum quantum yield of PSII was measured as (*F*~m~−*F*~0~)/*F*~m~=*F*~v~/*F*~m~ (Schreiber *et al.*, 1988).
PSI yield in leaves was measured as absorption changes at 820 nm induced by saturating pulses and far-red light (12 W m^−2^ as measured with a YSI Kettering model 65 A radiometer) in the absence or presence of actinic light as indicated using the DUAL-PAM100 ([@CIT0027]). The size of the intersystemic plastoquinone pool was calculated as the ratio of the areas induced by a single and multiple turnover flashes causing a single charge separation and the reduction of the entire plastoquinone pool, respectively ([@CIT0049]) Other parameters were calculated using the algorithms provided in the DUAL-PAM100 software (Walz).
Pigment analysis {#s11}
----------------
For pigment analyses, leaf samples were frozen in liquid N~2~ and either used directly for pigment extraction or stored at --80 °C for up to 2 weeks until further use. Pigments were extracted by grinding frozen leaf material in a mortar after addition of 1 ml of 100% acetone. After a short centrifugation, the supernatant was filtered through a 0.2 μm membrane filter (GE Healthcare, Little Chalfont, Buckinghamshire, UK) and then subjected to HPLC analysis. Separation and quantification of pigments were done by reversed-phase chromatography as described in [@CIT0013].
Metabolite analysis {#s12}
-------------------
Lipids and non-polar metabolites were extracted as described in [@CIT0022]. For LC-MS analysis, the Dionex Ultimate 3000 UHPLC (Thermo Fisher Scientific) in combination with a timsTOF (Bruker Daltonik) was used. The dry extract of 50 mg of material was resolved in acetonitrile:isopropanol (7:3) and injected on a C8 reversed phase column (Ultra C8 100×2.1 mm; Restek) with 300 µl min^−1^ flow at 60 °C. The solvents used are (A) water and (B) acetonitrile:isopropanol (7:3), both including 1% (v/v) ammonium acetate and 0.1% (v/v) acetic acid. The 26 min gradient started at 55% B, followed by a ramp to 99% B within 15 min. After a 5 min washing step at 99% B, the gradient was returned to 55% B and kept constant for 5 min equilibration.
For MS detection, an electrospray ionization (ESI) source was used in positive mode. Nitrogen was the dry gas, at 8 l min^−1^, 8 bar, and 200 °C. The timsTOF mass spectra were recorded in MS mode from 50 *m/z* to 1300 *m/z* with 40 000 resolution, 1 Hz scan speed, and 0.3 ppm mass accuracy. Compounds were annotated in a targeted approach using the specific mass (*m/z*) at retention time and the isotopic pattern. All data were acquired by otofControl 4.0. The evaluation was performed by DataAnalysis 5.1, ProfileAnalysis 2.3, and MetaboScape 1.0. Corticosterone (0.2 mg ml^--1^) was used as internal standard (IS) for normalization. Given values are relative ratios of the intensities×intensity of IS ^−1^×g FW^−1^.
Computational analyses {#s13}
----------------------
Sequences for PG18 from Arabidopsis and *Synechocystis* were obtained from TAIR (<https://www.arabidopsis.org>) and CyanoBase (<http://genome.microbedb.jp/cyanobase>), respectively. Homologs of PG18 from other species were collected from NCBI/BLAST (<http://blast.ncbi.nlm.nih.gov/Blast.cgi>) and Phytozome (<https://phytozome.jgi.doe.gov/pz/portal.html>). Phylogenetic trees were generated by using the CLC Main Workbench software (CLC bio, Aarhus, Denmark). Alignments were generated by using the algorithm provided by CLC Main Workbench (developed by QIAGEN Aarhus).
Results {#s14}
=======
PG18 is localized to PGs {#s15}
------------------------
To investigate the subcellular localization of PG18 experimentally, we transiently expressed a PG18--green fluorescent protein (GFP) fusion protein under the control of the 35S promoter in tobacco leaves. Leaves were infiltrated with *Agrobacterium* containing the construct, and protoplasts were isolated 2 d after the transfection. GFP fluorescence, as well as chlorophyll autofluorescence, was detected in isolated protoplasts with a confocal laser-scanning microscope. The GFP signal was found exclusively in chloroplasts, where it appeared to form punctate structures ([Fig. 1A](#F1){ref-type="fig"}).
{#F1}
To determine whether PG18 is a soluble protein or is associated with any of the chloroplast's membranes, total proteins isolated from leaves were separated into a membrane and a soluble fraction. PG18 was found exclusively in the membrane fraction by immunoblot analysis ([Fig. 1B](#F1){ref-type="fig"}). We then treated the membrane fraction with 1 M NaCl, 0.1 M Na~2~CO~3~, and 3 M urea to ascertain whether PG18 is attached to the membrane by hydrophobic or electrostatic interactions and could therefore be removed by any of these agents ([Fig. 1C](#F1){ref-type="fig"}). Incubation with buffer served as a negative control, and disruption of the membrane with 1% LDS was applied to achieve total solubilization of the membrane. Only the treatment with 1% LDS resulted in complete solubilization of PG18. However, a small amount of PG18 was found in the supernatant after treatment with 3 M urea ([Fig. 1C](#F1){ref-type="fig"}). To assess the efficiency of these treatments, the blot was probed with OE33 antiserum. OE33 is a peripheral membrane protein in the thylakoid lumen and is at least partially extracted by each of the agents mentioned above, as they all damage the integrity of the thylakoid membrane ([Fig. 1C](#F1){ref-type="fig"}) ([@CIT0005]). Cytochrome *f* (Cyt *f*) was used as an example of an integral membrane protein and, as expected, it is solubilized only by treatment with 1% LDS ([Fig. 1C](#F1){ref-type="fig"}). In [Fig. 1C](#F1){ref-type="fig"}, *pg18* protein extract of was loaded, showing specificity of the antiserum. These results were surprising, as they imply that PG18 behaves like an integral membrane protein, although its sequence appears not to contain any classical transmembrane domains. Several proteomics studies have identified PG18 in PGs ([@CIT0015]; [@CIT0043]; [@CIT0064]) and we therefore set out to test this further. To this end, we fractionated pea chloroplasts into PGs and thylakoid membranes. Pea plants were chosen instead of Arabidopsis for this experiment because they provide larger amounts of leaf material. The specific reactivity of the polyclonal antibody raised against *P. sativum* PG18 has already been established (data not shown). Thylakoid membranes and PGs were separated on a sucrose density gradient, and selected fractions were subjected to SDS--PAGE and subsequently probed with antisera against LHCA1 (chlorophyll *a*-*b* binding protein 1) as a thylakoid marker, plastoglobulin 35 (PGL35) as a PG marker, and PG18 ([Fig. 1D](#F1){ref-type="fig"}). As revealed by the distribution of the PG marker protein PGL35, the PG fraction was not contaminated with thylakoids, and the majority of PG18 was also detected in this fraction. Since PGs are known to be associated with the thylakoid membrane ([@CIT0001]), it is not surprising that small amounts of both PG18 and PGL35 are detected in the thylakoid fraction. Additionally, in order to demonstrate that PG18 is located at the surface of the PGs, we treated intact thylakoids and their associated PGs with trypsin ([Fig. 1E](#F1){ref-type="fig"}). As expected, PG18 was completely digested---as was the α-subunit of the ATP synthase, which is exposed on the stromal side of the thylakoids---while the luminal OE33 protein used as a control for thylakoid integrity remained intact. Taken together, the observed punctate structures of PG18--GFP, the subfractionation data, and the fact that PG18 has been assigned to PGs by MS analyses ([@CIT0064]; [@CIT0037]) led us to the conclusion that PG18 is indeed a PG-localized protein. Considering that none of the PG proteins identified so far harbors transmembrane domains, but are nevertheless most likely to be integrated into the lipid monolayer surface of PGs, the observation that PG18 behaves similarly to an integral membrane protein upon treatment with various chaotropic agents is compatible with the protein's localization to PGs.
Arabidopsis *pg18* mutants display a pale green phenotype, and alterations in thylakoid membrane structure and lipid content {#s16}
----------------------------------------------------------------------------------------------------------------------------
To investigate the function of PG18, we isolated a homozygous T-DNA insertion line. The mutant was viable on soil, but displayed a pale green phenotype under NL conditions (21 °C, 120 µmol photons m^−2^ s^−1^ light intensity, 16/8 h light/dark). Since it has been observed that PGs become more abundant under light stress ([@CIT0065]), we also exposed mutant plants to increased light intensities (IL; 250 µmol photons m^−2^ s^−1^), namely 17.5 d growth under NL conditions followed by 3.5 d under IL conditions ([Fig. 2A](#F2){ref-type="fig"}). The position of the T-DNA insertion as well as genotyping PCR is shown in [Fig. 2B](#F2){ref-type="fig"}. To ensure that the phenotype correlated with the disruption of the *PG18* gene, we complemented the mutant with a PG18--GFP fusion construct expressed under the control of a 35S promoter. A representative example of a complemented line is shown in [Fig. 2A](#F2){ref-type="fig"} (right panel). The knock out of PG18 was verified at the protein level by immunoblot analysis ([Fig. 2C](#F2){ref-type="fig"}). A specific band at 18 kDa was detected in the WT that was absent in the mutant. An antiserum against the chloroplast protein TIC110 was used as a loading control. In addition, proteins were extracted from the complemented mutant line and likewise probed with PG18 antiserum. A protein of 45 kDa was detected, corresponding to the expected size of PG18--GFP.
{#F2}
It has been reported that PG proteins are strongly expressed under increased light; however, PG18 was not among the proteins showing increased accumulation ([@CIT0064]). To verify this, we monitored the levels of PG18 in plants grown under NL and IL conditions. Consistent with the MS-based quantification in [@CIT0064], PG18 is equally abundant in NL and IL plants, while expression of the core PG protein PGL35 increases upon exposure to high light levels ([Fig. 2C](#F2){ref-type="fig"}).
Since the pale phenotype indicated a defect in chloroplasts, we proceeded to analyze the phenotype at the ultrastructural level ([Fig. 2E](#F2){ref-type="fig"}). Rosette leaves of 3-week-old WT, *pg18* mutant, and complemented plants grown under NL and IL conditions were analyzed. In comparison with the WT, the cross-sectional area of *pg18* chloroplasts was significantly smaller under both conditions ([Fig. 2F](#F2){ref-type="fig"}). When examined at higher magnification, the stroma lamellae in *pg18* mutant chloroplasts were shorter and appeared to be stacked, rather than present in single layers. Moreover, the stroma lamellae exhibited less branching between grana stacks, while the latter displayed a more compact organization and generally consisted of more layers in the mutant than in the WT or the complemented line. This effect was observed under both NL and IL conditions, although it was more pronounced under IL ([Fig. 2F](#F2){ref-type="fig"}, lower panels). Notably, under both light conditions, PG abundance per chloroplast was also reduced significantly in the mutant chloroplasts in comparison with the WT ([Fig. 2G](#F2){ref-type="fig"}). The number of PGs doubled under IL in the WT. Strikingly, this effect of exposure to IL was also observed in the mutant, in spite of the overall reduction in the numbers of PGs ([Fig. 2G](#F2){ref-type="fig"}).
As a next step, we aimed to analyze changes in the lipid content of the WT and *pg18* mutant. We analyzed the major thylakoid lipids monogalactosyldiacylglycerol (MGDG) and digalactosyldiacylglycerol (DGDG) and, albeit not statistically significant, we observed a slight reduction of DGDG under NL conditions in the mutant. A similar effect was caused by IL treatment already in the WT; however, no further reduction of these lipids was observed in the mutant under IL conditions ([Supplementary Fig. S1](#sup1){ref-type="supplementary-material"}). Sulfoquinovosyldiacylglycerol (SQDG) and the phospholipids phosphatidylglycerol and phosphatidylcholine remained essentially unchanged in the mutant ([Supplementary Fig. S1](#sup1){ref-type="supplementary-material"}).
Arabidopsis *pg18* mutant plants show symptoms of light stress {#s17}
--------------------------------------------------------------
To assess differences in thylakoid pigments that might be related to the mutant phenotype, the pigment content of *pg18* leaves was analyzed ([Fig. 3](#F3){ref-type="fig"}). Chl *a* and *b* were indeed reduced in the mutant, in agreement with its pale green phenotype (36% and 20% reduction for Chl *a* and *b*, respectively) under both growth light conditions. However, only the reduction in Chl *a* relative to the WT was significant ([Fig. 3A](#F3){ref-type="fig"}, left panel), resulting in a significantly decreased Chl *a*:*b* ratio in the mutant ([Fig. 3A](#F3){ref-type="fig"}, right panel).
{#F3}
Analysis of the carotenoid content revealed that β-carotene (Car) was significantly reduced in the *pg18* mutant compared with the WT, irrespective of the light conditions during growth. In contrast, the levels of lutein (Lut) as well as neoxanthin (Nx) were increased in the mutant ([Fig. 3B](#F3){ref-type="fig"}). Furthermore, the amount (VAZ pool size) and de-epoxidation state (DEPS) of the xanthophyll cycle pigments violaxanthin (Vx), antheraxanthin (Ax), and zeaxanthin (Zx) were analyzed, since both parameters are reliable indicators of light stress. Indeed, a significant increase of both the VAZ pool size ([Fig. 3C](#F3){ref-type="fig"}, left panel) and the DEPS ([Fig. 3C](#F3){ref-type="fig"}, right panel) was determined for *pg18* compared with the WT, with a more pronounced increase under IL conditions, supporting the view that the mutant is more susceptible to light stress. No difference in pigment accumulation was observed for the 35S::PG18-GFP line as compared with the WT (data not shown).
Photosynthetic performance is affected in *pg18* mutant plants {#s18}
--------------------------------------------------------------
The photosynthetic performance of both photosystems was analyzed in 3-week-old plants grown under NL and IL conditions using a pulse amplitude-modulated (PAM) fluorimeter. Compared with the WT, the *F*~0~ values were 3- and 2-fold increased in the *pg18* mutant grown under NL and IL conditions, respectively, indicative of a disturbed PSII function ([Supplementary Table S3](#sup1){ref-type="supplementary-material"}; [Supplementary Fig. S2A](#sup1){ref-type="supplementary-material"}). In accordance with this, the maximum PSII quantum yield expressed as the ratio of the variable to the maximum fluorescence (*F*~v~/*F*~m~) was reduced to \~66% (NL) and 56% (IL) in *pg18* as compared with the WT, again indicating defects of PSII in the mutant lines ([Supplementary Table S3](#sup1){ref-type="supplementary-material"}). The fluorescence dropped far below the *F*~0~ level during light induction in *pg18*, which is typical for mutants directly affected in PSII ([Supplementary Fig. S2A](#sup1){ref-type="supplementary-material"}). Both PSI and PSII yields were reduced in the *pg18* mutant, especially under increased light intensities ([Fig. 4A](#F4){ref-type="fig"}, B). Moreover, the theoretically deduced electron transport rate (ETR) was dramatically reduced by \>3-fold in NL and \~5-fold in IL in the mutant compared with the WT, which is an indication of an inefficient linear electron flow ([Fig. 4C](#F4){ref-type="fig"}). The non-photochemical quenching (NPQ) of excitation energy in the *pg18* mutant was twice as high as in the WT even under low light, which very probably reflects an increased proton gradient across the thylakoid membrane ([Fig. 4D](#F4){ref-type="fig"}). As can be seen from an image of whole plants recorded with an imaging PAM, the *F*~v~/*F*~m~ value was equally reduced in all rosette leaves of the mutant ([Fig. 4E](#F4){ref-type="fig"}). No difference from the WT was observed for the 35S::PG18-GFP line (data not shown).
{#F4}
As compared with the WT, the quantum yield of non-photochemical energy dissipation due to PSI donor Y(ND) site limitation was increased in the mutant, whereas the acceptor side limitation Y(NA) was decreased, indicating that the electron transport towards PSI is limiting the overall photosynthetic electron transport ([Supplementary Table S3](#sup1){ref-type="supplementary-material"}). We further investigated the amount of electrons present within the plastoquinone pool involved in the photosynthetic electron transport, since a substantial portion of the plastoquinone is known to be present in PGs ([@CIT0030]). Interestingly, the size of the plastoquinone pool shared by both photosystems was twice as high in *pg18* as in the WT ([Supplementary Table S3](#sup1){ref-type="supplementary-material"}). However, metabolite measurements performed with entire leaves showed no notable difference in the plastoquione amounts ([Supplementary Table S4](#sup1){ref-type="supplementary-material"}).
Next, we analyzed the ability to recover from photoinhibition by measuring the *F*~v~/*F*~m~ value in response to 2 h exposure to 1000 µmol photons m^−2^ s^−1^ and a subsequent 6 h dark recovery phase. Relative to the initial dark *F*~v~/*F*~m~ ratio, the high light-induced reduction (and thus photoinhibition of PSII) was the same for WT and *pg18* plants, and the recovery rate was almost comparable with that of the WT, demonstrating a rather efficient repair system in the *pg18* mutant ([Supplementary Fig. 2B](#sup1){ref-type="supplementary-material"}). Additionally, the translation efficiency was assessed by *in vivo* radiolabeling of WT and mutant *pg18* plants with ^35^S-labeled methionine for 20 min. The labeled proteins were separated by BN--PAGE. In line with the recovery experiment results, the translation as well as the complex assembly rate did not show differences ([Supplementary Fig. 2C](#sup1){ref-type="supplementary-material"}). Since the effect on the ETR and the quantum yield of both photosystems might be related to altered steady-state levels, stoichiometry, or assembly of the thylakoid membrane complexes in the mutant, we went on to investigate them in more detail.
Loss of PG18 has an impact on the accumulation of thylakoid membrane protein complexes {#s19}
--------------------------------------------------------------------------------------
In order to monitor the abundance of the thylakoid membrane proteins in the WT and the *pg18* mutant, we performed immunoblot analyses with antisera against subunits of the photosystems, the ATP synthase, and the Cyt *b*~6~*f* complex. The results revealed that components of the ATP synthase (α and γ subunits), and the PsaG, PsaD, PsaF, and LHCA2 subunits of PSI were notably reduced as compared with WT levels ([Fig. 5A](#F5){ref-type="fig"}, B). A slight reduction was also observed for the PSII core components D1 and CP47. However, mutant levels of Cyt *f* were unchanged relative to the WT. Unexpectedly, levels of LHCB2 were increased by \~20%. To analyze the assembly status of thylakoid complexes, thylakoid membranes were solubilized with 1% β-dodecylmaltoside and separated in BN gels, followed by SDS--PAGE in the second dimension. In accordance with the immunoblot results, the mutant *pg18* showed higher accumulation of LHCBs compared with the WT, and a slight reduction in subunits of ATP synthase and PSI ([Supplementary Fig. 3](#sup1){ref-type="supplementary-material"}).
{#F5}
Loss of the *Synechocystis* homolog of PG18 affects cyanobacterial fitness {#s20}
--------------------------------------------------------------------------
PGs are not exclusive to plants, but are also found in cyanobacteria as lipid droplets ([@CIT0054]). Therefore, we asked whether homologs of PG18 are present in other organisms, since other PG proteins---such as FBNs---have been identified in cyanobacteria ([@CIT0005]). Interestingly, PG18 turns out to be conserved from cyanobacteria to higher plants ([Supplementary Fig. 4A](#sup1){ref-type="supplementary-material"}; Supplementary Data S1, S2).
In order to investigate the function of the PG18 homolog in cyanobacteria, we generated a knock out mutant in *Synechocystis* by inserting a kanamycin resistance cassette into the *sll1769* ORF ([Supplementary Fig. 4B](#sup1){ref-type="supplementary-material"}). However, in this mutant (*synpg18*^−^), the ultrastructure of the thylakoid membrane is apparently normal ([Fig. 6A](#F6){ref-type="fig"}). Nevertheless, the growth rate of the *synpg18*^−^ strain was significantly reduced compared with the WT under photoautotrophic conditions, whereas no differences were observed under photoheterotrophic conditions ([Table 1](#T1){ref-type="table"}). In accordance with what was observed in Arabidopsis, the chlorophyll content of the *synpg18*^−^ mutant was reduced relative to the WT, as was photosynthetic performance---as reflected in the parameter *F*~v~/*F*~m~: 0.27 for *synpg18*^−^ as against 0.42 for the WT ([Table 1](#T1){ref-type="table"}). Since photosynthesis was affected in the *synpg18*^−^ mutant, we analyzed the accumulation of thylakoid membrane proteins. The *synpg18*^−^ mutant showed significantly lower levels of the PSI subunit PsaD and Cyt *f* of the Cyt *b*~6~*f* complex, whereas other subunits were present in normal amounts ([Fig. 6B](#F6){ref-type="fig"}).
######
Physiological characteristics of the *synpg18*^−^ mutant
Strain Doubling time (h) Chlorophyll content (μg OD~750~^−1^) *F* ~v~/*F*~m~
--------------- -------------------------------- -------------------------------------- ----------------
WT 8.15±0.40^*a*^ 17.44±0.28^*b*^ 2.22±0.34 0.41±0.01
*synpg18* ^−^ 8.05±0.21^a^ 19.34±0.13^*b*^ 1.79±0.10 0.27±0.02
^*a*,b^ Doubling times in the ^*a*^presence or ^*b*^absence of 5 mM glucose, measured under continuous illumination at 30 μmol photons m^−2^ s^−1^ and CO~2~-limiting conditions. The photosynthetic parameter *F*~v~/*F*~m~ was obtained from chlorophyll fluorescence measurements. Data are mean values ±SD of three independent replicates.
{#F6}
Discussion {#s21}
==========
Several lines of evidence indicate that PGs play an important role in thylakoid development. Their number decreases during de-etiolation, while mutants with defects in thylakoid formation often accumulate larger numbers of PGs ([@CIT0029]; [@CIT0003]). Their function may consist of providing metabolites, such as carotenoids and prenylquinones. Additionally, they may be responsible for providing triglycerides for membrane lipid synthesis.
In this study, we have addressed the role and localization of PG18. PG18 is a protein of unknown function that was identified in PGs of higher plants by MS ([@CIT0037]). With the help of a PG18-specific antiserum, we confirmed that PG18 is indeed localized to PGs. Surprisingly, despite the absence of obvious hydrophobic regions in its sequence, PG18 behaves as a membrane protein. A previous study on the fibrillin PGL34, which likewise does not contain typical transmembrane domains, revealed that most deletion variants tested failed to localize to PGs when transiently expressed as GFP fusions ([@CIT0059]). This suggests that the association of PG proteins with the PG monolayer might rely on their folded conformation rather than a particular hydrophobic domain.
Our characterization of a loss-of-function mutant of PG18 clearly demonstrates that the lack of this protein has a severe impact on plant fitness throughout development. Mutant plants show a light green phenotype and are smaller than the WT. Since PGs have been shown to accumulate under light stress ([@CIT0065]), we assessed whether the mutant line exhibits a stronger phenotype under such conditions. Mutants for other PG proteins typically show enhanced phenotypes when subjected to light stress ([@CIT0045]; [@CIT0063]; [@CIT0014]). However, the phenotype of *pg18* remained essentially the same under NL and IL conditions. This indicates that PG18 plays a more general role in chloroplast biogenesis. This notion is supported by the fact that *pg18* plants have smaller chloroplasts and fewer PGs per chloroplast than the WT. Notably, the number of PGs was increased under IL conditions in both WT and *pg18* chloroplasts, although the latter always contained fewer PGs per chloroplast. This indicates that PG18 is not involved in promoting PG accumulation under light stress.
The pale green leaves of *pg18* plants point to a reduction in chlorophyll content. This assumption was verified by analyzing the chlorophyll content of leaves from plants grown under NL and IL conditions, and the most significant reduction was noted in Chl *a*. The carotenoid fraction revealed a strong reduction in β-carotene in the *pg18* mutant compared with the WT under both NL and IL conditions. This reduction as well as the reduced Chl *a*/*b* ratio support an increased PSII/PSI ratio, as also shown by analysis of the protein levels. Moreover, the increase in Vx and Nx (on a chlorophyll basis) is in good agreement with the observed relatively higher amounts of LHCII.
Analysis of the pigments of the xanthophyll cycle in the *pg18* mutant showed that levels of the photoprotective pigments Ax and Zx were increased relative to the WT, and further enhanced under IL conditions. These pigments are synthesized from Vx under light stress and contribute to energy dissipation (NPQ) and thus to photoprotection of PSII ([@CIT0017]; Farber *et al*., 1997; [@CIT0023]). Interestingly, levels of the photoprotective prenyllipid α-tocopherol, which is stored in PGs, increased in the WT after IL treatment. This effect, however, was not observed in *pg18* when comparing NL and IL conditions ([Supplementary Table S4](#sup1){ref-type="supplementary-material"}). Taken together, *pg18* mutant plants showed symptoms of light stress even under normal light conditions, which probably explains the accumulation of larger grana stacks as well as more LHCII in their chloroplasts.
Chlorophyll fluorescence analyses showed that photosynthetic performance in general was affected, with both photosystems showing lower levels and quantum yields than the WT, and these deficits became somewhat more pronounced under increased light intensities. Interestingly, we observed elevated levels of phylloquinone in the mutant, which might result from reduced PSI levels, thus leading to an excess of phylloquinone ([Supplementary Table S4](#sup1){ref-type="supplementary-material"}). In turn, this shows that phylloquinone, which is stored in PGs, is not limiting. The severely increased *F*~0~ value in *pg18* mutants can partially be explained by the over-representation of LHCB2, which is not functionally connected to PSII but contributes to the fluorescence. The increased NPQ is in accordance with the reduced levels of the ATP synthase causing accumulation of lumenal protons and thus increased dissipation of heat via qE. This observation correlates with higher levels of Zx and Ax found in the mutant plants relative to the WT. These pigments are known to modulate NPQ by deactivating excited states in the PSII antenna, and also acting as antioxidants in the thylakoid membrane ([@CIT0020]; [@CIT0019]; [@CIT0041]). The decreased electron transport was also confirmed by the increased PSI donor side limitation, although levels of this photosystem were more reduced than those of PSII or the Cyt *b*~6~*f* complex, indicating strong down-regulation of PSII by increased NPQ. Furthermore, the increased size of the intersystemic electron pool indicates that levels of PQ molecules involved in photosynthesis are unchanged but that less active photosystems share the same PQ pool size as the WT. Complementation of the mutant line rescued both the photosynthetic activity and reduced pigment accumulation to WT levels (data no shown).
Moreover, to investigate the light sensitivity of PSII and the ability to repair it after photoinhibition via *de novo* synthesis and incorporation of D1 under IL conditions, we performed light stress recovery analysis. Strikingly, both the WT and *pg18* were equally affected by light stress, but the WT recovered slightly faster than *pg18*. Therefore, the pleiotropic effects on photosynthetic performance in *pg18* chloroplasts probably result from assembly defects or altered stoichiometry of the photosynthetic complexes and/or antenna proteins. In line with this, analysis of the accumulation of thylakoid proteins revealed a strong reduction in the ATP synthase and PSI complexes. In agreement with the reduction in PSI, levels of Chl *a* were sharply reduced in the *pg18* mutant compared with the WT, which is also reflected in the fact that the ratio of Chl *a* to Chl *b* remains essentially the same under NL and IL conditions in *pg18*. Furthermore, our finding that the stromal lamellae are shorter and less branched in the *pg18* mutant than in the WT ([Fig. 2E](#F2){ref-type="fig"}) correlates with the reduction in levels of PSI and ATP synthase, as less membrane area is available for their integration. In summary, the overall stoichiometry of the complexes in the thylakoid membrane is affected in the *pg18* mutant. Interestingly, there is a clear accumulation of LHC proteins in the mutant line, which is compatible with the fact that *pg18* plants accumulate more photoprotective LHC-binding pigments such as Zx ([@CIT0024]). This effect can also be observed in the high level of NPQ in the *pg18* mutant, which is further enhanced under IL, in accordance with the increased levels of Zx in the mutant relative to the WT under both NL and IL conditions. Hence, PG18 seems not to intervene directly in the assembly of any particular complex in the thylakoid membrane, but rather affects the composition of some of them, possibly by modulating the structural organization of the thylakoid membrane during its biogenesis.
Despite the lack of identifiable functional domains, suggesting that PG18 is not itself an enzyme, loss of PG18 has a significant impact on the composition and architecture of the thylakoid membrane. Considering that PGs play an important role in mobilizing lipids for incorporation into the thylakoid membrane ([@CIT0008]; [@CIT0052]), an alteration in PGs could be expected to affect thylakoid complexes, as has been observed when the lipid content of the thylakoid membrane is altered ([@CIT0066]; [@CIT0025]). The observed higher MGDG:DGDG ratio in the mutants correlated with the increased grana formation and stacking of thylakoid membranes ([@CIT0031]).
Levels of PG18 were not found to be specifically up-regulated under stress conditions, nor does it accumulate to a greater extent in PGs isolated after high light treatment ([@CIT0064]). These findings indicate that it is a constitutive component of PGs, possibly fulfilling a general role in PG maintenance or interacting with other PG proteins. This inference is supported by the observation that mutants for other PG proteins do not show a phenotype under normal conditions. Phenotypes only become manifest when PG mutants are exposed to stresses, such as high light intensities ([@CIT0045]; [@CIT0051]; [@CIT0063]; [@CIT0039]; [@CIT0002]; [@CIT0014]). Moreover, PG18 was also found in chromoplasts of red pepper ([@CIT0064]), suggesting that its function is not restricted to chloroplasts. It is noteworthy, however, that PG18 is phosphorylated, which might lead to conformational changes or otherwise have an impact on its activity, possibly depending on different developmental stages or stress conditions ([@CIT0061]; [@CIT0034]). Interestingly, co-expression analysis previously showed that PG18 is co-expressed with a number of components of the chloroplast redox network, such as STN7 and the thioredoxins ([@CIT0037]). Redox regulation is known to be important for regulation of photosynthesis, especially under short-term acclimation to different light conditions. It is therefore highly feasible that PGs, and PG18 in particular, are involved in regulation via redox-controlled processes.
In accordance with what we observed in Arabidopsis, loss of SynPG18 has similar effects on *Synechocystis* fitness. Although no alterations in thylakoid ultrastructure were observed in the mutant cyanobacteria, growth rate and chlorophyll content were lower than in the WT, as in the case of *A. thaliana*. The reduction in growth rate becomes manifest only when cyanobacteria were grown under photoautotrophic conditions, which is compatible with the assumption that SynPG18 is involved in photosynthetic performance. Additionally, *F*~v~/*F*~m~ was reduced in the *synpg18*^−^ mutant. Examination of the composition of the photosynthetic complexes revealed reduced amounts of PsaD and Cyt*f*, while other proteins such as D1 and ATP synthase subunit α were present in WT levels. This result differs slightly from what we observed in *A. thaliana*, where ATP synthase subunits were more strongly reduced and Cyt*f* was unchanged. This interspecies difference can probably be explained from an evolutionary perspective. It can be assumed that in cyanobacteria PG18 has a role in thylakoid biogenesis that has been conserved throughout evolution, while new features leading it to affect ATP synthase accumulation in green plants could have been acquired subsequently. This assumption is supported by the evidence that, in plants, PGs have multiple roles in development, senescence, and light stress, making them more complex than their counterparts in cyanobacteria.
In summary, investigation of the *pg18* mutant underlines the importance of PGs in the formation of thylakoid membranes and shaping their protein complex composition. Elucidating its exact role will be a challenging task to address in the future.
Supplementary data {#s22}
==================
Supplementary data are available at *JXB* online.
**Data S1.** Species and accession numbers.
**Data S2.** Protein sequence alignment of PG18 in based on the accessions given in Supplementary Data S1.
**Fig. S1.** Lipid analysis in the WT and *pg18*.
**Fig. S2.** Photosynthetic performance and photoinhibition/recovery in the WT and *pg18*.
**Fig. S3.** Loss of PG18 function affects PSI and ATP synthase complexes.
**Fig. S4.** Conservation of PG18.
**Table S1.** List of oligonucleotides used in this research.
**Table S2.** List of antisera used in this research.
**Table S3.** Photosynthetic parameters.
**Table S4.** Measurements of prenyllipids.
######
Click here for additional data file.
######
Click here for additional data file.
######
Click here for additional data file.
This work was supported by the DAAD/BECAS Chile (fund 57144001) to REC, the DFG (SFB-TR 175, projects A03, B05, and B06) to JM, JS, and SS, the DFG in the context of Research Unit FOR2092 (Ni390/9) to JN and SH, and the DFG (JA 665/12-1) to PJ. We are grateful to Tamara Hechtl for excellent technical assistance, and we would like to thank Stephanie Seifert for help with genotyping experiments.
FBN
: fibrillin
PG
: plastoglobule
PG18
: plastoglobular protein 18
| {
"pile_set_name": "PubMed Central"
} |
INTRODUCTION
============
The access, primary attribute of the quality in health care -- access to structures and care processes in a timely manner, receiving effective care[@B1] -- directly influences the quality of primary health care (PHC).
In the last ten years, there has been a significant increase in population coverage by the PHC, mainly by the Family Health Strategy model (*Estratégia Saúde da Família*). However, the decentralization in the implementation and management of the PHC allowed a great variation in the quality of the health centers around the country. In addition, the greater availability of PHC centers did not result in an expressive improvement that ensured universal and equitable health coverage[@B2].
Organizational changes are needed to address the issue of long delay in scheduling an appointment at PHC, which represents the inability of health centers to provide timely access and leads to avoidable negative experiences for users. The uncertainty about being seen or not causes distress, especially among those who believe that their health condition is progressively worsening. In addition, it contributes to the increase in the inappropriate use of emergency services[@B3].
Improvements to the scheduling model can increase access to PHC[@B4]. An effective scheduling process improves the work environment, quality of care, patient safety, satisfaction of health teams, timely access, and user experience[@B5]. In the Methods section of this article, there is a characterization of the three scheduling models used (advanced, carve-out and traditional access). Other authors have already described the main current scheduling models^5--7^.
The implementation of advanced access is the most common and cost-effective intervention to reduce the delay in scheduling a PHC appointment[@B8]. In addition, this model is associated with the improvement of the satisfaction of the health team[@B10] ^,^ [@B11], the longitudinality of patient care and safety[@B11], as well as the reduction of absenteeism[@B9] ^,^ [@B10], of inappropriate appointments in emergency services[@B10] ^,^ [@B11], of medical time dispensed in the risk classification to prioritize appointments[@B10] and of the negative work *backlog*, which consists of schedules resulting from the work not completed by the team in a same day, either because of the low supply or the organization of the work process itself[@B8] ^,^ [@B9].
In advanced access, the user can obtain a appointment within two business days (but can be scheduled for later if the user prefers)[@B12]; there is no distinction between urgent (unscheduled) and routine (scheduled) appointments[@B6], and the longitudinality of care is prioritized[@B12]. This system presupposes five principles: balance between supply and demand, reduction of negative backlog, improvement of interprofessional practices, contingency plan and revision of the scheduling system[@B6].
In Brazil, since 2014, several advanced access deployment experiments have been conducted locally. This intervention was not included in any national orientation, and several initiatives were undertaken by the PHC teams in an attempt of meeting local demands. In addition, it has not yet been evaluated by any study.
This article aims to investigate the relationship between the quality of PHC perceived by the user and the appointment scheduling model adopted in the health center. To this end, the study evaluated whether PHC teams that promote advanced access had a higher PHC score (in essential and derivative attributes) than those using another appointment scheduling model.
METHODS
=======
Population-based cross-sectional study held in the health district in Northern Florianópolis in 2012. It used the Primary Care Assessment Tool (PCATool-Brasil) for adults as the instrument for measuring the PHC quality.
Adults (≥ 18 years) living in the territory covered by the 11 health centers (HC) of the region (n = 598) were included. Interviewees who did not have as reference service one of the health centers under analysis (n = 189) were excluded.
The interviewers were community health agents (n = 83) trained for this purpose, under the supervision of nurses (n = 24) of the Family Health Strategy teams, which is the PHC model of organization recommended by the municipal health policy. To minimize gauging biases, the community health agents did not make the interview in their own working area, being allocated to the area of another health center. A pilot of the collection was held for training and clarification of doubts about the application of the instrument. Field collection lasted five months (from May to September 2012).
The data that subsidized the identification of the appointment scheduling model were obtained through contact with each PHC, which was classified into one of the following models:
1. Advanced access: approximately 65%--90% of daily medical appointments are reserved for unscheduled care[@B6] ^,^ [@B7]. Most prescheduled appointments result from positive work backlogs, representing patients who do not want the appointment on that day and those who are scheduled by the physician after evaluation of clinical and social criteria. The maximum delay in booking a medical appointment is two business days.
2. Weekly carve-out: 50% of daily medical appointments are reserved for the care of unscheduled care, whereas 50% are for the care of scheduled appointments. The maximum delay in booking a medical appointment is five business days.
3. Carve-out every 15 days: 50% of daily medical appointments are reserved for the unscheduled care, whereas 50% are for the care of scheduled appointments. The maximum delay to booking an appointment is ten business days.
4. Traditional: all daily medical appointments are pre-booked (supersaturated schedule). There is no reserve of vacancies for unscheduled care, which are embedded between the appointments, generating double-booking. In this model the average delay in booking a medical appointment is quite variable, sometimes taking longer than thirty days.
Instrument
----------
The PCATool-Brasil[@B13] evaluates the quality of PHC by the presence and extension of its essential (first contact access, comprehensiveness, longitudinality and care coordination) and derivative (family orientation and community orientation) attributes, through questions about health centers which can be answered by users, professionals or public administrators. This instrument was created based on the Donabedian quality assessment model, considering the measurement of aspects of health center structure, processes and outcomes[@B13].
Stein[@B14] describes the importance of using validated instruments to evaluate the quality of care in the PHC and improve its performance, besides stating that the PCATool is used by public administrators and researchers due to its excellent measurement properties. In Brazil it is currently the most widely used instrument to evaluate the effectiveness of family health teams[@B15]. The validation and use of this instrument in different countries, such as Canada, United States, Spain, China, Argentina and Brazil, shows its suitability in different sanitary and cultural contexts[@B16].
All key attributes of PHC are measured by the PCATool-Brasil from the perspective of individual experience, including components related to the structure and the care process. The responses are of Likert type, with the scale ranging from one ("absolutely not") to four ("absolutely yes") and the additional option nine ("I do not know"/"I do not recall"). It has 87 items divided into 10 components related to PHC attributes:
1. degree of affiliation with the health center;
2. first contact access (sub-dimension utilization);
3. first contact access (sub-dimension accessibility);
4. longitudinality;
5. coordination (sub-dimension integration of care);
6. coordination (sub-dimension information system);
7. comprehensiveness (sub-dimension services available);
8. comprehensiveness (sub-dimension services provided);
9. family orientation;
10. community orientation.
According to the validation instrument PCATool-Brasil for adults, scores are standardized for a scale from zero to 10, with values equal to or greater than 6.6 considered as high, which corresponds to responses for options three or four on the instrument's original scale. The standardization for the scale from zero to 10 is conducted as follows:
S
tan
d
a
r
d
i
z
e
d
s
c
o
r
e
=
(
Escore
\-
1
)
×
10
(
4
\-
1
)
The PHC quality score was calculated according to the PCATool-Brasil guidebook, adult version*,* of the Brazilian Ministry of Health. The general score (GS) of PHC, calculated as the mean of all attributes plus the degree of affiliation (mean value of essential and derivative attributes and degree of affiliation) was calculated[@B13].
Secondary Data
==============
Secondary data from InfoSaúde, a computerized system used by the Municipal Health Secretariat (SMS) of Florianópolis were used. They covered the period from January 1, 2011 to December 31, 2011. All 11 health centers in the Northern health district during the period analyzed were included in the study.
Data on the presence or absence of economically deprived areas were obtained from the Department of Health Geoprocessing of the SMS in Florianópolis. Areas classified as economically deprived comply with the following criteria: low family income, housing and infrastructure network precariousness, environmental precariousness and risk and precarious areas in land tenure, and urban equipment and services[@B17].
Data on the poverty ratio, defined as *per capita* income up to a minimum wage, were based on the nominal income of the population according to the SMS of Florianópolis. Other secondary data extracted from InfoSaúde were: panel size (population enlisted by health center), number of PHC teams per health center, number of medical appointments per year, and number of people attended per doctor per year. It should be noted that the study considered the period when some teams stood without a doctor during the period analyzed.
Sampling and Statistical Analysis
=================================
To calculate the sample size, the OpenEpi^®^ software was used. A total of 459 questionnaires were required. Considering percentage of loss plus the estimated sample size of 30%, 598 questionnaires were applied. The parameters used for this calculation were: 95% confidence level, 5% absolute precision, and 1.2 design effect for the cluster effect adjustment to estimate the proportion of users that would assign high PHC scores (≥ 6.6) for 50% of the evaluated health centers.
The household sampling process was by clusters stratified by health center and distributed proportionally by their panel size. Households were selected by systematic sampling by street and house.
Descriptive analysis was then performed with absolute frequency, percentage, mean and standard error. The multilevel analysis methodology was used to observe the relationship between the dependent variable (general PHC score) and the study factor (scheduling model adopted).
First, adjustments were made to the model using the dependent variable, the factor under study and the other independent variables in an individual way: age, race, poverty ratio (used in the study as a proxy income variable), panel size per PHC team, number of PHC teams, economically deprived area, number of medical appointments in the year and number of people assisted in the year. The independent variables that presented p-value less than 0.20 in this first model were chosen to compose the final multilevel model. Although the variable age was not statistically significant, it was maintained in the multivariate model for the purpose of adjusting estimates because of its conceptual relevance.
The multilevel model was adjusted according to the aggregates of information: individual (users of health centers) and health centers. The results were presented by the coefficient B, its respective confidence intervals (95%CI) and p-values. The significance level of 5%, bilateral, was used for all statistical analyses. The suitability of the model was verified using residue analysis and the presence of collinearity between the variables. The [Box](#t1){ref-type="table"} shows the two-level model: the first shows the characteristics of users and the second shows the characteristics of health centers.
BoxConceptual model used for multilevel analysis.Level 1Characteristics of users-- Age-- Sex-- Skin color-- Poverty ratio (income proxy variable)Level 2Characteristics of Health Centers-- Scheduling model-- Panel size by health team-- Presence of economically deprived areas-- Number of medical appointments in one year per health team-- Number of people served in one year per health teamGeneral PHC score (degree of orientation for PHC)[^3]
The interviewees participated voluntarily in the research and signed an informed consent form, from which they received a copy, according to Resolution 466/2012 of the National Health Council. The research project was approved by the Monitoring Committee of the Municipal Health Research Projects and by the Research Ethics Committee Involving Human Beings of the Universidade Federal de Santa Catarina, which issued a favorable opinion under number 1,635,663.
RESULTS
=======
A total of 598 users were interviewed. The proportion of users who attributed a high PHC score to the health centers evaluated was 46.45% (n = 190), in line with the proportion used to calculate the sample size. Approximately 70% of the people (n = 409) reported using health centers as a reference service for primary health care. This was the final sample of the subsequent analyses.
[Table 1](#t2){ref-type="table"} presents the description of health centers in the northern health district, with secondary data collected and general PHC score according to PCATool-Brasil. Only 36.36% of the health centers evaluated had a high general PHC score, the highest being the health center that used advanced access.
Table 1Description of health centers in the Northern sanitary district according to PCATool-Brasil in 2012. Florianópolis, state of Santa Catarina, 2017.Health centersScheduling modelNumber of health teamsPanel size by health team (inhabitants)Presence of economically deprived areasNumber of medical appointments in one year per FH teamNumber of people served in one year per health teamGeneral PHC score\* measured by PCATool-Brasil based on evaluation of users1Weekly35,249Yes2,9081,5256.482Advanced53,784No4,4331,5337.053Every 15 days34,579No2,6631,2705.394Weekly25,651Yes3,7601,2646.015Weekly23,581No3,1251,2507.236Traditional16,910No3,4801,1555.687Traditional14,114No2,9891,6066.108Traditional11,630No3,2315936.719Traditional12,828No3,4861,6966.8610Every 15 days22,746Yes2,5549876.1911Traditional14,160Yes2,9701,3626.09[^4][^5]
[Table 2](#t3){ref-type="table"} presents the distribution of the characteristics of the users according to the scheduling model. The general mean age of the interviewees was 47 years, with a standard deviation (SD) of 0.86. Regarding skin color, 91.9% declared themselves white.
Table 2Distribution of the characteristics of users according to the scheduling model. Florianópolis, state of Santa Catarina, 2017.CharacteristicGeneralScheduling modelTraditional (n = 63)Weekly carve-out (n = 160)Carve-out every 15 days (n = 90)Advanced access (n = 96)Mean (SD)Mean (SD)Min.-Max.Mean (SD)Min.-Max.Mean (SD)Min.-Max.Mean (SD)Min.-Max.Age (years)47.0 (0.86)48.4 (2.4)(16.0--85.0)46.2 (1.3)(16.0--80.0)46.5 (1.6)(18.0--89.0)48.5 (2.2)(20.0--78.0) n (%)Skin color White376 (91.9)59 (93.7)-143 (89.3)-85 (94.0)-89 (92.2)-Non-white33 (8.1)4 (6.3)-17 (10.7)-5 (6.0)-7 (7.8)-[^6]
[Table 3](#t4){ref-type="table"} shows the mean general PHC score in health centers according to the scheduling model. It was verified that the health center that adopted the advanced access presented high mean general PHC score, whereas the others presented low score.
Table 3Mean general score\* of primary health care measured by PCATool-Brasil of health centers according to the scheduling model. Florianópolis, state of Santa Catarina, 2017.Scheduling modelMeanStandard error95%CIAdvanced access7.040.496.09--8.00Weekly carve-out6.260.275.67--6.74Carve-out every 15 days5.870.355.18--6.57Traditional scheduling6.290.275.67--6.74[^7][^8]
[Table 4](#t5){ref-type="table"} shows the characteristics associated to the general PHC score, as well as the results of the adjusted models using the multilevel methodology. The variable poverty ratio was dichotomized by its median, whose result was 11%. Considering the models with adjustment of the independent variables to the outcome, it can be observed that the scheduling model and the panel size were shown to be associated with the mean general PHC score.
Table 4Characteristics associated to the general primary health care score in the adult users' perception users of health centers. Florianópolis, state of Santa Catarina, 2017.CharacteristicUnivariate model\*Multivariate model\*Beta95%CIpBeta95%CIpUsersAge (increase by age group every 10 years)0.04-0.05--0.130.400.04-0.05--0.130.41Skin color White0.19-0.33--0.700.48 Non-white0.00 Health centersScheduling model Weekly carve-out-0.49-1.60--0.610.38-1.41-2.53---0.300.01Carve-out every 15 days-1.16-2.33--0.010.03-2.36-3.61---1.100.00Traditional scheduling-0.89-1.98--0.190.11-2.64-4.24---1.050.00Advanced access0.00 0.00 Panel size per health team (every 1,000 individuals)-0.26-0.51--0.000.05-0.11-0.20---0.020.01Proportion of poverty (income) (median = 0.11) Up to 0.110.45-0.22--1.130.19 More than 0.110.00 Number of health teams (increase of 1 FH team)0.14-0.15--0.430.35 Economically deprived areas Presence0.07-0.85--0.980.88 Absence0.00 Number of appointments in one year per health team (every 100 appointments)0.04-0.02--0.100.20 Number of people served in one year per FH team (every 100 people)0.00-0.13--0.140.97 [^9]
In the multilevel multivariate model, there was a statistical significance of the difference in the general mean PHC score of the health center with advanced access in relation to the others. It also allowed the inference of an inversely proportional relationship between the delay in scheduling appointments and the mean general PHC score, using the advanced access as a parameter to compare the different scheduling models. PHC teams using the traditional model obtained a mean general PHC score worse than that of the carve-out model every 15 days, which in turn had a general mean PHC score worse than that of the weekly carve-out model (β = -2.64; β = -2.36; β = -1.41, respectively). Therefore, all these values of beta (β) were negative, that is, they represent lower general mean PHC scores when compared to the advanced access, keeping constant the age of the users and the panel size.
In addition, the results showed that the greater panel size of a PHC team, the lower their general PHC score, keeping constant the age and the scheduling model. The results show that the increase of 1,000 people reduces, on average, 0.11 in the general mean PHC score.
DISCUSSION
==========
The main result of this article was to identify the association between the scheduling model and the general mean PHC score in Florianópolis. Most of the studies on advanced access point to the increasing productivity and the reducing absenteeism in this scheduling model in comparison to the others[@B9], not addressing the quality of health services.
The multilevel analysis shows that the health center adopting the advanced access obtained a higher general mean PHC score than the health centers with other scheduling models. PHC users expect timely access and high-quality health services. The existing literature shows that advanced access is associated with improved quality of medical appointment[@B18], care provided[@B19] and clinical results of diabetic users[@B20], with a decrease in the delay of medical appointments^9,21--23^ and absenteeism[@B9] ^,^ [@B21].
It was also verified that a longer delay in scheduling appointments is inversely related to the general PHC score. Negative health outcomes are associated with longer delay periods in scheduling PHC appointment. In addition, O'Hare and Corlett[@B19], as well as Lukas et al.[@B24], reported an increase in user satisfaction after reducing delay in scheduling appointments, reinforcing the importance of timely access to PHC centers.
The panel size by PHC team was inversely related to the general PHC score. Large panels in PHC centers are associated with lower quality of disease prevention and health promotion activities, poorer management of chronic diseases, lower technical quality of care provided, and negative interference in effective access and longitudinality of care. However, there are no studies on the effect of panel size on health equity[@B25]. The new National Policy of Basic Care[@B26] (PNAB -- *Política Nacional de Atenção Básica*), of 2017, recommends a panel size of 2,000 to 3,500 users per family doctor. According to Murray and Tantau[@B6], it should be approximately 2,500 people. In England, Kiran et al.[@B27] suggest reducing this number per professional, from 2,500--3,550 to 1,800 people. In the USA, Peterson et al.[@B28] identified that about 50% of the total number of family doctors questioned (n = 11,231) and who devoted 81 to 100% of their time in direct clinical care had a panel of 1,501 to 3,000 users, most of whom (21.3%) reported having between 1,501 to 2,000 users. Masseria et al.[@B29] described the organization of PHC in 14 European countries and found that the average panel size per general practitioner is 2,000 users, except in Poland (4,161 users). Therefore, it is observed the indication of panels smaller than those recommended by the current PNAB in socioeconomic contexts less adverse, iniquitous and consequently pathogenic than the Brazilian ones[@B30].
A key requirement for the deployment of advanced access is the balance between supply and demand in PHC centers. To achieve this balance, two organizational measures are recommended: reduction of the delay in scheduling an appointment and adjustment of the panel size. Considering that in the international literature these two measures are associated with timely access, it seems that the association found between the advanced access and the quality perceived by the users is due to the fact that this scheduling model increases the timely access in the PHC centers evaluated.
Among the limitations of this study is the fact that it does not measure the number of appointments and of people attended by registered nurses, only medical appointments. Another limitation was not to evaluate among the different scheduling models of medical appointments: absenteeism, unmet demand, working time of the health team and time of professional training and qualification, as well as the workday of the professionals. These variables would be useful in the analysis and interpretation of the results found.
FINAL CONSIDERATIONS
====================
This study detected a positive effect of the advanced access to PHC quality. It is expected and understandable that the shift to scheduling models that are more agile and sensitive to the needs of users is positively related to the quality of care provided in PHC centers.
Organizational changes aimed at improving the performance and quality of PHC should include measures that are easily understood by the population and the health team and that have been successfully tested in similar scenarios. For this, it is fundamental to verify the relation between interventions and the presence and extension of the PHC attributes as a measure of PHC quality. Other researches in distinct Brazilian contexts and scenarios are necessary to corroborate the findings reported here.
[^1]: Authors' Contributions: Design and planning of the study: TBV, CDT, EH. Data collection, analysis and interpretation: TBV, CDT, LH. Elaboration or revision of the manuscript: TBV, SAR, CDT, EH, LH. All authors have approved the final version of the manuscript and assume public responsibility for its contents.
[^2]: Conflict of Interest: The authors declare no conflicts of interest.
[^3]: PHC: primary health care
[^4]: PCATool: Primary Care Assessment Tool; PHC: primary health care
[^5]: \* Score ranging from 0 to 10, representing the mean of the score among all individuals interviewed who reported having the health center evaluated as a referral service.
[^6]: SD: standard deviation; Min.-Max.: minimum--maximum
[^7]: PCATool: Primary Care Assessment Tool
[^8]: \* Score ranging from 0 to 10.
[^9]: \* Adjusted through multilevel methodology (individual-level and contextual-level variables).
| {
"pile_set_name": "PubMed Central"
} |
Introduction {#Sec1}
============
The prevalence of doping in sports and fitness is an issue of current concern for a healthy society, and in particular for all those involved in sports, for example for evaluating anti-doping policy measures \[[@CR1]--[@CR3]\]. Its evaluation for the general population, however, is not an easy matter, and even though it is virtually impossible to uncover the exact prevalence of a prohibited activity such as doping, various methods are available to expose parts of this particular problem. These include (i) laboratory-based chemical analyses (in biological matrices such as blood and urine), (ii) questionnaires, (iii) inferences from performances, and (iv) inferences from ego documents. Doping control is widely implemented at professional level (e.g., Olympic games and world championships) and advanced standardized methodologies already exist to detect a great variety of prohibited and controlled substances. However, data are scarce and few scientific articles have addressed the subject so far, with results showing a wide variance in the prevalence of doping at both amateur and professional level. For example, a study showed a prevalence of 2.6% among tertiary education students in Europe \[[@CR3]\], evaluated by anonymous questionnaires, and up to 8.2% in fitness centers in the Netherlands \[[@CR2]\], evaluated by randomized response technique. In the case of elite sports, the range is even larger and likely to be between 14 and 39% \[[@CR1]\].
Recently, analytical chemical methods have been developed that may play a more important role in this topic. Wastewater contains the excreted biomarkers of human metabolism that directly reflect the exposure and stressors placed upon an entire contributing community. The quantitative measurement of these specific biomarkers in wastewater from communities allows the averaged patterns of factors related to lifestyle, disease, and environment to be used for the assessment of community health \[[@CR4]\]. It has been shown that individual communities present different patterns with respect to the levels of various illicit drug residues in wastewater \[[@CR5]--[@CR7]\] and it is hypothesized that this will also concur for other biomarkers \[[@CR4]\]. This approach, known as wastewater-based epidemiology (WBE), has been also used to determine the use of other substances such as erectile dysfunction pharmaceuticals \[[@CR8]\], alcohol \[[@CR9]\], nicotine, and caffeine \[[@CR10]\] and the exposure of a population to pesticides \[[@CR11]\] or phthalate plasticizers \[[@CR12]\]. Hence, there is the clear potential to develop a wider range of innovative solutions to quantitatively assess patterns of factors related to health, fitness, and illness within populations, while also providing means of collecting data for epidemiological and socio-economic studies.
Professional athletes are tested by national and international organizations on the use of doping substances. Information on the exact amounts of doping substances that are actually being taken is mostly unknown. The same holds true for athletes visiting fitness centers. WBE can be a valuable complimentary approach for national doping authorities, as suggested by recent studies \[[@CR1], [@CR13]\]. It can give information on the usage of known compounds in a specified area or group, albeit at the (averaged) group level. The presence of substances that are commonly used either as doping/lifestyle drugs or for the treatment of diverse diseases, has already been identified in wastewaters from fitness centers and wastewater treatment plant (WWTP) influents and effluents \[[@CR14]\], providing a deeper insight into the origin and whereabouts of these compounds.
The present study was conducted to further investigate the application of the WBE approach for assessing the use of doping substances, targeting amateur athletes and other users in the general population at specific sport events. To this aim, a sensitive analytical methodology based on solid phase extraction (SPE) followed by liquid chromatography coupled to high-resolution mass spectrometry was developed and validated. The method was applied to 24-h composite wastewater samples collected at the entrance of three WWTPs and one pumping station while different sport events were taking place within the corresponding sewer catchment area.
Materials and methods {#Sec2}
=====================
Reagents and standards {#Sec3}
----------------------
The analytical standards included in the study were as follows: metandienone, metenolone, mibolerone, nandrolone, sibutramine, clomiphene, tamoxifen and anastrozole, obtained from TRC Toronto Research Chemicals Inc. (Ontario, Canada); ephedrine, norephedrine, trenbolone, clenbuterol, methylhexanamine, 2,4-dinitrophenol, and finasteride, obtained from Sigma-Aldrich (Stenheim, Germany). The chemical properties and structures of the analytes selected are compiled in Table [S1](#MOESM1){ref-type="media"} (see the Electronic Supplementary Material, ESM). The isotopically labeled internal standards (ILIS) used as surrogates for the quantification of their analogue native analytes were the following: methandrostenolone-d~3~, mibolerone-d~3~, trenbolone-d~5~, sibutramine-d~6~, 2,4-dinitrophenol-d~3~, clomiphene-d~5~, tamoxifen-d~5~, anastrozole-d~12~, and finasteride-d~9~, purchased from TRC; ephedrine-d~3~, norephedrine-d~3~, clenbuterol-d~9~, and methylhexanamine-d~4~, purchased from Sigma-Aldrich. The purity of the ILIS was verified (\> 99%) by means of the injection of increasing concentrations up to 160 μg L^−1^, the result being that even in high concentrations no trace of the native signal was detected. In addition, the deuterated standard atrazine-d~5~ was added as external standard to all vials (calibration and samples) prior to analysis in order to monitor the injection and ionization process.
Individual stock solutions were prepared from either the powdered substance or a 1 mg mL^−1^ ampoule solution, in methanol, at a level ranging from 60 to 160 mg L^−1^ for the native compounds and 20 mg L^−1^ for the ILIS, and stored at − 20 °C. Two mix stock solutions containing the native analytical standards and the ILIS separately were prepared at a final level of 1 mg L^−1^ in methanol. Working solutions were prepared by dilution to the desired concentration with methanol and stored at − 20 °C. Calibration curves were prepared daily by diluting with ultrapure water the appropriate mix (containing atrazine-d~5~ as external standard) to a final water/methanol (90:10, *v*/*v*) composition.
Methanol and acetonitrile HPLC-grade solvents, ammonium hydroxide solution (28--30%), and hydrochloric acid were supplied by Avantor Performance Materials B.V (Deventer, The Netherlands). Formic acid, FA (50% in water) was obtained from Fluka Analytical (Sigma-Aldrich, Stenheim, Germany). The ultrapure water was obtained by purifying demineralized water in an Elga Purelab Chorus ultrapure water system (High Wycombe, UK).
Glass fiber filters (type A/E, 1 μm) were purchased from Pall Corporation (Port Washington, NY, USA). SPE cartridges, built of a mixed-mode, reversed-phase/strong cation exchange, water-wettable polymer (Oasis MCX, 150 mg, 6 cm^3^) were obtained from Waters (Milford, MA, USA).
For the calibration of the Sciex TripleTOF mass spectrometer APCI positive and negative calibration solutions were obtained from AB Sciex (MA, US).
Sample collection {#Sec4}
-----------------
Four 15-day sampling campaigns were carried out at four different locations, targeting three different sport events. The 2-week sampling period was planned including samples from a period where no sport event was taking place to determine the background concentrations corresponding to excretions from the regular population. This allowed the comparison to the event period, where visitors and athletes would add up. The samples collected during the four sampling campaigns were stored in HDPE bottles at − 20 °C until analysis.
The targeted sport events and the sampling strategy are described below:
### Event A {#Sec5}
This relatively large 5-day Olympic sport event took place in a large size city (between 500,000 and 1000,000 inhabitants). More than 1000 professional athletes participated, all of who were monitored by the international anti-doping system in place as overseen by the World Anti-Doping Agency. In addition, 125,000 visitor tickets were sold to attend the different events. The sample collection was performed at the entrance of the WWTP, at a 10-km distance from the main event location. The 24-h composite samples were collected in volume proportional mode, with an average sampling frequency of 3.5 min.
### Event B {#Sec6}
The second was a relatively small 1-day bodybuilding event that took place in a small-size city (less than 100,000 inhabitants). It gathered 500 amateur athletes, coaches and volunteers, and more than 800 visitors. There were no anti-doping controls in place for this event. The sample collection was performed at the entrance of the WWTP, at a 3-km distance from the event location. The 24-h composite samples were collected in volume proportional mode, with an average sampling frequency of 12 min.
### Event C {#Sec7}
The third was a relatively large 2-day bodybuilding event that took place in a town (less than 100,000 inhabitants) close to a medium-size city (between 100,000 and 500,000 inhabitants). Over 100 amateur athletes participated and 8000 visitors attended. There were no anti-doping controls in place for this event. The sample collection was performed at two locations, at a pumping station closer to the source (in the town) and at the entrance of the WWTP serving the larger catchment area (that included both the town and the medium-size city). The distance from the event location to the sampling collection points were 5 and 12 km, respectively. At the pumping station, the composite samples were collected in time-proportional mode with a frequency of 5 min, whereas at the WWTP they were collected in volume proportional mode, with an average sampling frequency of 12 min.
Analytical methodology {#Sec8}
----------------------
### Sample treatment {#Sec9}
Fifty milliliters of homogenized sample were spiked at 200 ng L^−1^ with ILIS to act as surrogates during the sample handling and analysis and to correct for possible analyte losses and/or matrix effects. The samples were filtered using a 1-μm type A/E glass fiber filter and acidified to pH 2 to 3 with a solution of hydrochloric acid. Next, the sample was loaded onto a mix-mode cationic polymer-based cartridge (Oasis MCX) previously conditioned with 8 mL of methanol and 8 mL of ultrapure water acidified with 2% FA. After loading the samples, the cartridges were washed with 4 mL of ultrapure water acidified with 2% FA. Next, the cartridges were vacuum dried. Prior to elution, a second washing step with 4 mL of ultrapure water with 5% acetonitrile acidified with 2% FA was performed. Elution was done in two steps, using 4 mL of acetonitrile followed by 4 mL of acetonitrile with 5% NH~4~OH, both collected as one eluate. The eluate was evaporated to 250 μL by means of a Barkey optocontrol (Germany) with a gentle N~2~ stream (block temperature set at 300 °C), where after 250 μL of ultrapure water were added and evaporated again to 250 μL (adjusted by weight) and reconstituted to 0.5 mL of water/methanol 90:10 (*v*/*v*) with a 80:20 solution containing atrazine-d~5~ as external standard.
### Instrumental analysis {#Sec10}
Fifty milliliters of the sample extracts were injected into a UPLC (NEXERA X2 LC-30AD, Shimadzu Corporation, Kyoto, Japan) coupled to a time of flight high-resolution mass spectrometer (TripleTOF 5600+, AB Sciex, MA, US). The LC separation was performed on a XBridge BEH XP C18 column (Waters) with particle size 2.5 μm, and dimensions of 2.1 mm × 150 mm, preceded by a 2.0 mm × 2.1 mm I.D. Phenomenex SecurityGuard Ultra column (Phenomenex, Torrance, USA), at a constant flow rate of 0.250 mL min^−1^. Mobile phase solvents were ultrapure water and methanol both with 0.05% FA (*v*/*v*). The percentage of organic solvent changed as follows: 0 min, 20%; 12 min, 100%; 15 min, 100%; 16 min, 20%; 20 min, 20%. Between consecutive runs, the analytical column was re-equilibrated for 4 min.
The TripleTOF was set up to acquire the full scan in the range of 50 to 800 *m/z* as well as the full scan of the product ions of the target compounds, for which the retention time and optimal collision energy were pre-set. Mass calibration was performed with every batch run just prior to the sequence start. The system operated in positive ionization mode, with the ion spray voltage at 5 kV, source temperature at 500 °C and declustering potential adjusted to 70 V. For 2,4-dinitrophenol, the system was operated in negative mode, with the ion spray voltage at − 4.5 kV, source temperature at 500 °C and declustering potential adjusted to − 70 V. Gas 1, gas 2, and curtain gas were set at 40, 50, and 25 psi, respectively.
Data processing was performed with the MultiQuant 3.0 software, version 3.0.5373.0 (AB Sciex). Analyte concentrations were quantified from the sum of the acquired product ions relative signal (native divided by the corresponding deuterated analogue, when available; when it was not available, a closely resembling deuterated was selected). The acquisition and data processing parameters can be found in Table [S2](#MOESM1){ref-type="media"} (see the ESM). In addition, GraphPad Prism 5 was used for post statistical evaluation of the results.
Analyte concentrations were multiplied with 24-h flow rates to obtain daily loads. For comparison between different cities, daily loads were normalized to the numbers of inhabitants connected to the corresponding sewer system that were obtained from census data provided by the WWTP managers (see ESM Table [S3](#MOESM1){ref-type="media"}). The number of inhabitants used for the calculation was not corrected for the number of extra people attending each event, since the additional number did not change the normalized load with more than 1%, and we assumed that many visitors would be the people living in the cities, and therefore already accounted in the number of inhabitants.
### Method validation {#Sec11}
The method was validated in terms of linearity, limits of detection and quantification, precision intra-day and inter-day (repeatability), procedural recovery (accuracy), and matrix effects by analyzing wastewater spiked with selected analytes.
A calibration curve was established by analyzing spiked ultrapure water and wastewater with standard solutions at 10 different concentrations, ranging from 0 to 50 μg L^−1^, to investigate linearity. Limits of detection and quantification (LOD and LOQ, respectively) were defined as the concentration that provides signal-to-noise (S/N) values of 3 and 10 for the quantifier ion of each analyte. The values were calculated at the lowest point of the calibration curve. Intra-day and inter-day precision was assessed at two levels, 50 and 200 ng L^−1^, with six replicates per level, and during three non-consecutive days. Procedural recovery (%) was calculated as the ratio of the signal of the analyte spiked to a sample after sample treatment (*γ*) against the signal of the analyte spiked to the same sample before treatment (*β*): \[*γ*/*β*\] × 100. Matrix effect (%) was calculated as the ratio of the signal of solution *β* (where the signal of the native compound present in the used sample was subtracted) against the signal of the analyte spiked to ultrapure water (*α*): \[*β*/*α*\] × 100.
### Analyte stability {#Sec12}
It was important to assess the stability of the target compounds in wastewater, in-sewer and in-sample conditions, since the samples were stored at − 20 °C between 1 and 8 months before analysis. Besides, samples were kept in the autosampler at 4 °C during the 24-h cycle. In addition, in the case of the automated sampling at the pumping station (event C), there was sometimes a delay and samples had to be left longer in the autosampler (max 24 h).
To evaluate the stability, wastewater samples were spiked at 400 ng L^−1^ in order to assure quantification of the parent compound, even if it would degrade by tenfold. The stability test was performed in triplicate for each temperature and storage condition. The experiment was scheduled to last a month and a half, with six sampling points during that period, and using three different storage temperature conditions: freezer (− 20 °C), fridge (4 °C), and laboratory room temperature (20 °C). Samples were prepared and stored in polypropylene tubes.
This type of experiment allows measuring the total degradation that compounds might suffer in wastewater due to transformation processes in-sewer and storage; however, it is neither possible to differentiate the type of degradation (chemical, biological, or physical) nor to identify when (during in-sewer transport or storage) the degradation may have occurred \[[@CR15]--[@CR17]\].
Results and discussion {#Sec13}
======================
Selection of target compounds {#Sec14}
-----------------------------
The choice of substances was made in order to detect compounds relevant for actual doping use, such as those promoting muscle growth (anabolic steroids), increasing metabolism by burning fat (weight loss stimulants), or hiding the derived effects or preventing the detection of doping substances use (masking agents). Table [S1](#MOESM1){ref-type="media"} (see the ESM) presents the list of compounds, which it is not all-inclusive but fit-for-purpose for this research, with a description of their type of action and licit formulation (ATC code) when available. Not all of them are mentioned on the 2017 Prohibited List as published by the World Anti-Doping Agency or known to be abused in a fitness-setting. For example, finasteride is not prohibited but mentioned as a confounding factor used to alter athlete's steroid profile and 2,4-dinitrophenol, also not listed but considered relevant due to its use being associated with fatal incidents \[[@CR18]\].
Method development {#Sec15}
------------------
### Sample treatment {#Sec16}
Sample preparation plays an important role in the method development because wastewater is a very complex matrix and target analytes are expected at the low ng L^−1^ concentration level. Preparatory steps such as sample dilution and filtration to minimize matrix interferences as well as the different parameters involved in the concentration by SPE required optimization. For the optimization of the sample treatment, a pooled wastewater reference sample was used as matrix.
First, the effect of diluting the sample was investigated at four dilution levels in triplicate (1× (no dilution), 2×, 5×, 10×). Results showed no impact on the overall recoveries (defined as the comparison between the quantified concentration and the nominal spike value); therefore, dilution of the sample was not performed during the method development. Second, the effect of sample filtration was evaluated for four different filter materials: type GF/F glass microfiber filter (0.7 μm, *Ø* 47 mm, Whatman), regenerated cellulose (0.2 μm, *Ø* 47 mm, Satorius), type A/E glass fiber filter (1 μm, *Ø* 47 mm, Pall), and polyethersulfone (PES) membrane (0.2 μm, *Ø* 90 mm, Nalgene). Removing the particulate phase from the sample by filtering could have implications for the calculation of the daily loads for compounds having a high log *K*~ow~ (or *K*~oc~). For the compounds observed at levels above their LOQ, this holds only for sibutramine and metandienone. Their fraction sorbed could be substantial if the particle concentration would be more than 1 g L^−1^. We estimated/observed particulate phase concentrations in the samples invariably below 1 g L^−1^ and therefore the contribution of the compounds sorbed to the particle phase negligible. Type A/E glass fiber presented the most convenient combination of good recovery and being less affected by matrix effects and was therefore chosen.
Optimization of the SPE procedure was done with the aim of reaching good extraction recoveries for all the target analytes despite differences in physicochemical properties, and concentrating expected influent concentrations from low ng L^−1^ to the μg L^−1^ range. The pKa of the analytes indicated that the majority of species would be either neutral or positively charged at pH = 7 (see ESM Table [S1](#MOESM1){ref-type="media"}), except for 2,4-dinitrophenol which has pKa 4.09 and will therefore be almost entirely (89--100%) in its anionic form at pH 5--9. SPE cartridges Oasis HLB and MCX were selected for the optimization test because of their hydrophilic--lipophilic balance and cation exchange properties, respectively. In addition, C18 cartridges were included as they are commonly used for the analysis of doping substances in human urine. Our results showed that weight loss agents as well as anabolic steroids were hardly or not recovered with C18 cartridges. As a result, the use of the C18 sorbent was renounced. Regarding HLB and MCX, they provided satisfactory results overall, but MCX was selected due to its mixed-mode properties that allowed a better retention of weak bases as well as neutral compounds.
Once the MCX cartridge was chosen, the standard procedure recommended by the manufacturer was adjusted by optimizing the solvent composition used for the elution, evaluating the possibility of dividing the elution into two different steps, and adjusting the solvent volumes. The use of acetonitrile rather than methanol decreased the matrix interferences in the eluate, and adding 5% of NH~4~OH allowed higher recovery for those compounds with lower pKa (namely ephedrines). A two-step elution was then selected, with first 4 mL of acetonitrile (best for anabolic steroids) followed by 4 mL of 5% NH4OH in acetonitrile, combining these to a final eluate of 8 mL.
Finally, the effect of washing was investigated, including up to two steps, one before and/or one after the cartridge drying, with different combinations of ultrapure water and/or 2% FA and/or 5% acetonitrile. Results for anabolic steroids and masking agents improved when a washing step after drying the cartridge was incorporated. Therefore, the final procedure consisted of washing with 2% FA in ultrapure water before drying and another washing step with 5% acetonitrile and 2% FA in ultrapure water after drying.
### LC-MS/MS (HRMS) {#Sec17}
The sensitivity and selectivity of the analytical method was optimized by selecting the most appropriate LC column, the mobile phase composition and gradient, and the injection volume (for the LC part) and the product ions (for the MS part).
Three LC columns were tested: Xbridge C18 150 × 2.1 mm, 2.5 μm particle size, Kinetex 1.7 μm F5 100 A 150 × 2.1 mm, and Kinetex 1.7 μm Biphenyl 100 A 150 × 2.1 mm. The column selected was the Xbridge because it provided sufficient separation for the target compounds and its characteristics were suitable for a wide range of compounds in the case that a screening for suspects was applied. Both, acetonitrile and methanol with 0.05% FA were tested as organic solvent for the mobile phase composition. Acetonitrile provided lower backpressure in the LC column and better peak shape and sensitivity for methylhexanamine. However, the peak shape for three masking agents worsened, and clomiphene isomers were not resolved. To compromise the general peak shape methanol was selected, and to control the backpressure the flow was adjusted to 0.25 mL min^−1^. The mobile phase gradient was important because all anabolic steroids were eluting within the 10--13 min range. Modifications of the gradient were tested to improve the separation, but no clear improvements were observed while the run time increased substantially. In addition, the injection volume was set to 50 μL as a compromise between a higher signal while keeping a Gaussian peak shape. For the MS part, individual analytes were injected in order to obtain the optimal collision energy and the accurate mass of at least one product ion in addition to the accurate mass of the protonated ion. As can be seen in Table [S2](#MOESM1){ref-type="media"} (see the ESM), two transitions were acquired for all compounds, with the only exception of methylhexanamine and 2,4-dinitrophenol, for which only one product ion was used, due to the small molecule size, which made it troublesome to obtain more specific ions and/or with enough sensitivity.
### Analyte stability {#Sec18}
A summary of the results of the stability tests is presented in Table [1](#Tab1){ref-type="table"}, which shows the time points at which the change in concentration compared to the initial concentration is larger than 20%, for each compound and temperature. Figure [S1](#MOESM1){ref-type="media"} (see the ESM) presents the full data evaluation. In order to identify outlying values, for each compound, a stability model was generated based on a quadratic or linear fit, as has been previously done for other illicit substances \[[@CR19], [@CR20]\]. The statistical evaluation revealed that a linear fit was the preferred model in most cases except for sibutramine at 20 °C, trenbolone at − 20 °C, metandienone at 20 °C, and clomiphene at 4 and − 20 °C, where a quadratic fit was preferred. Outliers were identified by the best-fit model evaluation and excluded from the graphs. A variation of ± 20% was not considered significant as it was accepted as the method variability. Our results highlight the importance of immediately storing the samples in the freezer after collection (− 20 °C); otherwise, target compounds might undergo significant degradation. At − 20 °C, only for clomiphene a loss of 25% after 3 days was observed, which could be related to its low solubility in water. When stored at 4 °C, most analytes remained stable in the studied period; however, compounds such as metandienone experienced 50% decrease after 7 days, and metenolone, tamoxifen, and clomiphene suffered a severe degradation (up to 75%) after 3 days. Such degradation could also take place during the automated 24-h cycle of sample collection and might explain the non-detection of these compounds (see below). At room temperature (20 °C), compounds such as sibutramine and metandienone experienced a significant decrease of 50% after 1 week; 2,4-dinitrophenol after 3 days; and clomiphene, tamoxifen, and metenolone already after 1 day. Unexpectedly, storage at 4 or 20 °C led to higher concentrations than the nominal value for some compounds. This was the case of ephedrine or norephedrine that was found to be stable in this study but strongly degraded in another study with same temperature and pH conditions \[[@CR20]\]. A clear explanation was not found, but it could be due to the higher variability in measurements induced by changes in matrix components as a result of storage at higher temperature, or in the case of norephedrine, due to in-sample transformation from amphetamine \[[@CR21]\].Table 1Results of the stability tests conducted at three different temperatures. Results are expressed as the time point (in days) where the change from the initial concentration is higher than 20% for each compound20 °C4 °C− 20 °CNorephedrine30 \< 45\> 45\> 45Ephedrine0 \< 17 \< 14\> 45Methylhexanamine0 \< 10 \< 17 \< 14Clenbuterol30 \< 4530 \< 45\> 45Anastrozole30 \< 4530 \< 4530 \< 45Sibutramine1 \< 314 \< 3030 \< 45Trenbolone0 \< 11 \< 330 \< 45Nandrolone0 \< 10 \< 130 \< 45Metandienone3 \< 77 \< 14\> 45Finasteride0 \< 13 \< 7\> 45Clomiphene0 \< 10 \< 11 \< 3Mibolerone\> 45\> 45\> 45Metenolone0 \< 10 \< 1\> 45Tamoxifen0 \< 10 \< 1\> 452,4-Dinitrophenol1 \< 37 \< 14\> 45
### Method performance {#Sec19}
Table [2](#Tab2){ref-type="table"} summarizes the results of the method validation. The linearity of the response expressed by the regression coefficient (*r*) invariably showed a value above 0.99, without a significant difference between solutions prepared in ultrapure water and wastewater. For the further quantitative analyses, the calibration in ultrapure water was chosen. LODs ranged from 0.2 ng L^−1^ for anastrozole to 20 ng L^−1^ for mibolerone, trenbolone, and nandrolone. LOQs were below 30 ng L^−1^ except for mibolerone, for which only the high concentration level was successfully validated, and trenbolone and nandrolone, for which, depending on the wastewater tested (see ESM Table [S4](#MOESM1){ref-type="media"}), the LOQs slightly varied around 50 ng L^−1^. The precision, expressed as relative standard deviation (RSD%), remained between 2 and 15%, with the exception of drostanolone (33%). Procedural recoveries were in general satisfactory and ranged from 60 to 130%, with the exception of metenolone, which presented the highest recoveries (up to 160%). For this compound, a deuterated analogue was not available, and this might be the reason for such a high value. RSDs remained below 16% in all cases, which indicated low variability. Matrix effects ranged from 43 to 121%, with RSD up to 22%. Norephedrine, anastrozole, nandrolone, metenolone, and drostanolone seemed to be suppressed by the matrix, whereas the response of ephedrine was slightly enhanced. For clomiphene and tamoxifen, the matrix effect could not be evaluated since no signal was detected in the spiked ultrapure water sample, which corresponds to 100% matrix suppression.Table 2Method performance in terms of linearity, limits of detection and quantification, intra-day and inter-day precision, procedural recovery, and matrix effectsLinearityLimitsPrecisionRecovery^c^Matrix effect^c^*r*LODLOQIntra-day RSD (%) (*n* = 6)Inter-day RSD (%) (*n* = 18, day = 3)\[*γ*/*β*\] ± RSD (%) (*n* = 6)\[*β*/*α*\] ± RSD (%) (*n* = 6)ng L^−1^ng L^−1^50 ng L^−1^200 ng L^−1^50 ng L^−1^200 ng L^−1^50 ng L^−1^200 ng L^−1^50 ng L^−1^200 ng L^−1^Norephedrine0.99973104443132 ± 784 ± 982 ± 993 ± 7Ephedrine0.99973105243125 ± 9105 ± 9121 ± 1289 ± 5Methylhexanamine^a^0.9989725968863 ± 1570 ± 492 ± 1687 ± 12Clenbuterol0.998731043101092 ± 769 ± 1596 ± 7100 ± 5Anastrozole0.99880.20.74871292 ± 859 ± 1474 ± 887 ± 9Sibutramine0.99970.725387130 ± 11104 ± 1164 ± 1291 ± 22Trenbolone0.99232050455991 ± 970 ± 16101 ± 796 ± 10Nandrolone0.999420501211714104 ± 1499 ± 1589 ± 1877 ± 18Metandienone0.99937254264105 ± 885 ± 1195 ± 396 ± 9Finasteride0.99340.7226121384 ± 1061 ± 14107 ± 14102 ± 16Clomiphene0.99977254210596 ± 1073 ± 14n.a.n.a.Mibolerone^b^0.99862060~~--~~3~~--~~3~~--~~99 ± 4~~--~~111 ± 5Metenolone0.9986310441414160 ± 16147 ± 543 ± 1045 ± 16Tamoxifen0.9996725628398 ± 969 ± 12n.a.n.a.2,4-Dinitrophenol0.99959306513777 ± 1660 ± 2065 ± 788 ± 5LOD and LOQ obtained for pooled wastewater sample. Table [SI](#MOESM1){ref-type="media"}-4 presents LODs and LOQs specific for each of the sampling locations*n*.*a*. not available, compound non-detected in ultrapure water (*α*), equivalent to 100% matrix suppression^a^*n* = 5^b^Only one level was successfully validated for this compound^c^For explanation of symbols used in recovery and matrix effect calculations, see the text
Application to wastewater samples {#Sec20}
---------------------------------
### Local wastewater characteristics {#Sec21}
As already highlighted above, wastewater is a complex matrix. The method validation presented in the previous section corresponds to the performance obtained with an in-house reference sample made from pooled wastewater. However, the composition of wastewater is very much affected by geographical location and temporal characteristics. For this reason, the procedural recovery and the limits of detection and quantification were re-evaluated for each location sampled. To this end, with every sample batch, three samples were analyzed in duplicate one being spiked at 200 ng L^−1^. Results are presented in Table [S4](#MOESM1){ref-type="media"} (see the ESM).
The variability of the recoveries remained below 25% in all but three cases. Ephedrine showed high variability in the samples from WWTPs A and C due to higher noise signals in the chromatograms. Regarding the limits of detection and quantification, in general, they were in line (similar range) with those obtained with the reference-pooled wastewater, although some differences were observed. In the case of methylhexanamine and clenbuterol, lower LOQs were observed at all locations, whereas for ephedrine, anatrozole, finasteride, mibolerone, metenolone, clomiphene, tamoxifen, and drostanolone LOQs were higher at all locations. For the remaining compounds (norephedrine, sibutramine, trenbolone, nandrolone, and metadienone) LOQs were in the same range as those previously estimated in the reference-pooled wastewater.
### Results obtained for the wastewater influents {#Sec22}
The concentrations calculated in wastewater influents were transformed into influent loads (expressed in mg day^−1^, see Tables [S5](#MOESM1){ref-type="media"} to [S8](#MOESM1){ref-type="media"} in the ESM) by multiplying with the daily influent flow rates provided by the WWTP operators (presented in ESM Table [S3](#MOESM1){ref-type="media"}).
#### Event A {#FPar1}
The results obtained for the samples collected at the entrance of the WWTP A (ESM Table [S5](#MOESM1){ref-type="media"}) showed the presence of three weight loss compounds in all the samples: ephedrine with the highest loads, ranging from 70 to 122 g day^−1^, followed by norephedrine, ranging from 6.6 to 14 g day^−1^, and methylhexanamine, ranging from 7 to 13 g day^−1^. From the other weight loss products, sibutramine, was quantified in 4 out of the 15 samples analyzed, ranging from 776 to 2910 mg day^−1^, and detected in other 4 samples \< LOQ, and 2,4-dinitrophenol was quantified in 1 sample at 4300 mg day^−1^ and detected in other 1 sample \< LOQ. The anabolic steroid metandienone was quantified in 5 out of the 15 samples, ranging from 1480 to 2020 mg day^−1^, and detected in other 8 samples \< LOQ. Clenbuterol was detected in 1 sample \< LOQ. None of the remaining compounds were detected (i.e., no signal or signal below LOD).
Statistical evaluation of the loads during the event and the corresponding reference weekend using two-tailed Mann-Whitney test (*p* \< 0.05 significant difference) showed that the medians of both periods were significantly different for norephedrine (*p* = 0.03). This suggested an enhanced use of norephedrine during the event weekend. This stimulant is most commonly used to reduce body fat before a sport event by fast increasing human metabolism. However, it is also a metabolite of the recreational substance amphetamine \[[@CR21]\], and therefore, the increase could also be caused by an enhanced consumption of this illegal substance (and its degradation from wastewater \[[@CR20]\]).
#### Event B {#FPar2}
The results obtained for the samples collected at the entrance of the WWTP B (ESM Table [S6](#MOESM1){ref-type="media"}) showed the presence of three weight loss compounds in all the samples: ephedrine with the highest loads, ranging from 6 to 15 g day^−1^, followed by norephedrine (457 to 894 mg day^−1^), and methylhexanamine (254 to 710 mg day^−1^). From the other weight loss products, sibutramine was quantified in 1 out of the 15 samples analyzed, at 90 mg day^−1^, and was detected, albeit below LOQ, in another 10 samples, and 2,4-dinitrophenol was quantified in 3 samples ranging from 516 to 6430 mg day^−1^ and present at levels between LOD and LOQ in another 6 samples. The anabolic steroid metandienone was detected in all of the samples at levels between LOD and LOQ.
An increase in loads just prior to the event weekend can be observed for ephedrine, norephedrine, methylhexanamine, and 2,4-dinitrophenol. This increase in loads might be explained by the fact that the athletes arrived already on Friday or Saturday. Statistically evaluating the use of ephedrine, norephedrine, and methylhexanamine in the two weekends (and the days prior to the weekend) using a two-tailed Mann-Whitney test, a significant difference between the medians was observed for norephedrine and methylhexanamine (*p* = 0.03 in both cases) but not for ephedrine. In the case of 2,4-dinitrophenol, there was not sufficient data for a statistical evaluation. However, its use just prior to and during the sport event was obvious as it was found to be present at quantifiable concentrations only in those days (ESM Table [S6](#MOESM1){ref-type="media"}), while being \< LOQ during the reference weekend (see also section "[Normalized loads](#Sec23){ref-type="sec"}" below). We conclude that weight loss products had been used on the days prior to the sport event B. The case of 2,4-dinitrophenol was particularly noteworthy, because its use has been associated with severe adverse effects and even risk of death \[[@CR18]\].
#### Event C {#FPar3}
The results obtained for the samples collected at the pumping station C and the entrance of the WWTP C (ESM, Tables [S7](#MOESM1){ref-type="media"} and [S8](#MOESM1){ref-type="media"}) revealed again the presence of three weight loss compounds in all of the samples. The highest loads were observed for ephedrine, ranging from 4 to 33 g day^−1^ at the pumping station and 29 to 108 g day^−1^ at the WWTP entrance, respectively. Norephedrine loads ranged from 341 to 1400 and 4540 to 8790 mg day^−1^, respectively, and methylhexanamine loads from 594 to 2020 and 4450 to 125,000 mg day^−1^, respectively. From the other weight loss products, sibutramine was quantified in 6 out of the 14 samples analyzed in wastewater from pumping station C, ranging from 22 to 335 mg day^−1^, and detected in 2 more samples at a level between LOD and LOQ. In the influent of WWTP C, sibutramine was quantified in 5 out of the 16 samples analyzed, ranging from 295 to 1660 mg day^−1^, and detected in 6 more samples at concentrations between LOD and LOQ. 2,4-Dinitrophenol was quantified in 7 out of the 14 samples analyzed in wastewater from pumping station C, ranging from 255 to 9740 mg day^−1^, and detected in 5 more samples at a level between LOD and LOQ, and in the case of WWTP C 2,4-DNP was quantified in 6 out of the 16 influent samples analyzed, ranging from 6390 to 717,000 mg day^−1^, and detected in 10 more samples at a level between LOD and LOQ. Metandienone was quantified in 4 out of 16 samples from the influent of WWTC C, ranging from 761 to 1190 mg day^−1^. Clenbuterol, clomiphene, and nandrolone were just detected at levels between LOD and LOQ (clenbuterol, 1 sample at each location; clomiphene, 2 samples from pumping station C; nandrolone, 2 samples from WWTP C).
Statistical evaluation of the loads during the event and the corresponding reference weekend using two-tailed Mann-Whitney test (*p* \< 0.05 significant difference) showed that the medians of both periods and both locations were not significantly different for ephedrine, norephedrine, and methylhexanamine. Once again in the case of 2,4-dinitrophenol, there was not sufficient data for a statistical evaluation; however, its use just prior to and during the sport event was obvious (at both locations) due to the relatively high levels quantified during those days, while loads were much lower or below LOQs in the reference weekend (see also the next section).
### Normalized loads {#Sec23}
The comparison of the normalized loads per compound is presented in Fig. [1](#Fig1){ref-type="fig"}. In the case of norephedrine, ephedrine, and methylhexanamine, the trend is similar. Lower normalized loads correspond to the small-size city sampled at WWTP B. In the large and medium-size cities (A and C), the normalized loads appear to be similar. One characteristic that might support this difference between A and C against B is the relative number of students in each of the cities. Whereas in A and C, the percentage of residing students represents a 14% of the total population, in B it only represents 0.2% \[[@CR22]\]. In the past, many studies have revealed the use of stimulant drugs by college students for cognitive enhancement, specially the use of methylphenidate, the active pharmaceutical ingredient for the treatment of attention deficit hyperactivity disorder, but also other type of prescription and lifestyle stimulant drugs. A recent study in the Netherlands showed evidence of polydrug use with this purpose, although with lower prevalence than in other countries in Europe (1.6 versus 4.6--16%) \[[@CR23]\].Fig. 1Normalized loads (expressed in mg day^−1^ 1000 inh^−1^) of the most detected weight loss products (ephedrine, norephedrine, methylhexanamine, and 2,4-dinitrophenol) per sampling location
It is noteworthy that the normalized loads from the pumping station, which correspond to a small part (20%) of the larger city C, also match with the larger cities (A and C). This might be explained by the effect of the sport event. Whereas event A was held for professional athletes and official anti-doping controls were performed, events B and C were held for amateur athletes and no anti-doping measures were taken. Especially event C is also known among the amateur bodybuilding community, and not only the participants might use illicit substances but also the event visitors.
The normalized loads of 2,4-dinitrophenol clearly reveal an aberrantly high load during the B and C events. This compound was originally used in the manufacture of dyes, wood preservatives, and as a pesticide. However, another use was discovered as solution for rapid weight loss. Although it seems an effective solution to this end, it is highly toxic, and even small overdoses have been reported to result in death \[[@CR24], [@CR25]\]. The high loads found during the events (including the day before the events start, but not before, and not after) might indicate the use as weight loss, since such a sudden increase/decrease would not be expected in any other of its known uses, for example as a pesticide. It is therefore of great concern to discover its use, and measures should be taken to inform and bring awareness to the athletes community, not only restricted to anti-doping.
Finally, an interesting finding is the relative absence of detection of analytes from the group of anabolic steroids. However, it is important to stress that the targeted analytes in this method are the parent compounds, and they might not be the best biomarkers due to excretion as conjugates or in the form of metabolites or transformation products. Temperatures of the wastewaters were such that for a few compounds some degradation may have occurred in-sewer (see Table [1](#Tab1){ref-type="table"} and ESM Table [S3](#MOESM1){ref-type="media"}). Therefore, estimation of the use of these compounds in wastewater would require a more in-depth study on the most suitable biomarkers to be determined in this complex matrix.
Conclusions {#Sec24}
===========
Chemical analysis of wastewater can reveal the use of doping agents by the general population and during sport events. The results of the present study provided valuable information that can be of interest for anti-doping authorities. The analytical methodology developed in this work, based on the use of LC-MS/MS, allowed the detection and quantification in wastewater of doping substances used by the general population and amateur athletes attending the targeted events. Weight loss stimulants, namely ephedrine, norephedrine, and methylhexanamine, were found in high amounts. In addition the detection of 2,4-dinitrophenol is a major concern due to its known adverse effects. The results suggested the increase in loads of some substances possibly during the monitored sport events.
Further refinement of analysis to include metabolites and transformation products would provide valuable information, especially in the cases of compounds rapidly metabolized by the human body and therefore not present in the urine and wastewater (or present in minor quantities).
Electronic supplementary material
=================================
{#Sec25}
ESM 1(PDF 1028 kb)
**Electronic supplementary material**
The online version of this article (10.1007/s00216-017-0835-3) contains supplementary material, which is available to authorized users.
This project has been carried out with financial support of the World Anti-Doping Agency. Ana Causanilles acknowledges the European Union for her Early Stage Researcher contract as part of the EU-ITN SEWPROF (Marie Curie PEOPLE Grant No. 317205). The authors are grateful to the people and agencies that provided the wastewater samples.
Conflict of interest {#FPar4}
====================
The authors declare that they have no conflict of interest.
| {
"pile_set_name": "PubMed Central"
} |
Introduction
============
On December 31, 2019, the World Health Organization (WHO) was alerted to a case of pneumonia of unknown cause that originated from China \[[@ref1]\]. The causative pathogen was a novel coronavirus, severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), and the disease it causes is now referred to as coronavirus disease (COVID-19) \[[@ref1]\]. As numerous other individuals became afflicted, the WHO declared the outbreak of COVID-19 to be an international public health emergency \[[@ref1]\]. The problem escalated, with many individuals across different countries being afflicted; on March 11, 2020, the WHO increased their alert to a pandemic \[[@ref1]\]. The rapid escalation of the number of individuals infected, the increasing number of deaths, and the measures undertaken by governments have resulted in significant psychological impact not only on health care workers, but also on the public \[[@ref2]\].
In the four months after the discovery of the index case, several studies highlighted the psychological impact of COVID-19 on frontline health care workers and on members of the public. Li et al \[[@ref3]\] used a mobile phone app to administer the Chinese version of a vicarious traumatization questionnaire to 214 individuals from the public and 526 nurses (234 frontline nurses and 292 nurses not on the front line). They found that the vicarious traumatization scores were higher among the public and nonfrontline health care workers. Wang et al \[[@ref4]\] investigated the mental health status of 1120 members of the public living in China using both the Impact of Event Scale--Revised (IES-R) and the Depression, Anxiety, and Stress Scale (DASS-21). Notably, 53.8% of respondents reported the psychological impact to be moderate to severe, with 16.5% having moderate to severe depressive symptoms, 28.8% having moderate to severe anxiety symptoms, and 8.1% having moderate to severe stress. These studies focused on the immediate psychological impact of the COVID-19 pandemic; meanwhile, in their most recent study, Wang et al \[[@ref5]\] examined the longitudinal changes in the mental health of the general population in China. A total of 1738 participants were included, and they were administered the IES-R questionnaire and the DASS-21 at baseline and after 4 weeks. While there was a mean reduction in the overall scores across a period of 4 weeks, the mean scores for the IES-R scale were still high, suggesting the presence of posttraumatic stress disorder (PTSD) symptoms \[[@ref5]\]. It is evident from these studies that individuals experienced elevated levels of anxiety and depression in the acute phase, when they first became aware of the pandemic, and that this psychological distress persisted into subsequent weeks.
Since the onset of the pandemic, the priorities for government have understandably been the treatment of people infected with COVID-19 and steps to limit spread. It is evident in many countries that the need for clinical services has exceeded the supply, requiring the construction of temporary medical facilities and redeployment of staff. Beyond hospitals, the impact of COVID-19 is multifactorial; economies are being affected, individuals are suffering bereavement after loss of loved ones, and others are physically isolated and quarantined. Social distancing and lockdowns have created difficulties in accessing mental health services. Individuals with psychiatric disorders are likely to have increasing difficulty accessing conventional mental health services. Hao et al \[[@ref6]\] reported that mean PTSD, anxiety, depression, and insomnia scores were elevated in psychiatric patients compared to those in the general population. A need indeed exists for web-based mental health interventions that use digital tools. An editorial has also been published describing how web-based tools and social media have been used in China to support the mental health needs of frontline workers as well as of people who are infected or living in quarantine facilities \[[@ref7]\]. It is becoming apparent that technological tools such as SMS text messages, web-based interventions, mobile interventions, and conversational agents can help ameliorate psychological distress in the workplace and society. In an opinion paper, Zhou et al \[[@ref8]\] described the efforts of the Australian government to provide telemental health solutions to address the psychological impact of COVID-19. Unfortunately, the scope of this paper was limited to the identification of relevant services in Australia. In another recent paper, Cosic et al \[[@ref9]\] highlighted the potential of digital tools in dealing with the psychological distress associated with COVID-19 and how their prior experiences can be applied in developing relevant apps; however, the authors failed to identify any existing digital tools that individuals can use. To our knowledge, except for the paper by Zhou et al \[[@ref8]\], few publications have described how digital tools are being used to ameliorate psychological symptoms among individuals. Thus, our aim in this paper was to identify existing SMS text message, web-based, mobile, and conversational agents that the public can access to ameliorate the psychological symptoms they are facing during the COVID-19 pandemic.
Methods
=======
To identify digital tools that have been published specifically for COVID-19, a search was performed through PubMed and MEDLINE from the inception of the databases through June 17, 2020. The following search strings were used: "*NCOV* OR *2019-nCoV* OR *SARS-CoV-2* OR *Coronavirus* OR *COVID19* OR *COVID*" and "*mHealth* OR *eHealth* OR *text*". Given that the aim was to identify potential digital tools, the terms electronic health (*eHealth*), mobile health (*mHealth*), and *text* were used because these terminologies would identify all potential web-based, mobile, and SMS text message interventions.
Another search was conducted on PubMed and MEDLINE to identify existing digital tools for depression and anxiety disorders. To identify these tools, reviews of digital tools (SMS text messaging, web-based interventions, mobile apps, and conversational agents) were identified. Only articles in the English language were considered. A narrative synthesis of the identified tools was conducted.
A web-based search engine (Google) was used to identify if the cited web-based interventions could be accessed. A mobile app search engine, App Annie \[[@ref10]\], was used to identify if the mobile apps identified in the literature search were commercially available. All available data have been included in the manuscript.
Results
=======
Based on the search strategy, a total of 9829 articles were identified from the databases. Of these, 80/9829 (0.8%) were duplicated references. Upon further screening, a total of 24/9829 articles (0.2%) were identified as potentially relevant to COVID-19. Upon further examination of the full texts of these 24 articles, only one article described how an SMS text messaging intervention was applied to address mental health issues resulting from COVID-19 \[[@ref11]\]. The remaining articles did not describe or mention how the discussed SMS text message, web-based, or mobile interventions helped ameliorate the psychological symptoms associated with COVID-19.
In our search for existing digital tools to manage depression and anxiety disorders, a total of 5 articles were identified. Two web-based reviews were identified for depressive disorders, and another was identified for anxiety disorders. Of the identified web-based interventions, we managed to access 5 of the 6 listed websites (83%), namely Beating the Blues \[[@ref12]\], Living Life to the Full \[[@ref13]\], Deprexis \[[@ref14]\], moodgym \[[@ref15]\], and Interapy \[[@ref16]\]. One review of mobile interventions for depressive and anxiety disorders was identified. Of the identified mobile interventions, only 7/32 apps (22%) were available in commercial stores, namely Angesthjalpen, AnxietyCoach, SmartCAT, Headgear, MoodHacker, SuperBetter, and Thought Challenger. Another review highlighted conversational agents for psychiatric disorders. Of the 7 identified conversational agents, only 2 (29%) were commercially available. [Table 1](#table1){ref-type="table"} provides a summarized overview of the identified studies.
######
Overview of the studies identified in the literature search.
-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
Study Year Mechanism of delivery of digital tools Identified digital tools and prior evaluations Availability
------------------------------------- ------ ---------------------------------------------- ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- ------------------------------------------------------------------------------------------------------------------------------
Agyapong et al \[[@ref11]\] 2020 SMS text messaging (specific to COVID-19^a^) Text4Hope (specific to COVID-19) enables subscribers to receive 3 months of daily supportive SMS text messages with or without web links to web-based mental health resources. Only available to individuals living in Alberta, Canada.
Rodriguez-Pulido et al \[[@ref17]\] 2020 Web-based Interventions Beating the Blues is a web-based intervention for depressive disorder.\ Available to users in the United Kingdom.
\
\
\
Burger et al \[[@ref18]\] 2020 Web-based interventions Living Life to the Full (2 comparative trials were performed involving a total of 659 participants), Deprexis (6 comparative trials were performed involving a total of 1863 participants), and SHADE (3 comparative trials performed involving a total of 475 participants) were evaluated. moodgym was extensively evaluated, with a total of 11 comparative trials involving a total of 7294 participants. All the above interventions have been evaluated as websites that provide psychological therapy for depressive disorders. The websites for Living Life to the Full, Deprexis, and moodgym can be accessed.
Anderson et al \[[@ref19]\] 2019 Web-based interventions The Interapy program from the Netherlands was highlighted as a program that assisted individuals with symptoms of depression, panic disorder, posttraumatic stress disorder, and burnout. moodgym was also highlighted as a commercially available option for anxiety and depression. The websites for moodgym and Interapy can be accessed.
Miralles et al \[[@ref20]\]\ 2020 Mobile interventions\ 7 Cups, Be Good to Yourself, Bluewatch, Dcombat, Get Happy Program, Headgear, iCare Prevent, MedLink, Mobile Sensing and Support, Moodhacker, Moodivate, MyGamePlan, PRIME-D, Push-D, SocioEmpathy, SPSRS, SuperBetter, The Sound Advice, Thought Challenger, TODAC, Kokoro-App, Agoraphobia free, Stress Free, Angesthjalpen, AnxietyCoach, CBT Assistant, Challenger, Lantern, Psych Assist, Public Speech Trainer, SmartCAT, and GET.ON.PAPP have been previously evaluated and reported in published research. Headgear, MoodHacker, SuperBetter, Thought Challenger, Angesthjalpen, AnxietyCoach, and SmartCAT are commercially available.
\ \
Gaffney et al \[[@ref21]\] 2019 Conversational agents or chatbots Woebot, Tess, and eSMART-TH have been evaluated previously for depressive disorder, SABORI has been evaluated for psychological distress, and Tess has been evaluated for anxiety disorder. Woebot and Tess are commercially available.\
\
\
\
-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
^a^COVID-19: coronavirus disease.
Discussion
==========
Principal Findings
------------------
This review is one of the first to examine the literature for digital interventions that can be used by the general public as well as specific groups, such as workers and health care professionals, to ameliorate the psychological distress they are experiencing during the COVID-19 pandemic as well as specific symptoms such as panic buying \[[@ref22]-[@ref25]\]. The findings from our paper complement those of Zhou et al \[[@ref8]\], who highlighted tools in their article that can be accessed by individuals in Australia. Our review helped address some of the inherent limitations of the work by Zhou et al \[[@ref8]\], given that the authors only listed available resources without providing any evidence-based justification of the suggested interventions. We identified an SMS text message--based intervention that was designed to address the mental health needs of individuals in Canada. We managed to identify web-based, mobile, and conversational agents that are commercially available and have been previously validated by research.
Our review only identified one publication that describes how SMS text messaging technologies are used as a form of psychological support. As mentioned in the Introduction, numerous studies have been published that characterize the immediate and delayed psychological impact of COVID-19 on medical workers and members of the public \[[@ref2],[@ref4],[@ref5]\]. There is still a lack of evaluation of psychological tools to address the identified psychological concerns, namely heightened levels of stress, anxiety, and depression. Psychological distress occurs frequently in everyday life; however, during epidemics and pandemics such as COVID-19, the prevalence of distress is extremely high and existing mental health services are unable to function normally. This provides strong justification for the rapid identification of tools that have an evidence base and can be promoted rapidly to address the unmet psychological needs of individuals. Our research highlighted commercially available digital tools (web-based, mobile, and conversational agents) that individuals can access during the COVID-19 pandemic. It is challenging for members of the public to identify tools that have been proven to be clinically effective and are available commercially; hence, our review is important. We focused primarily on reviews of digital tools to identify such tools. Our methods helped address the limitations of the prior work by Zhou et al \[[@ref8]\], as they suggested tools but made no attempt to review the evidence base of those tools. The evidence-based websites, smartphone apps, and conversational agents we identified can help ameliorate symptoms of depression and anxiety. These tools have wide application; they can help individuals who are at risk of developing an illness or help individuals with pre-existing illness to cope with these symptoms. This is important because it is anticipated that it will be challenging for people to obtain appropriate psychiatric care as governments impose lockdowns and curb movement.
We identified a variety of commercially available tools; however, there may be limitations to some of these tools. Validations may have been conducted in certain localities or regions, and we cannot be sure that these tools will be as effective in other localities. However, we propose that a tool that has undergone validation, if only in a precise locality, is likely to be superior to a tool without any validation. We also recognize that accessing smartphone apps may be difficult in a different region or country.
It is evident from this paper that psychological tools to help individuals cope with heightened stress, anxiety, and depression due to COVID-19 are lacking. While some commercial tools are available, they are not without limitations. It is important for academic researchers, clinicians, and developers to work jointly to conceptualize tools that can be used by the general population to ameliorate their symptoms of psychological distress. It may also be valuable to consider participatory action research design when conceptualizing new tools to ensure that the created tools better meet the needs of individuals. It may be wise to consider modification of existing tools so that versatile tools can be rapidly deployed to meet the increasing need. In the interim period, while comprehensive treatment tools may not yet be available, we learned from the editorial by Liu et al \[[@ref7]\] that in China, helplines and social media platforms are being used to extend support to individuals who are experiencing psychological distress. Similarly, in other countries such as Singapore, the government has set up a mental health hotline to address the needs of the public and to refer at-risk individuals to appropriate mental health services. These clinical services provide some form of supportive therapy; however, there is still a need for tools that provide individuals with more comprehensive treatment, such as cognitive behavioral therapy for depression or anxiety.
Strengths and Limitations
-------------------------
The strength of this paper is that we examined the literature for digital tools that have been validated, which can help with depressive and anxiety symptoms. In addition, we conducted a search to determine if these tools were available commercially. It is possible, despite these strengths, that we have missed some tools; to mitigate this, we included recently published reviews, but we acknowledge that their search for digital interventions may not be as recent.
Conclusions
-----------
The COVID-19 pandemic has caused significant psychological distress. Commercially available digital tools may be useful for at-risk individuals or individuals with pre-existing psychiatric symptoms. The tools we identified may help address the psychological distress individuals are experiencing during the COVID-19 pandemic.
This work received no funding.
Authors\' Contributions: MZ and HS conceptualized the study. MZ worked on and wrote the initial draft of the manuscript. HS revised the initial draft. All authors approved the manuscript prior to submission.
Conflicts of Interest: None declared.
COVID-19
: coronavirus disease
DASS-21
: Depression, Anxiety, and Stress Scale
eHealth
: electronic health
IES-R
: Impact of Event Scale--Revised
mHealth
: mobile health
PTSD
: posttraumatic stress disorder
SARS-CoV-2
: severe acute respiratory syndrome coronavirus 2
WHO
: World Health Organization
| {
"pile_set_name": "PubMed Central"
} |
1. Introduction {#sec1-ijms-20-05315}
===============
Aloperine, a quinolizidine alkaloid isolated from *Sophora alopecuroides*, reportedly exhibits multiple pharmacological activities, including anti-inflammatory, antiallergenic, antiviral, antimicrobial, antinociceptive, and antitumor effects; moreover, it exhibits protective effects against renal and neuronal injuries as well as pulmonary fibrosis \[[@B1-ijms-20-05315],[@B2-ijms-20-05315],[@B3-ijms-20-05315],[@B4-ijms-20-05315],[@B5-ijms-20-05315],[@B6-ijms-20-05315],[@B7-ijms-20-05315],[@B8-ijms-20-05315],[@B9-ijms-20-05315],[@B10-ijms-20-05315],[@B11-ijms-20-05315],[@B12-ijms-20-05315],[@B13-ijms-20-05315],[@B14-ijms-20-05315],[@B15-ijms-20-05315],[@B16-ijms-20-05315],[@B17-ijms-20-05315],[@B18-ijms-20-05315],[@B19-ijms-20-05315],[@B20-ijms-20-05315]\]. Aloperine exerts antitumor effects against multiple human cancers, including multiple myeloma; colon, breast, prostate, and thyroid cancers; and osteosarcoma \[[@B1-ijms-20-05315],[@B7-ijms-20-05315],[@B12-ijms-20-05315],[@B15-ijms-20-05315],[@B18-ijms-20-05315],[@B20-ijms-20-05315]\]. In previous studies, our and other research groups have demonstrated that aloperine can suppress tumorigenesis, inhibit tumor cell proliferation, and induce cell cycle arrest and apoptosis in various human cancer cells \[[@B1-ijms-20-05315],[@B7-ijms-20-05315],[@B12-ijms-20-05315],[@B15-ijms-20-05315],[@B18-ijms-20-05315],[@B19-ijms-20-05315]\]. Studies investigating the mechanisms underlying the antitumor activity of aloperine have demonstrated the modulation of various signaling pathways in human tumors, including JAK/STAT3, Ras-Erk, and PI3K/Akt pathways, following aloperine treatment \[[@B1-ijms-20-05315],[@B7-ijms-20-05315],[@B13-ijms-20-05315],[@B15-ijms-20-05315],[@B19-ijms-20-05315],[@B20-ijms-20-05315]\]. Furthermore, an in vivo study has demonstrated the safety and efficacy of aloperine as a therapeutic agent \[[@B7-ijms-20-05315]\]. Although antitumor activities of aloperine have been reported in various studies, its detailed mechanism of action and possible further anticancer effects should be elucidated urgently.
Autophagy, a self-degradative mechanism, disassembles unnecessary or dysfunctional components in cells, thereby maintaining homeostasis and intracellular energy balance. Under stressful conditions, such as nutrient deprivation, hypoxia, or infection, autophagy is induced to eliminate stress and maintain homeostasis for cell survival \[[@B21-ijms-20-05315],[@B22-ijms-20-05315]\]. During autophagy activation, double-membrane vesicles (autophagosome or autophagic vacuole) are formed and specific or non-specific target cargos are recruited in the cytoplasm for disassembling and recycling unnecessary or dysfunctional cellular components via lysosomes (autolysosomes) \[[@B22-ijms-20-05315]\]. Currently, autophagy is implicated in various physiological responses and pathophysiological processes, including aging, metabolic disorders, neurodegenerative diseases, cardiovascular disorders, immune responses, carcinogenesis, and infectious diseases \[[@B23-ijms-20-05315]\]. Autophagy plays a dual role in cancer. It can promote cancer growth and survival by maintaining cellular energy production and eliminating stress, but it has also been recognized as a therapeutic strategy against cancer \[[@B24-ijms-20-05315]\]. However, there is insufficient evidence regarding the modulation of autophagy by aloperine. Therefore, the regulation of autophagy machinery and related mechanisms following aloperine treatment are worth exploring.
We have previously demonstrated anticancer activity of aloperine in human thyroid cancer cells through elevated caspase-dependent apoptosis via the PI3K/Akt signaling pathway \[[@B15-ijms-20-05315]\]. The PI3K/Akt/mTOR signaling pathway plays important roles in autophagy regulation \[[@B25-ijms-20-05315]\] and is a potential target for anticancer therapy. In the present study, we evaluated autophagy in multiple human thyroid cancer cells following aloperine treatment. Furthermore, we investigated the upstream signaling pathway involved in aloperine-mediated autophagy modulation.
2. Results {#sec2-ijms-20-05315}
==========
2.1. Aloperine Reduces Cellular Viability in Human Thyroid Cancer Cells {#sec2dot1-ijms-20-05315}
-----------------------------------------------------------------------
Various studies have demonstrated that aloperine can induce apoptosis in human cancer cells, including those of multiple myeloma; osteosarcoma; and hepatocellular, breast, colon, thyroid, and prostate cancer \[[@B1-ijms-20-05315],[@B7-ijms-20-05315],[@B12-ijms-20-05315],[@B15-ijms-20-05315],[@B18-ijms-20-05315],[@B19-ijms-20-05315]\]. In the present study, we confirmed the effects of aloperine on cellular viability in undifferentiated human thyroid cancer cells as well as in multidrug-resistant anaplastic and papillary thyroid cancer cells. The viability of KMH-2, 8505c, and IHH-4 cells treated with various concentrations of aloperine was significantly reduced in a dosage-dependent manner compared with that of the control cells ([Figure 1](#ijms-20-05315-f001){ref-type="fig"}). At 48 h post-incubation with aloperine, 50% cytotoxic concentration (CC~50~) for KMH-2, 8505c, and IHH-4 cells was 207.3, 268.4, and 169.4 µM, respectively, suggesting that KMH-2 and IHH-4 cells are more sensitive to aloperine comparing with 8505c cells. Whether this difference in these cells are due to genetic background or other reasons warrants further investigation.
2.2. Aloperine Promotes Autophagy Activation in Human Thyroid Cancer Cells {#sec2dot2-ijms-20-05315}
--------------------------------------------------------------------------
To evaluate whether aloperine can modulate cellular autophagy activity, human thyroid cancer cells were treated with aloperine at various concentrations. The expression of LC3-II, a biomarker for autophagosome formation, was determined using western blot. Elevated LC3-II expression was observed in KMH-2, 8505c, and IHH-4 cells following aloperine treatment ([Figure 2](#ijms-20-05315-f002){ref-type="fig"}). Moreover, LC3-II and p62 expression was decreased in KMH-2 and IHH-4 cells following treatment with aloperine at a high concentration or for a long time ([Figure 2](#ijms-20-05315-f002){ref-type="fig"}A,B,E,F), suggesting that aloperine induced autophagic flux in these cells. However, LC3-II and p62/SQSTM1 (p62) expression continued to increase in 8505c cells ([Figure 2](#ijms-20-05315-f002){ref-type="fig"}C,D), suggesting that aloperine blocked autophagic flux in these cells. Furthermore, immunofluorescence staining confirmed aloperine-induced autophagosome formation in KMH-2 and IHH-4 cells. Rapamycin elevated LC3 puncta in KMH-2 and IHH-4 cells ([Figure 3](#ijms-20-05315-f003){ref-type="fig"}). Aloperine treatment induced LC3 puncta, and co-incubation with 3-MA could reduce aloperine-mediated LC3 puncta ([Figure 3](#ijms-20-05315-f003){ref-type="fig"}) in KMH-2 and IHH-4 cells, suggesting that aloperine treatment promoted autophagosome formation in these cells following. To confirm the induction of autophagic flux in KMH-2 and IHH-4 cells by aloperine, the mRFP-EGFP-LC3 vector was transfected into KMH-2 cells and autophagosome and autolysosome puncta were located using confocal microscopy. Aloperine treatment promoted autophagosome formation and simultaneously enhanced autophagic flux in KMH-2 and IHH-4 cells ([Figure 4](#ijms-20-05315-f004){ref-type="fig"}). Moreover, chloroquine treatment could block the aloperine-mediated induction of autophagic flux. These results demonstrate that aloperine acts as an autophagy inducer in KMH-2 and IHH-4 cells but blocks autophagic flux in 8505c cells.
2.3. Modulations of Signaling Pathways with Aloperine in Human Thyroid Cancer Cells {#sec2dot3-ijms-20-05315}
-----------------------------------------------------------------------------------
To investigate the mechanisms underlying aloperine-mediated autophagy induction, the expression and activation of AMPK, Akt/mTOR, Erk, p38, and JNK signaling pathways were determined in KMH-2 and IHH-4 cells following aloperine treatment. The expression and activation of the AMPK pathway was reduced in both KMH-2 and IHH-4 cells ([Figure S1](#app1-ijms-20-05315){ref-type="app"}). Moreover, the activation of the Akt/mTOR and p70S6K pathways decreased in an aloperine dosage-related manner treatment in both cell lines, whereas the expression of LC3-II increased ([Figure 5](#ijms-20-05315-f005){ref-type="fig"}A). In addition, aloperine treatment suppressed overall mTOR expression ([Figure 5](#ijms-20-05315-f005){ref-type="fig"}A), suggesting that aloperine regulates mTOR translation. In addition, the expression of phospho-p38 and phospho-Erk was reduced ([Figure 5](#ijms-20-05315-f005){ref-type="fig"}B), whereas that of LC3-II was increased; however, the expression of phospho-JNK remained unchanged. These results suggest that aloperine modulates the Akt/mTOR, Erk, and p38 signaling pathways and activates autophagy in KMH-2 and IHH-4 cells.
2.4. Akt Signaling Pathway Contributes to Aloperine-mediated Autophagy Induction in Human Thyroid Cancer Cells {#sec2dot4-ijms-20-05315}
--------------------------------------------------------------------------------------------------------------
To address whether the Akt/mTOR, p38, and Erk signaling pathways are involved in the aloperine-mediated autophagy, KMH-2 and IHH-4 cells were pre-incubated with inhibitors including perifosine, SB203580, and PD98059. Inhibiting Akt pathway activation increased LC3-II expression in KMH-2 and IHH-4 cells ([Figure 5](#ijms-20-05315-f005){ref-type="fig"}A); this was also observed in aloperine-treated groups. Moreover, compared with aloperine alone, the combination of aloperine and perifosine suppressed phospho-Akt and increased LC3-II expression to a greater extent ([Figure 6](#ijms-20-05315-f006){ref-type="fig"}A), suggesting that Akt pathway inhibition contributes to aloperine-mediated autophagy induction. Moreover, treatment with SB203580 and PD98059 inhibited p38 and Erk pathway activation in KMH-2 and IHH-4 cells ([Figure 6](#ijms-20-05315-f006){ref-type="fig"}B,C), and aloperine treatment suppressed p38 and Erk pathway activation. Furthermore, combination of aloperine with PD98059 or SB203580 decreased LC3-II expression in KMH-2 and IHH-4 cells ([Figure 6](#ijms-20-05315-f006){ref-type="fig"}B,C), suggesting that aloperine-mediated autophagy activation is not regulated via Erk and p38 pathway modulation. However, the physiological significance of aloperine-mediated decrease in phospho-Erk and phospho-p38 warrants further investigation. Meanwhile, we confirmed autophagosome puncta formation in cells treated with a combination of aloperine and perifosine. Perifosine-increased Akt pathway inhibition significantly increased LC3 puncta formation ([Figure 7](#ijms-20-05315-f007){ref-type="fig"}). In addition, we confirmed the role of the Akt pathway in aloperine-mediated autophagy induction; a constitutively active-form construction of Akt was transiently transfected into KMH-2 and IHH-4 cells, and the activation of Akt and expression of LC3-II were monitored using Western blotting. The overexpression of active-form Akt in the cells treated with aloperine reduced the expression of LC3-II, demonstrating that the Akt signaling pathway is the upstream pathway involved in aloperine-mediated autophagy induction ([Figure 8](#ijms-20-05315-f008){ref-type="fig"}). These results demonstrate that aloperine may serve as an autophagy inducer via Akt/mTOR signaling pathway suppression in human thyroid cancer cells.
2.5. Aloperine-Mediated Autophagy Exerts a Cytotoxic Effect in Human Thyroid Cancer Cells {#sec2dot5-ijms-20-05315}
-----------------------------------------------------------------------------------------
To determine the role of aloperine-mediated autophagy modulation in human thyroid cancer cells, KMH-2 and IHH-4 cells were incubated with aloperine or combination with 3-MA or rapamycin, and the cellular viability was examined with CCK-8 assay. The cell viability was significantly reduced in the cells under aloperine treatment ([Figure 9](#ijms-20-05315-f009){ref-type="fig"}). However, the cellular toxicity was rescued in the cells under aloperine combination with 3-MA treatment ([Figure 9](#ijms-20-05315-f009){ref-type="fig"}). On the other hand, cells incubated with aloperine and rapamycin showed higher cytotoxicity ([Figure 9](#ijms-20-05315-f009){ref-type="fig"}). These data suggested that aloperine-mediated autophagy serve a cytotoxic activity in KMH-2 and IHH-4 cells.
3. Discussion {#sec3-ijms-20-05315}
=============
Aloperine, a natural alkaloid isolated from the herb *S. alopecuroides*, has been reported to exhibit anticancer activity in various human cancers \[[@B1-ijms-20-05315],[@B7-ijms-20-05315],[@B12-ijms-20-05315],[@B15-ijms-20-05315],[@B18-ijms-20-05315],[@B20-ijms-20-05315]\]. Previous studies have demonstrated that aloperine mediates cell proliferation, cell cycle inhibition and regulation, apoptosis induction, and tumor migration/invasion and suppression \[[@B1-ijms-20-05315],[@B7-ijms-20-05315],[@B12-ijms-20-05315],[@B15-ijms-20-05315],[@B18-ijms-20-05315],[@B19-ijms-20-05315],[@B26-ijms-20-05315]\]. In a previous study, we demonstrated that anticancer effects of aloperine in multiple human thyroid cancer cells via the suppression of cell proliferation and tumorigenesis as well as induction of cell cycle arrest and apoptosis \[[@B15-ijms-20-05315]\]. In this study, we evaluated autophagy regulation in IHH-4 (a multidrug-resistant papillary thyroid carcinoma), 8505C (an undifferentiated thyroid carcinoma), and KMH-2 (a multidrug-resistant anaplastic thyroid carcinoma) cells treated with aloperine. We also explored the mechanisms involved in aloperine-modulated autophagy activation.
We demonstrated that aloperine activates the autophagy machinery, promotes autophagosome formation, and increases autophagic flux in KMH-2 and IHH-4 cells ([Figure 2](#ijms-20-05315-f002){ref-type="fig"}A,B,E,F, [Figure 3](#ijms-20-05315-f003){ref-type="fig"}, [Figure 4](#ijms-20-05315-f004){ref-type="fig"}), suggesting its role as an autophagy inducer in these cells. Although LC3-II and p62 expression levels were elevated in 8505c cells following aloperine treatment, these levels did not further decrease with prolonged treatment ([Figure 2](#ijms-20-05315-f002){ref-type="fig"}C,D), suggesting that aloperine acts as an autophagy inhibitor in 8505c cells. However, further investigation is required to determine whether aloperine can simultaneously activate autophagy and block autophagic flux in 8505c cells. Alternatively, whether aloperine alters lysosomal functions, inhibits lysosomal proteolysis, or blocks the delivery of cargo to lysosomes needs further evaluation.
The PI3K/Akt/mTOR, AMPK, and MAPK/Erk signaling pathways play major roles in autophagy induction \[[@B27-ijms-20-05315]\]. In this study, we evaluated the expression and activation of AMPK, Akt/mTOR, Erk, p38, and JNK pathways in cells following aloperine treatment. We demonstrated that the aloperine-suppressed Akt/mTOR signaling pathway is the upstream mechanism for autophagy induction in KMH-2 and IHH-4 cells ([Figure 5](#ijms-20-05315-f005){ref-type="fig"}, [Figure 6](#ijms-20-05315-f006){ref-type="fig"}, [Figure 7](#ijms-20-05315-f007){ref-type="fig"} and [Figure 8](#ijms-20-05315-f008){ref-type="fig"}). Although our results show that aloperine can suppress the activation of the p38 and Erk pathways in KMH-2 and IHH-4 cells ([Figure 5](#ijms-20-05315-f005){ref-type="fig"}A,B), this suppression with a combination of aloperine and SB203580 or PD98059 treatment did not significantly increase LC3-II expression ([Figure 6](#ijms-20-05315-f006){ref-type="fig"}B,C). These results suggest that the aloperine-mediated inhibition of p38 and Erk pathways is not the underlying mechanism for aloperine-mediated autophagy induction. However, the physiological significance of aloperine-inhibited p38 and Erk signaling pathways in KMH-2 and IHH-4 cells warrants further investigation. In addition, we cannot rule out whether factors other than the Akt pathway are involved in aloperine-mediated autophagy induction.
Autophagy is a self-degradation process that clears unnecessary intracellular organelles and proteins, playing an important role in cellular homeostasis \[[@B22-ijms-20-05315],[@B23-ijms-20-05315],[@B27-ijms-20-05315]\]. Multiple compounds including clinical chemotherapeutic agents and/or natural products induce autophagy. However, such autophagy induction is either cytotoxic or cytoprotective \[[@B28-ijms-20-05315]\], demonstrating that autophagy induction in cancer cells may be more complicated. Although excessive and sustained autophagy may lead to cell death and tumor shrinkage, and autophagic cell death has been reported in numerous reports, the cytotoxic role of autophagy remains debatable because of insufficient data on autophagic cell death markers \[[@B29-ijms-20-05315]\]. In the present study, we demonstrated that aloperine can modulate the autophagy machinery and induce autophagosome as well as autophagic flux in human thyroid cancer cells. Blocking of autophagy with 3-MA or enhancing of autophagy with rapamycin in the aloperine-treated cells could reduce or enhance aloperine-mediated cytotoxicity. Therefore, we suggested that aloperine-mediated autophagy exerts a cytotoxic role in human thyroid cancer cells.
Autophagy activation by natural products has been considered a double-edged sword in determining the cell fate of human cancers, and the interplay between autophagy and apoptosis has recently been highlighted \[[@B30-ijms-20-05315]\]. Several proteins and signaling pathways, including the p53, Bcl-2, DAPK, Akt/mTOR, and JNK pathways, act as scaffolds in mediating the crosstalk between autophagy and apoptosis \[[@B30-ijms-20-05315],[@B31-ijms-20-05315]\]. In a previous study, we reported that aloperine induces caspase-dependent apoptosis through PI3K/Akt inhibition pathway in human thyroid cancer cells \[[@B15-ijms-20-05315]\]. In the present study, cells treated with aloperine exhibited autophagy induction through Akt/mTOR pathway suppression in human thyroid cancer cells. Therefore, we speculate that Akt/mTOR pathway inhibition induces both apoptosis and autophagy in human thyroid cancer cells following aloperine treatment. Overall, our findings indicate interplay of molecules in human thyroid cancer cells following aloperine treatment, which mediates both autophagy and apoptosis, and these molecules may be effective targets in developing anticancer therapies. In addition, aloperine exhibits multiple pharmacological activities, including anti-inflammatory, antiallergenic, antiviral, antimicrobial, and antinociceptive effects against renal and neuronal injuries as well as pulmonary fibrosis \[[@B1-ijms-20-05315],[@B2-ijms-20-05315],[@B3-ijms-20-05315],[@B4-ijms-20-05315],[@B5-ijms-20-05315],[@B6-ijms-20-05315],[@B7-ijms-20-05315],[@B8-ijms-20-05315],[@B9-ijms-20-05315],[@B10-ijms-20-05315],[@B11-ijms-20-05315],[@B12-ijms-20-05315],[@B13-ijms-20-05315],[@B14-ijms-20-05315],[@B15-ijms-20-05315],[@B16-ijms-20-05315],[@B17-ijms-20-05315],[@B18-ijms-20-05315],[@B19-ijms-20-05315],[@B20-ijms-20-05315]\]. Here, we demonstrate autophagy regulation by aloperine. Whether aloperine can protect against renal and neuronal injuries as well as pulmonary fibrosis through autophagy induction should be an interesting topic to explore in the future \[[@B32-ijms-20-05315],[@B33-ijms-20-05315]\].
4. Materials and Methods {#sec4-ijms-20-05315}
========================
4.1. Cell Line and Cell Culture {#sec4dot1-ijms-20-05315}
-------------------------------
Human thyroid cancer cell lines including KMH-2 (multidrug-resistant anaplastic thyroid carcinoma), 8505c (undifferentiated thyroid carcinoma), and IHH-4 (multidrug-resistant papillary thyroid carcinoma) were purchased from Japan Collection of Research Bioresources Cell Bank (JCRB, Japan). IHH-4 and KMH-2 cells were cultured in DMEM + RPMI (1:1) medium (GIBCO, Gaithersburg, MD, USA), and 8505c cells were cultured in MEM medium (GIBCO) supplemented with 10% FBS (Biological Industries, Kibbutz Beit Haemek, Israel) at 37 °C in a 5% CO~2~ incubator.
4.2. Cell Viability Assay {#sec4dot2-ijms-20-05315}
-------------------------
Cells (5 × 10^3^ cells/well) were seeded in 96-well cell culture plates, and adherent cells were incubated in a control medium containing 0.01% dimethyl sulfoxide (DMSO) or a medium containing aloperine (Selleck Chemicals, Houston, TX, USA). Cell viability was examined using the CCK-8 assay kit (Enzo Life Sciences, Farmingdale, NY, USA) following treatment for indicated time and at indicated dosages. Three independent experiments were performed.
4.3. Autophagosome Detection {#sec4dot3-ijms-20-05315}
----------------------------
Cells (without starvation) were treated with or without aloperine for various durations. The expression of autophagy marker LC3-II was examined using Western blotting. Furthermore, autophagosome formation was detected using immunofluorescence staining under a laser confocal scanning microscope (LSM800, ZEISS, Germany). To determine autophagic flux induction following aloperine treatment, p62 expression in cells was examined. GAPDH was used as a loading control in western blotting, and DAPI was used to label the nucleus.
4.4. Western Blotting {#sec4dot4-ijms-20-05315}
---------------------
Cells were cultured in 10-cm cell culture dishes and treated with aloperine. DMSO was used as a negative control. The whole cellular extract was subjected to sodium dodecyl sulfate--polyacrylamide gel electrophoresis, and the separated proteins were electrically transferred to a PVDF membrane (Millipore Corporation, USA). The membrane was blocked with primary antibodies (anti-LC3 Ab; Abcam, USA; anti-GAPDH Ab; GeneTex, USA; anti-AMPK α-1 Ab; Cell signaling, USA; anti-phosphorylated-AMPK α-1 (Thr 172) Ab; Cell Signaling, USA; anti-Akt Ab; Santa Cruz, USA; anti-phosphorylated-Akt (Ser 473) Ab; Santa Cruz, USA; anti-mTOR Ab; Cell Signaling, USA; anti-phosphorylated-mTOR (Ser 2448) Ab; Cell Signaling, USA; anti-p70S6K Ab; Cell Signaling, USA; anti-phosphorylated-p70S6K (Thr 389) Ab; Cell Signaling; anti-p62/SQSTM1 Ab; Abgent, USA; anti-Erk Ab; Cell Signaling, USA; anti-phosphorylated-Erk (Thr 202/204) Ab; Cell Signaling, USA; anti-JNK Ab; Cell Signaling, USA; anti-phosphorylated-JNK (Thr 183/Tyr 185) Ab; Cell Signaling, USA; anti-p38 Ab; Cell Signaling, USA; and anti-phosphorylated (Thr 180/Tyr 182) Ab; Cell Signaling Ab, USA) and was analyzed using the BioSpectrum 800 Imaging System (UVP, CA, USA).
4.5. Plasmid Transfection {#sec4dot5-ijms-20-05315}
-------------------------
To observe autophagosome and autolysosome formation in cells treated with aloperine, a pmRFP-EGFP-LC3 plasmid (purchased from Addgene, Watertown, MA, USA) \[[@B34-ijms-20-05315]\] was transfected with Lipofectamine 2000 (Thermo Scientific, Waltham, MA, USA) according to the manufacturer's instructions. To confirm the role of the Akt pathway in aloperine-mediated autophagosome induction, the pHRIG-Akt1 plasmid (a constitutively active-form construct of human Akt1, a myristoylated form of Akt-1, purchased from Addgene) and pBSSK^+^ (an empty vector used as a negative control) were transfected with Lipofectamine 2000 (Thermo Scientific, Waltham, MA, USA) according to the manufacturer's instructions.
4.6. Statistical Analysis {#sec4dot6-ijms-20-05315}
-------------------------
Data are presented as mean and standard error. Statistical differences were analyzed using one-way analysis of variance and Fisher's least significant difference test. Statistical significance was defined as *p* \< 0.05.
This work was supported by the grants from the Ministry of Science and Technology of the Republic of China, Taiwan (MOST 108-2314-B-705-001), Ditmanson Medical Foundation Chiayi Christian Hospital (Grant R107-24) and Chi Mei Medical Center (CLFHR10836). We thank Myron for the critical review of the manuscript.
######
Click here for additional data file.
Supplementary materials can be found at <https://www.mdpi.com/1422-0067/20/21/5315/s1>.
H.-I.Y. and H.-C.S.: Study design, original regulatory, funding, and manuscript preparation; S.-H.C.: In vitro laboratory work, data organization, and manuscript preparation; Y.-P.L.: Data organization and manuscript preparation; H.-H.C., T.-S.T., F.-P.K., C.-H.L., and C.-Y.H.: Manuscript preparation; Y.-R.L.: Study design, original regulatory, obtained funding, and manuscript preparation and writing. All authors read and approved the final manuscript.
The authors declare no conflict of interest.
{#ijms-20-05315-f001}
{#ijms-20-05315-f002}
######
Autophagosome monitoring in human thyroid cancer cells following aloperine treatment. (**A**) KMH-2 and (**B**) IHH-4 cells were incubated with 3-MA (5 mM), rapamycin (30 µM), or aloperine (200 µM) for 24 h, and autophagosome formation was examined using immunofluorescence staining. LC3 (**green**) was used to label autophagosomes and DAPI to label the nucleus (**blue**). DMSO was used as a negative control; 3-MA was used as the autophagy inhibitor and rapamycin as the autophagy inducer. Cells with elevated autophagosome signaling are labeled with an arrow. Autophagosome puncta in the cells were identified and quantified in 30 cells. \* *p* \< 0.05, \*\* *p* \< 0.01, compared with the DMSO group. ^\#\#^ *p* \< 0.01, compared with the aloperine group. Three independent experiments were performed, and results of one of these are shown.
{#ijms-20-05315-f004}
{#ijms-20-05315-f005}
{#ijms-20-05315-f006}
{#ijms-20-05315-f007}
{#ijms-20-05315-f008}
{#ijms-20-05315-f009}
[^1]: These authors contributed equally to this work.
[^2]: Hsiang-Hsun Chuang had passed away.
| {
"pile_set_name": "PubMed Central"
} |
1. Introduction {#sec1-ijms-21-00469}
===============
The liver is an organ that spans a large variety of different functions in the body: energy metabolism, detoxification, and production of serum proteins and bile, just to mention a few \[[@B1-ijms-21-00469]\]. Consequently, liver disease or liver toxicity that cause impaired liver functionality have severe effects on normal body functions. Liver diseases are a major burden to the public health and estimations show a global liver-related mortality rate of approx. 2 million patients per year \[[@B2-ijms-21-00469]\]. In addition, the liver is one of the organs with the highest susceptibility to drug toxicity, significantly contributing to the high attrition rates in current drug discovery processes \[[@B3-ijms-21-00469]\]. Therefore, there is a strong need for better and more predictive in vitro models for liver disease and toxicity. A general opinion within the field of safety assessment and disease modeling is that in vitro cell models will increasingly contribute to improve the mechanistic understanding of diseases and the prediction of adverse effects of drugs in humans in the future.
The gold standard for studying liver disease or hepatotoxicity in vitro are metabolically competent human primary hepatocytes (hphep), either freshly isolated or cryopreserved. However, the hphep have several shortcomings that limit their use. The shortage of relevant human donor material, the large donor variation, and the rapid loss of their functionality, e.g., of the expression of the drug metabolizing machinery, in culture \[[@B4-ijms-21-00469],[@B5-ijms-21-00469],[@B6-ijms-21-00469],[@B7-ijms-21-00469]\], are the most prominent problems. Furthermore, the liver consists of multiple cell types; hepatocytes and non-parenchymal cells, such as cholangiocytes, endothelial cells, hepatic stellate cells, and Kupffer cells. As a consequence, an in vitro liver model with capacity to display all hepatic functions needs to consist of several if not all hepatic cell types and not only hepatocytes. In addition to using multiple cell types, a 3-dimensional (3D) culture environment is beneficial for an in vitro model as it allows the formation of cell-cell contacts and the formation of cell polarity in hepatocytes as indicated by the presence of bile canaliculi \[[@B8-ijms-21-00469]\].
Nonetheless, functional hepatocytes will constitute the major part of a relevant in vitro liver model, and therefore, stem cell research has been focused on deriving mature hepatocytes from stem cells. Due to the many diverse features of hepatocytes, and also considering the fact that primary hepatocytes rapidly lose key functions when cultured in conventional 2D cultures in vitro, it is not surprising that hepatocytes are one of the most challenging cell type to derive from stem cells and further mature in vitro.
In order to explore the potential of a cell model and identify good-to-go areas for the cell model, a thorough characterization of the cells in question is necessary. This characterization needs to be done on multiple levels, preferably combining large-scale transcriptomics assessment with multiple functional assays \[[@B9-ijms-21-00469]\]. Importantly, hepatocytes exist in periportal and perivenous phenotypes in the liver lobe, commonly known as metabolic zonation \[[@B10-ijms-21-00469]\], and it is important to look at features for both these hepatocyte populations when assessing a hepatocyte cell model.
Human induced pluripotent stem cells (hiPSC) are a virtually unlimited source of cells and have the potential to differentiate into specialized cell types, which provides unique opportunities for usage in a wide range of applications. The possibility to convert hiPSCs into functional hepatocytes allows for novel opportunities in assay development and holds great potential for future pre-clinical and regenerative medicine applications. For example, hiPSC derived from patients suffering from, e.g., inherited metabolic diseases can be used for modeling these types of diseases in vitro \[[@B11-ijms-21-00469],[@B12-ijms-21-00469],[@B13-ijms-21-00469]\]. Combining the hiPSC technology with the constantly progressing genome editing technologies, e.g., CRISPR-Cas9, further increases the potential for disease modeling and therapeutic applications \[[@B14-ijms-21-00469],[@B15-ijms-21-00469]\].
The study presented here describes a thorough characterization of hiPSC-derived hepatocytes (hiPS-HEP), in order to assess their potential as in vitro tools for metabolism studies and disease modeling. The cells were subjected to a broad panel of characterization assays, ranging from transcriptomics analysis, protein expression, to multiple functional assays. The results show that the hiPS-HEP possess many adult hepatocyte features that can be maintained in conventional 2D cultures for over 2 weeks. Furthermore, primary hepatic stellate cells could be activated in 2D co-cultures with hiPS-HEP which, to the best of our knowledge, is the first time that this is reported. In addition, 3D co-culture spheroids were established, paving the way for using these for modeling non-alcoholic steatohepatitis (NASH). Importantly, the transcriptomics analysis revealed that a majority of genes (86%) are similarly expressed in all the three test groups. Taken together, this study highlights the potential of hiPSC-derived hepatocytes as a robust and reproducible source for hepatocyte-related in vitro models, e.g., metabolism-related liver diseases.
2. Results {#sec2-ijms-21-00469}
==========
In the present study, the maturation status and the functionality of the hiPS-HEP was compared to hphep by combining large-scale transcriptomics with multiple functional readouts covering a variety of important hepatocyte functions, as suggested by Schwartz et al. \[[@B9-ijms-21-00469]\] amongst others. In order to assess the reproducibility and robustness of the used differentiation protocol, two batches of hiPS-HEP derived from three different hiPSC lines were tested and compared to hphep from 3 different donors.
2.1. HiPSC-Derived Hepatocytes with Highly Uniform HNF4α Expression and Mature Hepatocyte Features {#sec2dot1-ijms-21-00469}
--------------------------------------------------------------------------------------------------
The hiPS-HEP cultures consisted of large cells with a polygonal cell shape and a dark cytoplasm ([Figure 1](#ijms-21-00469-f001){ref-type="fig"}C1), which is typical for hepatocytes. HiPS-HEP were plated at confluency after thawing and did not proliferate, indicating terminal differentiation. Importantly, expression of stemness markers such as Oct4 (*POUF51*) were as low in hiPS-HEP as in hphep (see heatmap in [Figure 1](#ijms-21-00469-f001){ref-type="fig"}B), which indicates the loss of the stemness features as a result of an efficient hepatocyte differentiation. For a first general assessment, the expression of typical hepatocyte-related markers such as albumin (*ALB*), asialoglycoprotein receptor 1 (*ASGPR1*), connexin 32 (*GJB1*), hepatocyte nuclear factor 4α (*HNF4α*), and α1-antitrypsin (*AAT*) were investigated in hiPS-HEP and these markers were found to be expressed at similar mRNA levels as in hphep ([Figure 1](#ijms-21-00469-f001){ref-type="fig"}A). As expected, the mRNA expression of these five markers decreased in the hphep after culturing for 24 h (hphep d0 vs. d1). Next, protein expression of these hepatocyte markers was investigated using immunocytochemistry (ICC; [Figure 1](#ijms-21-00469-f001){ref-type="fig"}C,D). Almost all cells in the hiPS-HEP and hphep cultures were immuno-positive for HNF4α ([Figure 1](#ijms-21-00469-f001){ref-type="fig"}C2 and C6, respectively; [Supplementary Figure S1](#app1-ijms-21-00469){ref-type="app"}). Quantification using CellC Cell Counting software showed that more than 92% of the cells in hiPS-HEP cultures derived from the three hiPSC lines ChiPSC12, ChiPSC18 and ChiPSC22 expressed HNF4α, counted in relation to 4′,6-diamidino-2-phenylindole (DAPI) staining ([Supplementary Figure S1](#app1-ijms-21-00469){ref-type="app"}). The few HNF4α-negative cells are also likely to be derived from definitive endoderm (DE) since these cultures are near-homogenous with 97% Sox17-immunopositive cells \[[@B16-ijms-21-00469]\].
In contrast to the highly uniform HNF4α expression, ASGPR1, α1-Antitrypsin (AAT), and Albumin (Alb) appeared to be more strongly expressed in a subpopulation in both hiPS-HEP and hphep cultures ([Figure 1](#ijms-21-00469-f001){ref-type="fig"}C3--C5 and C7--C9, respectively). The immunostaining of these three markers was comparable for hiPS-HEP derived from all three hiPSC lines in terms of staining intensity ([Supplementary Figure S1](#app1-ijms-21-00469){ref-type="app"}).
In agreement with *Albumin* mRNA expression levels and immunostainings, Albumin secretion was found to be at comparable levels in hiPS-HEP and hphep ([Figure 1](#ijms-21-00469-f001){ref-type="fig"}D). Importantly, Albumin secretion by hiPS-HEP was stable for over two weeks in culture (between day 4 and 20 post-thaw) indicating phenotypical stability. In order to evaluate urea secretion, another important liver specific function, the mRNA expression levels of key enzymes of the urea cycle and other related genes were compared in hphep and hiPS-HEP. Several urea cycle-related genes were expressed at comparable mRNA levels in hiPS-HEP and hphep ([Figure 1](#ijms-21-00469-f001){ref-type="fig"}B), whereas *ARG1*, *ASL, ASS1, CPS1, OTC,* and *SLC25A2* were expressed at significantly lower levels in hiPS-HEP compared to hphep directly after thawing (d0). When measuring urea secretion upon ammonium challenge, hiPS-HEP displayed lower urea secretion than hphep cultured for 1 day post-thaw (d1; [Figure 1](#ijms-21-00469-f001){ref-type="fig"}E) which may be due to the lower expression of *ARG1* in hiPS-HEP compared to hphep d1 ([Figure 1](#ijms-21-00469-f001){ref-type="fig"}B). Similarly, to the stable Albumin secretion, urea secretion was stable in hiPS-HEP between day 13 and 20 post-thaw ([Figure 1](#ijms-21-00469-f001){ref-type="fig"}E).
2.2. Expression of Drug Metabolizing Enzymes and Transporters {#sec2dot2-ijms-21-00469}
-------------------------------------------------------------
Expression and functionality of the drug metabolizing machinery, comprising phase I and phase II enzymes as well as transporter proteins, is of critical importance for the utility of an in vitro hepatocyte model in drug metabolism and hepatotoxicity studies but of less importance for disease modeling. Since several enzymes are known to be specific for fetal or adult hepatocytes, analyses of these enzymes can also aid to assess the grade of maturity of the cells.
We started by investigating the most important Cytochrome P450 (CYP) enzymes and found that several CYP enzymes were expressed in hiPS-HEP at similar mRNA levels as in hphep d1, e.g., *CYP2C19, 2C9, 3A4, 3A5*, and *3A7* ([Figure 2](#ijms-21-00469-f002){ref-type="fig"}A). In contrast to that, other enzymes such as *CYP1A2, CYP2B6,* and *CYP2D6*, were expressed at lower mRNA levels in hiPS-HEP compared to hphep d1 ([Figure 2](#ijms-21-00469-f002){ref-type="fig"}A). Importantly, the similar mRNA levels of the adult enzyme *CYP3A4* and the fetal enzyme *CYP3A7* in hiPS-HEP and hphep indicate an adult feature of hiPS-HEP. Since a low correlation between mRNA and activity levels for many phase I and II enzymes as well as transporters has been reported by Ohtsuki et al. \[[@B17-ijms-21-00469]\], we also assessed the enzyme activity levels by incubating with specific substrates for the different enzymes and measuring the formation of the specific metabolites with LC/MS. We found that hiPS-HEP had similar CYP1A, CYP3A, and CYP2C9 activities as hphep cultured for 20 h (including the activity assay; [Figure 2](#ijms-21-00469-f002){ref-type="fig"}D1--D3). In contrast, CYP2B6 and CYP2D6 activity levels were lower in hiPS-HEP compared to hphep ([Figure 2](#ijms-21-00469-f002){ref-type="fig"}D) which is in agreement with the mRNA expression ([Figure 2](#ijms-21-00469-f002){ref-type="fig"}A). Notably, CYP activities in hiPS-HEP were stable or even increasing for a period of 14 days, between day 4 and 19 ([Figure 2](#ijms-21-00469-f002){ref-type="fig"}D1--D6). A discrepancy between CYP2C19 and CYP2C9 mRNA and activity levels could be observed. *CYP2C19* mRNA expression is on the same level in the two cell types, whereas the activity is approximately 10 times lower in hiPS-HEP than in hphep ([Figure 2](#ijms-21-00469-f002){ref-type="fig"}A,D6). However, the LC/MS method does not seem to be as stable for detection of the CYP2C19 metabolite as for the other metabolites. Regarding CYP2C9, lower mRNA levels were observed in hiPS-HEP than in hphep but the activity levels were similar in both cell models ([Figure 2](#ijms-21-00469-f002){ref-type="fig"}A,D3). The discrepancies seen between mRNA and activity levels for some CYP enzymes are not unexpected. Ohtsuki et al. have reported previously that CYP2C9 and CYP2C19 are among the CYP enzymes with the lowest correlation between mRNA and activity levels \[[@B17-ijms-21-00469]\]. This clearly emphasizes that protein expression or preferably functional assays are essential when characterizing in vitro hepatocyte models.
Since drug metabolism is also performed by phase II enzymes, we investigated the expression of two classes of phase II enzymes, sulfotransferases (SULT) and Uridine 5′-diphospho-glucuronosyl transferases (UGT). All tested SULTs and UGTs were found to be expressed on similar mRNA levels between the two cell types ([Figure 2](#ijms-21-00469-f002){ref-type="fig"}B). In agreement with that, incubations with the substrate 7-OH-coumarin revealed similar or higher SULT and UGT activity levels in hiPS-HEP and in hphep cultured for 20 h (including the activity assay; [Figure 2](#ijms-21-00469-f002){ref-type="fig"}E,F). Similarly to the CYP activities, phase II enzyme activities were stable or even increasing in hiPS-HEP during a 14-day culture period (between day 4--19 post-thaw).
In addition to phase I and II enzymes, transporter proteins play an important role in xenobiotics metabolism. Therefore, we also investigated the mRNA expression of 19 transporters, including both uptake and efflux transporters. 11 transporters were expressed on comparable levels in both cell types, e.g., *ABCC2* (MRP2) and *SLC10A1* (NTCP), whereas eight were expressed at lower levels in hiPS-HEP, e.g., *ABCB11* (BSEP) and *SLCO1B1* (OATP1B1; [Figure 2](#ijms-21-00469-f002){ref-type="fig"}C). The protein expression of MRP2, NTCP, BSEP, and OATP1B1 were confirmed by immunostainings ([Figure 2](#ijms-21-00469-f002){ref-type="fig"}G1--G4), however, formation of bile canaliculi was not observed when performing a 5(6)-Carboxy-2′,7′-dichlorofluorescein diacetate (CDFDA) staining (data not shown) indicating that no cell polarity is formed in hiPS-HEP cultures.
Furthermore, in order to assess batch-to-batch variation coefficient of variation (CV) was calculated for [Figure 1](#ijms-21-00469-f001){ref-type="fig"}A (CV \< 8%) and for [Figure 2](#ijms-21-00469-f002){ref-type="fig"}A--C (CV \< 15% for the majority (87%) of these genes). This indicates little variation between batches derived from the same hiPSC line at different occasions and thus a stable differentiation procedure.
2.3. Glucose Metabolism and Insulin Signaling {#sec2dot3-ijms-21-00469}
---------------------------------------------
Another important and rapidly growing area of use for hepatocytes is disease modeling. For example, hepatocytes are crucial for modeling metabolic functions performed by the liver and for diseases such as non-alcoholic fatty liver disease (NAFLD) and non-alcoholic steatohepatitis (NASH) which are global, rapidly growing problems. In order to evaluate the utility of hiPS-HEP for modeling NAFLD/NASH, we started by investigating the energy metabolism of glucose and lipids as well as insulin signaling and inflammation in the hiPS-HEP.
First, we compared the mRNA expression of more than 40 genes central in glucose metabolism in hiPS-HEP and hphep and found the vast majority to be expressed at similar levels in both cell types ([Figure 3](#ijms-21-00469-f003){ref-type="fig"}A). Only eight of the genes were significantly differently expressed between hiPS-HEP and hphep d0 and/or d1. The genes expressed at lower levels in hiPS-HEP were fructose-1,6-biphosphatase 1 (*FBP1*), glutamic-oxaloacetic transaminase 1 (*GOT1*), pyruvate carboxylase (*PC*), and phosphoenolpyruvate carboxykinase 1 (*PCK1*), compared to hphep d0. On the other hand, Glycerol-3-phosphate dehydrogenase (*GPD2*), glycogen synthase 1 (*GYS1*), Oxoglutarate Dehydrogenase (*OGDH*), and hexokinase 2 (*HK2*) showed higher mRNA expression levels in hiPS-HEP compared to hphep d0 and/or d1. Interestingly, *PC, PCK1, FBP1, GOT1*, and *GPD2* are directly involved in gluconeogenesis, and *OGDH* has a strong influence on 2-oxoglutarate, which regulates the gluconeogenesis in the liver \[[@B18-ijms-21-00469]\]. Noteworthy is that *PC, PCK1,* and *FBP1* were also significantly lower expressed in hphep d1 than in hphep d0.
One important hepatocyte function in glucose metabolism is the ability to synthesize and store glycogen (glycogenesis). Therefore, Periodic Acid Schiff (PAS) staining was performed on both hiPS-HEP at day 12 after thaw and on hphep at day 1 after thaw. As shown in [Figure 3](#ijms-21-00469-f003){ref-type="fig"}C1--C3, glycogen was detected in a subset of hiPS-HEP (derived from all three hiPSC lines). Similarly, hphep also showed glycogen storage in a subpopulation of the cells ([Figure 3](#ijms-21-00469-f003){ref-type="fig"}C4). The presence of glycogen is in agreement with the finding that several key enzymes of glycogenesis are well expressed in hiPS-HEP: Hexokinase (*HK*), UDP-glucose pyrophosphorylase (*UGP2*), glycogen synthase (*GYS*), and Glycogen branching enzyme (*GBE1*; [Figure 3](#ijms-21-00469-f003){ref-type="fig"}A).
Another essential feature of metabolically functional hepatocytes is the physiological response to insulin. Thus, the mRNA expression of 19 genes involved in insulin signaling were compared in hiPS-HEP and hphep and no significant differences in expression levels of these 19 genes were found between the two cell types ([Figure 3](#ijms-21-00469-f003){ref-type="fig"}B), including key genes such as insulin receptor (*INSR*), insulin receptor substrate 1 and 2 (*IRS1, IRS2*), and Ak strain transforming (AKT) serine/threonine kinases 1 and 2 (*AKT1, AKT2*). In order to test if the hiPS-HEP respond to an insulin stimulus by phosphorylation of AKT kinase which is a key regulator in the insulin signaling cascade, the hiPS-HEP were incubated for 3 h in an insulin-free medium followed by an incubation for 10 min with 100 nM insulin. In the insulin-treated cells, a significant increase of phosphorylated AKT could be detected compared to untreated control cells, while the total AKT content was similar in treated and untreated cells, as shown by Western Blot ([Figure 3](#ijms-21-00469-f003){ref-type="fig"}D).
2.4. Cholesterol and Lipid Metabolism {#sec2dot4-ijms-21-00469}
-------------------------------------
Next, we assessed the mRNA expression levels of 29 genes involved in cholesterol metabolism and 47 genes involved in fatty acid metabolism in both hiPS-HEP and hphep. The vast majority of the cholesterol metabolism-related genes assessed were expressed at similar levels in hiPS-HEP and hphep, e.g., low-density lipoprotein (LDL) receptor (*LDLR*) and Proprotein convertase subtilisin/kexin type 9 (*PCSK9*) ([Figure 4](#ijms-21-00469-f004){ref-type="fig"}A). Only four of the investigated genes differed significantly between hiPS-HEP and hphep d0/1: ATP citrate lyase (*ACLY*) and apolipoprotein A1 and A4 (*APOA1; APOA4*), which were higher in hiPS-HEP, and apolipoprotein A5 (*APOA5*) which was lower expressed in hiPS-HEP ([Figure 4](#ijms-21-00469-f004){ref-type="fig"}A). Noteworthy, the important gene *APOB* is expressed on similar level in hiPS-HEP and hphep d0. Next, we tested if the hiPS-HEP were capable of taking up fluorescently labeled low density lipoprotein (LDL-DyLight). Already after a 3-h incubation, accumulation of LDL-DyLight could be observed in the cells ([Figure 4](#ijms-21-00469-f004){ref-type="fig"}C) and an even stronger accumulation could be seen after 24 h.
The assessment of 48 genes involved in fatty acid metabolism revealed that only three genes were expressed at different levels in hiPS-HEP and hphep d0 ([Figure 4](#ijms-21-00469-f004){ref-type="fig"}B). The exceptions were acetyl-CoA carboxylase beta (*ACACB*) and bile acyl-CoA synthetase (*SLC27A5*), which were both expressed at lower levels in hiPS-HEP, and protein kinase AMP-activated catalytic subunit alpha 2 (*PRKAA2*), which was expressed at higher levels in hiPS-HEP than in hphep d0. Additionally, three genes were significantly different in hiPS-HEP compared to hphep d1; *HACD3* and *PECR* (both higher in hiPS-HEP); and *PPARD* (lower in hiPS-HEP). Noteworthy, *PECR* was also higher in hphep d1 compared to hphep d0, and *PPARD* and SLC27A5 were also lower in hphep d1 compared to hphep d0.
2.5. Inflammatory Response and Induction of Steatosis {#sec2dot5-ijms-21-00469}
-----------------------------------------------------
Many genes involved in inflammatory response are expressed by hepatocytes and are of interest for disease modeling. [Figure 5](#ijms-21-00469-f005){ref-type="fig"}A shows the expression levels of 72 inflammation-related genes in hiPS-HEP and hphep d0/1. Noteworthy, most inflammation-related genes are expressed at low or medium levels in both hiPS-HEP and hphep d0/1, e.g., caspase 4 (*CASP4*), tumor necrosis factor (*TNF*), Toll like receptor 1, 3, 4 (*TLR1, TLR3, TLR4*) which may indicate a healthy, non-inflammatory status of both cell types under standard conditions ([Figure 5](#ijms-21-00469-f005){ref-type="fig"}A). However, a few genes are expressed at high levels in both cell types, e.g., *AIMP1, ATRN, F11R, ITCH, MIF, PLAA, PRDX5*, and *RELA*.
When comparing expression levels between hiPS-HEP and hphep d0/1, we found the majority of the genes to be similarly expressed ([Figure 5](#ijms-21-00469-f005){ref-type="fig"}A) which is similar to previous results ([Figure 3](#ijms-21-00469-f003){ref-type="fig"} and [Figure 4](#ijms-21-00469-f004){ref-type="fig"}). Only 12 out of the 72 genes differed significantly in expression levels between the two cell types: *CCL16, CRP, FOS, LTBR*, and *ORM1* were lower expressed in hiPS-HEP, whereas *AXL, CYBA, F2R, F2RL1*, and *RIPK2* were expressed at higher levels in hiPS-HEP.
Inflammation in the liver can, amongst others, be caused by fat accumulation. At first, fat is accumulated in the hepatocytes, a condition known as NAFLD, which can progress to NASH which is characterized by steatosis, inflammation, and fibrosis \[[@B19-ijms-21-00469]\]. To model these conditions in vitro, one prerequisite is that hepatocytes can take up free fatty acids and store as lipid droplets. Therefore, we incubated hiPS-HEP for 24 h with either 200 or 600 µM oleic acid (OA; coupled to fatty acid free bovine serum albumin (BSA)) or with fatty acid free BSA alone as vehicle control. Subsequently, the cells were stained for lipid accumulation using Oil Red O. Only a few lipid droplets were observed in the control cells ([Figure 5](#ijms-21-00469-f005){ref-type="fig"}B1) while the cells incubated with 200 and 600 µM OA showed a dose dependent increase in lipid droplets ([Figure 5](#ijms-21-00469-f005){ref-type="fig"}B3,B4, respectively). When we combined 200 µM OA with 1 µM Thapsigargin, a compound known to cause endoplasmic reticulum (ER) stress and to worsen steatosis \[[@B12-ijms-21-00469]\], a clear increase of lipid accumulation was observed compared to only 200 µM OA ([Figure 5](#ijms-21-00469-f005){ref-type="fig"}B2,B3, respectively).
Next, we investigated if inducing steatosis by treatment with OA and Thapsigargin caused an inflammatory response in the hiPS-HEP. Indeed, we found that the mRNA expression of the inflammatory marker tumor necrosis factor α (*TNFα*) was upregulated in OA/Thapsigargin-treated hiPS-HEP ([Figure 5](#ijms-21-00469-f005){ref-type="fig"}C). The strongest upregulation of *TNFα* was caused by 600 µM OA, followed by the combination of 200 µM OA and Thapsigargin compared to vehicle control ([Figure 5](#ijms-21-00469-f005){ref-type="fig"}C). Thus, this indicated that treatment with OA and Thapsigargin caused an inflammatory response.
2.6. Co-Cultures of hiPS-HEP and Hepatic Stellate Cells for Modeling NASH {#sec2dot6-ijms-21-00469}
-------------------------------------------------------------------------
In the advanced stage of NASH, fibrosis occurs in the liver, which is characterized by collagen deposited by activated hepatic stellate cells (HSC). Thus, for modeling NASH-related liver fibrosis, one needs to co-culture hepatocytes and HSC. Therefore, we adapted the culture conditions in a way that allowed co-culturing primary human HSC together with hiPS-HEP. In order to activate HSC, we treated the co-cultures with 20 ng/mL transforming growth factor β (TGFβ) for 4 days which strongly increased the protein expression of α-smooth muscle actin (α-SMA) ([Figure 5](#ijms-21-00469-f005){ref-type="fig"}D2), a characteristic marker for activated HSC, compared to untreated cultures ([Figure 5](#ijms-21-00469-f005){ref-type="fig"}D1). Since activated HSC are known to express collagens, we investigated collagen mRNA expression in the co-cultures after TGFβ treatment. In accordance with the observed increase in α-SMA expression upon TGFβ stimulation ([Figure 5](#ijms-21-00469-f005){ref-type="fig"}D2), we also observed increased *Collagen 1a1* and *5a1* mRNA expression in the TGFβ-treated co-cultures compared to hiPS-HEP cultures without HSC ([Figure 5](#ijms-21-00469-f005){ref-type="fig"}D3,D4). Monocultures of hiPS-HEP alone showed no *Collagen 1a1* expression irrespective of treatment ([Figure 5](#ijms-21-00469-f005){ref-type="fig"}D3), whereas *Collagen 5a1* was expressed in untreated hiPS-HEP monocultures and upregulated upon TGFβ treatment, however at lower levels than in the co-cultures ([Figure 5](#ijms-21-00469-f005){ref-type="fig"}D4). Since TGFβ has been reported to suppress *HNF4α* expression \[[@B20-ijms-21-00469]\] and *HNF4α* expression is crucial for maintaining a differentiated hepatocyte phenotype \[[@B21-ijms-21-00469]\], we investigated *HNF4α* expression both on mRNA and protein level in the co-cultures. In general, *HNF4α* mRNA expression was lower in co-cultures than in hiPS-HEP monocultures independent from TGFβ-treatment which was expected and can be explained with a decreased hiPS-HEP content in the co-cultures due to the addition of HSC ([Figure 5](#ijms-21-00469-f005){ref-type="fig"}D5). Importantly, TGFβ-treatment of the co-cultures tended to decrease *HNF4α* mRNA expression levels ([Figure 5](#ijms-21-00469-f005){ref-type="fig"}D5) but the observed differences were not significant. In accordance with this, fewer HNF4α-immuno-positive nuclei could be observed the TGFβ-treated cultures (compare [Figure 5](#ijms-21-00469-f005){ref-type="fig"}D1,D2).
Taken together, the co-cultures responded as expected to TGFβ treatment with activation of the HSC and upregulation of α-SMA and collagen expression. However, already the untreated control co-cultures displayed a weak α-SMA immunostaining ([Figure 5](#ijms-21-00469-f005){ref-type="fig"}D1) which indicated that a subpopulation of the HSC was already activated. This is in agreement with previous reports that HSC get activated in 2D cultures due to the stiff matrix \[[@B22-ijms-21-00469]\]. In order to avoid the HSC activation due to the 2D culture setting, we generated spheroids of hiPS-HEP and HSC. To this end, we seeded hiPS-HEP and HSC in ultra-low attachment plates, spheroids formed within 1--2 days and could be maintained for at least 35 days ([Figure 5](#ijms-21-00469-f005){ref-type="fig"}E1--E3). In future studies, the utility of the co-culture spheroids for NAFLD and NASH studies can be explored.
2.7. Similarities and Differences between hiPS-HEP and Hphep {#sec2dot7-ijms-21-00469}
------------------------------------------------------------
To further assess the similarities and differences between hiPS-HEP and hphep on gene expression level, a global analysis of the transcriptomics data was carried out. [Figure 6](#ijms-21-00469-f006){ref-type="fig"}A depicts the number of transcripts that are similarly expressed, with a coefficient of variation (CV) \< 0.1, throughout all samples between the different comparisons. A great majority of the transcripts (24,685 transcripts, 86%) are similarly expressed across all group comparisons. Not surprisingly, the highest number of similarly expressed genes (27,157) in the different sample group comparisons is found in hphep d0 vs. hphep d1. hiPS-HEP show a comparable number of similarly expressed genes when compared to hphep d1 and hphep d0 (26,492 and 26,244, respectively).
Using a generalized linear model (GLM) and the criterion described in the method section, in total 2866 differentially expressed genes (DEGs) were detected as significant in at least one of the sample group comparisons. [Figure 6](#ijms-21-00469-f006){ref-type="fig"}B displays the overlap of DEGs between the three group comparisons. Out of these DEGs, only 76 genes are differentially expressed between all sample group comparisons. A total of 2655 genes are differentially expressed when comparing hiPS-HEP to hphep, either at day 0 or day 1 ([Figure 6](#ijms-21-00469-f006){ref-type="fig"}B, areas a2--a7). Approximately 25% of these genes are also differentially expressed between hphep at day 1 and hphep at day 0 ([Figure 6](#ijms-21-00469-f006){ref-type="fig"}B, areas a2,a4,a6). To assess the biological relevance of the DEGs, a pathway over-representation analysis was carried out on the different groups of DEGs according to the overlapping areas (areas a1--a7) in [Figure 6](#ijms-21-00469-f006){ref-type="fig"}B. For a complete list of significant over-represented pathways, see [Supplemental Table S1](#app1-ijms-21-00469){ref-type="app"}. Among the 76 DEGs that differ between all three groups ([Figure 6](#ijms-21-00469-f006){ref-type="fig"}B, area a4) there is an over-representation for pathways such as, "Cytochrome P450---arranged by substrate type", "CYP2E1 reactions", "Phase I---Functionalization of compounds", "Phase II---Conjugation of compounds", as well as pathways related to bile acids and bile salts. These pathways are also over-represented in the comparison between hphep d1 and hphep d0 ([Figure 6](#ijms-21-00469-f006){ref-type="fig"}B, area a1), as well as in the overlap between hphep d1 vs. hphep d0 and hiPS-HEP vs. hphep d0 ([Figure 6](#ijms-21-00469-f006){ref-type="fig"}B, areas a2,a4). The genes *CYP2E1, CYP4A11*, and *CYP2C8*, associated with the CYP P450 pathways, have the following expression rank between the groups: hphep d0 \> hphep d1 \> hiPS-HEP. The same expression pattern is seen for the bile acid/bile salts related genes (*SLC27A5, SLCO1B1, AKR1C4*). The genes associated with "Phase II---Conjugation of compounds" have a more ambiguous expression, *ACSM5*: hphep d0 \> hphep d1 \> hiPS-HEP, *SULT2A1*: hphep d0 \> hiPS-HEP \> hphep d1, and *UGT2A3*: hiPS-HEP \> hphep d0 \> hphep d1.
The DEGs in hiPS-HEP compared to hphep d1 ([Figure 6](#ijms-21-00469-f006){ref-type="fig"}B, area a7) show an over-representation for "Attenuation phase" and "HSP1-dependent transactivation"; the associated genes are *HSPA1A, HSPA1B, DNAJB1*, with the ranking hphep d1 \> hphep d0 \> hiPS-HEP, and *CAMK2D* with an opposite expression ranking. When comparing hiPS-HEP to hphep d0 ([Figure 6](#ijms-21-00469-f006){ref-type="fig"}B, area a3), hiPS-HEP show a higher expression of genes related to glycogen synthesis and metabolism. For the comparison hphep d0 vs. hiPS-HEP and hphep d1 vs. hiPS-HEP ([Figure 6](#ijms-21-00469-f006){ref-type="fig"}B, area a5), the pathways "Laminin interactions", "Non-integrin membrane-ECM interactions", "ECM proteoglycans", and "Extracellular matrix organization" are over-represented with the expression ranking hiPS-HEP \> hphep d1 \> hphep d0. The DEGs in hphep d1 when compared to hiPS-HEP or hphep d0 ([Figure 6](#ijms-21-00469-f006){ref-type="fig"}B, area a6), are overrepresented in pathways related to respiratory electron transport and the citric acid cycle.
3. Discussion {#sec3-ijms-21-00469}
=============
The present study covers an in-depth characterization of hiPSC-derived hepatocytes including multiple phenotypic and functional aspects of relevance for several application areas, for example modeling of metabolic disease.
An essential requirement for the utility of stem cell-derived hepatocytes for all application areas is that cultures with a very high hepatocyte content can be robustly derived from large panels of stem cell lines. To evaluate the hepatocyte content in the hiPS-HEP cultures, we stained for the key hepatocyte transcription factor HNF4α and found that it was expressed in 92% of cells in the hiPS-HEP cultures, thus indicating a highly efficient differentiation result and a near-homogeneous culture regarding HNF4α expression. A similarly high differentiation efficiency, 93.8% HNF4α-positive cells, was previously shown when we screened a panel of 25 human pluripotent stem cell lines using an earlier version of the hepatocyte differentiation protocol used in this study to derive hiPS-HEP \[[@B16-ijms-21-00469]\]. Notably, the hepatocyte differentiation protocol worked for all lines without any modifications. In contrast to our findings, others have reported that their protocols either do not work for some hPSC lines making it necessary to select lines that appear more prone to differentiate into hepatocytes \[[@B23-ijms-21-00469]\], or that they require time- and labor-intensive adaptations of the protocol for individual lines \[[@B24-ijms-21-00469]\]. One great advantage of the hiPSC technology is the possibility to generate iPSC lines from many different donors, healthy and diseased, so that one can recreate the population diversity in vitro. However, this requires differentiation protocols that, without adaptations, generate highly pure hepatocyte cultures from virtually all lines.
Furthermore, the reproducibility of the differentiation protocol is of great importance so that different cell batches show the same characteristics. We found that different hiPS-HEP batches derived from the same hiPSC line showed overall low variation measured as coefficient of variation (CV). For example, expression of hepatocyte markers ([Figure 1](#ijms-21-00469-f001){ref-type="fig"}A) show CV \< 5% for all genes except for *ASGR1* (CV \< 8%), and expression of the drug metabolizing machinery show CV \< 15% for the majority (87%) of all genes in [Figure 2](#ijms-21-00469-f002){ref-type="fig"}A--C.
In contrast to the nearly uniform HNF4α-expression in hiPS-HEP, other markers or features are expressed only in a subpopulation of both hiPS-HEP and hphep, e.g., Albumin, α1-Antitrypsin, ASGPR1, and glycogen storage ([Figure 1](#ijms-21-00469-f001){ref-type="fig"} and [Figure 3](#ijms-21-00469-f003){ref-type="fig"}). One would expect to find such different populations of hepatocytes since this is well-described for the liver and known as metabolic zonation \[[@B10-ijms-21-00469]\]. In the liver lobe, hepatocytes exist in periportal and perivenous phenotypes and therefore, it is important to look at features for both populations, e.g., albumin expression (predominantly in periportal hepatocytes) and CYP expression (predominantly in perivenous hepatocytes), when assessing a hepatocyte cell model. However, we cannot rule out that the observed heterogeneity for maturation markers such as Albumin and ASGPR1 in the hiPS-HEP cultures could be due to different levels of maturity.
We assessed the maturation status of the hiPS-HEP by performing transcriptomics analysis and functional assays in comparison to adult human primary hepatocytes. Albumin secretion, which is one hallmark for adult hepatocytes, was measured at comparable levels in hiPS-HEP derived from line ChiPSC12 and hphep (CV \< 9%; [Figure 1](#ijms-21-00469-f001){ref-type="fig"}D) and with slightly higher secretion in hiPS-HEP derived from line ChiPSC18 and ChiPSC22. This was confirmed by comparable *Albumin* mRNA levels across both cell types (CV \< 9%; [Figure 1](#ijms-21-00469-f001){ref-type="fig"}A). Another characteristic of adult hepatocytes is high expression of the adult enzymes *CYP3A4* and low expression of the fetal enzyme *CYP3A7* which was the case in both cell types ([Figure 2](#ijms-21-00469-f002){ref-type="fig"}A). ASGPR1, a marker for stem cell-derived hepatocytes enriched for primary hepatocyte features \[[@B25-ijms-21-00469]\], showed also comparable mRNA levels and immunostainings in hiPS-HEP and hphep ([Figure 1](#ijms-21-00469-f001){ref-type="fig"}A,C3,C7) which suggested that hiPS-HEP have several adult hepatocyte features at levels comparable to hphep.
Two functionalities that require further improvement in hiPS-HEP are, (1) urea secretion and (2), the drug metabolizing machinery, comprising phase I and II enzymes as well as transporters. Urea secretion was clearly lower in hiPS-HEP than in hphep d1 ([Figure 1](#ijms-21-00469-f001){ref-type="fig"}E). The only urea cycle gene that was lower in hiPS-HEP than in hphep was *ARG1* ([Figure 1](#ijms-21-00469-f001){ref-type="fig"}B) which is the last in the chain of five key enzymes in the urea cycle \[[@B26-ijms-21-00469]\] and is likely the cause of the restricted urea production. Regarding phase I enzymes, the three main CYP enzymes, CYP1A, 3A, and 2C9, had comparable activity levels in hiPS-HEP and in hphep d1. CYP1A, 3A, and 2C9 combined metabolize about 60% of prescription drugs (CYP3A4 37%; CYP1A2 9%; CYP2C9 17%; \[[@B27-ijms-21-00469]\]), but in order to be useful for drug metabolism studies all enzymes need to be expressed at useful levels. Therefore, expression of CYP2B6, 2D6, and 2C19 needs to be improved in hiPS-HEP ([Figure 2](#ijms-21-00469-f002){ref-type="fig"}D). We observed a similar situation for transporters as for the CYP enzymes: some were well expressed, e.g., MRP2 (*ABCC2*) and NTCP (SLC10A1) and some need to be increased in hiPS-HEP, e.g., *OATP1B1* and *OATP1B3* ([Figure 2](#ijms-21-00469-f002){ref-type="fig"}C). In addition, formation of bile canaliculi was not observed, meaning that hiPS-HEP do not develop cell polarity with apical and basolateral membrane compartments. This limits the utility of the cells for applications requiring bile canaliculi formation, e.g., biliary excretion of drugs. Regarding phase II metabolism, we focused on UGTs and SULTs since they are responsible for more than half of all phase II metabolism of clinically used drugs \[[@B28-ijms-21-00469]\] and found SULT and UGT activity levels to be comparable in hiPS-HEP and hphep ([Figure 2](#ijms-21-00469-f002){ref-type="fig"}E,F).
Noteworthy, Albumin and urea secretion as well as CYP and phase II activities were stable over time in hiPS-HEP ([Figure 1](#ijms-21-00469-f001){ref-type="fig"}D--E; [Figure 2](#ijms-21-00469-f002){ref-type="fig"}D--F). This phenotypic stability of the hiPS-HEP is clearly a superior feature over hphep, which are known to quickly loose functions such as CYP activities in conventional 2D cultures \[[@B29-ijms-21-00469],[@B30-ijms-21-00469]\]. Many studies, for example for hepatitis virus studies and toxicity studies with repeated dosing, require a chronic treatment and a culture period of at least 2 weeks which is why phenotypic stability is of great importance for an in vitro hepatocyte model.
As expected, hphep display large variations of CYP activities between donors (see error bars for hphep in [Figure 2](#ijms-21-00469-f002){ref-type="fig"}D) which reflect the large inter-individual variation existing in the population (see e.g., \[[@B17-ijms-21-00469]\]). This variation is a major issue that researchers are facing when using primary hepatocytes for in vitro models since the amount of hphep from one donor is limited and researchers are therefore forced to switch to new donors that may have a very different phenotype potentially affecting their studies. Therefore, one main advantage of hiPS-HEP is that one can obtain virtually unlimited amounts of cells from one donor, and importantly, there is only very low variation between different batches of hiPS-HEP, as seen for example in the reproducible and robust expression of hepatocyte markers ([Figure 1](#ijms-21-00469-f001){ref-type="fig"}A), drug metabolizing genes ([Figure 2](#ijms-21-00469-f002){ref-type="fig"}A--C), CYP activities ([Figure 2](#ijms-21-00469-f002){ref-type="fig"}D), and the response to insulin ([Figure 3](#ijms-21-00469-f003){ref-type="fig"}D). Noteworthy, inter-individual variation in the CYP activity profile can be observed when comparing hiPS-HEP derived from different iPSC lines: e.g., hiPS-HEP derived from ChiPSC18 have higher CYP3A and 2D6 activities than hiPS-HEP derived from the other 2 hiPSC lines ([Figure 2](#ijms-21-00469-f002){ref-type="fig"}D2,D5), which is likely due to higher expression of the polymorphic genes CYP3A5 and 2D6 in ChiPSC18 ([Figure 2](#ijms-21-00469-f002){ref-type="fig"}A).
Energy metabolism is another key function of the liver. The results presented here indicate that hiPS-HEP have a comparable glucose and lipid metabolism as hphep, the gold standard ([Figure 3](#ijms-21-00469-f003){ref-type="fig"} and [Figure 4](#ijms-21-00469-f004){ref-type="fig"}). However, several genes related to gluconeogenesis, e.g., *PC, PCK1, FBP1, GOT1, GPD2,* and *OGDH*, are expressed on lower levels in hiPS-HEP ([Figure 3](#ijms-21-00469-f003){ref-type="fig"}A). A potential cause could be the relatively high glucose concentration in the culture medium which would also explain the downregulation of gluconeogenesis-related genes in hphep d1 compared to hphep d0 ([Figure 3](#ijms-21-00469-f003){ref-type="fig"}A).
One important area of use for hepatocytes is to model NAFLD, a common cause for chronic liver disease with a prevalence of 20--30% in Europe \[[@B31-ijms-21-00469]\]. NAFLD can progress to NASH which is associated with a higher risk of liver-related mortality and cardiovascular disease \[[@B32-ijms-21-00469]\], and further to cirrhosis with an increased risk for hepatocellular carcinoma \[[@B33-ijms-21-00469]\]. The long progression until clinical signs, the lack of good diagnostic tools as well as the absence of relevant preclinical models hampers the development of adequate treatments and drugs for NADFL/NASH, and highlights the need for novel translational in vitro models \[[@B34-ijms-21-00469]\]. Several animal models have been developed, however, due to the heterogeneous pathology few or none of these models represent the human situation accurately \[[@B35-ijms-21-00469]\]. Therefore, there is great and urgent need for an in vitro model for NAFLD/NASH.
Our data presented here suggest that hiPS-HEP could serve as such a potential NAFLD/NASH model. Most genes related to lipid and cholesterol metabolism are expressed at comparable levels as in hphep d0/1 ([Figure 4](#ijms-21-00469-f004){ref-type="fig"}). Only four genes differed significantly between hiPS-HEP and hphep d0/1, for example *APOA5*, which is involved in lipoprotein metabolism by interacting with LDL receptor \[[@B36-ijms-21-00469]\]. However, despite the lower *APOA5* expression in hiPS-HEP and in agreement with substantial expression of the *LDL receptor*, we found that hiPS-HEP accumulated labelled LDL upon a short incubation ([Figure 4](#ijms-21-00469-f004){ref-type="fig"}A,C). We also observed dose-dependent steatosis, which in turn lead to dose-dependent upregulation of the inflammatory marker *TNFα* ([Figure 5](#ijms-21-00469-f005){ref-type="fig"}B,C). Importantly, under standard culture conditions most inflammation related genes were expressed at low levels both in hphep and hiPS-HEP ([Figure 5](#ijms-21-00469-f005){ref-type="fig"}A), indicating a healthy status of the cells and no activation of an inflammatory response which is a prerequisite for the utility of the hiPS-HEP for modelling inflammation. Furthermore, hiPS-HEP responded in the anticipated manner to Thapsigargin-induced ER stress \[[@B12-ijms-21-00469],[@B37-ijms-21-00469]\], with both an increased lipid accumulation and an upregulated *TNFα* expression ([Figure 5](#ijms-21-00469-f005){ref-type="fig"}B,C). Taken together, these results indicate that hiPS-HEP are comparable to adult primary hepatocytes regarding glucose and lipid metabolism as well as inflammatory markers. The observed responses are essential for a relevant human in vitro model for NAFLD/NASH and highlight the potential use of hiPS-HEP for such models. Another important finding is that key drug targets for NAFLD/NASH treatment are well expressed in hiPS-HEP, e.g., the peroxisome proliferator−activated receptor-α (*PPARA*) and -δ (*PPARD*), which are targets for the drug elafibranor \[[@B38-ijms-21-00469]\], and the nuclear receptor FXR (*NR1H4*), which is a target for the drug obeticholic acid \[[@B39-ijms-21-00469]\]. All three genes are expressed on the same level in hiPS-HEP and hphep d0 ([Figure 4](#ijms-21-00469-f004){ref-type="fig"}B) which demonstrates the potential of hiPS-HEP for drug target studies for NAFLD/NASH.
Since the progression of NAFLD to NASH includes a contribution of inflammatory and fibrogenic response from non-parenchymal cells, hepatocytes need to be co-cultured with non-parenchymal cells, such as stellate cells and Kupffer cells, as reviewed in \[[@B35-ijms-21-00469]\]. To date, only a limited number of attempts using primary cells or cell lines (reviewed in \[[@B40-ijms-21-00469]\]) or hepatocytes derived from human pluripotent stem cells \[[@B12-ijms-21-00469],[@B41-ijms-21-00469]\] have been reported in the NASH field. However, in the two NASH studies utilizing stem cell-derived hepatocytes, those were not co-cultured with non-parenchymal cells \[[@B12-ijms-21-00469],[@B41-ijms-21-00469]\]. Two other studies described 2D co-cultures of several hiPSC-derived liver cell types: liver progenitor cells and HSC \[[@B42-ijms-21-00469]\], and liver progenitor cells, liver sinusoidal endothelial cells and HSC \[[@B43-ijms-21-00469]\], respectively, and reported a positive effect of the co-culture on the maturity of the hiPSC-liver progenitor cells but did not evaluate the utility of the co-cultures for modeling NASH. Another study reported modeling of liver fibrosis using the hepatoma cell line HepaRG and hiPSC-derived HSC \[[@B44-ijms-21-00469]\]. To the best of our knowledge, our study is the first one to describe co-cultures of hiPSC-derived hepatocytes and human primary HSC. Importantly, we found that α-SMA expression, a hallmark of activated HSC and fibrosis, was induced in HSC in 2D co-cultures by TGFβ treatment. Since HSC showed a weak activation in 2D even without TGFβ stimulation, we generated 3D-spheroids of hiPS-HEP and HSC which can be maintained for at least 35 days. The results presented here pave the way for further exploring the utility of these co-culture spheroids for NASH studies. The 3D co-culture setting may even further improve the hiPS-HEP functionality compared to 2D cultures, similarly to reports on 3D spheroids consisting of hphep and non-parenchymal cells \[[@B8-ijms-21-00469]\].
The global analysis of the transcriptomics data revealed very similar phenotypes of the three cell populations ([Figure 6](#ijms-21-00469-f006){ref-type="fig"}A). Differences were observed, amongst others, in phase I and II metabolism, indicating a lower functionality of hiPS-HEP and hphep d1 compared to hphep d0, which is in line with our gene expression and functional data on drug metabolism ([Figure 2](#ijms-21-00469-f002){ref-type="fig"}). Interestingly, in the DEGs between hiPS-HEP and hphep d0 and d1 "Attenuation phase" and "HSP1-dependent transactivation" are over-represented pathways which are related to cellular response to heat stress. It has been reported previously that cryopreservation and re-plating of hphep results in reduced ability to form cell-matrix and cell-cell interactions which triggers stress and heat shock responses \[[@B45-ijms-21-00469]\]. Our observation that DEGs involved in these pathways (*HSPA1A, HSPA1B*, and *DNAJB1*) are upregulated in hphep d0 and d1 compared to hiPS-HEP is in line with these reports. Furthermore, pathways connected to ECM and laminin interactions are over-represented in the DEGs between hiPS-HEP and hphep d0 and d1, with higher expression in hiPS-HEP. A possible explanation could be that hiPS-HEP are differentiated in 2D cultures whereas hphep come from a 3D environment. It would be interesting to investigate if these pathways are downregulated in hiPS-HEP 3D spheroids.
The identification of DEGs is of great importance since they can provide a basis for future improvements as the aim is to reach the same functionality and maturity in hiPS-HEP as in hphep d0. One incentive for such continuous efforts is that hphep are not useful as a cell source for gene editing and are limited for disease modeling of a specific genetic background. The advance of the hiPSC technology allows to routinely generate hiPSC lines from many individuals, including healthy persons and patients, and thus represent the whole population. Furthermore, genetic modification of hiPSC lines using for example CRISPR/Cas9, allows to study the molecular basis of diseases.
Taken together, the hiPS-HEP used in this study represent a homogenous cell population with a high similarity to hphep and a potential for modeling NAFLD/NASH. Our results highlight the potential of hiPSC-derived hepatocytes as a virtually unlimited cell source. In order to build more advanced cell models, to better recapitulate the human liver environment, to reflect the population diversity and to increase the predictability for disease modeling, advanced 3D culture systems including multiple cell types and microfluidics as well as panels of hiPSC lines derived from various ethnicities need to be developed and implemented.
4. Materials and Methods {#sec4-ijms-21-00469}
========================
4.1. Cell Culture {#sec4dot1-ijms-21-00469}
-----------------
Cryopreserved hiPS-HEPs derived from the hiPSC lines ChiPSC12, ChiPSC18, and ChiPSC22 were thawed, plated, and maintained according to the vendor instructions (Cellartis Enhanced hiPS-HEP v2 kits, Cat.no. Y10133, Y10134, Y10135, Takara Bio Europe, Gothenburg, Sweden, [www.takarabio.com](www.takarabio.com)). Enhanced hiPS-HEP v2 were maintained for up to 21 days post-thawing with media changes every 2 or 3 days using the Cellartis Enhanced hiPS-HEP Long-Term Maintenance Medium included in the kits.
For 2D co-cultures of Enhanced hiPS-HEP v2 and primary human stellate cells (HCS, BioIVT, Brussels, Belgium), cells were seeded on a Fibronectin coating (concentration 2 µg/cm^2^; Roche, Basel, Swiss) in a 2:1 ratio (hiPS-HEP: HSC) at a cell density of 400 K/cm^2^ using the regular plating medium included in the hiPS-HEP v2 kits. Instead of the regular HEP Long-Term Maintenance Medium the following maintenance medium was used for the co-cultures: Williams' Medium E supplemented with 0.1% PEST (15140--130, ThermoFisher, Stockholm, Sweden), 0.5% DMSO (D2650, Sigma, Darmstadt, Germany), 0.67% Cellartis Hepatocyte supplement (Y11149, Takara Bio), 2% Cellartis Hepatocyte Additives (Y11078, Takara Bio), 194 µM L-Proline (P5607, Sigma), and 173 µM L-ascorbic acid 2-phosphate sesquimagnesium salt (A8960, Sigma).
For 3D co-cultures of Enhanced hiPS-HEP v2 and HSC, cells were seeded in a 2:1 ratio (hiPS-HEP: HSC) in InVitroGRO CP medium (BioIVT) supplemented with 0.5% PEST, and 2% Cellartis Hepatocyte Additives (Y11078, Takara Bio) in Costar 96 well ultra-low attachment plates (7007, Corning, Amsterdam, The Netherlands) with 30,000 cells per well in 100 µL plating medium. At 24 h post-seeding, medium was changed to 3D spheroid maintenance medium as described by Rashidi and colleagues \[[@B46-ijms-21-00469]\]. At 24--48 h post-seeding, spheroids had formed and were maintained for up to 35 days with 50% medium changes every 2--3 days. Alternatively, spheroids could also be maintained with the 2D co-culture medium described above.
Cryopreserved hphep from three donors (Cat.no. M00995-P for male donors, F00995-P for female donors, BioIVT), selected based on high viability (see [Table 1](#ijms-21-00469-t001){ref-type="table"} for donor details), were thawed and plated according to the provider's instructions. For experiments that required the hepatocytes to be cultured for 24 h, medium was changed 4 h post-plating to fresh plating medium (InVitroGRO CP medium, BioIVT).
4.2. Transcriptomics Analysis {#sec4dot2-ijms-21-00469}
-----------------------------
### 4.2.1. Total RNA Extraction {#sec4dot2dot1-ijms-21-00469}
Two batches of Enhanced hiPS-HEP v2 from each hiPSC line (ChiPSC12, ChiPSC18, and ChiPSC22) were harvested on day 13 post-thawing by scraping the cells in the culture medium, centrifuging at 300× *g* for 5 min and frozen as dry cell pellets. Human primary hepatocytes (*n* = 3 donors) were harvested as dry cell pellets directly after thawing (day 0) and on day 1 post-thawing (day 1). Cell pellets were stored at −80 °C until RNA extraction. Total RNA was extracted from the cell pellets using the GenElute RNA/DNA/protein Plus Purification kit (E5163, Sigma Aldrich). RNA amounts were quantified using a NanoDrop ND-1000 (NanoDrop, <http://www.nanodrop.com>).
### 4.2.2. Gene Array {#sec4dot2dot2-ijms-21-00469}
The quality of the RNA and cDNA was verified using a 2100 Agilent Bioanalyzer. To measure the mRNA levels, cDNA was synthesized from the RNA samples applying the GeneChip WT PLUS Reagent Kit (Affymetrix, Stockholm, Sweden) and fragmented cDNA was hybridized at 45° Celsius for 16 h to whole transcript HuGene ST 2.0 arrays (Affymetrix, <http://www.affymetrix.com>) at SCIBLU Genomics (Lund University, Sweden). In total, 12 expression microarrays were run.
### 4.2.3. Data Processing {#sec4dot2dot3-ijms-21-00469}
Raw microarray data was imported into R (version 3.6.1, R Core Team, 2019, <https://www.r-project.org/>) and signal intensities normalized by the Robust Multichip Average method in the oligo package \[[@B47-ijms-21-00469]\]. The two biological replicates for each hiPSC line were summarized by mean expression. To remove transcripts with expression values close to background, probes with a log~2~ expression below 5 in all samples were removed. The resulting dataset contained 21,427 transcripts and 9 samples (3 hphep day 0, 3 hphep day 1, 3 hiPS-HEP). The transcripts were mapped to Human Genome Organization Gene Nomenclature Committee (HGNC) symbols using the HuGene 2.0 ST V1 NetAffx file from Affymetrix (NA36, genome build hg19, <http://www.affymetrix.com/analysis/index.affx>). The microarray data used in this study follow the MIAME standard and raw expression data are available at ArrayExpress (<https://www.ebi.ac.uk/arrayexpress/>), accession number E-MTAB-8286.
### 4.2.4. Statistical Analysis {#sec4dot2dot4-ijms-21-00469}
Inspection of transcriptomics data revealed an approximate gamma distribution, and thus statistical testing for differential expression was based on a generalized linear model (GLM) from the gamma family with a linear link. The GLM was fitted using genes as response variables and sample group (hphep day 0, hphep day 1, hiPS-HEP) as covariates. The primary human hepatocytes day 0 and day 1 samples were treated as paired. Statistical significance of differential expression was assessed with the likelihood ratio test. *p*-values were adjusted for multiple testing by Benjamini-Hochberg correction. Differentially expressed genes were identified using a combined criterion of adjusted *p*-value \< 0.05 and absolute log~2~ fold change \> 2.
Pathway over-representation analysis of the differentially expressed genes was carried out with Reactome Pathway Database \[[@B48-ijms-21-00469],[@B49-ijms-21-00469]\], using a criterion of a *p*-value \< 0.05, and at least two differentially expressed genes present for the identified over-represented pathways.
4.3. Immunocytochemistry {#sec4dot3-ijms-21-00469}
------------------------
Cells were stained as described previously in \[[@B7-ijms-21-00469]\]. Briefly, Enhanced hiPS-HEP v2 were fixed on day 12 post-thawing and hphep on day 1 post-thawing, by 15 min incubation with 4% Formaldehyde. Cells were stained with the following primary and secondary antibodies: rabbit anti-α1AT (A0012, DAKO, 1:200 dilution), rabbit anti-albumin (A0001, DAKO, 1:1000), mouse anti-ASGPR1 (MAI-40244, ThermoFisher 1:50), rabbit-anti-BSEP (purchased from Bruno Stieger, University Hospital Zurich, Zurich, Switzerland, 1:100); rabbit anti-HNF4α (SC-8987, SantaCruz Biotechnology, 1:300, Heidelberg, Germany), rabbit anti-MRP2 (SC-20766, SantaCruz Biotechnology, 1:50), rabbit anti-NTCP (Bruno Stieger, 1:400), rabbit anti-OATP1B1 (Bruno Stieger, 1:200), donkey anti-rabbit Alexa 594 IgG (A21207, ThermoFisher, 1:1000) or goat anti-mouse Alexa 488 (A11029, ThermoFisher, 1:1000). For nuclear counterstaining, DAPI (Sigma, D9542) was added during the incubation with the secondary antibodies (add 2 µL/mL of a 1 mg/mL DAPI stock in DMSO). Stainings were examined using an inverted fluorescence microscope (Eclipse Ti-U, Nikon, Amsterdam, The Netherlands), ANDOR Zyla sCMOS digital camera and the NIS-Elements software package (version 4.30). Technical control staining without primary antibodies were performed for all secondary antibodies and these were negative. Quantification of HNF4α positive cells was done in relation to DAPI stained cells using the CellC Cell Counting Software \[[@B50-ijms-21-00469]\].
Co-cultures of Enhanced hiPS-HEP v2 and HSC were permeabilized with 0.5% Triton X-100 solution in PBS for 1 h at room temperature (RT) and then incubated for 30 min at RT with blocking buffer containing 1% BSA and 10% normal goat serum (Sigma). Primary and secondary antibodies were diluted in blocking buffer. Cells were incubated for 2 h at RT (α-SMA) or overnight at 4 °C (HNF4α) with the primary antibodies and subsequently washed three times with PBS before incubating with the secondary antibodies for 1 h at RT. Next, cells were washed three times with PBS, stained for 5 min with DAPI diluted 1:25,000 in PBS and washed again. The following primary and secondary antibodies were used: mouse anti-α-smooth muscle actin (α-SMA; 1:1000; A5228, Sigma), mouse anti-HNF4α (1:200; ab41898, Abcam), goat anti-mouse Alexa Fluor 594 (1:300; ThermoFisher Scientific), and goat anti-mouse FITC (1:50; ThermoFisher Scientific). Fluorescent images were acquired on an Axio Observer Z1 Zeiss microscope (Zeiss, Breda, The Netherlands) and processed with ZEN 2.3 lite software (Zeiss).
4.4. Period Acid-Schiff (PAS) Staining {#sec4dot4-ijms-21-00469}
--------------------------------------
Glycogen storage was visualized by PAS staining of Enhanced hiPS-HEP v2 derived from C12, C18, and C22 (fixed on day 12 post-thawing) and hphep (fixed on day 1 post-thawing). Cells were stained with the glycogen assay kit (MAK016, Sigma) according to the manufacturer's protocol.
4.5. Albumin Secretion {#sec4dot5-ijms-21-00469}
----------------------
Albumin secretion from Enhanced hiPS-HEP v2 (derived from C12, C18, and C22) was analyzed on days 4, 6, 12, and 20 post-thawing and from hphep cells 24 h post-thawing. The culture medium was collected after 24 h of conditioning and Albumin content was analyzed with the Albuwell kit (Exocell, Philadelphia, PA) according to the manufacturer's protocol. The Albumin content in the medium was normalized to the assay duration (24 h) and the amount of protein per well.
For protein quantification, cells were washed once with DPBS (with Calcium and Magnesium) and lysed in 0.02 mM NaOH over night at 4 °C and stored at −20 °C until quantification using the Pierce BCA Protein Assay kit (ThermoFisher, Rockford, IL) according to the manufacture's instruction.
4.6. Urea Secretion {#sec4dot6-ijms-21-00469}
-------------------
On days 13 and 20 post-thawing, Enhanced hiPS-HEP v2 and hphep d1 were incubated with 5 mM ammonium chloride for 24 h. After 24 h, medium was collected, and urea secretion was analyzed with the QuantiChrom Urea Assay Kit (BioAssay Systems, Hayward, CA, USA). Urea content was normalized to the amount of protein per well (determined using the Pierce BCA Protein Assay Kit, see above) and the assay duration (24 h).
4.7. AKT Western Blot {#sec4dot7-ijms-21-00469}
---------------------
Phosphorylated AKT: Enhanced hiPS-HEP (derived from C18, on day 12 post-thawing) were incubated in insulin-free medium (phenol-red free Williams' medium E containing 0.1% PEST, 25 mM HEPES, 2 mM L-Glutamine) for 3 h and then treated for 10 min with 0 nM or 100 nM insulin. Phosphorylated AKT (Cell Signaling) and total AKT (Cell Signaling) were quantified by Western blot (NuPAGE 4--12% Bis-Tris Protein Gels, Thermo Fisher Scientific).
4.8. CYP Activity Assay {#sec4dot8-ijms-21-00469}
-----------------------
The CYP activities of Enhanced hiPS-HEP v2 were analyzed by performing a CYP activity assay at days 4, 12, and 19 after thawing and the results were compared to hphep cultured for 20 h. Briefly, the cells were carefully washed twice with pre-warmed Williams' medium E (Phenol-red free, +0.1% PEST). Then, the activity assay was started by adding 110 μL/cm^2^ culture area of pre-warmed Williams' medium E (phenol-red free) containing 0.1% PEST, 25 mM HEPES (H7523, Sigma), 2 mM L-Glutamine, and the probe substrate cocktail (see [Table 2](#ijms-21-00469-t002){ref-type="table"} below). After a 2-h incubation at 37 °C, 100 μL of the supernatant was collected and kept at −80 °C until LC/MS analysis. LC/MS analysis (performed at Pharmacelsus GmbH, Saarbrücken, Germany) was used to measure the formation of the specific metabolites Acetaminophen (CYP1A), OH-Bupropion (CYP2B6), 4-OH-Diclofenac (CYP2C9), 4-OH-Mephenytoin (CYP2C19), OH-Bufuralol (CYP2D6), and 1-OH-Midazolam (CYP3A). The metabolite concentrations were normalized to the amount of protein per well (determined using the Pierce BCA Protein Assay Kit, see above) and the assay duration (120 min). To be able to normalize the results between different LC/MS runs, a metabolite cocktail with known concentrations of all metabolites is included in every analysis batch.
4.9. Phase II Enzyme Activity Assay {#sec4dot9-ijms-21-00469}
-----------------------------------
The phase II enzyme activities of Enhanced hiPS-HEP v2 were analyzed day 4, 8, 12, and 19 after thawing by performing a phase II activity assay. Briefly, the cells were carefully washed twice with pre-warmed Williams' medium E (+0.1% PEST). Then the activity assay was started by adding 110 μL/cm^2^ culture area of pre-warmed Williams' medium E containing 0.1% PEST, 25 mM HEPES, 2 mM L-Glutamine, and 200 μM 7-OH-coumarin. After a 2 h incubation at 37 °C, 100 μL of the supernatant was collected and kept at −80 °C until LC/MS analysis. LC/MS analysis (performed at Pharmacelsus GmbH) was used to measure the formation of 7-OH-coumarin sulfate and 7-OH-coumarin glucuronide, specific metabolites for sulfotransferases and UDP-glucuronosyltransferases, respectively. The metabolite concentrations were normalized to the amount of protein per well (determined using the Pierce BCA Protein Assay Kit, see above) and the assay duration (120 min).
4.10. LDL Uptake {#sec4dot10-ijms-21-00469}
----------------
In order to determine the uptake of low-density lipoproteins (LDL), Enhanced hiPS-HEP v2 (derived from C12, C18, and C22), on day 6 after thawing, were incubated for 3 h with LDL-DyLight (Cat.no. 10011125, Cayman Chemical, Hamburg, Germany) diluted 1:100 in regular maintenance medium. Next, cells were washed once with DPBS (with Calcium and Magnesium) and immunofluorescence was recorded as described above under immunocytochemistry.
4.11. Fatty Acid Accumulation and Inflammatory Response {#sec4dot11-ijms-21-00469}
-------------------------------------------------------
Enhanced hiPS-HEP v2 were incubated for 24 h (on days 5--6 post-thawing) with Williams' medium E containing 0.1% PEST, 25 mM HEPES, and 2 mM L-Glutamine supplemented with either 200 µM oleic acid (OA, O1008, Sigma) coupled to 77 µM fatty acid-free BSA (FAF-BSA, A8806, Sigma), 600 µM OA coupled to 231 µM FAF-BSA, or only 231 µM FAF-BSA (as vehicle control). In addition, one group of cells was treated with 200 µM OA (coupled to 77 µM FAF-BSA) plus 1 µM Thapsigargin (T0933, Sigma). After the 24 h incubation, cells were either stained with Oil Red O or harvested for gene expression analysis. Oil Red O staining was performed using the Hepatic Lipid Accumulation/Steatosis Assay Kit (Cat.no. ab133131, Abcam) according to the protocol provided with the kit. Stainings were evaluated using a Zeiss AxioVert microscope, an Axicam 105 color camera and the ZEN2 software (all from Carl Zeiss, Jena, Germany). For gene expression analysis, cells were harvested in RNAprotect Cell Reagent (Cat No. 76526, Qiagen, Hilden, Germany). RNA preparation, cDNA synthesis, and qPCR were performed as described previously in \[[@B7-ijms-21-00469]\]. Gene expression was analyzed using the TaqMan Gene Expression Assays (Applied Biosystems, Foster City, CA): *TNFα* (Hs00174128_m1), and *CEBPα* (Hs00269972_s1) which served as a reference gene.
4.12. RNA Prep and RT-qPCR of Co-Cultures and 3D Spheroids {#sec4dot12-ijms-21-00469}
----------------------------------------------------------
Total RNA from co-cultures and from 3D spheroids was isolated using the Ambion RNAqueous RNA Isolation Kit (ThermoFisher). The manufacturer's supplied protocol was followed during the isolation and RNA concentrations were measured using the nanodrop ND-1000 spectrophotometer. Cellular RNA (maximum of 200 ng) was reverse transcribed into cDNA using the High Capacity RNA to cDNA kit (Applied Biosystems) in a total reaction volume of 20 µL consisting of 9 µL assay buffer, 1 µL enzyme mixture and added up to 20 µL with RNAse free H2O. Samples were incubated for 60 min at 37 °C and 5 min at 95 °C using a BioRad iCycler. qPCR samples were subsequently prepared by diluting cDNA samples 1:1 with RNAse free H2O and samples consisted of 2 µL diluted cDNA, 0.625 µL assay on demand (AoD, ThermoFisher) or 25 µM primers with 6.25 µL Taqman^®^ Gene expression mastermix (ThermoFisher) filled up to a total volume of 12.5 µL with RNAse free H2O. Samples were run at 50 °C for 2 min, 95 °C for 10 min and 40--45 cycles, each cycle consisting of 95 °C for 15 s and 60 °C for 1 min using the QuantStudio6 flex. The following TaqMan Gene Expression Assays were used: *Collagen 1a1* (Hs01076777_m1), *Collagen 5a1* (Hs00609133_m1), *HNF4α* (Hs00230853_m1), and *HPRT1* (Hs03929098_m1). The latter served as a reference gene.
We are grateful to Roxana Khatib Shahidi at The Lundberg Laboratory for Diabetes Research (Department of Molecular and Clinical Medicine, Sahlgrenska Academy at the University of Gothenburg, Gothenburg, Sweden) and Paula Metselaar and Daan 't Hart at TNO for technical assistance.
Supplementary materials can be found at <https://www.mdpi.com/1422-0067/21/2/469/s1>. Supplemental Figure S1 Representative pictures of immunocytochemical stainings of the hepatocyte markers Hepatocyte Nuclear Factor 4α (HNF4α), Asialoglycoprotein receptor 1 (ASGPR1), α1-Antitrypsin (AAT), and Albumin (Alb) in hiPS-HEP derived from ChiPSC12, ChiPSC18 and ChiPSC22, respectively, on day 12 post-thawing and hphep cultured for 24 h post-thaw. Scale bar 50 µm. Table S1 Results from the over-representation analysis on the identified DEGs. The table is separated according to the overlapping areas in [Figure 6](#ijms-21-00469-f006){ref-type="fig"}B.
######
Click here for additional data file.
Conceptualization, A.A., R.H., B.K.-M., J.S.; methodology, A.A., B.K.-M., R.H.; software, B.U.; validation, A.A., K.T., A.H., B.K.-M.; formal analysis, G.H., B.U.; investigation, A.A., K.T., A.H., B.K.-M.; resources, C.X.A., R.H., J.S.; data curation, B.U., G.H.; writing---original draft preparation, A.A., B.K.-M., B.U., G.H.; writing---review and editing, A.A., B.K.-M., B.U., G.H., K.T., C.X.A., A.H., R.H., J.S.; visualization, A.A., B.K.-M., B.U., G.H.; A.H., K.T., R.H.; supervision, J.S., B.K.-M., A.A., C.X.A., R.H.; project administration, J.S., B.K.-M., A.A., C.X.A., R.H.; funding acquisition, J.S., R.H. All authors have read and agreed to the published version of the manuscript.
This work was supported by the Systems Biology Research Centre at University of Skövde under grants from the Knowledge Foundation \[2014/0301, 2016/0293 and 2017/0302\], and Takara Bio Europe, Gothenburg, Sweden.
A.A., B.K.-M. and C.X.A. are employees at Takara Bio Europe. R.H. and K.T. are employees at TNO. The other authors declare no conflict of interest.
AAT
α1-Antitrypsin
Alb
Albumin
ASGPR1
Asialoglycoprotein receptor 1
CYP
Cytochrome P450
DEGs
Differentially Expressed Genes
hiPSC
Human induced pluripotent stem cells
hiPS-HEP
hiPSC-derived hepatocytes
HNF4α
Hepatocyte Nuclear Factor 4 α
hphep
human primary hepatocytes
HSC
Hepatic Stellate Cells
NAFLD
Non-Alcoholic Fatty Liver Disease
NASH
Non-Alcoholic Steatohepatitis
OA
Oleic acid
TGFβ
Transforming Growth Factor β
TNFα
Tumor Necrosis Factor α
{#ijms-21-00469-f001}
{#ijms-21-00469-f002}
{#ijms-21-00469-f003}
{#ijms-21-00469-f004}
{#ijms-21-00469-f005}
{#ijms-21-00469-f006}
ijms-21-00469-t001_Table 1
######
Donor demographics of the human primary hepatocytes.
Donor Age Gender Ethnicity Cause of Death Viability
--------- ----- -------- ------------------ -------------------------------------------------------------------- -----------
**KFF** 56 male Caucasian anoxia, 2nd to cardiac arrest 95%
**YEM** 46 female Caucasian intracerebral hemorrhage---stroke 87%
**MSW** 69 male African-American intracerebral hemorrhage, 2nd to cerebrovascular accident (stroke) 91%
ijms-21-00469-t002_Table 2
######
CYP activity substrate cocktail.
CYP Substrate Assay Concentration
---------- ------------- ---------------------
**1A** Phenacetin 10 μM
**2B6** Bupropion 10 μM
**2C19** Mephenytoin 50 μM
**2C9** Diclofenac 10 μM
**2D6** Bufuralol 10 μM
**3A** Midazolam 5 μM
[^1]: These authors contributed equally to this work.
[^2]: Present address: AstraZeneca Gothenburg, BioPharmaceuticals R&D, Late-stage Development, 43183 Mölndal, Sweden.
| {
"pile_set_name": "PubMed Central"
} |
Introduction
============
Over the last three decades, many issues have arisen that affect the sustainable use of antibiotics and antifungal drugs. These issues include serious side effects such as the nephrotoxicity that can occur with long term use of the antifungal amphotericin B ([@B20]), the emergence of multidrug-resistant fungal and bacterial strains ([@B44]; [@B23]) and the formation of bacterial and fungal biofilms that are persistent and respond poorly to currentantibiotics ([@B36]; [@B39]; [@B48]). These challenges to the current therapeutic choices have led to a demand for novel drugs for the treatment of infectious diseases ([@B63]; [@B12]; [@B48]). Naturally occurring antimicrobial peptides (AMPs) have promise as molecules for the control of infection. The unique mechanisms of AMPs mean that they are likely to be effective against microorganisms that have developed resistance to the small molecule antibiotics currently in the clinic ([@B46]; [@B3]). Furthermore, the plant and animal kingdoms house a huge array of AMPs; only a small percentage of which have been characterized at the functional level.
Small, cationic AMPs have attracted increasing research and clinical interest due to their novel mechanisms of action ([@B37]; [@B31]). They are ubiquitous in nature and are a part of the arsenal of innate immunity molecules that provide the first line of defense against invading pathogens in a potential host. AMPs are structurally diverse but often function by disrupting microbial membranes ([@B70]; [@B65]; [@B71]). They are also generally highly cationic, which facilitates interactions with negatively charged microbial membranes, and are often amphipathic, allowing them to insert into membranes and disrupt membrane integrity ([@B41]). They often have strong antimicrobial activity in the low micro or nanomolar range when tested *in vitro* in low salt media. However, many natural AMPs are not active *in vitro* at physiological salt concentrations (\>100 mM) ([@B1]) or in the presence of serum, limiting their potential for direct applications in the clinic ([@B26]; [@B42]).
Only a few naturally occurring salt-tolerant antimicrobial peptides have been identified, mostly from marine organisms ([@B35]; [@B22]; [@B19]). Salt-tolerant AMPs have also been generated through *in silico* rational design and peptidomimetic strategies. These include the insect cecropin-bee melittin hybrid (CEME) peptides ([@B50]; [@B25]), IMB1-3 which has a C-terminal signaling domain and an N-terminal killer domain ([@B38]) and a chimera of human β-defensins 1 and 3 ([@B55]). Even though these synthetic analogs are less sensitive to high salt (\>100 mM), they often have high haemolytic activity and poor pharmacokinetic properties such as serum binding and susceptibility to proteolytic cleavage. These attributes are major drawbacks that have severely hindered progress toward therapeutic applications ([@B68]; [@B30]). So far only a handful of these synthetic AMP analogs have been registered for clinical trials ([@B24]; [@B32]; [@B27]).
Plant defensins are one of the largest families of AMPs. They are usually cationic and defined by a cysteine stabilized alpha-beta (CSαβ) structural motif. The CSαβ motif is composed of a triple stranded β-sheet and an α-helix connected by three stabilizing disulphide bonds ([@B10]; [@B21]; [@B33]). A fourth disulphide bond tethers the N- and C-terminal, rendering the molecule pseudo cyclic. This structural motif conveys stability to extremes of pH and temperature as well as resistance to proteases. Like most other cationic AMPs, the characterized plant defensins are not active at physiological salt concentrations. We hypothesized that plant defensins with increased positive charge would retain activity at higher salt concentrations. To address this hypothesis we searched a database of ∼1200 plant defensin sequences ([@B56]), of which less than 2% have been characterized at the functional level ([@B57]), to identify those with high charge. We identified ZmD32, a defensin from *Zea mays* as one of the most highly charged (+10.1 at pH 7) plant defensins described to date ([Figure 1](#F1){ref-type="fig"}). We discovered that, unlike other plant defensins, ZmD32 is active against both bacteria and fungi and retains these activities in NaCl at concentrations of 100 mM and higher. However, the rate of cell killing by ZmD32 in the presence of elevated NaCl concentrations is slowed.
![Sequence analysis and purification of ZmD32. **(A)** The charge on defensins at pH 1--14 was estimated by running the sequences through PROTEIN CALCULATOR v3.4 (<http://protcalc.sourceforge.net/>). ZmD32 was more highly charged across all pHs than other defensins in the literature ([@B61]; [@B62]; [@B33]; [@B16]). **(B)** RP-HPLC trace of purified ZmD32 revealed a single peak. **(C)** SDS-PAGE analysis of ZmD32 alongside two other defensins NaD1 and NaD2. ZmD32 runs higher on the gel compared to the other two defensins despite the predicted molecular masses being similar (ZmD32 5529 Da, NaD1 5296 Da, NaD2 5255 Da). The mass of ZmD32 was confirmed by MALDI-TOF MS (data not shown). **(D)** Alignment of the sequence of ZmD32, NaD2 and NaD1 was performed using Clustal Omega ([www.ebi.ac.uk/Tools/msa/clustalo/](http://www.ebi.ac.uk/Tools/msa/clustalo/)). The symbol "^∗^" indicates amino acids common to all three sequences and ":" indicates amino acids with strongly similar properties and. amino acids with weakly similar properties. The charge of each defensin at pH 7 is presented to the right of the sequence. Positively charged residues are highlighted in blue, negatively charged residues in red and cysteines in gray. NaD2, which shares 78.7% sequence identity with ZmD32, was included as a comparator because the αNaD2 antibody was used to detect ZmD32 in subsequent experiments.](fmicb-10-00795-g001){#F1}
Materials and Methods {#s1}
=====================
Protein Sources and Maintenance of Microbial Strains
----------------------------------------------------
NaD1 and NaD2, were extracted from the floral tissues of *Nicotiana alata* as described previously ([@B33]) or by recombinant expression using a *Pichia pastoris* system followed by purification using cation exchange chromatography and reverse phase high performance liquid chromatography (RP-HPLC) ([@B34]). Peptide masses were confirmed by sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE) and matrix assisted laser desorption ionization time-of-flight (MALDI-TOF) mass spectrometry. ZmD32, NbD6, and PaD2 were all identified from our database of ∼1200 plant defensin sequences ([@B56]). ZmD32 (GenBank ID: AI665674.1) was selected by querying the database for the defensin with the highest charge at pH 7. NbD6 and PaD2 were identified based on a search of the database for sequences with similarity to ZmD32. The three defensin sequences retrieved from the data based were ordered from GenScript, codon optimized for *P. pastoris*, and were produced recombinantly using the pPINK expression system (Thermo Fisher Scientific), a modification of the *P. pastoris* system as described previously for other defensins ([@B29]).
The filamentous fungus *Fusarium graminearum* PH-1 was maintained on synthetic nutrient poor agar (0.1% KH2PO4, 0.1% KNO3, 0.1% MgSO4⋅7H2O, 0.05% KCl, 0.02% Glucose, 0.02% Sucrose, 2% agar) at 25°C. *Candida albicans* ATCC 90028, *Candida glabrata* ATCC 90030, *Candida tropicalis* ATCC 750 and clinical isolates of *Candida auris*, *Candida krusei*, and *Candida parapsilosis* (provided by Dr. Sarah Kidd at the National Mycology Reference Centre at SA Pathology, Adelaide, SA, Australia) were maintained on YPD agar plates (1% yeast extract, 2% peptone, 2% dextrose, 2% agar) and overnight cultures were grown in liquid YPD (1% yeast extract, 2% peptone, 2% dextrose) media at 30°C. Bacterial strains \[*Escherichia coli* TOP10, *Staphylococcus aureus* ATCC 9144, *Bacillus subtilis* (La Trobe University isolate), and *Pseudomonas aeruginosa* PAO1\] were all maintained on LB agar (1% tryptone, 0.5% yeast extract, 1% NaCl, 1.5% agar) plates and overnight cultures were grown in LB (1% tryptone, 0.5% yeast extract, 1% NaCl) medium at 37°C.
Antifungal and Antibacterial Assays
-----------------------------------
The 96-well microtitre plate assay described in ([@B9]) was used to test the effect of ZmD32 on the growth of both fungal and bacterial species. NaD1 was used as a standard alongside ZmD32 for comparison. Briefly, the microbes were diluted to an OD~600~ of 0.0002 in half-strength potato dextrose broth (½ PDB) (BD Biosciences) for *Candida* species, 5 × 10^4^ spores/mL in ½ PDB for *F. graminearum*, and to an OD~600~ of 0.01 in ½ PDB adjusted to pH 7 for *E. coli*, *B. subtilis*, *P. aeruginosa*, and *S. aureus*. Diluted microbes (90 μL) were then added to a twofold dilution series of plant defensin (10 μL), the optical density at 595 nm was measured at time 0 and 24 h. Growth was calculated as the difference in absorbance after the 24 h incubation, and data was analyzed in Microsoft Excel. For salt tolerance assays 100 mM NaCl was added to the culture medium. Anti-biofilm assays were performed as described in ([@B49]). Biofilms were generated by incubation of *C. albicans* in RPMI medium (Sigma) with 0.03% (w/v) L-glutamine and buffered to pH 7.0 with 0.165 M 4-morpholinopropanesulfonic acid (MOPS) at 37°C for 24 h. Prior to treatment with a twofold dilution series of defensin down from 300 μg/mL or a no defensin control, biofilms were washed three times with PBS (100 μL) before fresh RPMI medium was added. Presto Blue (Thermo Fisher Scientific) was used to measure the viability of the cells in the biofilm ([@B43]) after a 24 h treatment with defensin at 37°C. The significance of differences in IC~50~ for defensins between medium with and without salt was assessed using a two-tailed *t*-test with a cut-off of *p* = 0.05.
Cell Count Colony Forming Unit (CFU) Method
-------------------------------------------
*Candida albicans* ATCC 90028 cells were diluted to an OD~600~ of 0.1 in 1/2 PDB before 90 μL was transferred to each of five microfuge tubes containing 10 μL of 100 μM ZmD32 in Milli-Q water. Cells in the first tube (0 h) were diluted immediately to 1:1000 in ½ PDB and 100 μL was spread onto a fresh YPD agar plate and incubated at 30°C for 18--24 h. The cells in the remaining four tubes were incubated at 30°C with shaking (900 rpm) on a Thermomix Comfort (Eppendorf) heating block and plated at 30 min intervals until all the tubes were exhausted at 120 min. Cell viability in 100 mM NaCl was assessed following a similar process except the cells were plated every 2 h apart from the cells in the final tube, which were plated after 24 h incubation. Cell survival was determined by counting the colonies after overnight growth.
Membrane Lipid Binding Assay
----------------------------
Lipid binding assays were performed using Membrane Lipid Strips (Echelon Biosciences) as described in ([@B51]). ZmD32 binding to the lipids on the strips was detected using a polyclonal antibody generated in rabbits to the defensin NaD2 with which it shares 78.7% amino acid sequence identity ([@B16]). Cross reactivity of the antibody was determined by running 0.5 and 1 μg of ZmD32 and NaD2 on SDS-PAGE followed by Western blotting with the anti-NaD2 polyclonal antibody. The signals from equivalent concentrations of the two defensins were compared and used to determine the required antibody dilution needed to detect ZmD32 in the lipid binding assay. Lipid binding of NaD1 was performed as described in ([@B51]).
Anti-tumor Cell Assay With PC3, AHDF, and U937 Cell Lines
---------------------------------------------------------
Leukemia monocyte lymphoma (U937) cells and prostate cancer (PC3) cells were cultured in RPMI-1640 medium (Sigma) buffered to pH 7.2 with HEPES and supplemented with 5--10% (v/v) fetal calf serum (FCS), 100 U/mL penicillin (Sigma) and 100 μg/mL streptomycin (Invitrogen). Adult human dermal fibroblast (AHDF) cells were cultured in DMEM medium (Sigma) supplemented with 10% (v/v) FCS, 100 U/mL penicillin and 100 μg/mL streptomycin. All cell lines were cultured at 37°C in a humidified atmosphere containing 5% CO~2~. Cell viability assays were performed using an MTT assay as described in ([@B51]). A twofold dilution series of NaD1 or ZmD32 from 50 μM down was assessed. IC~50~ values were determined using GraphPad Prism 5 (GraphPad Software, CA, United States).
Human Red Blood Cell Lysis Assay
--------------------------------
Human blood was obtained from the Australian Red Cross Blood Service (Melbourne, VIC, Australia) under Material Supply Agreement 14-11VIC-03. Hemolytic activity of defensins was assessed as described in ([@B18]). The degree of red blood cell (RBC) lysis was measured by absorbance at 412 nm, relative to water-lysed RBCs.
NMR Structural Analysis
-----------------------
ZmD32 was dissolved in 10% D~2~O/H~2~O at a concentration of ∼1 mM and a pH of 3.5. NMR spectra were recorded at 298 K on a Bruker Avance III 600 spectrometer. Chemical shifts of backbone and sidechain resonances were assigned by analysis of 2D TOCSY (with an 80 ms MLEV-17 spin lock), NOESY (mixing time of 200 ms), ECOSY and natural abundance ^13^C and ^15^N HSQC experiments. Solvent suppression was achieved using excitation sculpting. Slowly exchanging amides were identified by slow D~2~O exchange and sensitivity of amide shifts to temperature. Spectra were processed using Topspin 3.5 (Bruker) then analyzed using CcpNmr Analysis ([@B69]). Chemical shifts were referenced to internal DSS.
Distance restraints were derived from NOESY spectra recorded in both 10 and 100% D~2~O and used to generate initial structures with CYANA. Additional restraints included disulphide bonds; hydrogen bonds; χ~1~ restraints from ECOSY and NOESY data; and backbone φ and ψ dihedral angles generated using the program TALOS-N ([@B59]). CNS was then used to generate a final set of 20 structures using torsion angle dynamics, and refinement and energy minimization in explicit solvent ([@B11]). Final structures were assessed for stereochemical quality using MolProbity ([@B13]).
Solvent Accessible Surface Potential Plot Analysis
--------------------------------------------------
The solvent accessible surface potential plot was generated using the APBS Tools (v2.1) ([@B4]) from the plugin menu of the Pymol ([@B14]) modeling, manipulation, and visualization program. To do this the structural information was saved as a PDB file and converted to PQR file format using the online PDB2PQR program ([@B15]). Essentially, the default input parameters were chosen except for the pH value, which was set to 7.4. The PROPKA program was chosen to assign protonation states at the chosen pH. The input data was submitted and the PDB2PQR calculated the force field parameters (charges and radii) and returned the result in PQR file format. The file was then saved and uploaded into Pymol and the APBS Tool option from the plugin pull down menu the was used to generate the Poisson-Boltzmann solvent accessible surface potential plot of each defensins.
The NaD2, NbD6, and PaD2 models were generated by homology modeling using SWISS-Model and the ZmD32 NMR structure as a template. The surface charge was then mapped as described above for NaD1 and ZmD32.
Results
=======
ZmD32 Inhibits Growth of Both Fungi and Bacteria
------------------------------------------------
A search of our database of approximately 1200 plant defensin sequences identified ZmD32 as the defensin with the highest charge (+10.1) at pH 7 notably higher than the typical +4 to +6 for most plant defensins ([@B58]). A tBLASTn query of the non-redundant NCBI data base aligned the ZmD32 mature defensin sequence to *Z. mays* mRNA EU952861 ([@B2]). The full length mRNA sequence indicates that *in planta* ZmD32 would be expressed as a class I defensin, that is with a N-terminal signal peptide but no C-terminal vacuolar targeting sequence. The high charge of the peptide was predicted to be retained across a wide pH range ([Figure 1A](#F1){ref-type="fig"}). ZmD32 was expressed in *P. pastoris*, purified by cation exchange and Reverse Phase-High Performance Liquid Chromatography (RP-HPLC) ([Figure 1B](#F1){ref-type="fig"}). The purity and mass of the peptide were confirmed by SDS-PAGE ([Figure 1C](#F1){ref-type="fig"}) and MALDI-TOF MS identified a single peak with a mass of 5530.6 Da (data not shown). This mass is in agreement with the expected mass of 5529.3 Da for the reduced defensin with an N-terminal alanine residue that remains after signal peptide cleavage.
ZmD32 exhibited antifungal activity at low micromolar concentrations. It inhibited the growth of *C. albicans*, *C. auris*, *C. glabrata*, *C. krusei*, *C. parapsilosis*, *C. tropicalis* and the filamentous fungal plant pathogen *F. graminearum* PH-1 with 50% inhibition concentrations (IC~50~) ranging from 0.5 to 4.0 μM. The level of activity of ZmD32 was comparable to that of the well-characterized antifungal plant defensin NaD1 (IC~50~ values of 0.4--2.8 μM) ([Table 1](#T1){ref-type="table"}).
######
Activity of ZmD32 against fungi and bacteria in low salt medium.
Species ZmD32 IC~50~ (μM) ZmD32 NaCl IC~50~ (μM) NaD1 IC~50~ (μM) NaD1 NaCl IC~50~ (μM)
-------------------------- ------------------- ------------------------ ------------------ -----------------------
*C. albicans* ATCC 90028 1.1 ± 0.5 3.0 ± 0.2\* 1.6 ± 0.4 \>10 μM
*C. auris* 3.4 ± 0.5 1.6 ± 0.4\* 2.9 ± 0.5 \>10 μM
*C. glabrata* ATCC 90030 1.2 ± 0.4 2.0 ± 1.0 2.8 ± 1.0 \>10 μM
*C. krusei* 0.9 ± 0.1 1.7 ± 0.4\* 1.2 ± 0.4 \>10 μM
*C. parapsilosis* 1.3 ± 0.6 1.5 ± 0.4 1.2 ± 0.3 3.4 ± 1.2\*
*C. tropicalis* ATCC 750 0.5 ± 0.1 0.7 ± 0.2 0.4 ± 0.1 1.3 ± 0.5\*
*F. graminearum* PH-1 1.0 ± 0.7 NA 1.0 ± 0.5 NA
*E. coli* 1 ± 0.2 0.3 ± 0.1\* 4.6 ± 1.6 \>10 μM
*S. aureus* ATCC 9144 1.5 ± 0.3 NA 5.0 ± 0.5 NA
*B. subtilis* 0.4 ± 0.2 NA 2.0 ± 0.2 NA
*P. aeruginosa* PAO1 1.7 ± 0.1 NA 4.3 ± 0.3 NA
Values are the mean ± standard deviation of three biological replicates.
∗
Indicates values that are significantly different from the IC
50
in medium without salt. NA indicates conditions that were not assessed
.
The broad spectrum antifungal activity of ZmD32 along with recent reports on the antibacterial activities of plant defensins ([@B53]) led to the hypothesis that ZmD32 would also be active against bacteria. ZmD32 inhibited the growth of *E. coli*, *B. subtilis*, *S. aureus*, and *P. aeruginosa* with IC~50~ values from 0.4 to 1.7 μM in ½ PDB, pH 7 ([Table 1](#T1){ref-type="table"}). ZmD32 was more active against both Gram-negative and Gram-positive bacteria than NaD1 (IC~50~ values 2.0--5.0 μM) ([Table 1](#T1){ref-type="table"}).
ZmD32 Retained Antimicrobial Activity in the Presence of Salt
-------------------------------------------------------------
Many plant defensins lose activity when the concentration of salt in the assay medium is raised. We examined whether the high charge on ZmD32 would allow this defensin to retain activity in media with elevated salt concentrations. ZmD32 inhibited *C. albicans*, *C. auris*, *C. glabrata*, *C. krusei*, *C. parapsilosis*, and *C. tropicalis* growth with IC~50~ values ranging from 0.7 to 3.0 in ½ PDB medium containing 100 mM NaCl. In contrast, NaD1 was only active against *C. parapsilosis* and *C. tropicalis* with IC~50~ values of 3.4 and 1.3 μM, respectively, in medium with added NaCl. ZmD32 also retained activity against *E. coli* when NaC1 concentrations were raised to 100 mM while NaD1 did not ([Table 1](#T1){ref-type="table"}). For *C. albicans* and *C. krusei* the IC~50~ for ZmD32 was slightly higher when the assay was performed with the addition of 100 mM NaCl, for *C. glabrata*, *C. parapsilosis*, and *C. tropicalis* the IC~50~ was not significantly different at 100 mM NaCl and for *C. auris* and *E. coli*. The activity of ZmD32 and NaD1 against *C. albicans* was also assessed in the presence of 5 mM MgCl~2~ or 2 mM CaCl~2~. NaD1 lost activity in the presence of both MgCl~2~ and CaCl~2~ while ZmD32 was active in both divalent metal salts but had an sevenfold increase in IC~50~ (data not shown).
The Rate of Fungal Cell Killing by ZmD32 Is Slower in 100 mM NaCl
-----------------------------------------------------------------
A cell survival assay was used to assess whether ZmD32 was fungistatic or fungicidal against *C. albicans*. There was a greater than 95% reduction in survival of *C. albicans* cells after a 30 min exposure to 10 μM ZmD32 and no viable cells after 2 h in ½ PDB medium without added salt ([Figure 2A](#F2){ref-type="fig"}). However, cell death was slower when the assay was performed in ½ PDB medium containing 100 mM NaCl ([Figure 2B](#F2){ref-type="fig"}). Most cells (92%) remained viable after 2 h exposure to 10 μM ZmD32 but viability decreased to 31 and 24% after 4 and 6 h, respectively. After 24 h no viable cells were detected.
{#F2}
ZmD32 Is Active Against Biofilms
--------------------------------
Biofilms are a crucial component of *C. albicans* infections. The activity of ZmD32 and NaD1 against established *C. albicans* biofilms was assessed using Presto Blue to measure cell viability. Biofilms were generated and assayed in RPMI medium, which contains a physiological concentration of sodium (132.1 mM). ZmD32 was active against *C. albicans* biofilms in RPMI but NaD1 was not ([Figure 3](#F3){ref-type="fig"}).
{#F3}
Anti-tumor Cell Activity Assays of ZmD32
----------------------------------------
To determine if ZmD32 was toxic to mammalian cells, the effect of ZmD32 on three human cell lines was tested and compared to NaD1. NaD1 has been reported to be active against tumor cell lines ([@B51]). Both ZmD32 and NaD1 were more active on the PC3 and U937 tumor cell lines than fibroblasts. However, NaD1 was much more active that ZmD32 ([Figure 4A](#F4){ref-type="fig"}--[C](#F4){ref-type="fig"}).
{#F4}
ZmD32 Exhibits Minimal Hemolytic Activity
-----------------------------------------
The effect of ZmD32 on the membrane integrity of human red blood cells was tested using samples from six blood donors. ZmD32 lysed only 1.9% of the red blood cells at the highest concentration (50 μM) tested, which was comparable to recombinant NaD1 which lysed about 1.5% at the same concentration ([Figure 4D](#F4){ref-type="fig"}).
Characterization of ZmD32 by NMR
--------------------------------
The solution structure of ZmD32 was determined by NMR spectroscopy. The NMR spectra had good amide dispersion and backbone resonances were fully assigned apart from residues Arg17 and Asn19 ([Supplementary Figure S2](#SM2){ref-type="supplementary-material"}). The secondary Hα shifts of ZmD32 indicated that the solution structure consists of both α-helix and β-strand elements ([Figure 5A](#F5){ref-type="fig"}).
{#F5}
The structure of ZmD32 was calculated with a total of 512 distance restraints, 74 dihedral angle restraints and 24 hydrogen bond restraints. Amide temperature coefficients and deuterium exchange experiments were used to identify residues taking part in hydrogen-bond interactions (S5, N22, C24, R25, T26, E27, F29, R33, F42, C43, T44, T45) further supporting the identified secondary structural elements. The following disulphide connectivities were also included as restraints in the structure calculations: Cys3-Cys47; Cys14-Cys34; Cys20-Cys41; and Cys24-Cys43. The two proline residues were both determined to adopt the *trans* conformation, as evidenced by strong H^δ^(*i*)Pro-H^α^ (*i*-1) signals in NOESY spectra and the ^13^C shifts of the C^β^ and C^γ^proline resonances. The resulting family of structures overlaid well, with a RMSD for the backbone atoms of 0.79 Å ([Figure 5B](#F5){ref-type="fig"}). Analysis of these structures revealed that 96% of the residues fall in the most favored regions of the Ramachandran plot and a mean overall MolProbity score of 1.6 indicates very good structural quality ([Supplementary Table S1](#SM3){ref-type="supplementary-material"}). ZmD32 adopts a typical CSαβ motif ([Figure 5C](#F5){ref-type="fig"}) with an α-helix spanning 10 residues from Arg17-Thr26 and a triple-stranded anti-parallel β-sheet (β1 = Thr2-Gln6; β2 = Gly31-Cys34; β3 = Cys41-His46). The loops connecting the β-strands to each other and to the α-helix are reasonably well-defined. The assigned chemical shifts of ZmD32 have been deposited in the BMRB (accession 30475) and structural coordinates have been deposited in the PDB (6DMZ).
ZmD32 is the most cationic plant defensin described with an overall charge of +10.1 at pH 7. The positively charged amino acids are located throughout the sequence ([Figure 5A](#F5){ref-type="fig"}) and distributed across the structure of the molecule ([Figure 5C](#F5){ref-type="fig"}). In order to gain further insight on how the charge may affect the activity of defensins in salt, a solvent accessible surface potential plot was generated for ZmD32 and NaD1. ZmD32 had a highly charged pocket that was not present on NaD1 and the surface of ZmD32 in general was more positively charged than NaD1 ([Figure 6](#F6){ref-type="fig"}).
{#F6}
Membrane Lipid Binding Assay
----------------------------
A number of plant defensins target lipids on fungal membranes ([@B52]; [@B51]; [@B6]). A protein-lipid overlay assay was there for used as a preliminary screen to determine whether ZmD32 interacts with lipids. ZmD32 bound to several phospholipids including phosphatidylinositol mono-/bis-/tri-phosphates, phosphatidic acid, phosphatidylserine, and cardiolipin ([Figure 7](#F7){ref-type="fig"}). Under the same conditions NaD1 bound preferentially to phosphatidylinositol-4,5-bisphosphate \[PI(4,5)P~2~\] but also bound to phosphatidic acid (PA) and cardiolipin.
{#F7}
Salt Tolerance of Defensins With High Sequence Similarity to ZmD32
------------------------------------------------------------------
As ZmD32 displayed attractive features for a putative antimicrobial therapeutic, namely retention of activity at physiological salt concentrations, broad-spectrum activity against microbes and minimal activity against human cells, we queried our database for defensins with sequences that were highly similar to ZmD32. This search returned three proteins, NaD2 (from *Nicotiana alata*), PaD2 (from *Parthenium argentatum*), and NbD6 (from *Nicotiana benthamiana*) with sequence similarity to ZmD32 of 85.1, 87.2, and 87.2% and charge at pH 7 of 4.9, 6.1, and 7.6, respectively ([Figure 8A,C](#F8){ref-type="fig"}). These three defensins were assessed for activity against *C. albicans* in media with and without the addition of 100 mM NaCl. All of the defensins were active in medium with no added salt; NbD6 was the most active while PaD6 and NaD2 had similar activity. NbD6 retained activity at 100 mM NaCl while PaD2 had some inhibitory activity that plateaued at around 50% growth inhibition compared to the untreated control ([Figure 8B](#F8){ref-type="fig"}). These four defensins are very similar in sequence, including the number and arrangement of cationic amino acids so we also compared the distribution of surface charge ([Supplementary Figure S1](#SM1){ref-type="supplementary-material"}). The cationic pocket on the surface of ZmD32 is not present in NaD2 or PaD2 and was present but not as charged in NbD6. This finding indicates that it is not just the presence of the cationic residues that leads to the formation of the cationic pocket but also the electrostatic effects of neighboring residues.
). The symbol "^∗^" indicates amino acids common to all three sequences and ":" indicates amino acids with strongly similar properties and. amino acids with weakly similar properties. The percent identity and similarity to ZmD32 calculated using the Ident and Sim tool in the Sequence Manipulation Suite (<http://www.bioinformatics.org/sms2/>) ([@B60]) is listed next to each sequence. Positively charged residues are highlighted in blue, negatively charged residues in red. **(B)** Activity of NaD2 (orange), NbD6 (purple), and PaD2 (red) against *C. albicans* in ½ PDB with (solid lines) and without (dotted lines) the addition of 100 mM NaCl. All of the defensins were active in medium with no added NaCl. In medium with 100 mM NaCl NbD6 retained activity, PaD2 had some activity but this plateaued at around 50% inhibition and NaD2 was inactive. Values are the average of three parallel replicates, error bars are standard deviation. Data is representative of three biological replicates. **(C)** The predicted charge of the three defensins along with ZmD32 was calculated using PROTEIN CALCULATOR v3.4 (<http://protcalc.sourceforge.net/>) and graphed vs. pH value.](fmicb-10-00795-g008){#F8}
Discussion
==========
One of the impediments to the development of antimicrobial peptides as pharmaceuticals is the substantial loss of activity at physiological salt concentrations. This loss of activity has been attributed to the disruption of the ionic interaction between cationic AMPs and anionic microbial membranes. ZmD32, the most positively charged defensin identified to date (+10.1 at pH 7) was active against *C. albicans* in both the yeast and biofilm form as well as other *Candida* species, the filamentous fungus *F. graminearum* and *E. coli* in media containing 100 mM NaCl. It also retained activity in the presence of MgCl~2~ and CaCl~2~, although with an increased IC~50~. ZmD32 is the first salt-tolerant antimicrobial plant defensin reported and one of only a handful of naturally occurring salt-tolerant antimicrobial peptides that have been described ([@B22]). Many of the other naturally occurring salt-tolerant AMPs were isolated from marine organisms that exist in a saline environment, providing an explanation for the evolution of salt-tolerant AMPs ([@B35]; [@B22]). The selective pressures for a salt-tolerant AMP in corn are less obvious. Three defensins with highly similar sequences to ZmD32 were assessed for salt-tolerant antifungal activity. Only NbD6 retained antifungal activity in a similar manner to ZmD32 when tested in high salt media. NbD6 was the most charged of the three defensins supporting the hypothesis that defensin charge correlates with activity in high salt media.
The antibacterial activity of ZmD32 and NaD1 adds to a growing list of antibacterial defensins ([@B53]; [@B54]; [@B67]), a protein family that is best known for antifungal activity ([@B64]; [@B65]; [@B45]). The lack of reported antibacterial activity for plant defensins likely reflects the medium used for antibacterial assays, Here the ½ PDB pH 7 medium has a NaCl concentration of less than 10 mM. This concentration is considerably lower than growth media often used in antibacterial assays such as Mueller Hinton, which is recommended by the Clinical and Laboratory Standards Institute, and has a Na^+^ concentration of 97 mM ([@B40]), or Lysogeny Broth (LB) which has a NaCl concentration of 171 mM. Only ZmD32 was active in ½ PDB after addition of 100 mM NaCl. NaD1 has been reported to have no activity against bacterial species such *E. coli* or *S. aureus* ([@B66]) but these earlier assays were conducted in LB. We anticipate many other plant defensins as well as other families of cationic AFPs reported to lack antibacterial activity will also be active in low salt media. Recent reports on the antibacterial activity of MtDef4 and MtDef5 further support the notion that plant defensins are more active against bacteria under low salt conditions, than previously reported ([@B53]; [@B67]).
Since ZmD32 had potent activity against fungal and bacterial cells regardless of salt levels, it was hypothesized that this defensin was acting by membrane disruption mechanisms. For example, non-specific disruption of phospholipid bilayers has been reported previously for non-selective membrane-lytic peptides ([@B7]). However, this mode of action is unlikely because ZmD32 had low activity on three human cell lines tested and did not significantly lyse red blood cells. Similarly, the increased charge on ZmD32 did not lead to increased activity against human cells because the mechanism of action of plant defensins against tumor cell lines involves formation of specific quaternary structures of defensins and lipids ([@B5]). The low activity of ZmD32 indicates that it does not form the structures for anti-tumor cell activity regardless of charge.
Although the exact mode of action of plant defensins against fungi varies, many share common features of an initial interaction with cell wall components, followed by an interaction with membrane components and entry into the cytoplasm ([@B45]). While the interaction of defensins with cell wall components remains a subject of investigation, the membrane binding targets have been largely identified as glycol-lipids, glyco-proteins, and phospholipids ([@B45]). In this work the ZmD32 lipid binding specificity was examined by a protein-lipid overlay assay. ZmD32 bound a number of phospholipids including PA, PS, cardiolipin, PtdIns(4)P PtdIns(4,5)P~2~, and PtdIns(3,4,5)P~3~ This broad spectrum lipid binding was surprising as ZmD32 has the same RGFRRR PA binding motif in loop 5 as MtDef4 ([@B52]) and NaD2 ([@B47]). Substitution of the native loop 5 sequence of NaD1 with RGFRRR switched the lipid binding preference of the resulting defensin from PI(4,5)P~2~ to PA ([@B8]). Binding of ZmD32 to phospholipids other than PA is likely to occur through sequences other than the defined loop 5 lipid-binding motif. It could arise from the attraction between highly cationic defensin and the negatively charged phosphates in the head groups of the phospholipids. This hypothesis is supported by the observation that ZmD32 lacks many of the PI(4,5)P~2~ binding residues found in the cationic grip of NaD1 (K4, K36, I37, L38) ([@B51]) but still binds well to PI(4,5)P~2~.
Lipid binding by NaD1 and other defensins occurs at NaCl concentrations of 100 mM and higher ([@B47]) and the tumor cell killing activity of NaD1, which is more dependent on lipid binding than the antifungal activity ([@B8]), also occurs at elevated concentrations of NaCl. The fact that only the antimicrobial activity of defensins is affected by salt concentration means that the loss of activity observed for most defensins in the presence of salt must be related to a structure that is common in microbes but absent from tumor cells and artificial lipid bilayers. The obvious hypothesis is that there are changes to microbial cell walls that affect the ability of defensins to access microbial membranes and exert their activity. This has been demonstrated for *C. albicans* where a shift to high salt induces a tightening of the cell wall matrix ([@B17]). Perhaps the additional charge and/or charge density on ZmD32 means that it is more attracted to the negatively charged membrane, meaning that the force of attraction is sufficient to drag it through the smaller pores of the fungal walls in the presence of NaCl. That is, the lipid binding specificity of ZmD32 does not result in salt tolerance, instead it is the magnitude and arrangement of the charge on the defensin that makes it retain activity in elevated salt concentrations.
The solvent accessible surface plot revealed a cationic surface with pockets of increased charged density in ZmD32. NaD1 has a similar cationic surface but lacked the pockets of increased charge. These pockets are in the region of loop 5, which in ZmD32 is rich in positively charged amino acids (RGFRRR). Comparison between the surface of ZmD32 and NaD2, which has the same loop 5 sequence, once again showed that ZmD32 had a more positively charged surface. These surface comparisons between defensins support that the charge density in loop 5 (RGFRRR) is necessary but not sufficient for antifungal action in high salt medium. Furthermore, the highly cationic region in structure ([Figure 6C](#F6){ref-type="fig"}) is formed by the basic residues R16, R17, R33, R35, and R39 that are orientated in the direction of the cationic pocket.
The demonstrated activity of ZmD32 against fungi, including biofilms, and bacteria in medium with NaCl concentrations in the physiological range, combined with minimal activity against human cells reveals that plant defensins can incorporate many properties desirable for development of novel antimicrobials. When combined with their known stability to heat, pH and proteolysis these properties provide a strong foundation for the development of plant defensins as antimicrobial therapeutics for use in the clinic. Their broad antifungal and antibacterial activities are particularly exciting as there is potential utility against polymicrobial infections.
Data Availability
=================
The datasets generated for this study can be found in Protein Data Bank, 6DMZ.
Author Contributions
====================
BK led experimental work and wrote the first draft of the manuscript. JM performed the antibiofilm assays and salt tolerance assays, and edited the manuscript. PH solved the NMR structure of ZmD32 and edited the manuscript. PQ performed the molecular modeling. DG-C performed the antibacterial assays. FL performed the hemolysis assays. TP performed the tumor cell assays. PV performed the protein expression. SV performed the ZmD32 like protein antifungal assays. KP performed the lipid binding assay. TS performed the database search. NW performed data analysis with the help of PV and SV. MH performed data analysis with the help of FL and TP and edited the manuscript. DC performed the NMR analysis. MA coordinated the experiments and data analysis, and edited the manuscript. MB coordinated the experiments and data analysis, and wrote the manuscript.
Conflict of Interest Statement
==============================
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
**Funding.** This work was funded by an Australian Research Council grant to MA and NW (DP150104386) and MA (DP160100309). JM is a Commonwealth Scientific and Industrial Research Organisation Science and Industry Endowment Fund STEM Fellow. DC is an Australian Research Council Australian Laureate Fellow (FL150100146).
We are grateful for access to the NMR Facility of the Queensland NMR Network.
Supplementary Material
======================
The Supplementary Material for this article can be found online at: <https://www.frontiersin.org/articles/10.3389/fmicb.2019.00795/full#supplementary-material>
######
Surface charge of ZmD32 like defensins. The structures of PaD2, NaD2, and NbD6 were modeled based on the NMR structure of ZmD32 using Swiss-Model ([@B28]) (<https://swissmodel.expasy.org/>). The surface charge was then mapped using the same methods as for ZmD32 and NaD1 in [Figure 6](#F6){ref-type="fig"} with blue indicating positive charge, red negative charge, and white neutral. The positive charge on the surface of ZmD32 is not present in PaD2 or NaD2 and is weaker in NbD6.
######
Click here for additional data file.
######
Fingerprint region of the **(A)** TOCSY and **(B)** NOESY spectra of ZmD32 recorded at 298 K at 600 MHz. The sequential connectivity pattern shown indicates the assignment of the peptide backbone resonances.
######
Click here for additional data file.
######
Statistical analysis of ZmD32 structures.
######
Click here for additional data file.
[^1]: Edited by: László Galgóczy, Hungarian Academy of Sciences, Hungary
[^2]: Reviewed by: Dilip Shah, Donald Danforth Plant Science Center, United States; Joel E. Lopez-Meza, Universidad Michoacana de San Nicolás de Hidalgo, Mexico
[^3]: ^†^These authors have contributed equally to this work
[^4]: This article was submitted to Fungi and Their Interactions, a section of the journal Frontiers in Microbiology
| {
"pile_set_name": "PubMed Central"
} |
Systems biology, a multidisciplinary field of research, is suddenly gaining new attention from scientists in both academia and industry. It involves the construction of quantitatively predictive models of biological systems with the aim of understanding, predicting and controlling physiological behaviour by integrating knowledge of interactions at molecular, cellular and population levels. This book draws together contributions from writers within different disciplines and demonstrates the value of bringing together theorists from computer science, engineering, physics, mathematics and statistics with experimentalists from biochemistry, chemistry, biology and physiology.
This collection of chapters gives a thorough introduction to systems biology, providing a clearly structured list of reasons why such a systems approach is required in biology. By using well chosen examples and illustrative applications, the book manages to convey some of the complicated concepts underpinning research in systems biology. It explains why this approach relies on an iterative cycle between theory, modelling and experiments using high-throughput techniques.
The book emphasises the diverse range of skills required to successfully contribute to research in systems modelling and calls for a coordinated multidisciplinary approach where experimentalists and theorists collaborate and share resources. It clearly demonstrates that this collaboration should go beyond the sharing of experimental datasets and models if the results of research in systems biology are to be used to understand biological systems at the predictive level needed for the diagnosis, prevention and treatment of medical disease. The growth of multidisciplinary doctoral training centres and recent availability of funding to encourage research in systems biology has the potential to deliver researchers with sufficient knowledge of both experimental and theoretical issues. The authors of this book are well known for their contributions to the field of systems biology and the integration of their extensive experience is perhaps the best way to form a comprehensive introduction to the advantages, disadvantages and challenges of systems modelling applied to cellular biology.
The four principle sections of the book are structured around general concepts, modelling approaches, models and reality and computational modelling. A clear account of the role of modelling in systems biology is provided, highlighting the philosophical arguments behind model selection and summarising the different kinds of modelling approaches available. It also describes important concepts such as nonlinearity, feedback loops, sensitivity analysis, parameter exploration and model consistency. By focusing on the interplay between complexity and robustness, it emphasises the importance of modelling the network structure and discusses concepts such as stability, fragility and evolvability. It also presents the advantages and disadvantages of modules when modelling complex systems. The book then proceeds to describe the wide range of possible modelling approaches; these include Bayesian inference, constraint-based models, nonlinear ordinary differential equations, qualitative models, stochastic models and spatially extended models. One of the book\'s strengths is that relevant applications and examples are presented throughout to help convey the essential concepts underlying each modelling approach. The applications include metabolic networks, molecular interaction networks, genetic regulatory networks, intracellular kinetics and reaction-diffusion systems. The next section describes the range of experimental data available and emphasises the many sources of observational uncertainty and measurement error that should be accounted for when modelling. It covers issues such as data acquisition, methods for identifying architecture and dynamics from experimental data, applications of control theory, synthetic gene regulatory systems and multilevel modelling. The book closes with a description of the computational techniques required when implementing a systems modelling approach. By discussing computational constraints, numerical simulation methods and software infrastructure, it is apparent that knowing the limitations and cost of increasing model size is important for both experimentalists and modellers.
\"Systems biology in cellular biology\" provides a great resource for researchers from different backgrounds who wish to contribute to systems biology. The notable collection of contributors helps to provide a comprehensive view of the latest developments in the exciting new field of systems biology. The book forms an impressive overview of the entire field which is suitable for students, teachers and researchers, whether they be experimentalists or theorists.
| {
"pile_set_name": "PubMed Central"
} |
All relevant data are within the manuscript.
Introduction {#sec001}
============
Both diabetic retinopathy (DR) and nephropathy are typical microvascular complications of diabetes mellitus (DM). It is well known that the prevalence of chronic kidney disease (CKD) and DR increases proportionally to the disease duration in type 2 DM \[[@pone.0220506.ref001], [@pone.0220506.ref002]\]. CKD and DR also share common risk factors such as smoking, poor glycemic control, systolic hypertension, or dyslipidemia \[[@pone.0220506.ref001], [@pone.0220506.ref003], [@pone.0220506.ref004]\]. Since both CKD and DR reflect similar pathogenesis and microvascular lesions, it is reasonable to assume that development of DR may predict development and progression of CKD.
However, the association between DR and CKD has not been well established in type 2 DM, and a few studies found that their association in type 2 DM patients is much weaker than that in type 1 DM patients \[[@pone.0220506.ref005], [@pone.0220506.ref006]\]. In addition, the role of DR as a predictor for development and progression of CKD is still controversial. The Microalbuminuria Collaborative Study Group found that DR was not an independent predictor of albuminuria \[[@pone.0220506.ref007]\], but other researchers found that the presence and severity of DR are still indicators for the risk of developing proteinuria \[[@pone.0220506.ref002], [@pone.0220506.ref008]\].
Recently, a few prospective cohort studies suggested that the presence of retinopathy is related to the development and progression of renal diseases in both diabetic and non-diabetic CKD patients \[[@pone.0220506.ref009]--[@pone.0220506.ref013]\]. However, there was no study evaluating the effect of DR severity upon CKD progression in type 2 DM patients. Therefore, we aimed to assess the value of DR severity to predict renal dysfunction and albuminuria progression in type 2 DM patients.
Materials and methods {#sec002}
=====================
Medical records from a total of 2,197 adult patients with type 2 DM who visited ophthalmology clinic in Kangnam Sacred Heart Hospital from August 2006 to February 2014 were screened for enrollment. Type 2 DM was diagnosed when the patient is older than 20 years old and presents with one of the followings according to World Health Organization diabetes diagnostic criteria: 1) fasting blood glucose ≥ 6.99 mmol/L, 2) 2-hour plasma glucose level after 75 g oral glucose tolerance test ≥ 11.1 mmol/L, 3) random plasma glucose level ≥ 11.1 mmol/L with diabetic symptoms, 4) glycated hemoglobin (HbA1c) ≥ 48 mmol/mol (6.5%)\[[@pone.0220506.ref014]\]. Among 2,197 patients, 547 patients were excluded from the analysis due to following reasons: 400 with no serial follow-up data over 1 year (272 with single eGFR measurement, 128 with eGFR follow-up measurement within a year (mean follow-up duration 199.4±103.3 days), 118 with advanced CKD (CKD stage 4 and 5), 17 type 1 DM, 49 without retinal exams, 8 hypertensive retinopathy, 6 loss of vision, 6 artificial eyes, and 1 retinal vein occlusion ([Fig 1](#pone.0220506.g001){ref-type="fig"}).
{#pone.0220506.g001}
CKD was defined by Kidney Disease: Improving Global Outcomes (KDIGO) clinical practice guideline based on estimated glomerular filtration rate (eGFR) and albuminuria \[[@pone.0220506.ref015]\]. The eGFR was calculated by CKD-EPI creatinine equation \[[@pone.0220506.ref016]\]. Serum creatinine was measured by IDMS-traceable method \[[@pone.0220506.ref017]\]. Annual eGFR decline rate was calculated from the difference of eGFR between recent follow-up and the initial visit divided by follow-up years. The degree of albuminuria was measured by random urine albumin-to-creatinine ratio (UACR). The CKD progression was defined based on one or more of the following: 1) decline in GFR category (≥90 \[G1\], 60--89 \[G2\], 45--59 \[G3a\], 30--44 \[G3b\], 15--29 \[G4\], \<15 \[G5\] mL/min/1.73m^2^) accompanied by a 25% or greater drop in eGFR from baseline, 2) sustained declined in eGFR of more than 5 mL/min/1.73m^2^/year \[[@pone.0220506.ref018]\]. Albuminuria progression was defined by one or more step progression in albuminuria (normo-albuminuria (UACR \< 30mg/g) to micro-albuminuria (30mg/g ≤ UACR \< 300 mg/g) or macro-albuminuria (UACR ≥ 300 mg/g), micro-albuminuria to macro-albuminuria) during follow-up.
The status of DR was evaluated by slit-lamp examination, indirect ophthalmoscopy and/or fluorescein angiography by ophthalmologists. To evaluate the effect of DR severity upon CKD progression, retinopathy was classified into following categories; no DR, non-proliferative DR (NPDR), and proliferative DR (PDR) \[[@pone.0220506.ref019]\]. Each eye was given a DR grade and the final DR severity was determined by the result of more severe eye. The progression of DR was defined by more than 1 step progression in DR grade of either eye or the development of PDR requiring photocoagulation or vitrectomy \[[@pone.0220506.ref020]\].
The following data were obtained from each patient: age, gender, duration of DM, use of angiotensin converting enzyme inhibitor (ACEi) or angiotensin receptor blocker (ARB) at initial visit, body mass index (BMI) at initial visit, serial measurements of serum creatinine, eGFR, UACR, DR status at the initial diagnosis and follow-up, other laboratory findings including plasma hemoglobin, serum albumin, total cholesterol, and HbA1c. The study was approved by the Institutional Review Board (2018-01-030). The written informed consent was waivered due to retrospective nature of the study.
Statistical analysis was conducted using the SPSS software version 20.0 (SPSS, Inc., Chicago, Ill., USA). For descriptive analysis, data were represented either as mean ± standard error or relative frequencies. For normally distributed variables, the Student t-test and one-way ANOVA were used for comparisons. Binary logistic regression analysis was used to evaluate the risk factors for CKD progression. We performed univariate analysis to find risk factors for CKD progression. To seek independent association between each risk factor and CKD progression, we performed multivariate analysis by entering all significant risk factors for CKD progression. The P-value \<0.05 was considered statistically significant.
Results {#sec003}
=======
Baseline characteristics of the subjects according to the status of diabetic retinopathy {#sec004}
----------------------------------------------------------------------------------------
A total of 1,592 patients were included in the analysis ([Table 1](#pone.0220506.t001){ref-type="table"}). Mean age was 57.9 ± 11.2 years old, and female was slightly predominant (n = 841, 52.8%). The mean follow-up period was 5.6±2.1 years. A total of 586 (36.8%) patients had DR at the initial visit. Among them, 384 (24.1%) had NPDR and 202 (12.7%) had PDR at either eye. Compared to the patients without DR, the patients with NPDR and PDR had lower BMI (24.6±3.5 and 23.5±3.4 vs. 25.4±3.6 kg/m^2^, p\<0.001), longer duration of DM (11.9±8.0 and 12.0±8.2 vs. 6.7±6.3 years, p\<0.001), higher level of HbA1c (8.3±1.8 and 8.5±2.1 vs. 7.5±1.7%, p\<0.001), lower hemoglobin level (13.0±1.8 and 12.3±1.8 vs. 13.5±1.6 g/dL, p\<0.001) and lower serum albumin level (4.3±0.5 and 4.2±0.6 vs. 4.4±0.4 g/dL, p\<0.001). In addition, patients with NPDR and PDR showed decreased eGFR (76.1±19.6 and 72.4±22.3 vs. 81.5±18.5 mL/min/1.73m^2^, p\<0.001) and increased amount of albuminuria (median 12.6 and 23.1 vs. 37.0 mg/g, p\<0.001) compared to those without DR.
10.1371/journal.pone.0220506.t001
###### Baseline characteristics of the subjects classified by diabetic retinopathy.
{#pone.0220506.t001g}
-------------------------------------------------------------------------------------------------------
Variables No DR\ NPDR\ PDR\ p-value
(n = 1,006) (n = 384) (n = 202)
--------------------------- -------------------- --------------------- ---------------------- ---------
Age (years) 57.5±11.2 59.3±10.9 57.3±11.2 0.016
Male (%) 470 (46.7) 175 (45.6) 106 (52.5) 0.252
Hypertension (%) 485 (50.2) 201 (54.6) 87 (43.5) 0.04
BMI (kg/m^2^) 25.4±3.6 24.6±3.5 23.5±3.4 \<0.001
DM duration (years) 6.7±6.3 11.9±8.0 12.0±8.2 \<0.001
ACEi or ARB (%) 471 (46.8) 202 (52.6) 97 (48.0) 0.154
sBP (mmHg) 128.8±18.7 131.8±20.2 133.2±22.6 0.026
dBP (mmHg) 76.9±12.7 78.9±13.2 79.0±15.1 0.057
HbA1c (%) 7.5±1.7 8.3±1.8 8.5±2.1 \<0.001
FPG (mg/dL) 143.2±59.9 153.6±68.0 171.4±96.3 \<0.001
Creatinine (mg/dL) 0.93±0.23 0.99±0.27 1.08±0.34 \<0.001
eGFR (mL/min/1.73m^2^) 81.5±18.5 76.1±19.6 72.4±22.3 \<0.001
UACR (mg/g) 12.6 \[7.0, 24.5\] 23.1 \[10.6, 98.9\] 37.0 \[16.5, 235.5\] \<0.001
Hemoglobin (g/dL) 13.5±1.6 13.0±1.8 12.3±1.8 \<0.001
Total cholesterol (mg/dL) 167.1±35.8 166.9±41.2 168.4±41.9 0.892
Albumin (g/dL) 4.4±0.4 4.3±0.5 4.2±0.6 \<0.001
-------------------------------------------------------------------------------------------------------
DR, diabetic retinopathy; NPDR, non-proliferative diabetic retinopathy; PDR, proliferative diabetic retinopathy; BMI, body mass index; DM, diabetes mellitus; ACEi, angiotensin converting enzyme inhibitor; ARB, angiotensin receptor blocker; sBP, systolic blood pressure; dBP, diastolic blood pressure; HbA1c, hemoglobin A1c; FPG, fasting plasma glucose; eGFR, estimated glomerular filtration rate; UACR, urinary albumin-to-creatinine ratio
Severity of diabetic retinopathy affects the rate of renal function decline {#sec005}
---------------------------------------------------------------------------
To assess whether DR severity affects subsequent renal function decline rate, we compared annual decline rate of eGFR among subjects with different DR severity. Since baseline renal function is a strong predictor of future eGFR decline, we divided subjects into groups based on baseline CKD stages according to GFR categories ([Fig 2](#pone.0220506.g002){ref-type="fig"}). Among the subjects with baseline CKD stage 1 (n = 517), patients with NPDR (n = 102) and PDR (n = 51) showed faster eGFR decline rate (-3.2±6.44 and -4.16±5.43 mL/min/1.73m^2^ per year) compared to those without DR (n = 364, -0.83±3.48 mL/min/1.73m^2^ per year, p\<0.001). The subjects in CKD stage 2 (n = 797) at baseline showed preserved renal function among no DR subjects (0.47±3.22 mL/min/1.73m^2^ per year) while rapid progression in NPDR and PDR (-1.91±5.71 and -3.97±6.08 mL/min/1.73m^2^ per year) patients (p\<0.001). The results were similar for those in CKD stage 3a (n = 188) and 3b (n = 90) showing preserved renal function in no DR subjects (1.24±3.91 and 0.5±4.34 mL/min/1.73m^2^ per year) while rapid progression in NPDR (-0.53±4.25 and -2.3±3.34 mL/min/1.73m^2^ per year) and PDR (-2.58±5.71 and -2.89±9.19 mL/min/1.73m^2^ per year) patients (p\<0.001).
{#pone.0220506.g002}
Diabetic retinopathy is a risk factor for CKD progression {#sec006}
---------------------------------------------------------
To evaluate the effect of DR severity upon CKD progression, 1,592 subjects were further divided into group of CKD progression (n = 311) and non-progression (n = 1,281, [Table 2](#pone.0220506.t002){ref-type="table"}). The subjects with CKD progression showed lower BMI (24.4±3.7 vs. 25.2±3.6 kg/m^2^, p = 0.014), longer duration of DM (11.2±8.2 vs. 8.0±7.1 years, p\<0.001), more frequent use of ACEi or ARB (55% vs. 46.8%, p = 0.009) and higher HbA1c (8.6±2.2 vs. 7.6±1.7%, p\<0.001). They also showed lower baseline eGFR (73.0±22.6 vs. 80.5±18.5 mL/min/1.73m^2^, p\<0.001), greater amount of albuminuria (median 13.5 vs. 36.7 mg/g, p = 0.001), lower plasma hemoglobin (12.5±2.0 vs. 13.4±1.6 g/dL, p\<0.001), and lower serum albumin level (4.1±0.6 vs. 4.5±0.4 g/dL, p\<0.001). The subjects with CKD progression also showed larger proportion of NPDR and PDR (37.0% and 32.2%) compared to those without CKD progression (21.0% and 8.0%, p\<0.001). The proportion of those with CKD progression increased as DR severity increased (9.5% vs. 29.9% vs. 49.5%, p\<0.001, [Fig 3](#pone.0220506.g003){ref-type="fig"}). The group with CKD progression showed higher proportion of DR progression (25.5% vs. 16.2%, p = 0.003).
![Proportion of the subjects with CKD progression according to initial diabetic retinopathy status.\
CKD progression was defined by one or more of the following: 1) decline in GFR category (≥90 \[G1\], 60--89 \[G2\], 45--59 \[G3a\], 30--44 \[G3b\], 15--29 \[G4\], \<15 \[G5\] mL/min/1.73m^2^) accompanied by a 25% or greater drop in eGFR from baseline, 2) sustained decline in eGFR of more than 5 mL/min/1.73m^2^/year. The proportion of the subjects with CKD progression increased as DR severity increased (9.5 vs. 29.9 vs. 49.5%, p\<0.001). GFR, glomerular filtration rate; DR, diabetic retinopathy; NPDR, non-proliferative diabetic retinopathy; PDR, proliferative diabetic retinopathy.](pone.0220506.g003){#pone.0220506.g003}
10.1371/journal.pone.0220506.t002
###### Risk factors associated with CKD progression.
{#pone.0220506.t002g}
------------------------------------------------------------------------------------------
Parameters CKD progression P- value
------------------------------------ -------------------- ---------------------- ---------
Age (years) 57.7±11.0 58.6±12.0 0.204
Male (%) 46.3% 50.8% 0.164
Hypertension (%) 50.3% 46.7% 0.275
BMI (kg/m^2^) 25.2±3.6 24.4±3.7 0.014
DM duration (years) 8.0±7.1 11.2±8.2 \<0.001
ACEi or ARB (%) 599 (46.8) 171 (55.0) 0.009
HbA1c (%) 7.6±1.7 8.6±2.2 \<0.001
Cr at baseline (mg/dL) 0.94±0.23 1.06±0.33 \<0.001
eGFR at baseline (mL/min/1.73m^2^) 80.5±18.5 73.0±22.6 \<0.001
UACR at baseline (mg/g) 13.5 \[7.4, 27.1\] 36.7 \[13.3, 226.3\] 0.001
Hemoglobin (g/dL) 13.4±1.6 12.5±2.0 \<0.001
Total cholesterol (mg/dL) 166.3±35.0 170.9±48.0 0.114
Albumin (g/dL) 4.5±0.4 4.1±0.6 \<0.001
DR at baseline (%)\ \ \ \<0.001
No DR\ 71.0%\ 30.9%\
NPDR\ 21.0%\ 37.0%\
PDR 8.0% 32.2%
DR progression (%) 16.2% 25.5% 0.003
------------------------------------------------------------------------------------------
eGFR, estimated glomerular filtration rate; BMI, body mass index; DM, diabetes mellitus; ACEi, angiotensin converting enzyme inhibitor; ARB, angiotensin receptor blocker; HbA1c, hemoglobin A1c; UACR, urinary albumin-to-creatinine ratio; DR, diabetic retinopathy; NPDR, non-proliferative diabetic retinopathy; PDR, proliferative diabetic retinopathy
Diabetic retinopathy is a risk factor for albuminuria progression {#sec007}
-----------------------------------------------------------------
To evaluate the effect of DR severity upon albuminuria progression, a total of 948 subjects with serial albuminuria data over 1 year were included in the analysis ([Table 3](#pone.0220506.t003){ref-type="table"}). A total of 644 subjects were excluded from the analysis for the following reasons: 546 without initial albuminuria data, 61 with initial macroalbuminuria (UACR≥300mg/g), and 37 without serial albuminuria data. Among 948 subjects, 180 patients showed albuminuria progression while 768 patients showed stable amount of albuminuria. The group with albuminuria progression showed longer duration of DM (9.2±7.4 vs. 7.7±6.9 years, p = 0.008), more frequent use of ACEi or ARB (67.8% vs. 50.9%, p\<0.001), higher HbA1c (8.2±2.2 vs. 7.5±1.7%, p\<0.001) and lower baseline eGFR (76.7±20.1 vs. 82.3±17.5 mL/min/1.73m^2^, p = 0.001) and higher amount of baseline UACR (median 13.6 vs. 18.5 mg/g, p = 0.014). As same as those with CKD progression, the group with albuminuria progression showed larger proportion of NPDR and PDR compared to those without progression (30.0% and 15.0% vs. 18.8% and 4.5%, p\<0.001). The proportion of the subjects with albuminuria progression increased as DR severity increased (14.4% vs. 27.1% vs. 43.5%, p\<0.001, [Fig 4](#pone.0220506.g004){ref-type="fig"}). However, proportion of DR progression during follow-up was not different between two groups (18.8% vs. 13.9%, p = 0.148).
{#pone.0220506.g004}
10.1371/journal.pone.0220506.t003
###### Risk factors associated with albuminuria progression.
{#pone.0220506.t003g}
----------------------------------------------------------------------------------------------
Parameters Albuminuria progression P- value
------------------------------------ ------------------------- --------------------- ---------
Age (years) 56.5±10.6 57.4±12.0 0.328
Male (%) 47.4% 40.0% 0.081
Hypertension (%) 46.9% 52.3% 0.206
BMI (kg/m^2^) 25.1±3.5 25.2±3.6 0.75
DM duration (years) 7.7±6.9 9.2±7.4 0.008
ACEi or ARB (%) 391 (50.9) 122 (67.8) \<0.001
HbA1c (%) 7.5±1.7 8.2±2.2 \<0.001
Cr at baseline (mg/dL) 0.92±0.21 0.98±0.26 0.006
eGFR at baseline (mL/min/1.73m^2^) 82.3±17.5 76.7±20.1 0.001
UACR at baseline (mg/g) 13.6 \[7.4, 27.4\] 18.5 \[10.1, 28.5\] 0.014
Hemoglobin (g/dL) 13.5±1.7 13.3±1.5 0.127
Total cholesterol (mg/dL) 163.8±34.0 164.0±35.2 0.941
Albumin (g/dL) 4.5±0.4 4.4±0.4 0.198
DR at baseline (%)\ \ \ \<0.001
No DR\ 76.6%\ 55.0%\
NPDR\ 18.9%\ 30.0%\
PDR 4.6% 15.0%
DR progression (%) 13.9% 18.8% 0.148
----------------------------------------------------------------------------------------------
BMI, body mass index; DM, diabetes mellitus; ACEi, angiotensin converting enzyme inhibitor; ARB, angiotensin receptor blocker; HbA1c, hemoglobin A1c; eGFR, estimated glomerular filtration rate; UACR, urinary albumin-to-creatinine ratio; DR, diabetic retinopathy; NPDR, non-proliferative diabetic retinopathy; PDR, proliferative diabetic retinopathy
Diabetic retinopathy is a prognostic factor for progression of chronic kidney disease {#sec008}
-------------------------------------------------------------------------------------
To assess whether DR severity is a prognostic factor for CKD progression, we performed binary logistic regression analysis. In the univariate analysis, low BMI, longer DM duration, use of ACEi or ARB at the initial visit, higher HbA1c, lower baseline eGFR, higher baseline UACR, lower plasma hemoglobin, lower serum albumin, presence of DR, and presence of DR progression were all related to the CKD progression ([Table 4](#pone.0220506.t004){ref-type="table"}, p\<0.05).
10.1371/journal.pone.0220506.t004
###### Binary logistic regression analysis for CKD progression.
{#pone.0220506.t004g}
Parameter Odds ratio Confidence interval p-value
-------------------------------- ------------ --------------------- ---------
Age, yr 1.007 0.996--1.019 0.204
Male 1.198 0.935--1.535 0.153
HTN 1.157 0.899--1.489 0.258
BMI, kg/m^2^ 0.943 0.899--0.988 0.014
DM duration, yr 1.055 1.039--1.072 \<0.001
Use of ACEi or ARB 1.391 1.084--1.784 0.009
HbA1c, % 1.308 1.227--1.395 \<0.001
Baseline eGFR, mL/min/1.73m^2^ 0.981 0.974--0.987 \<0.001
Baseline UACR, mg/g 1.002 1.001--1.003 \<0.001
Plasma hemoglobin, g/dL 0.743 0.689--0.801 \<0.001
Total cholesterol, mg/dL 1.003 1.000--1.006 0.057
Serum albumin, g/dL 0.186 0.14--0.247 \<0.001
NPDR (vs. no DR) 4.052 2.993--5.488 \<0.001
PDR (vs. no DR) 9.293 6.569--13.146 \<0.001
DR progression 1.769 1.230--2.545 0.002
CKD, chronic kidney disease; HTN, hypertension; BMI, body mass index; DM, diabetes mellitus; ACEi, angiotensin converting enzyme inhibitor; ARB, angiotensin receptor blocker; HbA1c, hemoglobin A1c; eGFR, estimated glomerular filtration rate; UACR, urinary albumin-to-creatinine ratio; NPDR, non-proliferative diabetic retinopathy; PDR, proliferative diabetic retinopathy; DR, diabetic retinopathy
When we performed multiple logistic analysis after adjusting for DM duration, use of ACEi or ARB, HbA1c, baseline eGFR, baseline UACR, plasma hemoglobin and serum albumin levels, baseline DR severity and DR progression status, DR severity was independently associated with CKD progression showing that NPDR has 2.9 times and PDR has 16.6 times higher risk for CKD progression compared to no DR group ([Table 5](#pone.0220506.t005){ref-type="table"}, p for trend \<0.001). However, DR progression during follow-up period was not an independent risk factor for CKD progression (p = 0.563).
10.1371/journal.pone.0220506.t005
###### Multiple logistic regression analysis for prediction of CKD progression.
{#pone.0220506.t005g}
Parameter Odds ratio Confidence interval p-value
------------------------------------ ------------ --------------------- ---------
Baseline eGFR \<60 mL/min/1.73m^2^ 2.068 1.062--4.027 0.033
Baseline UACR ≥30 mg/g 3.314 1.847--5.947 \<0.001
Albumin \<4.0 g/dL 2.840 1.495--5.394 0.001
NPDR (vs. no DR) 2.910 1.602--5.285 \<0.001
PDR (vs. no DR) 16.582 2.431--113.123 0.004
DR progression 1.213 0.631--2.330 0.563
Hemoglobin \<10 g/dL 1.849 0.619--5.520 0.271
HbA1c ≥7.0% 1.703 0.901--3.218 0.101
BMI \<25 kg/m^2^ 1.642 0.942--2.860 0.08
DM duration ≥10years 0.978 0.537--1.780 0.942
Use of ACEi or ARB 1.121 0.625--2.009 0.702
CKD, chronic kidney disease; HbA1c, hemoglobin A1c; NPDR, non-proliferative diabetic retinopathy; PDR, proliferative diabetic retinopathy; DR, diabetic retinopathy; BMI, body mass index; DM, diabetes mellitus; eGFR, estimated glomerular filtration rate; UACR, urinary albumin-to-creatinine ratio
Discussion {#sec009}
==========
The current study was performed to show the importance of baseline DR severity upon prospective renal function decline and albuminuria progression. Our study showed that DR was associated with longer duration of DM, poorer control of blood sugar, lower eGFR, greater amount of albuminuria, and poorer nutritional indices (serum albumin, BMI, and plasma hemoglobin). Interestingly, baseline DR severity was associated with progressive renal function decline and albuminuria progression. Even after adjusting for other risk factors, NPDR had 2.9 times and PDR had 16.6 times higher risk for CKD progression in type2 DM subjects. On the other hand, DR progression during follow up period did not affect future CKD progression.
It is well known that the association between DR and CKD is strong and the presence of CKD almost always accompany DR \[[@pone.0220506.ref021]\]. However, this association is known to be weaker in type 2 DM patients \[[@pone.0220506.ref005], [@pone.0220506.ref006]\]. A large amount of cross-sectional studies demonstrated that the presence of DR is associated with concurrent renal dysfunction \[[@pone.0220506.ref009], [@pone.0220506.ref012], [@pone.0220506.ref013], [@pone.0220506.ref022]\]. However, only a few studies have evaluated causal relationship between DR and CKD. Trevisan et al. demonstrated that the rate of renal function decline was larger with those who have retinopathy and proteinuria compared to those without retinopathy \[[@pone.0220506.ref023]\]. Another study evaluated the effect of retinopathy upon renal outcome in elderly group and showed that retinopathy even affects faster renal function decline in non-DM subjects \[[@pone.0220506.ref008]\]. There was also the study evaluating the effect of DR upon renal outcome compared to the non-diabetic hypertensive retinopathy \[[@pone.0220506.ref011]\]. However, there was no study to evaluate the renal outcome according to DR severity.
Our study is the first large-scale study to evaluate the effect of DR severity upon renal outcome. Previous association studies only described the association of the 'presence of DR' with either eGFR decline or proteinuria progression. However, our study is the first study to evaluate the effect of 'degree of DR' but not the presence of DR upon CKD progression.
There can be some possible explanation about why DR severity affects renal function deterioration. Firstly, DR and diabetic CKD are both microvascular complication which lead to extravasation and inflammation. Previous study by Matsuyama et al. suggested that pigment epithelium-derived factor, an inhibitor for angiogenesis, is significantly elevated in the type 2 DM patients with DR and CKD, which may indicate microvascular damage \[[@pone.0220506.ref024]\]. Another study by Yang et al. also suggested that retinal damage marker found in urine proteome reflect renal progression in type 2 DM patients \[[@pone.0220506.ref025]\]. However, other recent study by McKay et al. showed that mere retinal microvascular parameters including vascular caliber, tortuosity, and fractal dimension cannot predict renal outcome in type 2 DM patients \[[@pone.0220506.ref026]\]. Moriya et al. suggested that microalbuminuria together with DR result in glomerulosclerosis and renal progression \[[@pone.0220506.ref027]\]. However, our study suggested that DR severity independently affect renal function deterioration after adjustment for albuminuria status.
Meanwhile, DR progression was not an independent risk factor for further progression of CKD. This may be a part due to missing data. Only 1,422 patients had available follow-up retinal exam more than a year apart and another 170 patients with both PDR at the baseline were excluded from the data analysis. In addition, some subjects were excluded from the analysis due to advanced CKD at baseline (n = 108), macro-albuminuria at baseline (n = 61), and no available albuminuria data (n = 546). Therefore, whether DR progression may reflect the renal outcome should be assessed in the future in a large prospective cohort study.
This study has several limitations. The data in this study was collected retrospectively from a single center. Secondly, we did not collect 24-hour urine samples to quantify the amount of albuminuria. We used albumin-to-creatinine ratio in random urine to quantify the amount of albuminuria. In addition, there were some missing data regarding serial follow up of albuminuria. Thirdly, we neither perform kidney biopsy to see the severity of diabetic nephropathy nor unravel the possible mechanism underlying relationship between DR and CKD. Lastly, this study design limits inferences on causality. These subjects were from the Korean population and were visited in the hospital and the results cannot be generalized to general population and other ethnic groups. Therefore, prospective, cohort-based study should be warranted to reproduce and confirm our study finding.
In conclusion, DR was a prognostic factor for CKD progression in type2 DM patients. Therefore, clinicians must evaluate DR severity at the first visit and closely monitor renal function and albuminuria in the subjects with severe DR.
[^1]: **Competing Interests:**The authors have declared that no competing interests exist.
| {
"pile_set_name": "PubMed Central"
} |
Recent Controversies: Carcinogenic Cosmetics and Personal Care Products
=======================================================================
There have been numerous controversies surrounding cosmetics and their chemical constituents with carcinogenesis. Examples include associations of parabens and aluminum with breast cancer ([@pky012-B1],[@pky012-B2]), talc powder with ovarian cancer ([@pky012-B5]), and, most recently, hair dye with breast cancer ([@pky012-B11]). These articles have often been covered extensively by the media and have stoked consumer concerns. Several of these initial epidemiological associations have not been reproduced in larger cohort studies, as in the case of talc powder and ovarian cancer ([@pky012-B12],[@pky012-B13]), or follow-up toxicology analyses by other scientists or regulatory agencies ([@pky012-B14]). Given the ubiquitous use of cosmetics, oncologists often face questions from patients concerning the possible etiology of these products in relation to a cancer diagnosis. As one indication, the American Cancer Society publishes numerous "frequently asked questions" surrounding cosmetics and carcinogenesis for the public, suggesting the inherent need for such clarification to the lay public ([@pky012-B18]).
The Food and Drug Administration's Center for Food Safety and Applied Nutrition's Adverse Event Reporting System
================================================================================================================
The US Food and Drug Administration (FDA) made the Center for Food Safety and Applied Nutrition's Adverse Event Reporting System (CAERS) publicly available in late 2016 to increase transparency and encourage adverse event reporting from consumers related to cosmetics. A recent update of the database in 2017 includes adverse event reported to the FDA related to cosmetics submitted by consumers and health care providers from January 2004 to March 2017. A previous study of an earlier version of this database identified hair products, skin care products, and tattoos (nonpermanent) as the cosmetic products most often associated with adverse events ([@pky012-B19]). However, there was no further analysis of specific symptom complaints.
All CAERS reports for cosmetics from 2004 to March 2017 were extracted on May 30, 2017, and examined for duplicates or incomplete entries. Entries were then sorted based on the specific self-reported negative health effects. Products whose effects did not include cancer or cancer-associated search terms were excluded. Search terms included "cancer," "neoplasm," "leukemia," "mass," "adenoma," "lesion," "metaplasia," "carcinoma," "malignant," and "metastatic." Products associated with cancer were then categorized into one of five broader categories: talc powders, hair products, moisturizers, cleansers, tanning products, or miscellaneous. The miscellaneous category includes nail polish, oral products, deodorants, and makeup products. These were organized by number of total reports and individual cancer types. The majority of cancer-related reports for cosmetic products were "redacted" in the database under the Freedom of Information Act exemptions. These exemptions include records that are being compiled for law enforcement purposes, attorney work-products being used in preparation for litigation, or information that would interfere with pending law enforcement matters ([@pky012-B20]). These redacted reports were included in overall cancer reports but removed for product class-specific analyses. Available demographic data from reporters were collected. Our goal was to determine whether useful insights could be derived from this national database for cancer epidemiology and cosmetics.
CAERS Reports Associating Cosmetics With Cancer From 2004 to 2017
=================================================================
A total of 4427 cancer-related adverse events were reported to the FDA database associated with cosmetics ([Figure 1](#pky012-F1){ref-type="fig"}). This represents 41% of all cosmetic adverse events (10 726 total). The majority of cancer reports were associated with products whose names were redacted in CAERS (n = 4210, 95.1% of all cancer reports). Overall, the available demographic data of reporters were limited. The average age of all respondents with any adverse event was 48 years (range \< 1--95 years); 84% of respondents were female, 11% were male, and 5% did not indicate sex. Of all cases, including redacted reports, ovarian cancer reports dominated (n = 3992, 90%), followed by mesothelioma (n = 92, 2%) and then malignant neoplasm unspecified (n = 46, 1%). Of the nonredacted reports (n = 218) where product classes were available ([Figure 2](#pky012-F2){ref-type="fig"}), talc powders were most associated with a report of cancer, composing 70% of all cancer reports (n = 153). Within the cancers associated with talc powders, ovarian cancer composed the majority, with 141 cases (92%). Only 15 cases reported a specific ovarian cancer subtype (granulosa, serous cystadenocarcinoma, epithelial, and clear cell). The next most commonly reported cancers for talc powder included malignant mesothelioma (2.1%), malignant neoplasm (1.1%), skin cancer (1%), and uterine cancer (0.01%). Topical moisturizers consisted of 10.5% of nonredacted cancer reports. Specifically, moisturizers were attributed by reporters to skin cancer (n = 14, 11%), breast cancer (n = 7, 5%), and bone lesions (n = 2). Hair products were the third most common product class, composing 8.3% of total cancer cases and associated with nine different types of cancer. In total, 33 individual cancers were reported in the CAERS database. Ten of these cancers were associated with only one reported case and 15% of all the reported cancer-associated outcomes were nonspecific. These included reports of "neoplasm malignant," "mass," or "metastatic malignancy."
{#pky012-F1}
{#pky012-F2}
Limitations of the CAERS Database
=================================
Reports to the FDA can be done through physical mail, fax, or online with specific forms ([@pky012-B21]). Respondents must first choose to fill out the form as a consumer or health care professional. Then, each form first asks about the actual reaction (in which reporters can freely type out the adverse reaction in paragraph form), patient information (currently optional to list comorbidities and current medications), product information, and, within the health care professionals form only, concomitant product use. While some association between cancer and cosmetics are observed with these reports, there were several challenges with data extraction and interpretation due to the inherent limitations of the database, as noted previously ([@pky012-B19],[@pky012-B22]). First, the inclusion of nonspecific categories such as "mass" or "neoplasm malignant" makes these reports largely uninterpretable. Lack of further cancer subtyping also limits interpretation as different subtypes can have widely variant pathogenesis. For example, knowledge of the subtype of ovarian cancer (eg, serous, germ cell) is necessary to determine causal relationships. Furthermore, the lack of concomitant medical problems of respondents also severely limits any conclusions drawn for this database. For example, a consumer reporting skin cancer related to a cosmetic may also have had significant tanning bed use. Finally, the duration and frequency of a specific cosmetic and any concomitant cosmetic product use must be included as well. Without a record of these comorbidities, it is impossible to draw conclusions of causality from this database.
Challenges Linking Increased Cancer Risk to Cosmetics and Personal Care Products
================================================================================
Broadly, one of the biggest challenges facing the safety and regulation of cosmetic products is the limitations on the FDA to take action against cosmetic products of potential public health risk. The Food, Drug and Cosmetics Act (FDCA) passed in 1938 allows the FDA to remove harmful food products and regulate which drugs may be sold on the market; however, regulation of cosmetic products is limited to premarket testing of color additives ([@pky012-B23]). The FDCA does not require companies to report safety information on any other product components and only allows the FDA to recall products that are "adulterated" or "misbranded." Similarly, the Consumer Product Safety Act excludes cosmetic products. This lack of regulatory power significantly inhibits the ability to assess chemicals prior to their inclusion or develop appropriate metrics for hazardous ingredients.
The assessment of the carcinogenicity of chemicals and chemical mixtures currently relies on a combination of the use of laboratory studies using animal models and human epidemiological studies ([@pky012-B24]). Often, carcinogenicity is first established or highly suspected when positive laboratory results are obtained in one or more animal species ([@pky012-B25]). However, species-specific mechanisms that are not applicable in humans must be considered, and experimental parameters including the route of exposure, species, strain, sex, age, and duration of exposure must be taken into account when attempting to extrapolate findings from animals to humans ([@pky012-B24]). For instance, several studies have shown endocrine disruption in rats exposed to parabens ([@pky012-B26],[@pky012-B27]). However, while endocrine disruption is thought to be the mechanism behind parabens that cause breast cancer, several studies in humans have found no association ([@pky012-B15]).
In epidemiological studies, biomarkers have increasingly been employed in human studies to investigate links between exposures to exogenous chemicals and cancer risk ([@pky012-B28]). Because biomarkers capture molecular signatures of exposure, early effects, and susceptibility across the entire exposure-disease continuum, they are critical for assessing causality in cancer research ([@pky012-B29]). Taken together with data obtained from laboratory animal studies, results from molecular epidemiology can help to build multilayer evidence for causal claims.
However, given the vast numbers of untested chemicals that are currently in commerce in the United States, it is clearly not feasible to conduct costly and time-consuming laboratory animal studies and/or epidemiology studies on all chemicals ([@pky012-B30]). As a result, cost-effective and high-throughput screening methods are critically needed that can help pinpoint suspected carcinogens for subsequent follow-up evaluation. Major federal research initiatives that are currently working to address this need include the US Environmental Protection Agency's ToxCast Program ([@pky012-B31]), the National Toxicology Program's High Throughput Screening Initiative ([@pky012-B32]), and the interagency Tox21 Initiatives ([@pky012-B33]), which are focused on developing rapid screening methods for testing large numbers of chemicals for toxicity. Carcinogenicity testing has also recently begun to incorporate more efficient in vivo and in vitro mechanism-based screening that focuses on early biological indicators of toxicity, as opposed to targeting cancer end points ([@pky012-B34]). This understanding could ultimately increase the capacity of high-throughput carcinogenicity testing to help identify candidate carcinogens for further evaluation. Fundamentally, it is challenging to prove causality for an individual chemical applied topically over many years with a specific cancer within the context of exposures to hundreds of other confounding chemicals and the complex multifactorial etiology of many cancers.
Media Attention and Cosmetics
=============================
Recently, talc baby powder has received substantial media attention. Within the CAERS database, most ovarian cancer reports happened after 2015, coinciding with a peak of hundreds of class-action lawsuits filed against Johnson and Johnson for their talcum products ([@pky012-B35]). This peak was sustained in 2016, when Johnson and Johnson paid hundreds of millions in damages and settlements related to talc baby powders. Such outside influences likely prompt recall bias and obscure the causal relationships observed within this database. The media itself is also subject to reporter bias in regard to scientific studies relating to cosmetics and carcinogenesis. Highlighting positive associations makes for better headlines. A survey of 937 members of the Society of Toxicology showed that 80% of respondents believed that popular media overstate products risks ([@pky012-B36]).
Cancer Associations Identified in CAERS: Support for or Lack of Collateral Biomedical Evidence
==============================================================================================
The concern for talc products initially arose because of its chemical similarity to asbestos, a chemical shown to cause ovarian cancer in occupational settings ([@pky012-B37]). Hypotheses on the mechanism behind talc powder's carcinogenicity include the impact of estrogen and/or prolactin on macrophages and the inflammatory response to talc ([@pky012-B41]). Macrophages and monocytes have previously been shown to play a role in scavenging talc ([@pky012-B42]). Experiments in pregnant mice have shown estrogen-mediated impaired macrophage response to titanium dioxide exposure, a molecular similar in structure to talc ([@pky012-B43]). There have also been several studies showing migration of talc through the vagina to the ovaries ([@pky012-B5]). However, at least one of these studies was later disproven due to contamination from talc-containing surgical gloves ([@pky012-B44]), and several more recent occupational studies have favored the migration of inhaled talc particles from the lung to the ovary ([@pky012-B45]).
Although early studies showed an association between epithelial ovarian cancer and life-long talc powder use, the majority of these studies were case--control studies limited by small sample size, nonsignificant odds ratios, and recall bias ([@pky012-B5],[@pky012-B9],[@pky012-B10]). Another study showed that only 14% of women with ovarian cancer had any talc exposure ([@pky012-B8]). Subsequent investigation has included two prospective trials showing no increased risk between talc use and ovarian cancer ([@pky012-B12],[@pky012-B13]). Recent pooled analyses showed a weak but statistically significant association between talc and serous ovarian cancer, but no association with duration or frequency of use ([@pky012-B46],[@pky012-B47]). Of note, the meta-analysis was limited by notable variation in study designs.
While skin moisturizers made up only 10.5% of total cancer cases, they were the product class with the highest number of skin cancer reports (n = 14). Little available biomedical evidence exists in the literature for such an association. A single study in mice demonstrated a higher rate of skin cancer development when topical moisturizers were applied to the skin prior to Ultra-violet B irradiation. The mechanism for such an effect is unclear but has been postulated to involve moisturizer-mediated inflammation and proliferation of DNA-damaged skin ([@pky012-B48]). However, studies have shown that the application of moisturizers prior to radiation inhibited carcinogenesis ([@pky012-B49]). Few studies or associations have been observed in humans. Breast cancer was the second most common cancer reported with moisturizers. This association may possibly be explained by studies linking breast cancer with parabens (a common skin product preservative), which have been covered extensively by the media ([@pky012-B2],[@pky012-B3]).
Various circumstances of exposure to hair dyes have been associated in some studies with increased risk of leukemia, non-Hodgkin lymphomas, and bladder cancer ([@pky012-B50]). Concerning the association between risk of cancer from both occupational and, typically lesser, consumer exposures, the International Agency for Research on Cancer's (IARC's) comprehensive review of the literature found "limited evidence" for hair dye carcinogenicity in occupational exposure and "inadequate evidence" for carcinogenicity following personal use ([@pky012-B53]). There are data indicating that long-term occupational exposure of hairdressers to the aryl amines contained in hair dyes was associated with later development of bladder cancer ([@pky012-B50]). Since the IARC evaluations were made, additional studies have been published, including, for example, data concerning occupational ([@pky012-B54]) and personal exposure ([@pky012-B55]). One other study has indicated that personal use of darker hair dyes and relaxants was linked to estrogen receptor--positive (ER+) and ER- breast cancer ([@pky012-B11]). Some studies have demonstrated mutagenicity of hair dye chemicals, particularly arylamine p-phenylenediamine (PPD) in conjunction with hydrogen peroxide, in animal models through the formation of reactive oxygen species ([@pky012-B56],[@pky012-B57]). Moreover, PPD acetylation in the skin is thought to play a role in carcinogenesis over time. In the case of bladder cancer, arylamines are again implicated through activation of the cytochrome p450 system that leads to DNA-binding metabolites ([@pky012-B58]). Despite these data, several meta-analyses have been equivocal, with some showing a positive association with bladder and hematopoietic cancers but not breast cancer ([@pky012-B54],[@pky012-B55],[@pky012-B59]).
Hair products outside of hair dyes have also been implicated in increased cancer risk. In 2011, the FDA issued warning letters to the manufacturers of certain heated hair treatments, Brazilian blow-out, due to the release of formaldehyde with heat. Formaldehyde is a known carcinogen associated with lung and hematologic malignancies ([@pky012-B63]). Of the nine unique cancers associated with hair products in the CAERS database, only Chronic lymphocytic leukemia, lung cancer, and breast cancer have any supporting reports in the literature. Other cancers reported in the database included thyroid and skin lesions, for which little outside evidence exists.
Policy Implications for Cosmetics Safety and Cancer Risk
========================================================
The CAERS database has the potential to become a useful cancer epidemiological tool. As it stands, there are significant limitations. The primary need is for broader participation from all stakeholders including physicians and manufacturers. Currently, manufacturers are not required by law to forward adverse events related to cosmetics. As an example, a manufacturer of a hair product had received more than 21 000 adverse event reports directly from consumers that were not forwarded to the FDA. At the time, the FDA had only received 127 reports ([@pky012-B67]).
As previously mentioned, the data included within the CAERS database contain considerable gaps, particularly in regard to patient demographics and cancer specifics. There are several potential improvements. First, there is a need for more specific cancer subtyping whenever possible. Second, report data would be of greater benefit if reporters were compelled to provide 1) family history, 2) comorbid conditions, and 3) relevant personal behavioral characteristics (eg, smoking history, drug use, alcohol use, tanning bed use). Finally, to potentially control for reporter bias, consumers can be asked to note whether their reports were triggered by a media report.
The above improvements would enable CAERS to be a better surveillance tool. Emerging concerns could be identified earlier, which could facilitate follow-up scientific studies, site visits, and cross-referencing with carcinogens listed by other agencies such as the Environmental Protection Agency, National Toxicology Program, and the International Agency for Research on Cancer ([@pky012-B68]). If a potential risk is identified through CAERS, the FDA and other regulatory bodies could better focus scarce resources on further investigation. Consideration should then be given by manufacturers to add warning labels, change formulations, or alter the recommended frequency of use of cosmetic products when there is demonstration of potential risk. Such warning labels have been suggested for talc powder ([@pky012-B69]). However, the FDA's current position is that there is insufficient evidence to warrant this.
Ultimately, CAERS and other databases alone will not be enough to ensure consumer safety. Rather, they should be viewed as an initial step in part of a larger need for greater FDA authority over the cosmetics industry ([@pky012-B19]). This includes required mandatory manufacturer registration with the FDA and the need for greater funding for regulatory activities. In 2017, the Office of Cosmetics and Colors, the division of the FDA charged with enforcing labeling of cosmetic products, operated with an annual budget of only \$13 million to regulate the \$62 billion US cosmetics industry ([@pky012-B22]).
Conclusions
===========
With the ongoing media attention surrounding cosmetics and carcinogenesis, oncologists will continue to face questions from concerned patients. Nonspecific cancer subtyping, lack of comorbid medical condition information, and the risk of reporter bias all limit the current FDA database for cosmetics. Better and broader data collection is necessary if the CAERS database is to become a useful cancer epidemiological tool that can highlight emerging concerns and direct scarce regulatory resources to promote public safety and allay consumer fears. Concomitant investments in toxicology, biomarker discovery, and regulatory science are needed.
Funding
=======
Dr. Xu recognizes support from the Foglia Family Foundation and National Institutes of Health grant T32AR060710.
Notes
=====
Affiliations of authors: Division of Cancer Epidemiology and Prevention, Department of Preventive Medicine (WEF), and Department of Dermatology (SX), Northwestern University Feinberg School of Medicine (SLJ, EC, MK), Chicago, IL.
The authors have no conflicts of interest to declare.
[^1]: See the Notes section for the full list of authors' affiliations.
| {
"pile_set_name": "PubMed Central"
} |
Few attempts have been made to model learning of sensory-motor control using spiking neural units. We trained a 2-degree-of-freedom virtual arm to reach for a target using a spiking-neuron model of motor cortex that maps proprioceptive representations of limb position to motor commands and undergoes learning based on reinforcement mechanisms suggested by the dopaminergic reward system. A 2-layer model of layer 5 motor cortex (M1) passed motor commands to the virtual arm and received proprioceptive position information from it. The reinforcement algorithm trained synapses of M1 using reward (punishment) signals based on visual perception of decreasing (increasing) distance of the virtual hand from the target. Output M1 units were partially driven by noise, creating stochastic movements that were shaped to achieve desired outcomes.
The virtual arm consisted of a shoulder joint, upper arm, elbow joint, and forearm. The upper- and forearm were each controlled by a pair of flexor/extensor muscles. These muscles received rotational commands from 192 output cells of the M1 model, while the M1 model received input from muscle-specific groups of sensory cells, each of which were tuned to fire over a range of muscle lengths. The M1 model had 384 excitatory and 192 inhibitory event-based integrate-and-fire neurons, with AMPA/NMDA and GABA synapses. Excitatory and inhibitory units were interconnected probabilistically. Plasticity was enabled in the feedforward connections between input and output excitatory units. Poisson noise was added to the output units for driving stochastic movements. The reinforcement learning (RL) algorithm used eligibility traces for synaptic credit/blame assignment, and a global signal (+1=reward, -1=punishment) corresponding to dopaminergic bursting/dipping. Eligibility traces were spike-timing-dependent, with pre-before-post spiking required. Reward (punishment) was delivered when the distance between the hand and target decreased (increased) \[[@B1]\].
RL learning occurred over 100 training sessions with the arm starting at 15 different initial positions. Each sub-session consisted of 15 s of RL training from a specific starting position. After training, the network was tested for its ability to reach the arm to target from each starting position, over the course of a 15 s trial. Compared to the naive network, the network post-training was able to reach the target from all starting positions. This was most clearly pronounced when the arm started at a large distance from the target. After reaching the target, the hand tended to oscillate around the target. Learning was most effective when recurrent connectivity in the output units was turned off or at low levels. Best overall performance was achieved with no recurrent connectivity and moderate maximal weights. Although learning typically increased average synaptic weight gains in the input-to-output M1 connections, there were frequent reductions in weights as well. Our model predicts that optimal motor performance is sensitive to perturbations in both strength and density of recurrent connectivity within motor cortex and that therefore the wiring of recurrent connectivity during development might be carefully regulated.
Acknowledgements
================
research supported by DARPA grant N66001-10-C-2008.
The authors would like to thank Larry Eberle (SUNY Downstate) for Neurosim lab computer support; Michael Hines (Yale) and Ted Carnevale (Yale) for NEURON simulator support.
| {
"pile_set_name": "PubMed Central"
} |
Hemozoin \[heʺmo-zoʹin\]
========================
From the Greek *haima* ("blood") + *zoon* ("animal"), hemozoin ([Figure](#F1){ref-type="fig"}) is a pigment produced by malaria parasites from hemoglobin in the host's red blood cells. This pigment was first observed by Johann Heinrich Meckel in 1847 in the blood and spleen of a mentally impaired person. In 1849, Rudolf Virchow made the connection to malaria, but it was initially believed that it was produced in the patient's spleen as a part of the immune response to malaria. In 1880, Charles Louis Alphonse Laveran observed pigmented parasites in the blood of an Algerian soldier and realized that the parasites, not the patient, produce "malaria pigment." The term "hemozoin" was coined by Louis Westenra Sambon.
{#F1}
*Suggested citation for this article*: Etymologia: hemozoin. Emerg Infect Dis. 2016 Feb \[*date cited*\]. <http://dx.doi.org/10.3201/eid2202.ET2202>
| {
"pile_set_name": "PubMed Central"
} |
Background {#s1}
==========
The Star-fruit plant (family: Oxalidaceae; species: Averrhoa carambola L.) is widely distributed around the world, especially in tropical countries such as India, Malaysia, Indonesia, and Philippines. This Star-fruit plant belongs to the genus, Averrhoa, which contains 5 species, namely A. bilimbi, A. dolichocarpa, A. leucopetala, A. microphylla and A. carambola. However, A. carambola is widely cultivated on a commercial scale \[[@R01]\]. Averrhoa carambola is considered the most important species and cultivated extensively in South-east Asia and Malaysia \[[@R02],[@R03]\]. In addition, it is a popular fruit in the United States, Australia and South Pacific Islands market \[[@R04]\] . Star-fruits are fleshly, crunchy, juicy and slightly tart, acidic and sweet in the taste. This fruit is known to have high antioxidant property that efficiently scavenge free radicals as well as helps in hypoglycemic and hypocholesterolemia treatments \[[@R05], [@R06],[@R07]\]. Star-fruits are also commonly used in preparation of juice, pickles and salads. However, it can be eaten raw and used for cleaning utensils; because, it helps in removing the rust caused by iron oxidation. Star-fruits are well known for the oxalic acid content in it which gives an adverse effect when consumed by uremic patients \[[@R08], [@R09],[@R10]\]. The aim of this review article is to highlight the nutritional, medicinal and toxicological attributes of the Star-fruit.
Methodology {#s2}
===========
Botanical description {#s2a}
---------------------
Averrhoa carambola is a slow growing species of woody plants; it is multi-stemmed with short trunk and best grows up to 6 to 9 m in height. It has a bushy appearance with many branches producing a broad, rounded crown and a trunk base which can reach up to 15cm in diameter \[[@R03],[@R11]\]. In addition, the tree has leaflets that fold together at nightfall and shows sensitivity to light and shock, such as abrupt movements of the leaves. Star-fruits plant produces small clusters of red, lilac or purple flowers containing five petals. Fruits are usually small and dark green in colour when unripe; but, fruits turn yellow in colour when they are fully ripe. Usually, star-fruits are fleshy with 5 longitudinal ridges or angles ([Figure 1A](#F1){ref-type="fig"}), and are crunchy crisp in texture. Furthermore, the fruits are star-shaped when cut horizontally ([Figure 1B](#F1){ref-type="fig"}); hence, the fruit is called as Star-fruit \[[@R12],[@R13]\]. The firmness and colour of the star-fruits changes with its development as shown in [Table 1](#T1){ref-type="table"}. The taxonomical classification of Star-fruit plant and common names of star-fruits are given in [Table 2](#T2){ref-type="table"}and [3](#T3){ref-type="table"}, respectively.
Star-fruit plant varieties {#s2b}
--------------------------
In Malaysia, A. carambola is a commercial cultivar, and its fruits are widely marketed in all the states and exported mainly to the Europe. In Malaysia, four states namely, Selangor, Negeri Sembilan, Pahang, and Johor are cultivating Star-fruits \[[@R14]\]. Nineteen (19) varieties of star-fruit are registered under the Department of Agriculture, Malaysia. However, out of these 19 varieties, only two varieties are popular as the best commercial clones, namely 'Belimbing Besi' (B10) and 'Belimbing Madu' (B17) \[[@R15],[@R16]\]. Besides Malaysia, the United States (USA) also cultivates Star-fruit plants for its fruits \[[@R03]\]. Taiwan has its own collection of Star-fruit plant accessions, such as 'Mih Tao', 'Dah Pon', 'Tean Ma' and in Thailand, 'Fwang Tung'\[[@R03]\] . Registered and widely accepted superior clones (varieties) of the Star-fruit plants are depicted in the [Table 4](#T4){ref-type="table"}.
Nutritional attributes of star-fruits {#s2c}
-------------------------------------
The star-fruit is a good source of various minerals and vitamins ([Table 5](#T5){ref-type="table"}). Star-fruits are also a rich source of natural antioxidants such as L-ascorbic acid (Vitamin C) and Gallic acid, which aid in scavenging reactive oxidative species (ROS) \[[@R17]\]. The literature shows that Star-fruits are a good source of magnesium, potassium, phosphorous, as well as β-carotene, and vitamin C, which are common antioxidants \[[@R18]\]. The presence of antioxidants like iron, zinc and manganese in the fruits aid in strengthening the immune system \[[@R19]\]. In addition, the presence of high amounts of fibres in fruits aids in absorbing glucose and retarding the glucose diffusion into the blood stream; as a result, it helps in controlling blood glucose concentration \[[@R20]\]. The Starfruit intake also exhibits hypo-cholesterolemic and hypolipidaemic effect as it enhances the removal of cholesterol, lipid, and bile acid through the excrement \[[@R21]\].
Medicinal properties {#s2d}
--------------------
Nowadays, herbal medications are becoming popular worldwide as an alternative therapy to drug medication. In addition to food source, Star-fruits are also considered as herbs in many parts of Brazil, China, India, and Malaysia as well as in Taiwan, and widely used in Ayurvedic and traditional Chinese Medicine \[[@R22], [@R23]\] preparations as remedy for fever, sore throat, cough, asthma, chronic headache, and skin inflammations. The phytochemical and pharmacological studies suggest that the extracts of Star-fruit plant leaves, fruits and roots contains saponins, flavonoids, alkaloids and tannins \[[@R24], [@R25]\] which are known to confer antioxidant and specific healing properties \[[@R26]\]. The major vitamins and acids found in star-fruits are highlighted in [Table 6](#T6){ref-type="table"}.
Antioxidant property {#s2e}
--------------------
Studies reported that Star-fruits contain proanthocyanins which serves as an antioxidant besides Vitamin C and Gallic acid \[[@R07]\] . The main purpose of antioxidants is to scavenge ROS such as peroxides. Usually, fatty acids are susceptible to oxidative damage by peroxides and hyperperoxides \[[@R27]\]. Consumption of Star-fruits is helpful in removing toxins from the body and aids the immune system in guarding against cancer, ROS damage and lipoperoxidation \[[@R28]\].
A source of water insoluble fibres {#s2f}
----------------------------------
Usually, when consuming Star-fruit juice, often the fibre's residual parts of fruits are excluded from the beverage. In spite of this, Star-fruit contains approximately 60% of cellulose, 27% of hemicelluloses and 13% of pectin \[[@R29]\]. It indicates that star-fruit is indeed rich in insoluble fibres fractions (IFF). The insoluble fibres have the ability to retain water more than cellulose; thus called as 'water insoluble fibre fractions' or WIFF. WIFFs actually aids in smooth movement of the bowels and has the capability of lowering blood glucose by slowing down the absorption of carbohydrate in our body \[[@R29],[@R30]\]. In addition, the fibres also facilitate in lowering the total cholesterol level in the body by promoting hypoglycaemic effect. Consuming the fruit-juice together with the fibres (called as smoothie) does help in removing lipids through the excrement, and thus lowering the risks of cardiovascular diseases. It has also been reported that Star-fruit extracts do have selective anti-brain-tumour activity \[[@R31],[@R32]\].
Anti--inflammatory and anti--microbial property {#s2g}
-----------------------------------------------
Research findings of Cabrini et al \[[@R17], [@R33]\] indicate that antiinflammatory activity of Star-fruit extracts help in lowering the skin inflammatory condition. Researchers induced a skin inflammatory condition akin to eczema using croton-oil on a mice model. When ethanolic extracts of Star-fruit plant leaves were applied on the skin, it resulted in reduced inflammation and gradually reduced eczema in the mice \[[@R08]\]. In addition to this, the extracts in various concentrations were found to inhibit the growth of Staphylococcus aureus (MBC of 15.62mg/ml) and Klebsiella spp. (MBC of 125mg/ml) \[[@R09]\]. Extracts were also effective against Escherichia coli, Pseudomonas aeruginosa and Bacillus cereus \[[@R08],[@R10],[@R34]\].
Anti-ulcer property {#s2h}
-------------------
Traditionally, star-fruits are used to relieve stomach discomfort or any ulcer-like disorders. The research findings of Cabrini et al \[[@R17], [@R33]\] demonstrated that extracts of Star-fruit plant leaves have anti-ulcerogenic properties. The extracts contain terpenoids (diterpenes and triterpenes), flavonoids and mucilage, which are known to have the anti-ulcer activity. The mucilage provides a lining to the gastro-intestinal mucosa, thus helping to avoid damages due to gastritis \[[@R35]\].
Toxicological effects {#s2i}
---------------------
Star-fruits do possess many magnificent properties. However, this fruit also poses threat to health as it exudes toxic effects in high uremic patients or patients with chronic renal disease due to its high oxalate content \[[@R36],[@R37]\]. Patients with renal disease are unable to secrete toxic substances out of their body efficiently; as a result of it, they are affected adversely by the oxalates \[[@R38]\]. The first toxicological effect was demonstrated on mice model by Muir and Lam \[[@R39]\]. Variable dosages of the fruit extracts were prepared and injected into the mice through intra-peritoneal injection, and fruit extracts exceeding 8g/kg provoked convulsions and death in the mice \[[@R35]\]. Further analysis of the test reports showed that Star-fruit juice with oxalate content was responsible for the death of rats. Chronic renal failure patients had high mortality rate after consuming the Star-fruits \[[@R40],[@R41]\]. It was noted that these patients had symptoms of hiccups, mental confusions or disturbance in consciousness and vomiting before succumbing to death. Reports also suggest that uraemic patients experienced nephrotoxic and neurotoxic effects when they consumed Star-fruit \[[@R42]\]. Most of the patients were able to recover after immediate hemodialysis, which spanned for weeks but some experienced total renal failure, causing death. Even though star-fruits have many documented nutritional and medicinal benefits; but, due to the oxalate and caramboxin content in the fruits, it is considered toxic to patients experiencing renal problems.
Perspective {#s2j}
-----------
Indisputably, fruits are very important in our daily diet for various health benefits. However, some fruits may contain high amounts of unique secondary metabolites, which are hazardous to our health. Star-fruit plants are cultivated commercially in tropical countries for their fruits. This fruit have several medicinal properties; hence, it is used medicinally for many years in Ayurvedic treatments. Star-fruits contain various antioxidants which are considered medicinally important and beneficial for the health. However, the negative part of this fruit is that it produces oxalic acid and caramboxin, which are toxic to uremic patients. It can cause death if consumed in sufficient quantities by those experiencing renal failure. Thus, more public awareness about oxalate poisoning on uremic patients should be promoted. It will help to avoid adverse side effect of star-fruits in high uremic patients. It is very important that the public is well educated on the benefits as well as the hazardous effects of the star-fruits.
Modern techniques and molecular biology knowledge should be utilized to understand gene expression in star-fruits. A systematic 'transcriptomics' study of the star-fruit will help us to elucidate the genes expressed in it, and the genes involved in biosynthesis of oxalic acid. Furthermore, transcriptomics study will be helpful to discover the novel genes expressed in the star-fruits. If we identify and characterize the key genes involved in oxalic acid and CBX biosynthesis, then the genetic alteration of the Star-fruit genome can be considered as one of the strategies to knock-out the oxalic acid and CBX synthesis in fruit specific manner. Various post-transcriptional gene silencing (PTGS) techniques such as antisense, inverted-repeats and or intron-spliced inverted repeats \[[@R43]\] can be utilized in genetic engineering of the Star-fruit plant to knock-out the biosynthesis of oxalic acid and CBX. If we develop genetically engineered Star-fruit plants which will produce fruits devoid of oxalic acid and CBX then it will certainly help to increase not only its economic value but nutritional value also.
Conclusion {#s3}
==========
Star-fruit is a good source of nutritionally and medicinally important natural products beneficial for human health. However, due to the oxalate and caramboxin content in the fruits, it is toxic to patients with renal problems. Transcriptomics study of the Star-fruit will help us to understand genes expressed in it and the discovery of novel genes could help in designing a novel strategy for genetic engineering of this plant to knock-out the oxalate and caramboxin biosynthesis in fruit-specific manner to enhance its nutritional quality.
Disclosure {#s4}
==========
Authors attest that there are no conflicts of interest to declare.
Authors are grateful to the Ministry of Education Malaysia of Malaysian Government for research funding \[FRGS/1/2013/ST03/AIMST/02/1\].
**Edited by P Kangueane**
**Citation:**Muthu *et al.* Bioinformation 12(12): 420-424 (2016)
CBX
: caramboxin
MBC
: minimum bactericidal concentration
PTGS
: Post-Transcriptional gene silencing
TCM
: Traditional Chinese Medicines
###### The developmental stages, firmness and colour of Averrhoa carambola L. fruits
Stage Firmness Colour
----------- -------------- --------------------
Young Firm Texture 100% green colour
Half-Ripe Firm Texture Yellowish green
Ripe Soft Texture 100% Yellow colour
###### Taxonomical classification of Star-fruit plant
Taxonomy Classification
--------------- ----------------
Division: Spermatophyta
Sub-Division: Angiospermae
Class: Dicotyledonae
Sub-Class: Polypetalae
Series: Disciflorae
Order: Geraniales
Family: Oxalidaceae
Genus: Averrhoa
Species: carambola
###### Common names of Star-fruit (Averrhoa carambola L.) in different dialect
Language Name
------------ --------------------------------
Bengali Kamranga
English Star-fruit, Chinese Gooseberry
Haiti Zibline
Hindi Kamrakh
Indonesian Blimbing
Malay Belimbing
Tamil Thambaratham
Sanskrit Kamaranga
Venezuela Tamarindo
###### Registered varieties of Star-fruits plant, Averrhoa carambola L.
No Variety Registry No Variety name Year Registered Country
--------------------------------------------------------------------------------------------------------------------- --------------------- ------------------------------ ----------------- ----------
1 B1 Yong Toh Yin 1937 Malaysia
2 B2 MAHA 66 1966 Malaysia
3 B3 Foo Red 1966 Malaysia
4 B4 Sg.Besi 1 1966 Malaysia
5 B5 Sg. Besi 2 1966 Malaysia
6 B6 Sg. Besi 3 1966 Malaysia
7 B7 Sg. Besi 4 1966 Malaysia
8 B8 Sg. Besi 5 1968 Malaysia
9 B9 NA 1968 Malaysia
10 B10 Belimbing Besi \* 1968 Malaysia
11 B11 Chan Yong 1 1968 Malaysia
12 B12 Belimbing Madu Chan Yong 2\* 1968 Malaysia
13 B13 Istana Perak 1 1971 Malaysia
14 B14 Istana Perak 2 1971 Malaysia
15 B15 Istana Perak 3 1971 Malaysia
16 B16 NA NA Malaysia
17 B17 Belimbing Madu\* 1988 Malaysia
18 B18 NA 1989 Malaysia
19 B19 NA 1989 Malaysia
20 NA Golden Star\* 1965 USA
21 NA Newcomb 1965 USA
22 NA Thayer 1965 USA
23 NA Arkin 1965 USA
24 272065 Mih Tao\* 1963 Taiwan
25 NA Dah Pon 1963 Taiwan
26 NA Tean Ma 1963 Taiwan
27 NA Fwang Tung\* 1973 Thailand
\*indicates the commercially cultivated superior clones of Star-fruit plant; NA, not available; USA, United States.
###### Minerals found in the star-fruits (Averrhoa carambola L.) \[12\]
Mineral Amount \[mg/100g fruit\]\*
------------------------- ----------------------------
Sodium (Na) 3.8 - 3.85
Potassium (K) 167.13 - 168.0
Calcium (Ca) 6.37 - 6.40
Phosphorous (P) 17.87 - 17.88
Magnesium (Mg) 11.85 - 12.05
Iron (Fe) 0.34 - 0.45
Copper (Cu) 0.19 - 0.45
Zinc (Zn) 0.29 - 0.51
Manganese (Mn) 0.04 - 0.52
Selenium (Se) Not Detectable
\*on a dry weight basis
###### Carotene, vitamins and acids found in mature star-fruits (Averrhoa carambola L.)
Name Amount \[mg/100g Star-fruit weight\]\*
------------------------- ----------------------------------------
Carotene 0.003 - 0.55
Tartaric acid 4.37
Oxalic acid 9.6
Ketoglutaric acid 2.2
Citric acid 1.32
Vitamin B1 & B2 0.12
Vitamin C 25.8
\*on a dry weight basis
{#F1}
| {
"pile_set_name": "PubMed Central"
} |
1. Introduction {#s0005}
===============
DNA methylation studies have been traditionally concerned with 5-methylcytosine (5mC), mostly in the context of CpG dinucleotides, which until recently was the only epigenetic modification of DNA with known biological function. The rediscovery of 5-hydroxymethylcytosine (5hmC) in mammalian DNA has opened up an exciting new area of epigenetic research concerning the potential role of 5hmC in both normal development and disease progression [@b0005; @b0010]. 5hmC has been proposed to be an intermediate in an active DNA demethylation pathway and may also be a stable epigenetic modification in its own right, contributing unique regulatory functions to the epigenome [@b0015].
Current methods for profiling 5hmC at a genome-wide level rely largely on sequencing-based protocols that are both costly and time-consuming. In addition, they require a high degree of expertise with the chosen assay for successful implementation. The Infinium HumanMethylation450 BeadChip is a widely-used, robust and reliable tool for DNA methylation profiling [@b0020; @b0025], suitable for high-throughput sample processing and is therefore the platform of choice for current epigenome-wide association studies [@b0030; @b0035]. The 450K array interrogates over 480,000 CpG sites in the human genome, making it an appropriate tool for 5hmC detection as this modification has been reported to occur exclusively in the CpG context in both mammalian embryonic stem cells and frontal cortex tissue [@b0040; @b0045; @b0050]. However, current workflows for this microarray rely on sodium bisulfite conversion of DNA, which discriminates between methylated and unmethylated cytosine bases within the genome but cannot differentiate between 5mC and 5hmC [@b0055].
The recent development of oxidative bisulfite (oxBS) chemistry presents an opportunity to adapt bisulfite-based 5mC profiling technology for 5hmC detection [@b0045]. Here we describe low- and high-input protocols for genome-wide profiling of 5hmC by coupling oxBS chemistry and the Infinium 450K BeadChip (oxBS-450K). Alternative approaches also using oxBS or Tet-assisted bisulfite (TAB) chemistries have been developed by Field et al. [@b0060] and Chopra et al. [@b0065], respectively. In brief, oxBS involves a selective oxidation step prior to bisulfite conversion of genomic DNA that converts 5hmC residues to 5-formylcytosine (5fC) with high efficiency. 5fC behaves in a manner analogous to unmodified cytosine under bisulfite conditions, allowing only true 5mC positions to be detected as cytosine after oxidative bisulfite conversion and PCR amplification of the DNA sample. Subtraction of oxBS-generated methylation profiles from standard BS-only methylation profiles allows for the detection of hydroxymethylated cytosine positions within the genome. Whilst traditional bisulfite conversion leads to potential overestimation of 5mC levels in a given sample due to the presence of 5hmC, the oxBS-450K method generates a more accurate 5mC profile by removing the confounding factor of 5hmC.
Global 5hmC levels are known to vary considerably between tissue types and have been found to be highest in the brain [@b0070; @b0075] and so the distribution of 5hmC in this tissue may be of particular interest. Here, we show that the Infinium 450K can be used to successfully detect 5hmC in three distinct genomic DNA samples (two brain and one whole blood) by comparison of BS- and oxBS-450K datasets and provide validation of our reported 5hmC signal using independent technologies.
2. Materials and methods {#s0010}
========================
2.1. DNA samples {#s0015}
----------------
All procedures followed are in accordance with the ethical standards of the Helsinki declaration (1964, 2008) of the World Medical Association. Human brain genomic DNA from a 33 year-old male was obtained from AMSBIO (catalogue \#D1234035, lot \#B701205; referred to as 'brain1'). Human frontal cortex DNA was collected according to Ethical Approval obtained from the Berkshire Local Research Ethics Committee (REC 08/H0505/165), with informed written consent (referred to as 'brain2'). Pooled (male and female) whole blood DNA was obtained from blood donations of volunteers from the UCL Cancer Institute.
2.2. Oxidative bisulfite (oxBS) and bisulfite-only (BS) conversion workflows {#s0020}
----------------------------------------------------------------------------
We tested two protocols (low and high input) and report the results for both to highlight critical steps and provide insights into their dynamics.
### 2.2.1. Low-input oxBS-450K protocol {#s0025}
1 μg DNA per sample was processed using a trial version of the TrueMethyl® 24 kit (CEGX). Samples were split evenly into two aliquots of ∼500 ng processed through either the BS-only or oxBS conversion workflow. Each aliquot was subjected to an initial buffer exchange step using a spin column format and eluted in ultra-pure water. The full eluate (∼22 μl per sample) was carried forward. Samples were denatured using the provided denaturing solution for 30 min at 37 °C, in a total reaction volume of 24 μl and immediately taken forward to oxidation. 1 μl of the provided oxidant solution was added to each sample undergoing the oxBS workflow only (1 μl of ultra-pure water was added to BS-only samples for mock oxidation). All samples were incubated for 30 min at 40 °C. Oxidised samples were equilibrated to room temperature before proceeding immediately to bisulfite conversion. Bisulfite reagent solution was prepared as described in the TrueMethyl® protocol and 170 μl was added to each 25 μl oxidation reaction mixture. Finally, 5 μl of bisulfite additive was added to each reaction, bringing the total volume to 200 μl. All reactions were incubated using bisulfite-specific thermal cycling conditions (see [Supplementary information](#s0085){ref-type="sec"}). Converted DNAs were purified using the provided spin columns and 4 μl of each sample eluate was used as input into the Infinium 450K array.
### 2.2.2. High-input oxBS-450K protocol {#s0030}
4 μg DNA per sample was processed using the TrueMethyl® protocol for 450K analysis (Version 1.1, CEGX™). The protocol was performed as described, with 2 minor alterations. First, the initial shearing step of DNA was performed using a Bioruptor (Diagenode™). The DNA was diluted into a total volume of 50 μl in 1.5 ml microtubes and sonicated for 30 s on setting 'H'. 10 μl was set aside for visualisation on a 2% agarose gel to confirm correct DNA fragmentation (all ⩽10 kb). Secondly, the remaining 40 μl (∼3.2 ug) of sheared DNA was split into 2 × 20 μl aliquots and purified using Agencourt AMPure XP® beads using a modified protocol (<http://www.cambridge-epigenetix.com/uploads/files/TrueMethyl24_UGuide.pdf>, TrueMethyl® 24 kit user guide, CEGX™). Each sample was eluted in 20 μl ultra-pure water and then processed either through the oxBS or BS-only workflow as outlined in the TrueMethyl® user guide. 7 μl of the final eluate was used as input into the Infinium 450K array, made possible by adding 1 μl of 0.4 N NaOH to each sample rather than the standard 4 μl of 0.1 N NaOH when setting up the MSA4 plate.
2.3. 450K BeadChip processing and data analysis {#s0035}
-----------------------------------------------
450K BeadChips were processed by UCL Genomics following the manufacturer's recommendations and data was analysed using the Chip Analysis Methylation Pipeline (ChAMP) implemented in *R* and available from Bioconductor [@b0080]. Briefly, raw IDAT files were used as input data and probes were filtered by their raw intensity values using a detection *p*-value threshold of 0.01. Probes corresponding to the X and Y chromosomes were removed from the dataset as both male and female samples were being analysed. The data was then normalised using the SWAN method [@b0085], producing a final dataset containing 432,056 probes and their corresponding beta values for each sample for further downstream analysis. The *R* package *pvclust* was used for multiscale bootstrap resampling (1000 replications) for hierarchical cluster analysis.
2.4. Oxidative bisulfite pyrosequencing (oxBS-pyroseq) {#s0040}
------------------------------------------------------
50 bp windows containing at least 2 probes showing a hydroxymethylation level of 30% or higher (based on the AMSBIO brain dataset) were selected for validation by oxBS-pyroseq. Primers were designed using PyroMark Assay Design 2.0 software (see [supplementary information](#s0085){ref-type="sec"} for sequences and probe IDs). Two aliquots of AMSBIO human brain DNA underwent either oxidative bisulfite or bisulfite-only conversion (as detailed in Section [2.2](#s0020){ref-type="sec"}) and were PCR amplified using a Pyromark PCR kit (Qiagen, cat. \#978703) under the following thermocycling conditions: 95 °C for 15 min, (94 °C for 30 s, 66 °C (−0.5 °C per cycle) for 30 s, 72 °C for 30 s) × 13 cycles, (94 °C for 30 s, 56 °C for 30 s, 72 °C for 30 s) × 50 cycles, 72 °C for 10 min. Assays were run on a PyroMark Q96 ID system and results were analysed using Pyro Q-CpG 1.0.9 software. Methylation scores at each CpG site were normalised against a linear calibration curve and 5hmC levels calculated by subtraction of oxBS from BS-only methylation calls.
2.5. Mass spectrometry {#s0045}
----------------------
500 ng of genomic DNA was incubated with 5 U of DNA Degradase Plus (Zymo Research) at 37 °C for 3 h. The resulting mixture of 2′-deoxynucleosides was analysed on a Triple Quad 6500 mass spectrometer (AB Sciex) fitted with an Infinity 1290 LC system (Agilent) and an Acquity UPLC HSS T3 column (Waters), using a gradient of water and acetonitrile with 0.1% formic acid. External calibration was performed using synthetic standards, and for accurate quantification, all samples and standards were spiked with isotopically labelled nucleosides. 5mC and 5hmC levels are expressed as a percentage of total cytosines.
3. Results and discussion {#s0050}
=========================
Global 5hmC levels are known to be lower than 5mC levels and more varied between different tissues [@b0070; @b0075]. To test the oxBS-450K protocol, we therefore selected tissues expected to have high (two types of brain tissue) and low (whole blood) levels of 5hmC and confirmed their percentage by quantitative LC--MS to be 0.039% for whole blood and 0.924% and 1.107% for the two brain samples, respectively. In comparison, the corresponding 5mC levels were ∼4% in all three samples.
3.1. Identification of 5hmC {#s0055}
---------------------------
5hmC detection was achieved by identifying differentially methylated CpG sites between the BS- and oxBS-treated replicates within each sample set. The beta value associated with each probe on an array reflects the methylation level at that particular location on a scale of 0--1, where 0 is unmethylated and 1 is fully methylated.
Beta values resulting from BS treatment represent the total methylation score (5mC and 5hmC) as both cytosine modifications undergo bisulfite conversion to uracil with comparable efficiency. In contrast, beta values resulting from oxBS treatment represent only the 5mC level at the corresponding probe locations. Therefore normalised beta values were used to calculate delta beta (Δ*β*) values for each probe by subtraction of the oxBS beta value from the BS-only beta value. The Δ*β* score is a reflection of the 5hmC level at each particular probe location.
Unsupervised hierarchical clustering of this dataset showed a clear separation between blood, brain1 and brain2 and also between treatment options, with both BS-only and oxBS-treated replicates of each sample clustering in pairs ([Fig. 1](#f0005){ref-type="fig"}A). The corresponding density profiles show a clear reduction of methylation signals after oxBS treatment ([Fig. 1](#f0005){ref-type="fig"}B).
Using this method, both positive and negative Δ*β* scores were calculated. Positive Δ*β* values represent potential sites of 5hmC; negative Δ*β* values, or 'hypermethylation' in the oxBS-treated sample, were not expected as the oxidation reaction is unidirectional and thus are likely to reflect background noise generated by this method. [Fig. 2](#f0010){ref-type="fig"}A shows the Δ*β* distribution across all probes for the brain1 sample, displaying significantly more positive Δ*β* scores than negative, as expected. Less than 3% of all negative Δ*β* scores correspond to a 5hmC level of more than 10%, however over 30% of all positive Δ*β* scores correspond to a 5hmC level over 10%. In addition, [Fig. 2](#f0010){ref-type="fig"}B shows that the largest Δ*β* scores are associated with the most significant *p*-values and that significance falls away as the Δ*β* values approach zero, as expected. Furthermore, frequency plots of all probes according to their associated *p*-values ([Fig. 3](#f0015){ref-type="fig"}) show that the majority of Δ*β* values above zero represent significant differences in methylation score (*p* ⩽ 0.05). In contrast, the *p*-values associated with negative Δ*β* scores are distributed across the entire range (0--1), with increasing numbers of probes associated with higher *p*-values. This is consistent with the idea that the negative Δ*β* scores represent false differences in methylation score between the paired BS-only and oxBS datasets.
There are various ways to define an appropriate Δ*β* threshold to identify a probe-set of significantly hydroxymethylated cytosines, including shrinkage estimation of variances, but this approach requires each sample to be run in multiple (3--4) replicates resulting in extra costs [@b0060]. As the amount of available genomic DNA is often limiting, we based our strategy on running two replicates per sample and opted for using a conservative threshold of Δ*β* ⩾ 0.3, or minimum 30% 5hmC, followed by validation. Using this approach, we identified 6578 and 7692 probes in our brain1 and brain2 samples, respectively, in contrast to just 801 probes in whole blood DNA. Inclusion of a third technical replicate of the brain1 sample, however, increased the statistical power of the dataset by allowing the Δ*β* calculations to reach statistical significance. We were able to use a BH-adjusted *p*-value threshold of 0.05 to call 5hmC, which dramatically increased the probe number because any significant Δ*β* values in the range of 0--30% 5hmC could now be identified alongside the more highly hydroxymethylated probes, resulting in the detection of 64,720 significant probes in total. Of these, only 117 sites were associated with negative (but nevertheless significant) Δ*β* scores, corresponding to a false discovery rate of less than 0.002. 5hmC levels fell within the range of 4--59%, with a mean level of 18% 5hmC across all positive 5hmC probe sites whilst all negative Δ*β* scores corresponded to less than 20%.
While the low-input oxBS-450K protocol only requires 1 μg starting DNA, the high-input protocol requires 4 μg DNA but results in much improved sensitivity and reproducibility. Using the same detection *p*-value threshold of 0.01 as for the low-input protocol to filter probes, we found that the technical replicates showed an improved correlation of *R*^2^ \> 0.99 compared to *R*^2^ \> 0.95 for the low-input protocol, comparable to current 'gold-standard' bisulfite conversion protocols. Comparison of BS-only and oxBS datasets using just two replicates per sample was sufficient to call 165,495 5hmC sites (BH-adjusted *p*-value \< 0.05) with just 627 false positive probes associated with a negative Δ*β* score ([Fig. 4](#f0020){ref-type="fig"}). 5hmC levels were detected in the range of 4--57%, with a mean 5hmC level of 15%. In contrast, negative Δ*β* values all fell below 20%.
Analysis of the significant but negative Δ*β* probes between the low- and high-input protocols for the brain1 sample suggests that the locations of these negative sites are random: 117 sites were called by the low-input (3-replicate) method compared to 627 with the high-input method, yet only 8 negative Δ*β* sites were found in common between the two (the increase in negative probe number is a reflection of an overall increase in the number of significant probes identified using the high-input method). Despite showing statistically significant *β* value differences, these pseudo false positives cannot reflect true differences in methylation level at these probe sites and should be filtered out before further downstream data analysis.
Comparison of the 5hmC-containing probes identified by the different protocols shows that overlap is high, supporting our conclusion that these are sites of genuine 5hmC. 83.6% of the 5hmC-containing sites initially called by the low-input, 2-replicate method remained significantly hydroxymethylated with the addition of a third replicate (5498 of 6578). This figure increases to 94.2% when comparison is made to the high-input, 2-replicate dataset. Moreover, only 3.5% of the 5hmC probes called by the low-input method fail to reach significance when either replicate number or DNA input is increased, suggesting that the false positive rate of our low-input, 2-replicate method is fairly low ([Supplementary Fig. 1](#s0085){ref-type="sec"}).
We conclude that the use of the high-input oxBS-450K protocol on just two replicates results in detection of 5hmC at over twice the number of probes compared to the low-input protocol using three replicates. However, both the mean and range of 5hmC levels detected is very similar in both cases, suggesting that using lower amounts of input DNA does not necessarily affect the quality of the array data produced, but does severely limit the number of probes whose Δ*β* values reach statistical significance. The correlation matrix shown in [Fig. 5](#f0025){ref-type="fig"} is intended to help guide potential users on which protocol to use.
3.2. Distribution of 5hmC {#s0060}
-------------------------
We used our probe sets of significant 5hmC sites to investigate the distribution of 5hmC within different genomic features, compared to the overall distribution of probes present on the array. The following categories were used for classification of each probe: 3′UTR region, 5′UTR region, 1st exon, 200 bp window upstream of the TSS (TSS200), 1500 bp window upstream of the TSS (TSS1500), intergenic region or gene body. A permutation test was performed to calculate percentage enrichment for 5hmC within each feature ([Fig. 6](#f0030){ref-type="fig"}) and 5hmC was found to be significantly enriched in the gene body in all three DNA samples. This is consistent with previous studies concerning the genomic distribution of 5hmC [@b0065; @b0095]. Conversely, 5hmC was found to be significantly depleted from promoter regions (defined as within 1500 bp upstream of TSS) and from CpG islands ([Supplementary Fig. 2](#s0085){ref-type="sec"}), which agrees with several published studies that describe an inverse correlation between 5hmC and CpG density at the promoter [@b0045; @b0065; @b0100].
3.3. Validation of 5hmC {#s0065}
-----------------------
We used two methods (LC--MS and oxBS-pyroseq) to validate 5hmC identified with oxBS-450K. On the global level, we used quantitative LC--MS to confirm that each of our samples did indeed contain 5hmC levels suitable for detection with oxBS-450K. Using the average of two replicates, we established that the 5hmC levels of both brain samples were similar, at around 1% of total cytosine bases (0.92% in brain1 and 1.11% in brain2), whilst the whole blood sample contained only low levels of 5hmC, at 0.04% of total cytosine as expected. In contrast, 5mC levels in all three samples were broadly consistent at around 4% of total cytosine. The overall pattern of total 5hmC content per sample correlated well with the observed 5hmC signal from the 450K array data, where we observed ten-fold higher numbers of probes with significant 5hmC in the brain as compared to whole blood ([Supplementary Fig. 3](#s0085){ref-type="sec"}).
On the single cytosine level, we used quantitative oxBS-pyroseq to validate selected 5hmC sites identified with oxBS-450K in our brain1 sample. The hydroxymethylation status of six individual CpG sites across three genomic regions was confirmed using this assay ([Supplementary Fig. 4](#s0085){ref-type="sec"}), with 5hmC levels differing by less than 10% between the 450K array and pyrosequencing assay. In addition, our validation results suggest that there is little difference in the accuracy of 5hmC calling depending on the 450K protocol used, as both low-input and high-input methods generated 5hmC levels within 10% of the pyrosequencing readout.
4. Conclusions {#s0070}
==============
We have shown that oxidative bisulfite chemistry can be combined with the 450K DNA methylation microarray to reliably detect 5hmC and 5mC in the human genome. The oxBS-450K protocol presented here results in highly reproducible technical replicates, comparable to current 'gold-standard' bisulfite conversion kits. Our analysis shows that 5hmC can be detected at statistically significant levels at over 30% of all CpG sites interrogated on the array, with a mean 5hmC level of 15% in a human brain sample. Whilst the use of an optimised protocol offers an improvement in the correlation of technical replicates and allows a higher proportion of 5hmC sites to be called with statistical significance, the amount of input DNA required may prove limiting for many users. However, validation using oxBS-pyroseq suggests that accurate 5hmC levels, albeit at fewer sites, can nevertheless be called using a lower input of 1 μg DNA for oxBS-450K analysis.
Appendix A. Supplementary data {#s0085}
==============================
Supplementary Fig. 15hmC probe overlap between the three analysis methods for brain1. Overlap is shown for the 6578 probes called as hydroxymethylated using the low-input (2 replicate) protocol. Of these, 81.3% remained significantly hydroxymethylated both when the replicate number was increased and when the experiment was repeated using the high-input protocol. Only 3.5% of these probes were unable to reach significance when either replicate number was increased or when higher amounts of input DNA were used.Supplementary Fig. 2SuProbe distribution according to CpG context. (A) Distribution of all probes after filtering (432,056 total) on 450K BeadChip according to CpG context. (B) Distribution of hydroxymethylated probes (6578 total) in brain1 sample according to CpG context. (C) Relative enrichment of 5hmC probes in differing CpG contexts. Significant enrichment of 5hmC is seen in 'open sea' regions of low CpG density, although the most striking observation is the depletion of 5hmC at CpG islands.Supplementary Fig. 3LC--MS quantification of 5mC/5hmC content in 3 DNA samples. The fraction of 5mC appears consistent across all samples, in contrast to the 5hmC content, where nearly 30-fold higher levels are seen in the 2 brain samples as compared to whole blood. This pattern of increased levels of 5hmC in brain is reflected in the 450K dataset (low-input protocol results are shown), where the number of CpG sites containing the highest levels of 5hmC (⩾30%) are around 10-fold higher in brain versus blood.Supplementary Fig. 4Validation of 5hmC calls using a pyrosequencing assay. 5hmC levels at 6 individual CpG sites distributed across three distinct genomic regions (indicated by orange, purple and yellow shading, respectively) were validated in the brain1 sample. Results are plotted as % 5hmC at each CpG locus as measured by pyrosequencing, low-input 450K and high-input 450K assays.Supplementary data 1Supplementary Table 1. Bisulfite conversion thermal cycler conditions. Supplementary Table 2. Pyrosequencing PCR primers and associated probe IDs.
SS was supported by a MRC CASE studentship (G1000411). Research in the Beck lab was supported by the Wellcome Trust (99148), a Royal Society Wolfson Research Merit Award (WM100023) and EU-FP7 projects IDEAL (259679), EPIGENESYS (257082) and BLUEPRINT (282510). We also wish to acknowledge helpful discussions with the TrueMethyl 450K Forum, particularly Toby Ost, Christine Clark, Sarah Field, Dario Beraldi and Jonathan Mill.
{#f0005}
{#f0010}
{#f0015}
{#f0020}
{#f0025}
{#f0030}
| {
"pile_set_name": "PubMed Central"
} |
INTRODUCTION {#s1}
============
We previously reported that simultaneously stimulating CD3 on T cells and CD40 on B cells augments the antitumor reactivity of tumor-draining lymph node (TDLN) cells \[[@R1]\]. Furthermore, we demonstrated that *in vivo* sensitized and *in vitro* activated TDLN B cells mediate tumor regression in cancer adoptive immunotherapy \[[@R2]\]. In hosts that received whole body irradiation to delete lymphoid cells, the subsequent transfer of activated B cells had significant antitumor effects on established tumors \[[@R2]\]. This observation was made in a weakly immunogenic 3-methylcholanthrene-induced murine fibrosarcoma MCA 205 model and in a poorly immunogenic murine melanoma D5 model that are both syngeneic to B6 mice \[[@R2]\]. In a murine 4T1 model of breast cancer syngeneic to Balb/c mice, we reported that the transfer of LPS/anti-CD40- activated 4T1 TDLN B cells significantly reduced the induction of spontaneous 4T1 pulmonary metastases, and these effector B cells could directly kill 4T1 tumor cells \[[@R3]\]. Together, these studies demonstrated that transferred effector B cells can act independently in eliciting tumor regression *in vivo* in several murine tumor models syngeneic to hosts with different genetic backgrounds.
Interleukin 2 (IL-2) is a pleiotropic cytokine that stimulates T-cell proliferation; enhances NK cytolytic activity, induces the differentiation of Tregs, and causes activation- induced cell death \[[@R4], [@R5]\]. However, the effect of IL-2 on B lymphocytes is not well defined. In addition, CXCR4 is a chemokine receptor specific for stromal-derived-factor-1(SDF-1), and is also known as CXCL12, a molecule with strong chemoattractant properties for lymphocytes \[[@R6], [@R7]\]. Furthermore, a property of cytotoxic lymphocytes is their expression and release of powerful toxins, including the pore-forming protein perforin \[[@R8], [@R9]\]. While perforin is known to be a cytolytic protein found in the granules of cytotoxic T lymphocytes (CTLs) and natural killer cells \[[@R10], [@R11]\], its role in B cells is unknown. In this present study, we examined new mechanisms contributing to direct B cell-mediated antitumor immunity, including the impact of IL-2, the CXCR4/CXCL12 pathway and perforin in mediating tumor regression after the adoptive transfer of B effector cells.
RESULTS {#s2}
=======
Inhibition of pulmonary metastases by TDLN B cells is enhanced with IL-2 administration in adoptive immunotherapy {#s2_1}
-----------------------------------------------------------------------------------------------------------------
Although interleukin-2 was originally described as a "T cell growth factor", we have found that it can significantly enhance the antitumor immunity of the B effector cells in adoptive therapy. In order to investigate the role of IL-2 in B cell-mediated adoptive immunotherapy, we examined the efficacy of transferred TDLN B cells given in a suboptimal dose (1 × 10^6^ cells/mouse) in conjunction with or without IL-2 administration. Fourteen days after 4T1 tumor cells were injected into the mammary fat pad, mice were administered with activated TDLN B cells alone or TDLN B cells plus IL-2. Fourteen days later, mice were euthanized to quantify pulmonary metastases. A suboptimal dose of B cells alone showed no efficacy, but B cells plus IL-2 administration i.p. significantly inhibited the induction of spontaneous pulmonary metastases (Figure [1](#F1){ref-type="fig"}, Expt. 1). However, IL-2 alone had no therapeutic effect compared to PBS-treated controls (Figure [1](#F1){ref-type="fig"} Expt. 2). These experiments indicated that exogenous IL-2 administration augmented the therapeutic efficacy of transferred effector B cells.
{#F1}
In follow-up experiments, we investigated whether IL-2 receptor (IL-2R) was expressed on activated B effector cells. We purified TDLN B cells for this purpose. Unsorted 4T1 TDLN cells before purification are composed of approximately 30% CD19^+^ B cells and 60% CD3^+^ T cells, while cell purification enriched the CD19^+^ B cells to \> 95% (Figure [2A](#F2){ref-type="fig"}). We then tested IL-2R (CD25) expression on the purified 4T1 TDLN B cells (Figure [2B](#F2){ref-type="fig"}). Before B cell activation and expansion, about 10% of the B cells expressed CD25. B cell activation and expansion almost doubled the number of B cells expressing CD25, resulting in almost 20% of the B cells expressing CD25. These results suggest that IL-2 most likely interacts with TDLN B cells by directly binding to IL-2R. Figure [2A](#F2){ref-type="fig"} represents the effective purification of B cells from TDLN cells, which we have reported in a few of our previous publications \[[@R2], [@R3], [@R12]\]. Repeated experiments shown in Figure [2B](#F2){ref-type="fig"} allowed us to generate a bar graph with p values, indicating that B cell activation and expansion (A/E) significantly (p\<0.05) augmented the CD25 (IL-2R)-expressing B cells (Figure [2C](#F2){ref-type="fig"}).
{#F2}
IL-2 modulates antitumor humoral responses in the host treated with adoptively transferred TDLN B cells {#s2_2}
-------------------------------------------------------------------------------------------------------
To determine the mechanism(s) that IL-2 enhances the antitumor immunity of 4T1 TDLN B effector cells in adoptive therapy of 4T1-bearing animals, we collected host spleens at the end of the treatments. Splenic T and B cells were enriched and subsequently activated *in vitro* with anti-CD3/anti-CD28 or LPS/anti-CD40 respectively. Lastly, culture supernatants were collected for antibody detection. While we found no antibody in T cell cultures as expected, we found significantly (p\<0.01) higher IgG production by B cells isolated from animals treated with B cells + IL-2 compared with B cell alone treatment groups (Figure [3A](#F3){ref-type="fig"}).
{ref-type="fig"}. Binding experiments were repeated 3 times. **C.** Binding to 4T1 cells by IgG from culture supernatants of spleen B cells harvested from the 4T1 bearing mice treated with 4T1 TDLN B cells alone or 4T1 TDLN B cells plus IL-2 administration.](oncotarget-07-60461-g003){#F3}
To test the specificity of the antibody, we assessed the binding of the immune supernatants of cultured splenic B cells against 4T1 tumor cells. As shown in Figure [3B](#F3){ref-type="fig"}, in three replicate experiments, we observed that splenocytes from hosts subjected to B cell therapy plus IL-2 administration produced antibody which bound to 4T1 cells (3.2-3.9%) more than the antibody produced by splenocytes from hosts subjected to B cell therapy alone (1.0-1.3%), and was statistically significant (p\<0.01, Figure [3C](#F3){ref-type="fig"}). These data demonstrated enhanced systemic humoral responses against 4T1 in the hosts treated with TDLN B cells + IL-2 therapy. While we did not analyze the isotype of the IgG in this study, we have previously reported that the major isotype was IgG2b \[[@R13]\]. This IgG2b binds specifically to relevant tumor cells. Similarly, the binding was low (from 1.4% to 5.4% in multiple assays), but this binding was tumor specific since their binding to an irrelevant tumor target was 0-0.3%. More importantly, such binding was immunologically significant because such binding resulted in complement dependent cytotoxicity \[[@R13]\].
4T1 TDLN B cell-mediated cytotoxicity of 4T1 tumor cells involves the CXCR4/CXCL12 pathway {#s2_3}
------------------------------------------------------------------------------------------
Interaction of chemokine and chemokine receptors represents a key mechanism in the attraction and recruitment of immune cells to cancer cells \[[@R14], [@R15]\]. We examined the expression of several chemokine receptors on purified 4T1 TDLN B cells by flow cytometry, e.g. CXCR2, CXCR4, CCR5, CCR7, and CCR10. We found that CXCR2 and CCR7 were undetectable, and CCR5 or CCR10 expressed on \<5% B cells (data not shown). However, CXCR4 expression was comparatively high on purified 4T1 TDLN B cells and continued to be high after LPS/anti-CD40 activation/expansion (12-15%, Figure [4A](#F4){ref-type="fig"}). Statistically, CXCR4 expression on B cells did not change significantly (p\>0.05) after B cell activation and expansion (A/E) in the absence of IL-2 (Figure [4B](#F4){ref-type="fig"}). We then performed additional experiments comparing the CXCR4 expression on B cells in the presence or absence of IL-2. As shown in Figures [4C](#F4){ref-type="fig"} and [4D](#F4){ref-type="fig"}, CXCR4 expression was significantly increased when B cells were cultured with the addition of IL-2 (300 IU/ml). The culture conditions were the same as those used for Figure [4A](#F4){ref-type="fig"}. CXCR4 expression was 14-19% without the use of IL-2 (Figure [4C](#F4){ref-type="fig"}). However, in the presence of IL-2, CXCR4 expression was increased to 25-34%. The increased CXCR4 expression on B cells in the presence of IL-2 was statistically significant (p\<0.05) compared with that in the absence of IL-2 (Figure [4D](#F4){ref-type="fig"}).
{ref-type="fig"} and double stained with PE-anti-CD19 and FITC-anti-CXCR4 before and after activation/expansion (A/E) respectively. **B.** CXCR4 expression on 4T1 TDLN B cells before and after A/E without IL-2. **C.** Comparison of CXCR4 expression on 4T1 TDLN B cells with *vs*. without IL-2. **D.** IL-2 significantly increased expression of CXCR4 on 4T1 TDLN B cells. **E.** 4T1 tumor cells express CXCL12. Data from two independent experiments of flow cytometry assays are shown. **F.** IL-2 significantly augmented CXCL12 production by cultured 4T1 cells as detected in the culture supernatant using ELISA.](oncotarget-07-60461-g004){#F4}
CXCR4 is the receptor for CXCL12 (SDF-1) which has been shown to mediate target cell chemotaxis by cells expressing CXCR4 \[[@R15]\]. We detected the expression of intracellular CXCL12 on 4T1 cells and found that approximately 20-30% of 4T1 cells expressed CXCL12 (Figure [4E](#F4){ref-type="fig"}, Expt 1 & 2). To provide more evidence that IL-2 augments the TDLN B cell killer/effector function via the CXCR4/CXCL12 axis, we performed experiments to compare the CXCL12 production by 4T1 tumor cells in the presence or absence of IL-2. In Figure [4F](#F4){ref-type="fig"}, we cultured 4T1 cells with various doses of IL-2, collected the culture supernatants, and tested the production of CXCL12 using ELISA. IL-2 significantly enhanced the production of CXCL12 by 4T1 cells in an IL-2 dose dependent manner.
To ascertain the chemotactic effects of CXCL12 on TDLN B cell migration, transwell assays were performed. TDLN B cells were purified and used for the chemotaxis assays as described in the Methods section. Compared with the control group (complete medium supplemented with 10% FBS) at each time point, the addition of CXCL12 to the medium in the lower chambers resulted in significant (p \< 0.01) migration of TDLN B cells from the upper chamber to the lower chamber with time (Figure [5A](#F5){ref-type="fig"}).
{#F5}
Righi et al. have previously shown that AMD3100, a highly specific CXCR4 antagonist, blocks the binding of CXCR4 to CXCL12 \[[@R16]\]. We added AMD3100 to block the interaction of CXCL12 with CXCR4 during the LDH release assay while testing B cell cytotoxicity against 4T1 tumor cells. For this assay, effector B cells were generated from 4T1 TDLN B cells as described in the Methods section. Target cells (4T1) were plated with B cells at various effector to target cell ratios with or without the addition of AMD3100 (2.5 μg/ml) as indicated in Figure [5B](#F5){ref-type="fig"}. AMD3100 significantly (p\<0.05) inhibited B cell-mediated 4T1 tumor cell death at the E:T ratios of 10:1 and 30:1 in 2 of 2 experiments performed. These results suggest that the interaction of CXCR4 on B cells and its ligand CXCL12 produced by 4T1 cells is involved in the direct cytotoxicity of 4T1 cells by B cells.
We have recently reported that TDLN B cells directly kill 4T1 tumor cells via the Fas/FasL pathway \[[@R12]\]. To examine whether B cell-mediated 4T1 tumor cell death involves both Fas/FasL and CXCR4/CXCL12 pathways, we added anti-FasL and AMD3100 in the same experiment. As shown in Figure [5C](#F5){ref-type="fig"}, anti-FasL alone significantly (p\<0.05) suppressed the killing of activated B cells on 4T1 tumor cells at different E:T ratios; AMD3100 alone also significantly (p\<0.05) inhibited such killing, which was consistent with our observations in the previous experiments. Of note, anti-FasL and AMD3100, when used simultaneously, resulted in significantly greater inhibition of the 4T1 tumor cell cytotoxicity by 4T1 TDLN B cells compared to anti-FasL or AMD3100 alone (p\<0.05). These results support the conclusion that the cytotoxicity of 4T1 tumor cells by effector B cells involves both Fas/FasL and CXCR4/CXCL12 pathways.
The role of perforin in B cell-mediated cytotoxicity of tumor cells {#s2_4}
-------------------------------------------------------------------
Collectively, the above experiments demonstrated that B cells could directly kill tumor cells via cell-cell contact. It is well known that perforin, a small secreted pore-forming granule protein, can result in cell death after it is introduced into the target cell \[[@R17], [@R18]\]. To examine the potential role of perforin in B cell-mediated cytotoxicity of 4T1 cells, we collected 4T1 TDLN T cells and B cells, and activated T cells with anti-CD3/anti-CD28 and B cells with LPS/anti-CD40 respectively as described in the Methods section. We collected the culture supernatants and detected perforin produced by T cells and B cells (Figure [6A](#F6){ref-type="fig"}). Figure [6A](#F6){ref-type="fig"} indicates that perforin is released not only by T cells but also by the activated TDLN B cells. To provide experimental evidence that IL-2 augments the TDLN B cell killer/effector function via the perforin pathway, we performed experiments to compare the perforin production by 4T1 TDLN B cells in the presence and absence of IL-2. As shown in Figure [6B](#F6){ref-type="fig"}, addition of IL-2 to the B cell culture significantly (p\<0.01) increased the production of perforin. To further define the cytotoxic mechanisms of B cells, we compared B cell-mediated cytotoxicity of 4T1 cells with direct cell contact (i.e. mixed cell cultures) vs. no cell contact (i.e. transwell cultures) (Figure [6C](#F6){ref-type="fig"}). The starting 4T1 cell number was 400 as indicated. At the 4T1: B cell ratio = 1:10, about 50% of the 4T1 cells (approximately 200 cells) were killed by B cells in mixed culture. However, in the transwell experiment, about 75% (approximately 300 cells) survived, and about 25% of the 4T1 cells (approximately 100 cells) were killed. These experiments thus suggest that soluble factors, such as perforin, could lead to partial lysis of 4T1 cells as shown in the transwell experiment, but cell contact is needed for optimal B cell-mediated cytotoxicity of 4T1 cells as shown in the mixed culture experiment.
{#F6}
To better assess the cell death of 4T1 cells, we examined one of the cell death pathway molecules, cleaved caspase 3, by flow cytometry to verify the data assessed by trypan blue in Figure [6C](#F6){ref-type="fig"}. Figure [6D](#F6){ref-type="fig"} shows that, in transwell experiments as described in Figure [6C](#F6){ref-type="fig"}, cleaved caspase 3- positive cells was 25%-27% of the total 4T1 cells in 3 of 3 independent experiments performed. Furthermore, in the cell-cell contact assay, also as described in Figures [6C](#F6){ref-type="fig"} and Figure [6E](#F6){ref-type="fig"} shows that 13-17% of the mixed cells were CD19^−^, e.g. 4T1 cells. More 4T1 cells expressed cleaved caspase-3 (33-36% of the total 4T1 cells) than those 4T1 cells in the transwell experiments as shown in Figure [6D](#F6){ref-type="fig"}.
In Figure [5](#F5){ref-type="fig"}, we measured the cytotoxicity of 4T1 tumor cells by B cells in the presence of anti-FasL and the CXCR4-specific inhibitor AMD3100, and found that both Fas/FasL and CXCR4/CXCL12 pathways are involved in B cell-mediated killing of tumor cells. To examine the role that perforin may have, we performed additional transwell experiments in the presence of anti-FasL, AMD3100 and anti-perforin at the same time. Based on Figure [5C](#F5){ref-type="fig"}, anti-FasL demonstrated a very specific reactivity as its isotype IgG control showed no non-specific effects in 2 of the 2 experiments performed. In addition, Lehmann et al described a very specific inhibitory effect of anti-perforin antibody on granule-mediated killing of K562 cells when an isotype control antibody was used \[[@R19]\]. B cells were co-cultured with 4T1 at the ratios of 10:1 in transwells as described in the Methods section. Anti-FasL, AMD3100 and anti-perforin were added either separately or together as indicated in Figure [7](#F7){ref-type="fig"}. After 12h, viable 4T1 cells were counted in the lower chambers. As shown in Figure [7A](#F7){ref-type="fig"}, when B cells and tumor cells were co-cultured with cell-cell contact, anti- FasL, anti-perforin and AMD3100 alone each blocked the 4T1 cell killing significantly compared with the control (not blocked). In contrast, without cell contact (Figure [7B](#F7){ref-type="fig"}), only anti-perforin alone blocked the 4T1 cell killing significantly compared with the control (not blocked). In combination, anti-FasL+anti-perforin; anti-perforin+AMD3100; and anti-FasL+AMD3100 blocked the 4T1 cell killing significantly compared with the use of each blocking agent alone with cell-cell contact (Figure [7A](#F7){ref-type="fig"}). By contrast, without cell contact (Figure [7B](#F7){ref-type="fig"}), while anti-FasL+anti-perforin and anti- perforin+AMD3100 blocked the 4T1 cell killing significantly compared with anti-FasL or AMD3100 alone, anti-FasL+AMD3100 demonstrated no significant difference in blocking the 4T1 cell killing compared with the use of anti-FasL or AMD3100 alone. Finally, the anti-FasL+anti-perforin+AMD3100 triple-blocked group showed significantly increased blocking effect on 4T1 cell killing compared with the double blocking using any 2 of the 3 combinations in the cell-cell contact assays (Figure [7C](#F7){ref-type="fig"}). However, without cell contact (Figure [7D](#F7){ref-type="fig"}), this triple-blocked group revealed significantly increased blocking effect on 4T1 cell killing only when compared with the double blocking using anti-FasL+AMD3100. Collectively, these results indicate that effector B cells directly kill tumor cells in cell-cell contact via the Fas/FasL and CXCR4/CXCL12 pathways as well as perforin, while without cell contact, perforin secreted by B cells could also lead to tumor cell cytotoxicity.
{#F7}
DISCUSSION {#s3}
==========
We previously reported that purified TDLN B effector cells could kill tumor cells directly involving the Fas/FasL pathway and are regulated by IL-10 \[[@R12]\]. Like T cells, cytotoxic B cells may need to modulate multiple signaling pathways and are regulated by multiple cytokine/chemokines to mediate optimal cytotoxicity against tumor cells. In the current study, we further defined the mechanisms involved in B cell-mediated killing of tumor cells and found that activated B effector cells killed tumor involving the CXCR4/CXCL12 and perforin pathways as well as the Fas/FasL interaction, and can be augmented by IL-2.
Our data revealed that IL-2 can enhance B cell killer/effector function. It has been well documented that T cells from tumor or tumor draining lymph nodes, after activation *in vitro*, are effective in cancer immunotherapy and could kill tumor cells more effectively *in vivo* in combination with IL-2 administration \[[@R1], [@R3], [@R13], [@R20], [@R21]\]. In this study, we found that effector B cells also express IL-2R and that exogenous IL-2 administration significantly augmented the suppression of 4T1 lung metastasis. These data demonstrate for the first time that IL-2, can also serve as an adjuvant to adoptively transferred TDLN B cells *in vivo* to enhance their antitumor immunity. In this study, we have shown that B cells express IL-2R (CD25). Our *in vitro* experiments demonstrated that IL-2 modulated expression of CD25 on B cells after activation /expansion (A/E) before adoptive transfer; enhanced CXCR4 expression on B cells; augmented CXCL12 production by 4T1 tumor cells, and increased production of perforin by B cells. In addition, we found significantly more binding to 4T1 tumor cells by IgG containing culture supernatants of splenic B cells collected from the hosts subjected to B cell therapy plus IL-2 administration vs. B cell therapy only. Collectively, these data indicate the direct effects of IL-2 on B cells. As a corollary, IL-2 administration in mice may stimulate CD4^+^ T cell proliferation in tumor-bearing mice, and it is well-known that the Th2 cells can provide help for B cell function \[[@R22], [@R23]\]. In this report, we would like to focus on our new findings that IL-2 may act upon B cells directly. Figure [8](#F8){ref-type="fig"} represents a summary diagram showing the role of IL-2 in B cell anti-tumor immunity.
{#F8}
We have previously reported that activated TDLN B cells mediate tumor cytotoxicity, the Fas/FasL pathway \[[@R12]\]. In the current study, we investigated other mechanisms by which activated B cells kill tumor cells. The CXCR4/CXCL12 axis represents a major mechanism in tumor growth and metastasis, and its role in the cancer cell-tumor microenvironment has recently been studied regarding cancer progression and the attraction/activation of leukocytes \[[@R14], [@R24], [@R25]\]. While it was reported that CXCR4 may be expressed on both lymphocytes and cancer cells \[[@R15]\], we found that 4T1 TDLN B cells expressed CXCR4. In addition, the 4T1 tumor cells overexpress CXCL12. Functionally, we found that CXCR4 blockade using a CXCR4-specific inhibitor, AMD3100, significantly blocked B cell-mediated killing of 4T1 tumor. Of note, blockade of FasL and CXCR4 simultaneously inhibited B cell-mediated killing of tumor cells in an additive manner, indicating that both Fas/FasL and CXCR4/CXCL12 axes are involved in the killing of tumor cells by B cells.
Perforin is a cytolytic protein that binds to the tumor cell plasma membrane to form pores leading to its destruction \[[@R26], [@R27]\]. Perforin is found in cytotoxic T cells and natural killer cells. In this study, we found that effector B cells release perforin which can mediate tumor cell killing. Our study also found that effector B cells can kill tumor cells via cell-cell contact utilizing the Fas/FasL and CXCR4/CXCL12 pathways as well as perforin, while without cell contact, perforin secreted by B cells can also lead to tumor cell cytotoxicity. While perforin has been extensively studied in T cells, no previous report has described its role in B cell-mediated cell killing. In addition, we demonstrated that IL-2 could significantly (p\<0.01) enhance perforin production by activated B cells. Together, these data support our conclusion that IL-2 modulates several pathways contributing to the killing of tumor cells by B effector cells.
In summary, exogenous IL-2 administration significantly enhanced the antitumor reactivity of adoptively transferred effector B cells. Purified B effector cells kill tumor cells directly and such killing involves not only the Fas/FasL pathway, but also the CXCR4/CXCL12 pathway and perforin. These studies further identify the mechanisms by which effector B cells mediate antitumor reactivity which can be applied to future cancer immunotherapies.
MATERIALS AND METHODS {#s4}
=====================
Mice {#s4_1}
----
Female BALB/c mice were purchased from the Jackson Laboratories, Bar Harbor, ME. They were maintained in a pathogen-free environment and used at age 7 weeks or older. Principles of laboratory animal care (NIH publication No. 85-23, revised 1985) were followed. The animal protocols were approved by the University of Michigan Laboratory of Animal Medicine.
Murine tumor cells {#s4_2}
------------------
The 4T1 cell line is a mammary carcinoma syngeneic to BALB/c mice (kindly provided by Dr. M. Sabel, University of Michigan). Inoculating 4T1 cells into the mammary fat pad induces the development of spontaneous pulmonary metastases. 4T1 cells were maintained *in vitro* in complete medium (CM).
Tumor draining lymph nodes (TDLNs) {#s4_3}
----------------------------------
In order to induce TDLNs, 1 × 10^6^ 4T1 tumor cells in 0.1 ml PBS were injected subcutaneously (s.c.) into the lower flanks of BALB/c mice. Nine days after 4T1 cell inoculation, the draining inguinal lymph nodes were collected. The TDLNs were processed as previously described \[[@R12]\].
T cell and B cell activation and expansion {#s4_4}
------------------------------------------
CD19^+^ B cells were purified from the TDLN cells or splenocytes using anti-CD19- coupled microbeads and the MACS separator (MiltenyiBiotec. Inc. Auburn, CA). CD3^+^ T cells were purified from the TDLN or splenocytes using anti-CD3-coupled microbeads. B cells were activated with lipopolysaccharide (LPS, Sigma-Aldrich, Atlanta, GA) plus anti-CD40 (FGK45) mAb ascites in complete medium at 37°C with 5% CO~2~ for 3-4 days. T cells were activated with immobilized anti-CD3 and anti-CD28 mAbs in CM containing IL-2 (Prometheus Laboratories, San Diego, CA). The culture supernatants were collected and stored at −20°C for future experiments.
Flow cytometry analysis {#s4_5}
-----------------------
Cell surface expression of CD19, CD25 and CXCR4 and intracellular expression of CXCL12 and cleaved caspase-3 were analyzed by immunofluorescence assay. All fluorescein isothiocyanate (FITC)-or phycoerythrin (PE) -or peridinin chlorophyll protein (PerCP) - conjugated antibodies (FITC rat anti-mouse CD19, PE rat anti-mouse CD19, PE rat anti-mouse CD25, FITC rat anti-mouse CXCR4, and PerCP rat anti-mouse CXCL12) and matched isotype controls were purchased from BD Biosciences (San Jose, CA). Alexa Fluor® 488-conjugated Rabbit anti-mouse/human cleaved caspase-3 antibody and matched control were purchased from R&D system (Minneapolis, MN). To measure intracellular CXCL12 expression, 1 million 4T1 cells were incubated with 2 μl/ml Leukocyte Activation Cocktail (PMA/ionomycin/Golgiplug, BD Pharmingen, San Jose, CA) and 0.67 μl/ml Golgistop (BD Pharmingen, San Jose, CA) in 6- well plate for 4-6 hours at 37°C with 5% CO~2~. After fixation and permeabilization with Fixation/Perm Buffer (eBioscience, Inc., San Diego, CA), the cells were stained with PerCP rat anti-mouse CXCL12 mAb. In transwell assays, after 6 hours, 4T1 cells in the lower chamber were fixed and permeabilizated, followed by staining with anti-mouse cleaved caspase-3 mAb. In cell-cell contact assays, mixed B cells and 4T1 cells were harvested after incubation for 6 hours, surface stained with anti-CD19, and then fixed and permeabilizated by eBioscience Fixation and Permeabilization fixed kit, followed by staining with anti-mouse cleaved caspase-3 mAb. The 4T1 cells were gated from the CD19 negative population. Flow cytometry was performed on a LSRII flow cytometer (BD Biosciences). BD FACSDiva software (version 7.0) was used for all flow cytometry analysis.
Adoptive TDLN B cell therapy of 4T1 cancer {#s4_6}
------------------------------------------
Healthy Balb/c mice were inoculated with 5×10^4^ 4T1 cells into the mammary fat pad to induce spontaneous pulmonary metastases. Fourteen days after tumor inoculation, the tumor- bearing mice were treated with tail vein injection of activated 4T1 TDLN B cells. Commencing on the day of the effector B cell transfer, intraperitoneal (i.p.) injections of IL-2 (40,000 IU) (Prometheus Laboratories, San Diego, CA) were administered in 0.5 ml of PBS and continued twice daily for 8 doses. About 2 weeks after B cell transfer, all mice were sacrificed and lungs were harvested for enumeration of spontaneous pulmonary metastatic nodules. At the same time, spleens were collected for purification of splenic T and B cells as described above.
For 4T1 TDLN B cell cytotoxicity in the presence of anti-FasL antibody (Biolegend Inc., San Diego, CA) and/or AMD3100 (Sigma, Atlanta, GA), the effector B cells were generated as described above and cultured with 4T1 cells with the admixture of 20 μg /ml anti-FasL and/or 2.5 μg /ml AMD3100 to block FasL and/or CXCR4; respectively.
IgG binding by immune supernatant {#s4_7}
---------------------------------
4T1 cells were incubated with the immune supernatants collected from the cultured B cells with equal quantities of IgG followed by incubation with the 2^nd^ antibody: FITC-conjugated anti-mouse IgG. The concentration of total IgG was measured by ELISA quantitation kits (Bethyl laboratories, Montgomery, TX). The binding of supernatant antibody to 4T1 cells was assessed using flow cytometry.
Chemotaxis assay {#s4_8}
----------------
In transwell experiments, 5×10^5^ 4T1 TDLN B cells in 200 μl were added to the upper chamber of a transwell (insert pore size, 5 μm; Millipore, Billerica, MA). CXCL12 (100 ng/ml, R&D Systems, Minneapolis, MN) was added to the lower chamber in a volume of 900 μl. After 2, 4, 6, and 8h of incubation at 37°C, cells that migrated into the lower chamber were enumerated using a hemocytometer. The chemotactic index (%) represents the ratio of cells migrating in the presence of chemokines to the number of cells migrating in the absence of cytokines at each time point analyzed in triplicate.
LDH cytotoxicity assay {#s4_9}
----------------------
Cell cytotoxicity was assessed by measuring the release of cytoplasmic lactate dehydrogenase (LDH) into cell culture supernatants according to the manufacturer\'s protocol (CytoTox 96 Non-Radioactive Cytotoxicity Assay, Promega, Madison, WI). For TDLN B cell cytotoxicity, effector B cells were generated from 4T1 TDLN B cells using LPS/anti-CD40 as described above. Target cells were plated in triplicates in a 96-well U-bottom tissue culture plate (5000 cells/well) and co-incubated with TDLN B cells at effector to target cell ratios of 1:1, 3:1, 10:1 and 30:1. After 12 hours of incubation, cells were centrifuged and 50 μl supernatant from each well was transferred to a fresh 96-well plate, 50 μl of the substrate mix was added and incubated at room temperature in the dark for 15 to 30 min. Before LDH measurement, 50 μl of stop solution was added to each well. Maximal release of LDH was performed by incubating the target cells with Lysis Solution (Promega, Madison, WI). Target cells without effector cells were used as a spontaneous release control. Absorbance was measured at 490 nm using a 96-well plate reader.
Perforin release assessment {#s4_10}
---------------------------
T cells and B cells purified from 4T1 TDLNs were activated as described above. Culture supernatants were then collected for perforin release detection using ELISA kits (Cloud-Clone Corp, Houston, TX).
The cytotoxity and blockage assay {#s4_11}
---------------------------------
In cell-cell contact assays, 4T1 cells were plated in triplicates in a 12-well tissue culture plate (1×10^5^ cells/well, 1ml) and co-incubated with 4T1 TDLN B cells (1×10^6^ cells/well, 1ml). In 0.4 μm transwell assay, 1×10^6^ 4T1 TDLN B cells were added to the upper chamber of a transwell, and 1×10^5^ 4T1 cells were added to the lower chamber. Anti-FasL(20μg/ml), anti-perforin (5μg/ml) and AMD3100 (2.5μg/ml) were added either alone or in combination as indicated in the experiments. After 12 hours of incubation, trypan blue staining was used to count the viable 4T1 cells either in the mixed co-culture or in the lower chamber of the transwell. The suspended B cells in the medium of the lower chambers were removed, and the well was washed by PBS for 3 times. B cells were distinguished from 4T1 cells by their significant different morphological characteristics under the microscope.
Statistical analysis {#s4_12}
--------------------
Statistics were analyzed using the mean ± standard error. The significance of differences in numbers of metastatic nodules and cell lysis was determined using Student\'s t-test. *P*\<0.05 was considered statistically significant between the experimental groups. For comparison of more than 2 groups, statistical significance was determined using a one-way ANOVA followed by a Bonferroni multiple-group comparison test. Statistical analyses were performed with Graph- Pad Prism 6.0.
**CONFLICTS OF INTEREST**
There are no financial or commercial conflicts of interest.
**GRANT SUPPORT**
This work was supported in part by the Gillson Longenbaugh Foundation; National Natural Science Foundation of China (30971112), National Natural Science Foundation of China (81072170) and The First Plan of Science and Technology Program of Hefei City (\#25 Technology Bureau of Hefei 2011).
TDLN
: tumor-draining lymph node
CM
: complete medium
s.c.
: subcutaneous
i.v.
: intravenously
i.p.
: intraperitoneal
[^1]: Joint senior authors
| {
"pile_set_name": "PubMed Central"
} |
Background {#Sec1}
==========
Cervical cancer is one of the common cancers in women worldwide. According to the World Health Organization, the incidence of cervical cancer ranks first in 28 countries and it is the leading cause of female deaths in 42 countries and it is estimated that 570,000 new cases and 311,000 deaths may occur worldwide in 2019 \[[@CR1]\]. In China, cervical cancer ranked 8th among the top 10 malignant tumors in 2015, accounting for 2.83% of all cancers. The incidence of cervical cancer in rural areas is higher than that in cities (6.34 and 2.68%, respectively) and the average age of onset of cervical cancer is 51 years, primarily in the 40--50 years age range. Cervical cancer rarely occurs before 20 years of age \[[@CR2]\].
Cervical cancer includes squamous cell carcinoma (SCC), adenocarcinoma (AC), adenosquamous carcinoma (ASC), neuroendocrine cancer, undifferentiated cancer, and other pathological types. SCC accounts for approximately 70% and AC accounts for approximately 25% cases \[[@CR3]\]. High-risk human papilloma virus (HPV) is a significant risk factor for cervical cancer. At present, more than 100 different subtypes of HPV have been found and 54 types able to infect the reproductive tract mucosa. High-risk subtypes include HPV16, 18, 31, 33, 35, 39, and 45. HPV16 and 18 subtypes are most closely related to cervical cancer \[[@CR4]\]. In recent years, HPV vaccines and screenings have reduced the incidence of cervical cancer in developed countries. However, the age-adjusted incidence rates of AC increased by 0.5 to 3% per annum in Europe \[[@CR5]\]. It is still controversial as to whether the pathological types of cervical cancer have an impact on the prognosis of patients. Some studies have shown that SCC and AC have the same overall survival (OS) \[[@CR6], [@CR7]\], while some data suggest that patients with AC of the cervix had poorer OS and disease-free survival (DFS) than patients with SCC \[[@CR3], [@CR8]--[@CR10]\].
In the current National Comprehensive Cancer Network (NCCN) guidelines, the standard treatment methods revealed no differences between SCC and AC of cervical cancer. However, the prognosis and clinical characteristics of SCC cervical cancer differs from that of AC cervical cancer. It has been reported that patients with AC tend to be younger and Caucasian, and they are more commonly diagnosed in the early stages of disease \[[@CR11]\].
Some studies have shown that, when compared with radiotherapy alone, concurrent radiotherapy combined with platinum-based chemotherapy can improve survival \[[@CR12]--[@CR17]\]. Since 1999, concurrent chemoradiotherapy has become the standard treatment approach for locally advanced cervical cancer regardless of the histological subtype of the disease. In recent years, some studies have suggested that total hysterectomy after neoadjuvant chemotherapy can improve survival outcomes of cervical cancer \[[@CR18], [@CR19]\]. This may be because neoadjuvant chemotherapy has the potential to eradicate micrometastases and could reduce systemic failures, in addition to facilitating local control by surgical resection. However, other studies have shown that the effect of neoadjuvant chemotherapy plus surgery does not show superiority compared to concurrent chemoradiotherapy \[[@CR20], [@CR21]\].
Therefore, we conducted this retrospectively study to analyze the data of patients with cervical cancer in Surveillance, Epidemiology and End Results (SEER) database to compare the survival outcomes of patients with SCC and those with AC among locally advanced cervical cancer patients as well as factors that are related to the prognosis of cervical cancer.
Methods {#Sec2}
=======
Patients {#Sec3}
--------
Information on patients diagnosed with cervical cancer between 2010 and 2015 was extracted from the SEER-18 Regs Custom database using SEER\*Stat software, version 8.3.6. We limited our sample to the period 2010--2015, as the information about some variables only are available from 2010. We included cervical cancer patients in our study using the following parameters: only one primary malignancy; International Federation of Gynecology and Obstetrics (FIGO,2009 version) stages IB2 to IVA; pathological biopsy confirmed SCC and AC; no distant metastases; aged 20 to 69 years; active follow-up with complete date; and complete clinicopathological information. Since the patient information in the SEER database is deidentified and publicly available, our study was exempt from Institutional Review Board approval. This study was conducted in concordance with the Helsinki Declaration.
Variables for each patient included: race, age, TNM stage, histology grade, T stage, N stage, pathological subtype, treatment strategy, survival time, and vital status. We restaged TNM stage according to the FIGO classification (2009 version). American Joint Committee on Cancer 7th edition (AJCC7) provided updated guidelines on the grading of cervical cancer. Patients were treated either with radiation alone, radiotherapy combined with chemotherapy, and/or surgery. Radiotherapy primarily included external radiation, brachytherapy, or external radiation combined with brachytherapy. The primary endpoint of this study was OS.
Statistics analyses {#Sec4}
-------------------
The baseline characteristics between cervical cancer patients in the AC and SCC groups were compared with chi-square (χ^2^) or continuity correction tests. A Cox proportional hazards model was used to perform univariate and multivariate analyses of the various factors affecting OS. As the baseline characteristics were different between two groups, propensity score matching (PSM) was performed with a ratio of 1:1. The matching covariates included race, age, TNM stage, histology grade, radiation sequence with surgery, radiotherapy and chemotherapy. Comparisons between the SCC and AC groups was evaluated by a two-side log-rank test before and after matching and OS was estimated using the Kaplan-Meier method. All analyses were performed using R statistical software (version 3.6.1, package includes survminer, rms, foreign, tableone, broom, matchlt, R Foundation) for statistical analysis. Hazard ratio (HR) and 95% confidence intervals (CIs) were compared between two groups. Two-sided *p* \< 0.05 was defined as statistically significant.
Results and conclusions {#Sec5}
=======================
Patient baseline characteristics {#Sec6}
--------------------------------
There were 4131 eligible patients who were diagnosed with stage IB2 to IVA cervical cancer between 2010 to 2015, including 3385 (81.9%) patients in the SCC group and 746 (18.1%) patients in the AC group. The detailed characteristics of patients in the SCC and AC groups before and after propensity score matching are shown in Table [1](#Tab1){ref-type="table"}. There were significantly more patients between the ages 40 and 49 in the AC group than in the SCC group (28.5% in the SCC group and 33.9% in the AC group, *P* = 0.02). The SCC and AC groups differed significantly when comparing the distribution of well differentiated (grade I) and undifferentiated (grade IV) cancer (4.2 and 1.5% in SCC group, 18.5 and 4.0% in AC group, *P* \< 0.001). There were more patients with FIGO stages IIB - IVA in the SCC group (82.8% in SCC group, 71.9% in AC group, P \< 0.001). The median survival time of patients in the SCC and AC groups was 33.6 (range, 0--83 months) and 35.8 months (range, 0--83 months), respectively (*P* = 0.013). Table 1Characteristics of cervical cancer patients with AC and SCC before and after matchingCharacteristicsBefore matchingAfter matchingAC (*n* = 746)SCC (*n* = 3385)*P* valueAC(n = 676)SCC(*n* = 676)P valueage (%)0.020.872 \< 40158 (21.2)840 (24.8)146 (21.6)143 (21.2) 40--49253 (33.9)965 (28.5)222 (32.8)236 (34.9) 50--59199 (26.7)947 (28.0)183 (27.1)179 (26.5) 60--69136 (18.2)633 (18.7)125 (18.5)118 (17.5)race (%)\< 0.0010.763 Black59 (7.9)568 (16.8)58 (8.6)62 (9.2) White596 (79.9)2476 (73.1)538 (79.6)527 (78.0) Other91 (12.2)341 (10.1)80 (11.8)87 (12.9)grade (%)\< 0.0010.928 I138 (18.5)142 (4.2)80 (11.8)77 (11.4) II313 (42.0)1610 (47.6)306 (45.3)297 (43.9) III265 (35.5)1583 (46.8)264 (39.1)276 (40.8) IV30 (4.0)50 (1.5)26 (3.8)26 (3.8)FIGO stage (%)\< 0.0010.923 Ib2142 (19.0)366 (10.8)118 (17.5)121 (17.9) IIa125 (3.4)81 (2.4)21 (3.1)14 (2.1) IIa242 (5.6)135 (4.0)35 (5.2)38 (5.6) IIb160 (21.4)732 (21.6)142 (21.0)143 (21.2) IIIa15 (2.0)75 (2.2)14 (2.1)11 (1.6) IIIb346 (46.4)1837 (54.3)331 (49.0)333 (49.3) IVa16 (2.1)159 (4.7)15 (2.2)16 (2.4)sequence (%)\< 0.0010.974 No surgery318 (42.6)2011 (59.4)308 (45.6)302 (44.7) RPTS55 (7.4)133 (3.9)45 (6.7)48 (7.1) RAS365 (48.9)1218 (36.0)315 (46.6)317 (46.9) RBAS8 (1.1)23 (0.7)8 (1.2)9 (1.3)Radiation (%)0.5270.25 Beam radiation328 (44.0)1487 (43.9)294 (43.5)323 (47.8) brachytherapy82 (11.0)328 (9.7)70 (10.4)70 (10.4) BRB336 (45.0)1570 (46.4)312 (46.2)283 (41.9)Chemotherapy (%)0.0170.865 Yes652(87.4)3060(90.4)596(88.2)599(88.6) No94 (12.6)325 (9.6)80 (11.8)77 (11.4)status (%)0.055 alive535 (71.7)2303 (68.0)480 (71.0)486 (71.9)0.763dead211(28.3)1082(32.0)196(29.0)190(28.1)time (mean (SD))35.8(21.4)33.6(21.7)0.01335.0(21.1)33.7 (21.5)0.269AC adenocarcinoma, SCC Squamous cell carcinoma, FIGO the International Federation of Gynecology and Obstetrics, RPTS Radiation prior to surgery, RAS Radiation after surgery, RBAS Radiation before and after surgery, BRB Combination of beam with brachytherapy. Grade I Well differentiated,gradeII Moderately differentiated,gradeIII Poorly differentiated,gradeIV Undifferentiated;SD Standard Deviation
Survivals of patients in SCC and AC groups {#Sec7}
------------------------------------------
To balance the baseline characteristics, 676 cervical patients with AC were matched 1:1 to patients with SCC. The baseline characteristics were similar between two groups after matching (Table [1](#Tab1){ref-type="table"}). The OS curves of patients with SCC and AC are shown in Fig. [1](#Fig1){ref-type="fig"}. The 5-year OS of patients in the SCC and AC groups were 16 and 18% (*P* = 0.67) before matching and were 16 and 17% after matching (*P* = 0.53). Fig. 1The overall survival (OS) of patients with squamous cell carcinoma (SCC) and adenocarcinoma (AC) (A) before propensity score matching (PSM) (B) after PSM
Univariate and multivariate analysis {#Sec8}
------------------------------------
In Table [2](#Tab2){ref-type="table"}, we present the results of the univariate analysis showing that histology was not a significant factor for OS. The univariate COX hazard ratio (HR) was 1.02 (95% confidence interval (CI) 0.93--1.12, *P* = 0.69). As shown in Table [2](#Tab2){ref-type="table"}, the primary univariate factors affecting OS were: FIGO stage IVA (HR = 0.74, 95% CI 0.57--0.98, *p* = 0.03), chemotherapy (HR 0.85, 95% CI 0.75--0.97, *p* = 0.01), race (HR = 1.20, 95% CI 1.03--1.39, *p* = 0.02) and external radiation combined with brachytherapy (HR 0.93, 95% CI 0.86--1.00, *p* = 0.06). Multivariate analysis found the same associated factors (Fig. [2](#Fig2){ref-type="fig"}). Table 2Univariate analysis of factors associated with OSVariablesUnivariate (OS)HR (95%CI)P valuehistology AC1.00 (reference) SCC1.02 (0.93--1.12)0.69age (%) \< 401.00 (reference) 40--491.03 (0.93--1.13)0.61 50--590.97 (0.88--1.08)0.59 60--691.01 (0.90--1.14)0.81race (%) Black1.00 (reference) White1.06 (0.95--1.18)0.28 Other1.20 (1.03--1.39)0.02grade (%) I1.00 (reference) II1.07 (0.93--1.24)0.36 III1.00 (0.86--1.15)0.95 IV1.12 (0.82--1.53)0.47FIGO stage (%) Ib21.00 (reference) IIa11.02 (0.81--1.27)0.88 IIa21.17 (0.96--1.43)0.12 IIb0.98 (0.87--1.10)0.72 IIIa1.07 (0.80--1.42)0.64 IIIb1.00 (0.90--1.11)0.94 IVa0.74 (0.57--0.98)0.03sequence (%) No surgery1.00 (reference) RPTS0.88 (0.73--1.05)0.16 RAS0.93 (0.87--1.01)0.10 RBAS1.11 (0.72--1.73)0.63Radiation (%) Beam radiation1.00 (reference) brachytherapy0.98 (0.86--1.11)0.70 BRB0.93 (0.86--1.00)0.06Chemotherapy Yes1.00 (reference) No0.85 (0.75--0.97)0.01AC adenocarcinoma, SCC Squamous cell carcinoma, FIGO the International Federation of Gynecology and Obstetrics, RPTS Radiation prior to surgery, RAS Radiation after surgery, RBAS Radiation before and after surgery, BRB Combination of beam with brachytherapy, grade I Well differentiated,gradeII Moderately differentiated,gradeIII Poorly differentiated,gradeIV Undifferentiated;CI indicates confidence interval, HR hazard ratio
Subgroup analysis {#Sec9}
-----------------
For these analyses, we stratified patients by treatment and performed 1:1 matching on pathology. The detailed characteristics of subgroups after propensity score matching are shown in Table [3](#Tab3){ref-type="table"} and Table [4](#Tab4){ref-type="table"} .After matching, 15 cervical cancer patients with AC were matched with 15 patients with SCC (Fig. [3](#Fig3){ref-type="fig"}a) in the radiation only group. We showed that the OS of patients with SCC was significantly worse than that of patients with AC (*P* = 0.015). In the radiation combined with chemotherapy subgroup, we defined 282 matched pairs. The OS of patients receiving radiation combined with chemotherapy did not differ significantly by pathology (*P* = 0.62) (Fig. [3](#Fig3){ref-type="fig"}b). Figure [3](#Fig3){ref-type="fig"}c shows the OS comparing 49 matched subgroups treated with radiotherapy combined with surgery. Within this subgroup, the median survival time was shorter for AC patients, although these results were not statistically significant. Finally, we analyzed radiotherapy combined with chemotherapy plus surgery in 305 pairs and found that the OS of SCC and AC was very similar (*P* = 0.92) (Fig. [3](#Fig3){ref-type="fig"}d). Table 3characteristics of radiotherapy alone group and radiotherapy combined with chemotherapy group after matchingCharacteristicsRadiotherapy aloneCTRTAC (*n* = 15)SCC(n = 15)P valueAC(*n* = 282)SCC(n = 282)P valueage (%)0.8330.901 \< 402 (13.3)2 (13.3)41 (14.5)41 (14.5) 40--495 (33.3)6 (40.1)98 (34.8)100 (35.5) 50--594(26.7)2 (13.3)79 (28.0)84 (29.8) 60--694(26.7)5 (33.3)64 (22.7)57 (20.2)race (%)0.6480.969 Black2 (13.3)4(26.7)27 (9.6)27 (9.6) White13 (86.7)11 (73.3)36 (12.8)38 (13.5) Other0 (0.0)0 (0.0)219 (77.6)217 (76.9)grade (%)1.0000.991 I0 (0.0)0 (0.0)33 (11.7)34 (12.1) II4(26.7)5 (33.3)116 (41.1)114 (40.4) III11 (73.3)10 (66.7)124 (44.0)126 (44.7) IV0 (0.0)0 (0.0)9 (3.2)8 (2.8)FIGO stage (%)0.9700.012 Ib22 (13.3)1 (6.7)45 (16.0)25 (8.9) IIa11 (6.7)1 (6.7)5 (1.8)5 (1.8) IIa20 (0.0)1 (6.7)11 (3.9)7 (2.5) IIb2 (13.3)2 (13.3)77 (27.3)68 (24.1) IIIa1 (6.7)1 (6.7)10 (3.5)9 (3.2) IIIb7 (46.7)7 (46.7)127 (45.0)146 (51.7) IVa2 (13.3)2 (13.3)7 (2.5)22 (7.8)Radiation (%)0.7660.545 Beam radiation11 (73.3)11 (73.3)100 (35.5)91 (32.3) brachytherapy2 (13.3)3 (20.0)146 (51.7)159 (56.4) BRB2 (13.3)1 (6.7)36 (12.8)32 (11.3)status (%)1.0000.542 alive9 (60.0)9 (60.0)181 (64.2)173 (61.3) dead6(40.0)6(40.0)101(35.8)109(38.7)time(mean(SD))40.4 (32.5)26.1(18.6)0.15033.5 (21.6)31.8 (22.5)0.353AC adenocarcinoma, SCC Squamous cell carcinoma, FIGO the International Federation of Gynecology and Obstetrics, BRB Combination of beam with brachytherapy. CTRT radiotherapy combined with chemotherapy, grade I Well differentiated,gradeII Moderately differentiated,gradeIII Poorly differentiated,gradeIV Undifferentiated;SD Standard DeviationTable 4characteristics of radiotherapy plus surgery group and radiotherapy combined with chemotherapy and surgery group after matchingCharacteristicsradiotherapy plus surgeryCTRT plus surgeryAC(*n* = 49)SCC(n = 49)P valueAC(*n* = 305)SCC(n = 305)P valueage (%)0.7500.291 \< 409 (18.4)9 (18.4)84 (27.5)78 (25.6) 40--4917 (34.7)14 (28.6)101 (33.1)93 (30.5) 50--5916 (32.7)15 (30.6)74 (24.3)95 (31.1) 60--697 (14.3)11 (22.4)46 (15.1)39 (12.8)race (%)1.0000.56 Black3 (6.1)3 (6.1)17 (5.6)16 (5.2) White2 (4.1)2 (4.1)37 (12.1)29 (9.5) Other44 (89.8)44 (89.8)251 (82.3)260 (85.2)grade (%)0.5020.927 I1 (2.0)1 (2.0)33 (10.8)35 (11.5) II30 (61.2)25 (51.0)156 (51.1)153 (50.2) III17 (34.7)23 (46.9)109 (35.7)112 (36.7) IV1 (2.0)0 (0.0)7 (2.3)5 (1.6)FIGO stage (%)0.3780.066 Ib218 (36.7)17 (34.7)54 (17.7)34 (11.1) IIa13 (6.1)0 (0.0)11 (3.6)7 (2.3) IIa24 (8.2)3 (6.1)19 (6.2)9 (3.0) IIb5 (10.2)12 (24.5)57 (18.7)60 (19.7) IIIa1 (2.0)1 (2.0)2 (0.7)2 (0.7) IIIb16 (32.7)15 (30.6)159 (52.1)190 (62.3) IVa2 (4.1)1 (2.0)3 (1.0)3 (1.0)Radiation (%)0.6590.988 Beam radiation34 (69.4)34 (69.4)145 (47.5)144 (47.2) brachytherapy11 (22.4)13 (26.5)136 (44.6)136 (44.6) BRB4 (8.2)2 (4.1)24 (7.9)25 (8.2)status (%)1.000 alive40 (81.6)40 (81.6)234 (76.7)236 (77.4)0.923 dead9(18.4)9(18.4)71(23.3)69(22.6)time (mean (SD))37.9 (22.0)39.4(22.1)0.74937.3 (20.6)37.4(21.0)0.939AC adenocarcinoma, SCC Squamous cell carcinoma, FIGO the International Federation of Gynecology and Obstetrics, BRB Combination of beam with brachytherapy. CTRT radiotherapy combined with chemotherapy, grade I Well differentiated,gradeII Moderately differentiated,gradeIII Poorly differentiated,gradeIV Undifferentiated;SD Standard DeviationFig. 2Multivariate analysis of factors associated with OSFig. 3Treated with radiotherapy alone the overall survival (OS) of patients with squamous cell carcinoma (SCC) and adenocarcinoma (AC) (A); treated with radiotherapy and chemotherapy the OS of patients with SCC and AC(B), treated with radiotherapy and surgery the OS of patients with SCC and AC(C); treated with radiotherapy, surgery and chemotherapy the OS of patients with SCC and AC(D)
Discussion {#Sec10}
==========
Adenocarcinoma is a small pathological subtype in cervical cancer. In recent years, while the incidence of cervical cancer has generally declined \[[@CR22]\], the incidence of AC has increased \[[@CR5]\]. However, using the same treatment as for SCC may not be appropriate for AC patients \[[@CR23]\]. According to NCCN guidelines, the standard treatment for locally advanced cervical cancer is radiotherapy and concomitant chemotherapy using cisplatin, a protocol that was developed in SCC patients. As a result, there has been increasing interest in determining optimum treatments for AC. There are still considerable debates about the prognostic patterns between patients of AC and SCC as well as the proper treatment based on subgroup. ASC and AC cancers together represented 21.3, 22.9 and 24.1% of cases for the years 1988--1993, 1994--1999 and 2000--2005, respectively \[[@CR10]\]. A study based on the SEER database by Vijaya Galic, et al. showed that the incidence of AC and ASC was increasing 1--2% every 6 years \[[@CR9]\].
With this increasing pattern, researchers are increasingly interested in cervical AC. The literature shows that there is still a significant debate regarding the prognosis of patients with AC and SCC of the cervix. Some studies have shown that SCC of the cervix has a similar OS as that of AC \[[@CR6]\]. Other studies have shown that AC is less sensitive to radiotherapy than SCC of the cervix, especially in patients with poor prognostic factors \[[@CR8]\]. Therefore, in this study, we collected cervical cancer data from the SEER database to identify predictive factors associated with cervical cancer, and survival outcomes in patients with AC and SCC.
In 2011, a previous study by Katanyoo, et al. included a total of 423 patients with stages IIB--IVA and showed that the 5-year OS rates of AC and SCC were 59.9 and 61.7% (*p* = 0.191), respectively. When all prognostic factors were adjusted, clinical staging was the only factor that influenced OS \[[@CR6]\]. In 2014, Rose, et al. performed a retrospective analysis of 1671 patients with cervical cancer and suggested that when treated with radiotherapy alone, the OS of AC was worse than that of SCC. When the treatment was radiotherapy and concomitant platinum-based chemotherapy, the OS of patients is similar \[[@CR11]\]. In our study, we saw similar patterns. However, in 2018, Hu, et al. compared the treatment outcomes between SCC and AC of the cervix after definitive radiotherapy or concurrent chemoradiotherapy and showed that, when compared with SCC of the cervix, AC affected younger women, was more aggressive, and had more para-aortic metastatic lymph nodes. The 3-year disease-free survival, OS, local control rate, and distant control rate were worse for AC when compared to SCC \[[@CR3]\]. In the same year, a propensity score matching study by Yin, et al. enrolled 181 locally advanced cervical cancer patients who were treated with intensity modulated radiotherapy/volumetric modulated arc therapy and concurrent chemotherapy. The results showed that after a 1:1 ratio PSM, the 5-year OS, DFS, locoregional failure-free survival, and distant metastasis-free survival in the locally advanced cervical cancer of AC were 46.0, 43.3, 70.0, and 45.4%, respectively. These results were significantly lower than the corresponding rates of 90.0, 75.8, 96.6, and 78.8% in the matched locally advanced cervical cancer patients of SCC, respectively (*p* \< 0.05) \[[@CR24]\]. These studies showed a lower OS of AC in cervical cancer.
However, in the present study, we found that histology is not an independent factor of OS. Before PSM, the 3.5-year OS rates in the AC group were 44 and 18%, and the rates were 39 and 16% in the SCC group, respectively (*p* = 0.67). After PSM, the 3-year and 5-year OS of the AC group was 43 and 17%, compared to 40 and 16% in the SCC group (*P* = 0.53). There was no significant difference in OS between AC group and SCC group, regardless of matching status, leading to different conclusions from the previous studies. When comparing these studies, we did not exclude the impact of different treatments on the prognosis of the two pathological types. To make the results stronger, we divided the treatments into four groups and analyzed the difference between the AC and SCC groups.
In the subgroup analysis, when treated with radiotherapy alone, we found that the OS of patients with SCC was worse than that of patients with AC (*P* = 0.015), in contrast with other studies. The number of people in the analysis was small and the power to detect the difference between AC and SCC may have been too low. When we compared the treated with radiotherapy combined with chemotherapy or radiotherapy combined with chemotherapy plus surgery, the OS of patients with SCC and AC was similar (*P* = 0.62, *P* = 0.92). When treated with radiotherapy combined with surgery, the median survival time and the 3-year and 5-year OS of patients with AC was worse than patients with SCC. However there was no significant difference between two groups (*P* = 0.71). In univariable and multivariable analysis, the factors which had the strongest association with survival outcomes was FIGO stage, chemotherapy and external radiation combined with brachytherapy. The difference of study period, number of patients, criteria of enrolled patients, statistical methods and treatment modalities may help us understand the difference of our results when compared to the literature.
This study has some certain limitations. First, we used retrospective studies and we may have unintentional selectivity bias. Second, because the SEER database only recorded the sequence of radiotherapy and surgery, it did not record the sequence of surgery and chemotherapy, and we could not judge whether neoadjuvant therapy or adjuvant therapy is used in the treatment of cervical cancer patients. There was no recorded objective response time and disease-free survival time after treatment; therefore, we could not compare the difference between SCC and AC. In the future, we may need prospective clinical studies to confirm our conclusions.
Conclusions {#Sec11}
===========
Compared with SCC of the cervix, AC has similar OS when treated with radiotherapy combined with chemotherapy and/or surgery but with better OS when treated with radiation alone. Histology is not an independent factor of OS. The factors affecting the prognosis of cervical cancer are the FIGO stage, chemotherapy, race and external radiation combined with brachytherapy.
**Publisher's Note**
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Tian Tian and Xing Gong contributed equally to this work.
Thanks for The Third Affiliated Hospital of Kunming Medical University supporting the writing of this manuscript.
Conception and design: Tian Tian and Xing Gong; (II) Administrative support: Yiqin Ai; (III) Provision of study materials or patients: Tian Tian and Xing Gong; (IV) Collection and assembly of data: Tian Tian and Xing Gong; (V) Data analysis and interpretation: All authors; (VI) Manuscript writing: All authors; (VII) Final approval of manuscript: All authors.
None.
The datasets generated and/or analysed during the current study are available in the Surveillance, Epidemiology, and End Results (SEER) database repository, <https://seer.cancer.gov>.
Informed consent was obtained from all individual participants included in the study. Since the patient information in the SEER database is deidentified and publicly available, our study was exempt from Institutional Review Board approval. This study was conducted in concordance with the Helsinki Declaration.
Not applicable.
The authors declare that they have no competing interests.
| {
"pile_set_name": "PubMed Central"
} |
Introduction
============
Self-ligating brackets have been introducted for their advantages in orthodontic treatment: they are able to reduce unwanted friction ([@B1]), eliminate the requirement for elastomeric ligatures ([@B2]), insure more certain archwire engagement ([@B3]) and offer faster archwire removal and ligation ([@B4]). When orthodontists and surgeons collaborate in the exposure and orthodontic alignment of unerupted teeth, it could be useful to bond self-ligating bracket also onto ectopic teeth in order to reduce friction and speed orthodontic movement. In these cases it is difficult to maintain ideal working conditions and blood contamination during bonding can occur ([@B5]). Bonding of orthodontic brackets with composite resin adhesives requires a dry field of operation. Previous studies evaluating the effect of blood contamination on the bond strengths of ligh tcured composites showed a significant reduction in bond strength values ([@B5]-[@B10]). In fact when blood affects the bond, etched enamel becomes wet, most of the porosities become plugged, and resin penetration is impaired; this results in resin tags of insufficient numbers and lengths ([@B11]).
Nowadays in literature there are no published studies that compared shear bond strength of self ligating orthodontic bracket under bood contamination.
Accordingly, the aim of the present investigation was to measure and compare shear bond strength and adhesive remnant index score of three different self-ligating orthodontic brackets onto dry and blood-contaminated enamel. The null hypothesis of the study was that there is no significant difference in shear bond strength values and debond locations among the various groups.
Material and Methods
====================
Two hundred and forty freshly permanent extracted bovine mandibular incisors were collected from a local slaughterhouse and stored in a solution of 0.1% (wt/vol) thymol. The criteria for tooth selection included intact buccal enamel with no cracks caused by extraction and no caries. The teeth were cleansed of soft tissue and embedded in cold-curing, fast-setting acrylic (Leocryl, Leone, Sesto Fiorentino, Italy). Metal rings (15-mm diameter) were filled with the acrylic resin and allowed to cure, thus encasing each specimen while allowing the buccal surface of enamel to be exposed. Each tooth was oriented so that its labial surface was parallel to the shearing force. Teeth were randomly divided in 12 groups of 20 specimens.
Three different orthodontic self-ligating maxillary central incisor brackets were tested: Smart Clip (3M, Monrovia, Calif), Quick (Forestadent, Pforzheim, Germany), Damon 3MX (Ormco, Glendora, Calif). Brackets were tested under 4 different enamel surface conditions: ([@B1]) dry, ([@B2]) blood contamination before priming, ([@B3]) blood contamination after priming, ([@B4]) blood contamination before and after priming.
Before bonding, the labial surface of each incisor was cleaned for 10 seconds with a mixture of water and fluoride-free pumice in a rubber polishing cup with a low-speed handpiece. The enamel surface was rinsed with water to remove pumice or debris and then dried with an oil-free air stream.
Bonding procedures are described in ([Table 1](#T1){ref-type="table"}). Teeth were etched with 37% phosphoric acid gel (3M Unitek, Monrovia, California) for 30 seconds, followed by thorough washing and drying. A thin layer of primer (Ortho Solo; Ormco, Glendora, California) was applied on the etched enamel, and then the brackets were bonded with a resin (Transbond XT, 3M Unitek, Monrovia, California) near the center of the facial surface of the teeth with sufficient pressure to express excess adhesive, which was removed from the margins of the bracket base with a scaler before polymerization. To achieve reproducible conditions, the teeth treated under moistened conditions were contaminated with fresh human blood from 1 female donor; the blood was applied with a brush onto the labial surfaces until they were totally contaminated.
Table 1Bonding procedures for the different enamel surface conditions.
Brackets were then light-cured with a visible light-curing unit (Ortholux XT, 3M Unitek, Monrovia, California) for 10 seconds on the mesial side of the bracket and for 10 seconds on the distal side (total cure time 20 seconds). After bonding, all samples were stored in distilled water at room temperature for 24 hours and then tested in a shear mode with an universal testing machine (Model 3343, Instron, Canton, Massachussetts). Specimens were secured in the lower jaw of the machine so that the bonded bracket base was parallel to the shear force direction.
Specimens were stressed in an occlusogingival direction at a crosshead speed of 1 mm per minute, as in previous studies ([@B12]-[@B14]). The maximum load necessary to debond or initiate bracket fracture was recorded in newtons and then converted into MPa as a ratio of newtons to surface area of the bracket. After bond failure, the bracket bases and the enamel surfaces were examined under an optical microscope (Stereomicroscope SR, Zeiss, Oberkochen, Germany) at 10x magnification. The adhesive remnant index (ARI) was used to assess the amount of adhesive left on the enamel surface ([@B15]).
This scale ranges from 0 to 3. A score of 0 indicates no adhesive remaining on the tooth in the bonding area; 1 indicates less than half of the adhesive remaining on the tooth; 2 indicates more than half of the adhesive remaining on the tooth; and 3 indicates all adhesive remaining on the tooth. The ARI scores were used as a more complex method of defining bond failure site among the enamel, the adhesive, and the bracket base.
Statistical analysis was performed with Stata 11.0 software (Stata, College Station, Texas). Descriptive statistics, including the mean, standard deviation, median, and minimum and maximum values were calculated for all groups.
An analysis of variance (ANOVA) test was applied to determine whether significant differences in debond values existed among the groups. The Tukey test was used as post-hoc. The chi-square test was used to determine significant differences in the ARI scores among the different groups. Significance for all statistical tests was predetermined at P \< 0.05.
Results
=======
Descriptive statistics for the shear bond strength (MPa) of the different brackets are illustrated in ([Table 2](#T2){ref-type="table"}, Fig. [1](#F1){ref-type="fig"}). The normality of the data was calculated using the Kolmogorov-Smirnov test. The analysis of variance showed the presence of significant differences among the various groups (P \< 0.05). Post-hoc test showed that brackets Smart Clip, Quick and Damon 3MX bonded onto dry enamel (groups 1, 2 and 3) presented the highest shear bond strength (P \< 0.001). The three different brackets bonded under blood contamination before priming (groups 4, 5 and 6) or after priming (groups 7, 8 and 9) showed significantly lower shear bond strength (P \< 0.01) and no significant differences among them (P \> 0,05). Finally brackets bonded under blood contamination before and after priming (groups 10, 11 and 12) had lower bond strengths (P \< 0.05), and showed no significant difference among them (P \> 0.05).
Table 2Descriptive statistics (in MPa) of shear bond strengths of the 12 groups tested (each group consisted of 20 specimens).
Figure 1Mean shear bond strengths (MPa) of the three different brackets under the four different testing conditions (no contamination; blood contamination before priming; blood contamination after priming; blood contamination before and after priming).
The results of ARI Scores are illustrated in ([Table 3](#T3){ref-type="table"}). The chi-square test reported a higher frequency of ARI score of "2" for uncontaminated enamel groups (1,2 and 3) (P \< 0.05) and exhibited no significant difference among them (P \> 0.05). When blood contamination occurred before priming (Groups 4, 5 and 6) or after priming (groups 7, 8 and 9) a significantly higher frequency of ARI Score of "1" (P \< 0.05) was reported. When blood contamination was applied before and after priming Smart Clip and Quick brackets (groups 10 and 11) showed significantly higher frequency of ARI Score of "1" (P \< 0.05), whereas Damon 3MX brackets (group 12) showed significantly higher frequency of ARI Score of "0" (P \< 0.01).
Table 3Frequency of distribution of adhesive remnant index scores (%).
Discussion
==========
The null hypothesis of the study has been rejected. In the present investigation self ligating brackets bonded onto dry enamel had significantly higher shear bond strength values than other groups. When contamination occurred before or after priming significantly lower shear bond strength values were recorded. Lowest shear strength values were obtained when blood contamination occurred both before and after priming. In literature self ligating brackets has been tested only onto dry enamel ([@B16]-[@B18]). To our knowledge there are no published studies that evaluated bond strength of self-ligating brackets onto blood contaminated enamel. Previous investigations that evaluated shear bond strength of blood contaminated conventional orthodontic brackets showed significant reduction of strength values compared to brackets bonded onto uncontaminated enamel ([@B6]-[@B10],[@B14],[@B19]).
Fluid contamination during bonding can lead to premature failure of the bond because when etched enamel becomes wet, the plugging of its porosities can affect resin penetration ([@B20]). This results in resin tags of insufficient number and length ([@B11]). In fact blood seems to be a physical barrier that impedes the mechanical retention of the adhesive to the etched tooth.
Enamel surface contamination can occur at different critical times of the bonding procedure: after the tooth surface has been etched, after the primer has been applied and after both procedures ([@B14]). In the present investigation no significant differences were found among groups bonded after blood contamination before priming (groups 4, 5 and 6) and among groups bonded after blood contamination after priming (groups 7, 8 and 9). When brackets were bonded under blood contaminated enamel before and after priming (groups 10, 11 and 12) unsignificant (groups 10 and 12) and significant (group 11) further strength values reduction has been reported. These results agree with those previously reported in other investigations using conventional brackets bonded with different adhesive systems on blood-moistened enamel surfaces ([@B14],[@B19]).
In the present investigation permanent bovine lower incisors were used. Extracted human teeth are becoming difficult to obtain due to recent progress in conservative dental treatment. In order to find a substitute for human teeth the use of bovine lower incisors for dental experimental studies has been proposed for two main reasons. First, it is easier to obtain a sufficient number of sound bovine teeth than human teeth. Second, the bigger surface area of bovine lower incisors allows more accurate preparation. Bovine teeth, derived from animals of similar genetic lineage and dietary environment, might show higher homogeneity of mineral composition than different human teeth, which are collected from various donators with diverse dietary or fluoride supplementation ([@B21]). Only lower incisors were collected because they have a minimal curvature when compared to other bovine teeth, that allows a more accurate bracket placement ([@B21]).
Previous authors evaluated adhesive strength to bovine enamel and found no statistically significant difference between bovine and human teeth with any of the materials used, although the mean values were always slightly lower with bovine teeth ([@B22]). In fact, bovine enamel is similar to human enamel and generally the teeth of all mammals appears to be very similar on a histochemical and anatomic basis ([@B23]). Despite the differences, bovine enamel has been reported to be a reliable substitute for human enamel in bonding studies with similar or slightly lower strength values ([@B22],[@B24],[@B25]). Therefore, since bovine enamel has been reported to have similar physical properties, composition, and bond strengths than human enamel, it can be mimicking human teeth in general so it can be considered as a reliable substitute for human enamel in bonding studies ([@B21]-[@B27]).
A minimum bond strength of 6 to 8 MPa has been reported to be sufficient for most clinical orthodontic needs, because these values are considered able to withstand masticatory and orthodontic forces ([@B28],[@B29]). In the present study, the bond strengths of the three different brackets tested bonded onto dry enamel surface were above these limits. When brackets were used on blood-contaminated enamel, the minimum requirement was achieved only in groups 2 (brackets Smart Clip under blood contamination before priming) and 7 (brackets Quick under blood contamination after priming). All the other contaminated groups showed mean shear strength values under the minimum required for orthodontic purposes. This is in agreement with previous investigations that evaluated shear bond strength of conventional orthodontic brackets bonded onto blood contaminated enamel ([@B14],[@B19],[@B30]).
Moreover in the present investigation ARI Scores has been recorded. Uncontaminated enamel groups reported a higher frequency of ARI score of "2". Under blood contamination, brackets bonded with blood contamination before priming and groups with blood contamination after priming showed significantly higher frequency of ARI Score of "1". When blood contamination was applied before and after priming Smart Clip and Quick brackets showed significantly higher frequency of ARI Score of "1", whereas Damon 3MX brackets showed significantly higher frequency of ARI Score of "0". This is in agreement with previous investigations that evaluated ARI Scores of orthodontic brackets bonded onto blood contaminated enamel ([@B14],[@B10],[@B19]) that all showed lower scores for blood contaminated groups ([@B5],[@B14],[@B19],[@B30]).
The present study demonstrated that non-contaminated enamel surfaces showed highest bond strengths for all self-ligating brackets tested. Under blood-contamination all groups showed significantly lower shear bond strength values. Lowest shear bond strength values were recorded when blood application occurred before and after priming. Moreover non-contaminated groups showed higher frequency of ARI Score of "2". For blood contaminated groups, orthodontic self-ligating brackets showed higher frequency of ARI Score of "1" and "0".
The Authors wish to thank Leone, Forestadent, 3M/Unitek and Ormco for providing the materials tested in this study.
| {
"pile_set_name": "PubMed Central"
} |
I[NTRODUCTION]{.smallcaps} {#sec1-1}
==========================
Studies have demonstrated that although the prevalence of chronic diseases in the elderly population of China is high, resulting in a heavy disease burden, the awareness, treatment and prevention of chronic diseases are generally poor in China.\[[@ref1][@ref2][@ref3][@ref4]\] Although knowledge concerning the prevention of hypertension has been most extensively popularized, the awareness of this information in economically poor regions and rural areas is \<10%.\[[@ref3][@ref4]\] The control rate of hypertension in rural areas of China is only 2.3%, which is lower than that in rural India (8.6%).\[[@ref3][@ref4]\] However, the burden of chronic diseases will continuously increase due to the aging of the Chinese population. In addition, the high disabling potential of chronic neurological diseases such as dementia will also cause an enormous social and economic burden because of the huge population base of elderly individuals in China. Therefore, the prevention and control of chronic diseases in China are very important.
The prevention and control of chronic diseases have been the focus of military health services in elderly veteran communities. Related tasks have been performed for more than 30 years, and a complete prevention and control system has been established. Thus, is the chronic disease awareness status among elderly veterans higher than that of the elderly in the general population? Can the years of experience in the prevention and control of chronic diseases in elderly veteran communities provide valuable references? Currently, most Chinese veterans have entered the advanced aging stage. The huge burden caused by dementia and other chronic disabling neurological diseases (CDND) has gradually attracted our attention, and information and education about disease prevention have gradually been introduced. However, does the current awareness of strategies for the prevention and control of these diseases correspond to their high degree of disablement? To address this issue, a survey of the popularization and awareness status of chronic disease prevention knowledge was submitted to elderly veteran communities in Beijing to summarize the experiences and shortcomings of these individuals regarding their knowledge about the prevention and control of chronic diseases. The survey results are expected to provide a valuable reference for the future prevention and control of chronic diseases among the elderly in the general population.
M[ETHODS]{.smallcaps} {#sec1-2}
=====================
Research design {#sec2-1}
---------------
A cross-sectional cluster sampling method was applied to survey veterans living in elderly veteran communities in Beijing. The study protocol was reviewed and approved by the Ethics Committee of Chinese PLA General Hospital, and the investigation was initiated after obtaining informed consent from the subjects or their guardians.
Study subjects {#sec2-2}
--------------
Veterans were included if they met the following criteria: (1) At least 60 years old; (2) continuous residence in a veteran community in Beijing for at least 1 month; and (3) a history of working in the military system before retirement. Most Chinese veterans live in fixed veteran communities composed of stable elderly populations after retirement. The fixed healthcare management system, including community outpatient clinics, provides support and professional healthcare services to these veterans and stores long-term medical records. All of the above characteristics of veterans reduce the rate of missing data and facilitate epidemiological research. The working staff in the veteran communities and the spouses of the retired veterans were not included in this study.
Investigation of the popularization and awareness status of chronic diseases {#sec2-3}
----------------------------------------------------------------------------
Using a unified questionnaire, via face-to-face interviews, the baseline characteristics of the veterans were collected, and the popularization and awareness status of chronic disease prevention knowledge was investigated by qualified geriatric neurology graduate students after standardized training. The survey covered five CDND, including dementia, Alzheimer\'s disease (AD), Parkinson\'s disease (PD), sleep disorder and cerebrovascular disease (CVD), and three common chronic diseases (CCD), including hypertension, diabetes, and coronary heart disease (CHD); CVD was considered as a common chronic neurological disease. The survey investigated the current status of the veterans' awareness of the disease name, their knowledge about the prevention and treatment of these chronic diseases and the approaches used by the veterans to access this information, including media (books, newspapers, magazines, radio, and television), word of mouth (verbal communication among the elderly), and health care professionals (hospital seminar and medical staff in the elderly veteran communities). The prevalence of the above chronic diseases in the veterans was also recorded.
Statistical analysis {#sec2-4}
--------------------
EpiData 3.1 software (The EpiData Association, Odense, Denmark) developed by Lauritsen JM and Bruus M. was used to establish a database, and SPSS 19.0 (SSPS Inc., Chicago, IL, USA) was used for the statistical analysis of the data. The differences in the total awareness rates of various chronic diseases among veterans were compared with that of hypertension using the McNemar paired Chi-square test. Alternatively, the awareness rates of veterans with or without different chronic diseases were compared to that of hypertension using the Chi-square test. The demographic differences in the awareness and the veterans' approaches to access knowledge about a chronic disease were compared using the Chi-square test, the Fisher exact test, and the Chi-square trend test. Differences between groups were considered statistically significant when *P* \< 0.05. Due to the varying number of subjects who responded regarding their awareness of different chronic diseases, the sample sizes of each surveyed item were recorded; only complete data were analyzed, and any subjects for whom data were missing were omitted.
R[ESULTS]{.smallcaps} {#sec1-3}
=====================
A total of 3473 veterans aged ≥60 years living in 44 elderly veteran communities in Beijing completed this survey in 2008; their average age was 77.97 ± 5.10 years. Among them, 30.8% were in the oldest old category (aged ≥80 years), 94.4% were males, and their average education duration was 9.93 ± 4.18 years.
The awareness status of strategies for the prevention of chronic diseases {#sec2-5}
-------------------------------------------------------------------------
The prevalence of hypertension and CHD among veterans was higher than 60.0%, which was the highest among CCD; the prevalence of PD, AD and dementia were low among CDND, whereas the prevalence of CVD and sleep disorders were relatively high \[[Table 1](#T1){ref-type="table"}\].\[[@ref5]\] The awareness statuses of the disease name and of strategies for the prevention of CDND in all surveyed veterans, those with CDND, and those without CDND were significantly worse than those with CCD such as hypertension. The awareness rates for hypertension, CHD, and diabetes were approximately 100%; hypertension displayed the highest awareness rate. The awareness rate for CVD was the highest among CDND and was close to that for hypertension and the other CCD. The awareness rate for AD was the lowest at \<10%, followed by those for sleep disorders, PD, and dementia. The differences in the awareness rates between the veterans with CDND and those with hypertension were statistically significant \[[Table 1](#T1){ref-type="table"} and [Figure 1](#F1){ref-type="fig"}\].
######
Prevalence rates of chronic diseases and awareness rates of prevention knowledge about chronic diseases among veterans (%(*n*/*N*))
Disease type Prevalence rate Awareness rate of disease name Awareness rate of prevention knowledge
----------------- ------------------ -------------------------------- ---------------------------------------- --------------------- --------------------
Dementia 6.5 (224/3423) 98.3 (3327/3384)\* 67.1 (114/170)\* 89.4 (2853/3192)\* 88.2 (2986/3384)\*
AD 3.4 (117/3423) 7.3 (246/3381)\* 3.3 (3/92)\* 4.3 (141/3267)\* 4.3 (144/3381)\*
PD 2.0 (70/3473) 80.3 (2718/3385)\* 76.1 (51/67)\* 61.8 (2052/3267)\* 62.1 (2103/3385)\*
Sleep disorders 35.8 (1237/3454) 66.6 (2254/3385)\* 54.8 (660/1205)\* 51.0 (1112/2179)\* 52.3 (1772/3385)\*
CVD 20.1 (699/3473) 99.5 (3365/3381)^†^ 96.1 (619/644)^†^ 93.6 (2561/2736) 94.1 (3180/3381)\*
Diabetes 24.1 (837/3470) 99.7 (3371/3380) 98.0 (789/805) 92.6 (2384/2574)^†^ 93.9 (3173/3380)\*
CHD 64.5 (2238/3468) 99.8 (3373/3379) 96.9 (2097/2165)^†^ 94.2 (1141/1211) 95.9 (3240/3379)\*
Hypertension 62.3 (2163/3470) 99.9 (3377/3381) 98.2 (2065/2102) 95.1 (1216/1278) 97.0 (3281/3381)
\**P* \< 0.001; ^†^*P* \< 0.01; the *P* value reflects differences in the awareness rates of various chronic diseases relative to hypertension among veterans. The numbers in parentheses correspond to the numbers of veterans with the disease or who knew the disease name and prevention knowledge divided by the total numbers of veterans. AD: Alzheimer's disease; PD: Parkinson's disease; CVD: Cerebrovascular disease; CHD: Coronary heart disease.
{#F1}
Differences in the awareness rates of chronic diseases based on the demographic characteristics {#sec2-6}
-----------------------------------------------------------------------------------------------
The awareness statuses of the disease name and prevention and treatment knowledge about chronic diseases varied with age, years of education, and gender. Except for hypertension and CHD, the awareness rates of the name of chronic diseases in the oldest-old group (aged ≥80 years) were significantly lower than those in the younger elderly group (aged 60--79 years). The awareness rates of the prevention and treatment knowledge about dementia, CVD, and hypertension among the oldest-old group were remarkably lower than those among the younger elderly group. Except for hypertension, with increasing years of education, the awareness rate of the name of chronic diseases exhibited an increasing trend. Except for AD and PD, awareness status of the prevention and treatment knowledge about chronic diseases increased significantly with increasing years of education based on the Chi-square trend test. Except for dementia, the awareness status of the name of chronic diseases among females was significantly higher than that among males. Additionally, the awareness statuses of the prevention and treatment knowledge about AD among females were significantly higher than those among males based on the Chi-square test \[Tables [2](#T2){ref-type="table"} and [3](#T3){ref-type="table"}\].
######
Awareness rates of chronic disease names among veterans based on different demographic characteristics
Disease type Age (%(*n*/*N*)) Education duration Gender (%(*n*/*N*))
----------------- ------------------- -------------------- --------------------- ---------------- ---------------------- ------------------ -------------------
Dementia 98.9 (2354/2380) 96.9 (973/1004)^‡^ 15/18 96.6 (677/701) 99.0 (2625/2652)^‡^ 98.3 (3138/3191) 97.9 (189/193)
AD 8.0 (191/2376) 5.5 (55/1005)^†^ 0/18 2.7 (19/700) 8.5 (225/2650)^‡^ 6.6 (210/3188) 18.7 (36/193)^‡^
PD 84.4 (2008/2380) 70.6 (710/1005)^‡^ 11/18 67.0 (470/701) 84.0 (2229/2653)^‡^ 79.5 (2538/3192) 93.3 (180/193)^‡^
Sleep disorders 69.6 (1657/2380) 59.4 (597/1005)^‡^ 8/18 58.2 (408/701) 68.9 (1829/2653)^‡^ 65.8 (2101/3192) 79.3 (153/193)^‡^
CVD 99.8 (2372/2376) 98.8 (993/1005)^‡^ 17/18 99.3 (695/700) 99.6 (2640/2650)\* 99.5 (3172/3188) 100.0 (193/193)
Diabetes 99.9 (2375/2377) 99.3 (996/1003)^†^ 18/18 99.6 (696/699) 99.8 (2644/2650) 99.7 (3179/3188) 100.0 (192/192)
CHD 99.9 (2375/2377) 99.6 (998/1002) 18/18 99.3 (693/698) 100.0 (2649/2650)^†^ 99.8 (3181/3187) 100.0 (192/192)
Hypertension 100.0 (2376/2377) 99.7 (1001/1004) 18/18 99.6 (697/700) 100.0 (2649/2650)\* 99.9 (3185/3189) 100.0 (192/192)
\**P* \< 0.05; ^†^*P* \< 0.01; ^‡^*P* \< 0.001; the *P* value reflects differences in the awareness rates of various chronic diseases among veterans grouped by age, education level and gender. The Chi-square trend test was used to explore the educational differences in the awareness rate of chronic diseases among veterans. The numbers in parentheses correspond to the numbers of veterans who knew the disease name divided by the total numbers of veterans. AD: Alzheimer's disease; PD: Parkinson's disease; CVD: Cerebrovascular disease; CHD: Coronary heart disease.
######
Awareness rates of chronic disease prevention knowledge among veterans based on different demographic characteristics
Disease type Age (%(*n*/*N*)) Education duration Gender (%(*n*/*N*))
----------------- ------------------ -------------------- --------------------- ---------------- --------------------- ------------------ -----------------
Dementia 90.7 (2134/2353) 87.6 (852/973)^†^ 13/15 88.3 (597/676) 90.3 (2370/2625) 89.7 (2813/3137) 91.5 (173/189)
AD 58.6 (112/191) 58.2 (32/55) 0/0 10/19 58.7 (132/225) 54.8 (115/210) 80.6 (29/36)^†^
PD 78.3 (1572/2008) 74.8 (531/710) 8/11 75.3 (354/470) 77.9 (1737/2229) 77.1 (1958/2538) 80.6 (145/180)
Sleep disorders 78.2 (1296/1657) 79.7 (476/597) 6/8 74.0 (302/408) 79.7 (1458/1829)\* 78.7 (1653/2101) 77.8 (119/153)
CVD 95.4 (2264/2372) 92.2 (916/993)^‡^ 14/17 93.5 (650/695) 94.9 (2505/2640)\* 94.5 (2998/3172) 94.3 (182/193)
Diabetes 94.4 (2239/2373) 94.0 (934/994) 14/18 92.4 (641/694) 94.9 (2508/2642)^‡^ 94.2 (2991/3175) 94.8 (182/192)
CHD 96.1 (2281/2374) 96.1 (959/998) 14/18 94.5 (654/692) 96.6 (2560/2649)^‡^ 96.1 (3056/3180) 95.8 (184/192)
Hypertension 97.7 (2320/2374) 96.1 (961/1000)^†^ 14/18 95.3 (662/695) 97.9 (2592/2648)^‡^ 97.2 (3094/3182) 97.4 (187/192)
\**P* \< 0.05; ^†^*P* \< 0.01; ^‡^*P* \< 0.001; the *P* value reflects differences in the awareness rates of various chronic diseases among veterans grouped by age, education level and gender. The Chi-square trend test was used to explore the educational differences in the awareness rate of chronic diseases among veterans. The numbers in parentheses correspond to the numbers of veterans with prevention knowledge divided by total numbers of veterans. AD: Alzheimer's disease; PD: Parkinson's disease; CVD: Cerebrovascular disease; CHD: Coronary heart disease.
Approaches used to access knowledge about the prevention of chronic diseases {#sec2-7}
----------------------------------------------------------------------------
Regarding the approaches used to access knowledge about CCD such as hypertension, diabetes, CHD and CVD, media was the most frequently selected mode of communication, displaying the highest rate of nearly 80%. The rate of the use of health care professionals was 46.1--58.6%, which was similar to or higher than that of the word of mouth \[48.6--51.9%; [Table 4](#T4){ref-type="table"}\]. Compared with CCD such as hypertension, among the approaches to access knowledge about CDND, the rates of the use of health care professionals were significantly reduced to 10.6--28.2%. Excluding AD, the rates of the use of the word of mouth for CDND were 56.5--76.5%, which were higher than those for CCD. Additionally, the rates of the use of media for CDND were 46.2--56.2%, which were lower than those for CCD \[Tables [4](#T4){ref-type="table"}, [5](#T5){ref-type="table"}, and [Figure 2](#F2){ref-type="fig"}\].
######
Rates of the use of different approaches to access prevention knowledge about chronic disabling neurological diseases among veterans
Items Age (%(*n*/*N*)) Education duration Gender (%(*n*/*N*)) Total (%(*n*/*N*))
---------------------------- ------------------ -------------------- --------------------- -------------------- --------------------- ------------------ ----------------- ------------------
Dementia
Media 48.7 (1039/2134) 40.0 (339/848)^‡^ 5/13 42.6 (254/596) 47.1 (1116/2367) 45.9 (1289/2811) 52.0 (89/171) 46.2 (1378/2982)
Word of mouth 75.1 (1602/2134) 80.2 (680/848)^†^ 10/13 80.9 (482/596) 75.5 (1786/2367)^†^ 76.7 (2157/2811) 73.1 (125/171) 76.5 (2282/2982)
Health care professionals 11.0 (234/2134) 9.7 (82/848) 1/13 8.7 (52/596) 11.1 (262/2367) 10.2 (288/2811) 16.4 (28/171)\* 10.6 (316/2982)
AD
Media 78.4 (87/111) 80.6 (25/31) 0/0 8/10 79.2 (103/130) 77.0 (87/113) 86.2 (25/29) 78.9 (112/142)
Word of mouth 23.4 (26/111) 25.8 (8/31) 0/0 2/10 24.6 (32/130) 23.0 (26/113) 27.6 (8/29) 23.9 (34/142)
Health care professionals 31.5 (35/111) 16.1 (5/31) 0/0 3/10 27.7 (36/130) 23.9 (27/113) 44.8 (13/29)\* 28.2 (40/142)
PD
Media 53.5 (840/1571) 50.4 (267/530) 3/8 47.2 (167/354) 53.9 (935/1735)\* 52.8 (1033/1958) 51.7 (74/143) 52.7 (1107/2011)
Word of mouth 71.4 (1121/1571) 74.2 (393/530) 8/8 78.2 (277/354) 70.7 (1227/1735)^†^ 72.3 (1415/1958) 69.2 (99/143) 72.1 (1514/2011)
Health care professionals 11.7 (184/1571) 12.1 (64/530) 0/8 11.0 (39/354) 11.9 (207/1735) 11.3 (222/1958) 18.2 (26/143)\* 11.8 (248/2011)
Sleep disorders
Media 55.6 (719/1294) 58.1 (276/475) 4/6 52.0 (157/302) 57.0 (830/1455) 56.2 (929/1653) 56.9 (66/116) 56.2 (995/1769)
Word of mouth 56.0 (724/1294) 58.1 (276/475) 4/6 58.6 (177/302) 56.0 (815/1455) 56.6 (936/1653) 55.2 (64/116) 56.5 (1000/1769)
Health care professionals 14.5 (187/1294) 13.9 (66/475) 0/6 12.9 (39/302) 14.6 (212/1455) 13.8 (228/1653) 21.6 (25/116)\* 14.3 (253/1769)
CVD
Media 79.0 (1787/2262) 72.9 (666/913)^‡^ 8/14 73.2 (475/649) 78.4 (1962/2501)^†^ 77.1 (2312/2997) 79.2 (141/178) 77.3 (2453/3175)
Word of mouth 50.4 (1139/2262) 52.5 (479/913) 6/14 53.9 (350/649) 50.3 (1258/2501) 51.0 (1528/2997) 50.6 (90/178) 51.0 (1618/3175)
Health care professionals 46.7 (1056/2262) 45.7 (417/913) 8/14 47.5 (308/649) 46.1 (1153/2501) 46.7 (1399/2997) 41.6 (74/178) 46.4 (1473/3175)
\**P* \< 0.05; ^†^*P* \< 0.01; ^‡^*P* \< 0.001; the *P* value reflects differences in the use of different approaches to access prevention knowledge about chronic disabling neurological diseases among veterans grouped by age, education level and gender. The Chi-square trend test was used to explore the educational differences in the use of different approaches to access prevention knowledge about chronic disabling neurological diseases among veterans. The numbers in parentheses correspond to the numbers of veterans who used the indicated approach divided by the total numbers of veterans. AD: Alzheimer's disease; PD: Parkinson's disease; CVD: Cerebrovascular disease; CHD: Coronary heart disease.
######
Rates of the use of different approaches to access prevention knowledge about common chronic diseases among veterans
Items Age (%(*n/N*)) Education duration Gender (%(*n/N*)) Total (%(*n/N*))
---------------------------- ------------------ -------------------- ------------------- ------------------ --------------------- ------------------ ---------------- ------------------
Diabetes
Media 78.3 (1753/2239) 73.0 (681/933)^†^ 10/14 73.6 (471/640) 77.6 (1945/2508) 76.7 (2296/2993) 77.1 (138/179) 76.7 (2434/3172)
Word of mouth 51.4 (1151/2239) 52.9 (494/933) 7/14 54.4 (348/640) 51.3 (1286/2508) 52.0 (1555/2993) 50.3 (90/179) 51.9 (1645/3172)
Health care professionals 46.7 (1045/2239) 44.8 (418/933) 8/14 50.5 (323/640) 45.1 (1130/2508)^†^ 46.4 (1389/2993) 41.3 (74/179) 46.1 (1463/3172)
CHD
Media 81.0 (1846/2279) 75.1 (719/957)^‡^ 9/14 76.0 (496/653) 80.1 (2049/2557)\* 79.3 (2422/3055) 79.0 (143/181) 79.3 (2565/3236)
Word of mouth 49.5 (1127/2279) 48.5 (464/957) 6/14 51.1 (334/653) 48.7 (1245/2557) 49.1 (1500/3055) 50.3 (91/181) 49.2 (1591/3236)
Health care professionals 58.1 (1325/2279) 59.7 (571/957) 9/14 62.6 (409/653) 57.6 (1472/2557) 59.0 (1801/3055) 52.5 (95/181) 58.6 (1896/3236)
Hypertension
Media 81.6 (1891/2318) 75.3 (722/959)^‡^ 9/14 75.0 (495/660) 81.0 (2099/2590)^†^ 79.8 (2467/3093) 79.3 (146/184) 79.7 (2613/3277)
Word of mouth 48.3 (1119/2318) 49.4 (474/959) 7/14 51.1 (337/660) 48.0 (1243/2590) 48.8 (1508/3093) 46.2 (85/184) 48.6 (1593/3277)
Health care professionals 56.9 (1318/2318) 55.3 (530/959) 8/14 59.5 (393/660) 55.6 (1440/2590) 56.7 (1754/3093) 51.1 (94/184) 56.4 (1848/3277)
\**P* \< 0.05; ^†^*P* \< 0.01; ^‡^*P* \< 0.001; the *P* value reflects differences in the use of different approaches to access prevention knowledge about common chronic diseases among veterans grouped by age, education level and gender. The Chi-square trend test was used to explore the educational differences in the use of different approaches to access prevention knowledge of chronic diseases among veterans. The numbers in parentheses correspond to the numbers of patients who used the indicated approach divided by the total numbers of veterans. CHD: Coronary heart disease.
{#F2}
The approaches used to access knowledge about most chronic diseases were not significantly affected by age, years of education or gender \[Tables [4](#T4){ref-type="table"} and [5](#T5){ref-type="table"}\]. Except for AD, PD, and sleep disorders, the proportions of the oldest-old group that obtained knowledge about chronic diseases using media were significantly lower than those of the younger elderly group. The proportion of the oldest-old group that obtained prevention and treatment knowledge about dementia through word of mouth was significantly higher than that of the younger elderly group. The approaches used to access prevention and treatment knowledge about dementia, PD, CVD, diabetes, CHD, and hypertension were affected by the number of years of education. The proportions of females who accessed prevention and treatment information about dementia, AD, PD, and sleep disorders using health care professionals were significantly higher than those of males \[Tables [4](#T4){ref-type="table"} and [5](#T5){ref-type="table"}\].
D[ISCUSSION]{.smallcaps} {#sec1-4}
========================
This study demonstrated that the chronic disease awareness status among elderly veterans with or without chronic diseases was relatively good, but the awareness status of strategies for the prevention of CDND was far lower than that of CCD. The health care professionals played a limited role in the popularization of knowledge about CDND. The propagation of knowledge about CDND greatly relied on nonmedical professional media and peer education by word of mouth; thus, the accuracy and guiding significance of the information obtained are not assured.
Compared with CCD, the awareness status of CDND, particularly dementia, among elderly veterans was significantly lower. Although the prevalence of CDND was lower than that of CCD, CDND are the primary chronic disabling diseases in the elderly population, resulting in the heaviest disease burden, including huge costs.\[[@ref2][@ref6]\] However, regardless of the income level of a country (high or middle income), public awareness of dementia is generally poor, and this limited understanding delays the early diagnosis and treatment of patients.\[[@ref6]\] Although the prevention and control of chronic diseases in elderly veteran communities are implemented earlier than in the general population, the poor awareness status of CDND has maintained a huge gap with respect to its heavy social and economic burden. Therefore, extensive popularization should be implemented to raise awareness of these diseases, thus contributing to the efficient treatment of CDND and reducing the burden of these diseases.
Regarding the approaches used to access information about the prevention of CDND, the rates of the use of health care professionals were only 10.6--28.2%, whereas those for CCD were approximately 50%. Compared with CCD, the role of health care professionals in the dissemination of information about CDND was significantly lacking, and this weakness should be strengthened. According to a report from World Health Organization (WHO), regardless of economic development level of a country, the awareness of dementia is generally lacking not only among the public but also among health care and social service providers.\[[@ref6]\] Some physicians even look down on dementia patients and their families, increasing the patients' sense of shame and hindering the efficient treatment of patients.\[[@ref6]\] Therefore, the WHO advocates that countries worldwide should learn from the experience of the national dementia programs such as those of Japan and the UK to improve the awareness status of dementia among health care workers and the public.\[[@ref6]\] The health care professionals in geriatric neurology should provide professional training to medical personnel working in veteran communities and general population to improve their capability of preventing and controlling CDND.
Although the Chinese government had strengthened chronic disease prevention program, the awareness rates of CCD were generally low.\[[@ref7][@ref8]\] The awareness rate of hypertension was 70--80% in Beijing and Shanghai, whereas that in low economic status regions and rural areas was lower than 10%, and the awareness rates for CHD and CVD among those of low economic status were \<1%.\[[@ref3][@ref4][@ref9][@ref10][@ref11][@ref12][@ref13][@ref14]\] The awareness status for CDND was worse than that for hypertension regardless of the economic development level. The awareness rates of the early symptoms and nursing knowledge of dementia were only approximately 15% and 5%, respectively.\[[@ref15][@ref16][@ref17]\] The awareness rates of prevention knowledge about CCD, dementia and CDND among elderly veterans were significantly higher than those among the general elderly population. Notably, the veterans without chronic diseases also exhibited a higher awareness status of chronic diseases, improving the efficient diagnosis and effective treatment of chronic diseases. Since veteran communities were established more than 30 years ago, the prevention and control of CCD and CDND have been the focus of health care for veterans of advanced age.\[[@ref18][@ref19]\] This survey also documented the excellent accomplishments regarding the prevention and control of chronic diseases among veterans.
The proportions of elderly veterans who accessed knowledge about CCD or CDND from health care professionals were approximately 50% and 10%, respectively. This finding indicated that medical personnel played an effective role in the popularization of chronic disease prevention knowledge. Although the proportion of the oldest-old group and the prevalence of chronic diseases among veterans were significantly higher than those among the elderly in the general population,\[[@ref1][@ref3][@ref5][@ref8][@ref9][@ref10][@ref14][@ref20][@ref21][@ref22]\] the overall health and functional status of veterans were significantly superior to those of the elderly in the general population.\[[@ref5][@ref21][@ref22][@ref23][@ref24]\] These results reflected the higher awareness status of chronic diseases among veterans and the active provision of education about chronic diseases by medical personnel. Therefore, referring to the experience in veteran communities, education about chronic diseases in the general population by medical personnel should be strengthened to improve the currently poor awareness status of chronic disease prevention and control in China.
Similar to the findings in the general population,\[[@ref25]\] the most common approach used by elderly veterans to access prevention knowledge of chronic diseases was nonmedical professional media. The overall level of education among veterans was significantly higher than that among the elderly in the general population,\[[@ref21][@ref22]\] and more than 90% of veterans regularly read books and newspapers, watch TV, and listen to the radio.\[[@ref26]\] Accordingly, media plays an important role in the control of chronic diseases among veterans. However, information propagated through the media should be released after revision by healthcare professionals to avoid the dissemination of inaccurate information.
Concerning the popularization of the prevention knowledge about CDND, the role of media was not as significant as that about CCD, whereas the role of word of mouth was nearly the leading approach for the dissemination of information about CDND. These results suggested that media did not sufficiently address CDND and that the popularization of this knowledge through media should be strengthened. Although the word of mouth may not provide accurate information, numerous leisure activities at veteran communities\[[@ref26]\] enable frequent communication between veterans, such that prevention knowledge about chronic diseases can rapidly spread through peer education by word of mouth. Because the knowledge about CDND is highly technical, media and word of mouth should be guided to rapidly spread truthful prevention information to improve the currently poor awareness status of CDND and their prevention.
There are some limitations to this study. The subjects were predominantly male veterans with advanced age and high education. Few female subjects, young subjects or subjects with low education were included. Areas with poor medical conditions were not addressed. This study only investigated the awareness status of the prevention knowledge about chronic diseases, and the status of the correct understanding and the control of these diseases among veterans should be further investigated.
**Edited by:** Xin Chen
**Source of Support:** The study was supported by grants from the Special Health Research Foundation of the Health Department of the PLA General Logistics Department (No. 07BJZ04, 11BJZ09, and 12BJZ46).
**Conflict of Interest:** None declared.
| {
"pile_set_name": "PubMed Central"
} |
Introduction {#Sec1}
============
The ubiquity of NPs in daily life as well as their utilization as versatile tools in biomedicine has received increased attention among scientists in the field of nanotechnology. In this context, particularly non-metal oxide NPs, such as amorphous silica (SiO~2~) and polymeric NPs are considered for a broad range of applications, such as additives to drugs, cosmetics, varnishes, and food, or as tools for DNA delivery^[@CR1]^, the identification of biomarkers^[@CR2]^, and drug delivery^[@CR3]^.
As true for many nanomaterials silica NPs often aggregate in cell culture media. This depends not only on the surface functionalization but also on the transfer protocol from the initial synthesis/storage medium (usually water or ethanol) or the dried state to the cell culture medium^[@CR4]^. In many studies, the aggregation state in cell culture medium is not controlled at all and hence, impedes a comparison of different cell culture studies.
The advances in polymer science allow producing well-defined nanoparticle formulations in a wide range concerning size, shape, and sub-particle morphology^[@CR5],\ [@CR6]^. For polymer NPs polyesters such as polylactides, poly(ε-caprolacton), polyglycolides and their copolymers are widely used as long-circulating drug carriers due to their biodegradability and low toxicity *in vivo* ^[@CR7]--[@CR10]^. Polystyrene, because of its low cost and availability, is mainly used for *in vitro* cell uptake experiments as a model^[@CR11],\ [@CR12]^. Complex NPs or nanocapsules are thought to be ideally suited for applications in biomedical application as marker and targeted drug-delivery system^[@CR13]^. The great variety of parameters and the lack of standardized protocols often preclude a comprehensive understanding of the factors governing the biological impact of nanoparticles in cells *in vitro*. In view of the different experimental conditions and read-outs under which cytotoxicological investigations of NPs have been performed it is not surprising that the results related to the biological impact of these NPs on cells are rather inconsistent and hence inconclusive. Such inconsistencies prevent the elucidation of fundamental and molecular rules which govern the potential pathophysiological responses to NP exposure in living organisms.
A major obstacle for the safe development and exploitation of nanotechnology is the lack of standardized procedures to investigate potential adverse effects of nanomaterials. As demonstrated for the two selected examples, silica and polymeric NPs, hazard assessment remains inconclusive due to differences in nanomaterial properties, cell types or lines, read-outs, and preparation of nanoparticle suspensions. Furthermore, it is already known that although different laboratories may have used the same assay and nanomaterial for testing in the same cell line, variations in the experimental procedures still lead to inconsistent results. In view of this situation we have decided to perform a carefully controlled round-robin experiment involving a number of different laboratories in Germany. In this intercomparison experiment we studied quasi-monodisperse NPs with a diameter in a similar size range (TEM diameter: 55 ± 2 nm and 74 ± 11 nm, hydrodynamic diameter about 120 nm) but of different chemical composition i.e. silica and polystyrene NPs, two of the most frequently used nanomaterials. The particles used were synthesized and characterized in our laboratories using both transmission electron microscopy (TEM) and dynamic light scattering (DLS). As a cell model we choose the human HepG2 cell line, since the liver is one of the most critical targets of NPs, especially in the field of nanomedicine. As read-outs for cell viability we decided to include the most widely used assays to monitor i) cellular dehydrogenase activity and ATP-levels as well as ii) LDH release. In addition, we monitored the release of interleukin-6 (IL-6) and interleukin-8 (IL-8) as markers of inflammation. To measure nanoparticle uptake in cells confocal laser scanning microscopy and flow cytometry were employed. The collaborating laboratories involved in this intercomparison study were all part of the German DFG Priority Program SPP1313 "Biological Responses to Nanoscale Particles"^[@CR14]^ and involved laboratories atThe Freie Universität Berlin (Lab 1)The University Clinic of Mainz and the Max-Planck-Institute of Mainz (Lab 2)The University of Jena (Lab 3)The University of Duisburg-Essen (Lab 4)The Karlsruhe Institute of Technology (Lab 5).
Materials and Methods {#Sec2}
=====================
Chemicals {#Sec3}
---------
**A**ll chemicals used in this synthesis of SiO~2~ NPs were applied as received without further purification: hydrofluoric acid (Fluka, 40--45%), ammonia solution (Carl Roth, *puriss*., 25%), 3-(aminopropyl)trimethoxysilane (APS, Sigma-Aldrich, 97%), ethanol (Berkel AHK, *puriss*., 100%), fluorescein isothiocyanate (FITC, Fluka, 99%), Igepal CO-520 (Sigma-Aldrich, 99%), cyclohexane (Carl Roth p.a., ≥99.5%), and tetraethyl orthosilicate (TEOS, Sigma-Aldrich, ≥99.0%). Styrene (99%, Merck, Germany) and acrylic acid (AA, 99%, Aldrich) were freshly distilled under reduced pressure and stored at −18 °C before use. Other reagents and solvents were commercial products and were used without further purification: n-hexadecane, HD (Sigma-Aldrich), initiator 2,2′-azobis(2-methylbutyronitrile), V59 (Wako Chemicals, Germany), sodium dodecyl sulfate, SDS (99%, Sigma-Aldrich). Ultrapure water with a measured resistance of 18.2 MΩ∙cm was obtained from a Millipore purification system (Milli-Q^®^ Academic). Dulbecco's Modified Eagle Medium (DMEM) with 10% fetal calf serum (FCS) (Life Technologies) was used from the consortium stockpile as described above. Standard cell culture media (Dulbecco's Modified Eagle Medium (DMEM) with 10% fetal calf serum (FCS) were purchased from Invitrogen. The cell culture medium and the HepG2 cells (CLS- Cell Lines Service, Germany) were purchased by Lab 3 and distributed to all groups in the consortium.
Synthesis and characterization of SiO~2~ NPs for cell culture experiments {#Sec4}
-------------------------------------------------------------------------
The dye-labelled silica NPs used in this study were synthesized in Lab 1 (Berlin).
### Synthesis of the dye-label {#Sec5}
The coupling product of APS and FITC was synthesized under argon by a modification of the synthetic route described by Imhof *et al*.^[@CR15]^. Dye-labelled SiO2 NP cores covalently labelled with this coupling product were prepared by a modified microemulsion synthesis, as described before^[@CR4]^ (see reaction scheme in Fig. [1A](#Fig1){ref-type="fig"}). The resulting particles were purified three times using repeated sedimentation/redispersion cycles to remove any unbound dye. Subsequently, a SiO~2~ shell was grown on the labeled SiO~2~ cores by a Stöber-like growth process and the resulting particles were again purified by repeated sedimentation/redispersion cycles. The reaction vessels used for reactions yielding SiO~2~ NPs were etched before synthesis with hydrofluoric acid (8 vol.-%) in order to remove insoluble SiO~2~ residues from the glass surface. Subsequently, the vessels were rinsed with ultrapure water. An ultrasonic bath (Bandelin Sonorex, RK 512 H, 860 W) was used for redispersion of sedimented NPs.Figure 1Features of the NPs used in this study. (**A**) Scheme of the synthesis route of the fluorescein isothiocyanate (FITC) labeled SiO~2~ NPs coated with a dye-free SiO~2~ shell (SiO2/FITC NPs). (**B**) Schematic illustration of the synthesis of the polymer NPs, synthesized via miniemulsion synthesis. (**C**) Transmission electron micrographs of the SiO2/FITC and (**D**) transmission electron micrographs of the polymer NPs.
### Transfer of dye labelled NPs from ethanol to DMEM/FCS {#Sec6}
The transfer process of the dye labeled SiO~2~ NPs from ethanol into a standard cell culture medium (DMEM)/FCS serum solution was carried out in the dark. Ethanol was removed by centrifugation (533 g, 1 h, T = 20 °C). The nanoparticle sediment was redispersed in ultrapure water using ultrasonification (Bandelin Sonorex, RK 512 H, 860 W) for 10 min in a small round-bottom centrifuge tube (BRAND, 780762, 1.5 mL centrifuge tube) or a large (50 mL) conical-bottom polypropylene standard centrifuge tube. The centrifuge tubes should be so deeply immersed into the ultrasonic bath that the dispersion inside the tubes is completely surrounded by the bath liquid. This procedure is repeated thrice to guarantee that the ethanol content was negligible (\<0.5 vol-%) and no ethanol-related cytotoxicity can occur in the subsequent cell exposure experiments.
### Characterization of silica NPs {#Sec7}
SiO~2~ NP formulations were characterized with respect to their size by TEM and DLS in Lab 1 (Berlin) and Lab 4 (Duisburg-Essen). The particles were stored and handled in an argon atmosphere until they were transferred to the cell culture medium and added to cells. TEM images (Fig. [1C](#Fig1){ref-type="fig"}) were acquired using SiO~2~ NPs from ethanol dispersion. A NP diameter of 55 ± 2 nm was determined from the quantitative evaluation of these images. Zeta potential and dynamic light scattering measurements were performed using a Delsa Nano C from Beckman Coulter at a fixed scattering angle of 165° revealing a Zeta potential of −52 ± 3 mV in water. The non-filtered NPs were measured in ethanol, ultrapure water, and DMEM/FCS with 10% FCS at 25 °C and 37 °C. The particle concentration of all samples was 0.5 g/L for all dynamic and electrophoretic light scattering measurements. Complementing the cell experiments, all NPs used in this work were additionally incubated at 37 °C for 6, 12, 24, 48 and 72 hours with DMEM (containing 10% fetal calf serum) and then characterized by DLS. In these measurements a self-designed dynamic light scattering (DLS) setup was used in Lab 4^[@CR16]^. Briefly, this setup includes the 532 nm line of a DPSS-Laser (Polytec; Gem, Nd:YAG) for illumination. Scattered light was detected at a 90° angle by a photomultiplier tube (Perkin und Elmer; MP-943), processed by a hardware-correlator (Correlator Ltd.; flex99r-12) and evaluated using a self-developed LabView routine.
Synthesis and characterization of polymer NPs {#Sec8}
---------------------------------------------
Dye-labeled polymer NPs used in this work were synthesized and functionalized in Lab 2 (Mainz).
### Synthesis of the dye-label {#Sec9}
The fluorescent dye Bodipy-1, a polymerizable derivative of borondipyrromethene, was used for labelling the polymer NPs and was synthesized according to Garcia-Moreno^[@CR17]^. Bodipy-1, which has the maximum of absorption and emission at 523 and 536 nm, respectively, is part of the polymer and thus no leakage of the dye from the nanoparticle can occur.
### Synthesis of dye-labelled NPs {#Sec10}
Polymeric NPs were synthesized by the free-radical mini-emulsion polymerization process as described before^[@CR18],\ [@CR19]^. This process is schematically depicted in Fig. [1B](#Fig1){ref-type="fig"}.
### Characterization of polymer NPs {#Sec11}
TEM images of the functionalized polystyrene NPs were acquired after drop casting of the NP dispersion onto a carbon-coated copper grid u (Fig. [1C](#Fig1){ref-type="fig"}). A NP diameter of 74 ± 11 nm was determined from the quantitative evaluation of these images. Hydrodynamic particle size and size distribution were determined by DLS using a Nicomp 380 sub-micrometer particle sizer (Nicomp Particle Sizing Systems, U.S.A.) at 90° scattering geometry. Zeta potential measurements were performed in a 1 mM KCl solution at pH 7.2 and 23 °C using the Smoluchowski equation^[@CR20]^ with a detected value of −49 +/− 9 mVmV. The surface charge density was determined by means of polyelectrolyte titration. Particles were titrated with 10^−3^ M cationic polyelectrolyte polydiallyldimethyl ammonium chloride (PDADMAC). The measurements were carried out with 10 mL of nanoparticle sample in aqueous solution with a solid content of 1 g/L. In addition, the polymer NPs were also incubated at 37 °C for up to 72 h with DMEM (containing 10% fetal calf serum) and then characterized in Lab 4 (Essen).
HepG2 cell cultures {#Sec12}
-------------------
HepG2 cells, derived from human liver carcinoma (CLS- Cell Lines Service, Germany) were used between passages 20 and 30 and cultured in DMEM:Ham's F12 with L-glutamin supplemented with 10% FCS (Life Technologies) at 37 °C in a 5% CO~2~ humidified environment. No antibiotics were used. At 85% confluence, cells were harvested after washing with Hank´s BSS using 0.25% Trypsin/0.02% EDTA (3 min, 37 °C) and sub-cultured in 75 cm^2^ flask, slides or 96-well plate with a density of 1.8 \* 10^4^ cells/cm^2^. They were allowed to grow for 24 h before nanoparticle exposure. The cells were tested for mycoplasma contamination routinely via PCR. The HepG2 cultures were expanded by one laboratory and distributed to all other ones. Handling and other parameters were standardized among the different partners of the inter-laboratory experiment.
### Cell viability assays {#Sec13}
Cell viability assays were performed under standardized protocols using HepG2 cells in Lab 2, 3 and 5. The tested nanoparticle concentrations ranged from 0 to 500 µg/mL. Cells were treated for the incubation times 6, 12, 24, 48 and 72 h.
### Cellular dehydrogenase activity and cellular ATP-levels {#Sec14}
The metabolic dehydrogenase activity and the relative cellular ATP-levels of HepG2 cells were assessed after exposure to different nanoparticle concentrations in Lab 3 (Jena). Free NPs were removed by washing. The cellular dehydrogenase activity was measured by reduction of MTS tetrazolium salt to the water-soluble formazan product by living cells at an absorbance of 490 nm. To measure cellular ATP-levels the ATPLite assay (PerkinElmer) was carried out according to the manufacturer's instructions. The determined dehydrogenase activities and cellular ATP-levels were normalized to non-treated control cells.
### Cellular dehydrogenase activity, LDH release and release of interleukin 6 and 8 {#Sec15}
Determination of dehydrogenase activity by AlamarBlue reduction, the LDH release and the release of interleukin 6 and 8 of HepG2 cells were measured in Lab 5 (Karlsruhe) as also described previously^[@CR21]^. After treatment of HepG2 cells with NPs, samples were centrifuged (5 min at 300 g). The supernatants were removed and used for determination of LDH and cytokine content. AlamarBlue reagent, diluted 1:20 (v/v) with HBSS, was added to the cells and incubated for 2 h. Aerosil 200 NPs (hydrophilic, Evonic) were used as controls for cytotoxicity.
Detection of internalized NPs {#Sec16}
-----------------------------
Experiments to investigate the intracellular localization of NPs were performed in Lab 2 (Mainz) and Lab 3 (Jena) using confocal laser scanning microscopy (CLSM). For these experiments the HepG2 cells were seeded in ibiTreat µ-slides (IBIDI, Germany) (Lab 2, Mainz) or in glass chamber slides (BD, Germany) (Lab 3, Jena). After 24 h, NPs were added to the medium at a concentration up to 500 µg/mL. Before CLSM imaging cells were washed with PBS. Live cell images were taken in Lab 2 (Mainz) with a commercial setup (LSM SP5 STED Leica Laser Scanning Confocal Microscope, Leica). There, the cell membranes were stained with CellMaskOrange (2.5 mg·mL^−1^, Invitrogen), and the cell nucleus with DraQ5 (2.5·10^−6^ M, Biostatus). In Lab 3 (Jena), for microscopy observations with the LSM 510 Meta (Carl Zeiss MicroImaging GmbH) cell membrane was stained with Wheat Germ Agglutinin Alexa Fluor® 633 conjugate (Invitrogen) and the nucleus with Hoechst 33258 (Applichem) after fixation of cells with 4% formaldehyde. Fluorescent particles were detected at 533--570 nm. For a control experiment to investigate the clustering and adhesion of the NPs to differently coated culture slides, the SiO~2~ NPs were suspended in complete culture medium by pipetting and then they were added to the slides without any cells for supplementary microscopy studies.
Flow cytometry {#Sec17}
--------------
Flow cytometry was used for the detection of nanoparticle uptake and cell viability in Lab 2 (Mainz). For these investigations, fluorescent SiO~2~ or polymeric particles were added to the cell culture medium of HepG2 cells. After the selected times of nanoparticle incubation, adherent cells were detached from the culture vessel, washed with buffer (PBS, Gibco, Germany), incubated with 28.6 µg/mL/ 7-aminoactinomycin (7-AAD). Flow cytometry measurements were performed on a CyFlow ML using FlowMax 2.57 software (Partec, Germany). The FL1 channel (excited by a 488 nm laser line, emission filter 527 ± 12 nm) was used to analyze the uptake of NPs by utilization of a forward/sideward scatter plot, thereby excluding cell debris. For 7-AAD, the events in the cell gate were analyzed on a FL1/FL6 dot-plot and three different populations (viable, apoptotic, dead) were determined. To normalize the median fluorescence values the factor of particle fluorescence intensity was determined, according to Hocherl *et al*.^[@CR22]^.
Statistical analyses {#Sec18}
--------------------
The results are expressed as means ± standard deviation of several independent experiments as indicated in the legends. The significance of difference between two mean values was assessed by a two-tailed Student's *t*-test. A *p-value* \< 0.05 was considered to be statistically significant.
All data generated or analyzed during this study are included in this published article (and its Supplementary Information files).
Results {#Sec19}
=======
Physico-chemical nanoparticle characterization {#Sec20}
----------------------------------------------
In addition to the thorough NP characterization following the synthesis procedure, we carried out DLS measurements after incubating the NPs at 37 °C for 6, 12, 24, 48, and 72 h with DMEM, containing 10% fetal calf serum (see Table [1](#Tab1){ref-type="table"} and supporting information).Table 1Hydrodynamic diameters of the used nanoparticle formulations incubated in complete culture medium, measured via DLS in Lab 4 (Duisburg-Essen) for variable time periods.DLS measurementd~hydr~(t = 0 d) \[nm\]d~hydr~ (t = 1 d) \[nm\]d~hydr~ (t = 2 d) \[nm\]d~hydr~ (t = 3 d) \[nm\]SiO~2~ NPs125 ± 6132 ± 7130 ± 7122 ± 6Polymer NPs118 ± 6117 ± 6118 ± 6121 ± 6These results show that all NPs were indeed colloidally stable under the conditions and on the timescales of our experiments. Moreover, determination of hydrodynamic diameter of NPs immediately after transfer in complete culture medium was consistent among the laboratories (Supplementary Table [1](#MOESM1){ref-type="media"}).
Comparison of results from different cell viability assays {#Sec21}
----------------------------------------------------------
The effects of NPs on cell viability of HepG2 cells were investigated in Labs 2, 3 and 5 using either SiO~2~ or polymer NPs, standardized protocols with regard to concentrations, incubation times etc. Furthermore, the same batch of HepG2 cells and cell culture medium including fetal calf serum was used by all participants. The different laboratories showed similarities of results, varying sensitivities among the used cytotoxicological test systems but also several discrepancies, particularly due to the absence of harmonization of very specific technical details: In addition to the conditions during shipping and transfer mentioned above, it turned out that SiO~2~ NPs tend to aggregate on polymer coated slips (ibidi, Germany) compared to regular glass cover slips.
### Silica NPs {#Sec22}
Incubation of HepG2 cells with SiO~2~ NPs lead to only a slight reduction in cell viability in the tested concentration range. The 7-AAD-staining, which demonstrates acute cell death, showed no effect of the SiO~2~ NP, as seen by a 100% vitality at all times after their incubation and for all concentrations (Fig. [2A](#Fig2){ref-type="fig"}).Figure 2Vitality of HepG2 cells after incubation with SiO~2~-NP measured as determined by different laboratories and read-outs (each partner performed one assay). HepG2 cells were incubated with increasing concentrations of the SiO~2~-NP for different incubation times (up to 72 h). (**A**) Cytotoxicity measurements by flow cytometry (7-AAD staining). (**B**) Effect of SiO~2~-NP on metabolic dehydrogenase activity as detected via MTS-assay. (**C**) Determination of cellular dehydrogenase activity as detected via AlamarBlue assay. (**D**) LDH release after nanoparticle exposure. In (**C**,**D**) 250 µg/mL Aerosil 200-NPs und 20 ng/mL TNF-α were additionally used as controls. All data were normalized to control values (no particle exposure or Triton X-100), which were set as 100%. (**A**) N = 3; B: N = 1, 6 parallels; (**C**,**D**) N = 3, technical duplicates, means and standard deviation (\*p \< 0.05; \*\*p \< 0.01, \*\*\* \< 0,001 in comparison to untreated controls).
Via MTS assay, which determines the dehydrogenase activity as a marker of cell metabolism, we observed a slight decrease in the dehydrogenase activity after 24 h at the highest nanoparticle concentrations, but after 48 and 72 h of incubation we observed a recovery in dehydrogenase activity instead (Fig. [2B](#Fig2){ref-type="fig"}). This reduction in cell metabolism contrasts with the LDH release, which was not elevated after 24 h and 48 h (Fig. [2D](#Fig2){ref-type="fig"}) compared to the non-treated control sample. Also the AlamarBlue test showed at all concentrations and incubation times that cellular dehydrogenase activities were not decreased (Fig. [2C](#Fig2){ref-type="fig"}), which indicated a low impact of the SiO~2~ NPs on cell metabolism. The comparison of the different assays indicates that the MTS assay is more sensitive than 7-AAD staining, AlamarBlue reduction and LDH release. Hence, by using only one test, to a certain extent mild effects on cell viability could be missed. Only a slight release of IL-6 and IL-8 occurred after exposure to the highest concentration (500 μg/mL) of SiO~2~ NPs (Fig. [3A and B](#Fig3){ref-type="fig"}). This low IL-6 and IL-8 release demonstrated the good biocompatibility of the SiO~2~ NPs, because no inflammatory processes were induced. This result also supports the findings of the vitality assays displayed in Fig. [2](#Fig2){ref-type="fig"}.Figure 3Release of IL-6 (**A**) and IL-8 (**B**) after treatment of HepG2 cells with increasing concentrations of SiO~2~-NPs after 24 and 48 h. 250 μg/mL Aerosil 200 NPs und 20 ng/mL TNF-α were used as positive controls. N = 3, technical duplicates, means and standard deviation (\*p \< 0.05; \*\*p \< 0.01 in comparison to untreated controls).
### Polymer NPs {#Sec23}
Overall and in contrast to SiO~2~ NPs the different test systems showed less consistent results in relation to the polymer NPs. The dehydrogenase activity and the cellular ATP levels decreased consistently in a concentration and time dependent manner (Fig. [4A and B](#Fig4){ref-type="fig"}) indicating weak cytotoxic effects at the highest polymer nanoparticle concentration. The LDH release was slightly enhanced at the highest concentration after 24 and 48 h (Fig. [4D](#Fig4){ref-type="fig"}). In contrast, the AlamarBlue reduction increased slightly at the highest exposure concentration in comparison to the untreated control (Fig. [4C](#Fig4){ref-type="fig"}).Figure 4Vitality of HepG2 cells after incubation with polymer-NPs as determined by different laboratories and read-outs. HepG2 were incubated with increasing concentrations of the polymer-NPs for different incubation times (up to 72 h). (**A**) Effect of polymer-NP on metabolic dehydrogenase activity as detected via MTS-assay. (**B**) Relative cellular ATP-levels were detected by ATPLite assay. (**C**) Cellular dehydrogenase activities measured via AlamarBlue assay. (**D**) LDH release after nanoparticle exposure. In (**C**,**D**) 250 µg/mL Aerosil 200-NPs und 20 ng/mL TNF-α were additionally used as controls. All data were normalized to control values (no particle exposure or Triton X-100), which were set as 100%. (**A**,**B**) N = 1 with 6 parallels; (**C**,**D**) N = 3 with 2 parallels, means and standard deviation (\*p \< 0.05; \*\*p \< 0.01, \*\*\* \< 0,001 in comparison to untreated controls).
In conclusion, minor alterations in cellular metabolism are provoked by exposure to polymeric NPs. However, the weak cytotoxic effect monitored by the LDH assay at the highest concentration was not confirmed by 7-AAD staining (Supplementary Figure [1](#MOESM1){ref-type="media"}). The IL-6 and IL-8 measurements indicated, similar to the experiments with SiO~2~ NPs, no inflammatory effects of the polymer NPs (Fig. [5](#Fig5){ref-type="fig"}). We hypothesize that the cells counter regulate an increase in apoptotic cell death, as supported by MTS, ATP and LDH assays, with an increase of the proliferative activity (AlamarBlue). According to this hypothesis, the results of the various test systems would be consistent and reveal a much more complex image than one test alone could show.Figure 5Release of IL-6 (**A**) and IL-8 (**B**) after treatment of HepG2 cells with increasing concentrations of polymer-NPs after 24 and 48 h. 250 μg/mL Aerosil 200-NPs und 20 ng/mL TNFα were used as controls. N = 3, technical duplicates, means and standard deviation (\*p \< 0.05; \*\*p \< 0.01 in comparison to untreated controls).
Uptake of NPs by HepG2 cells {#Sec24}
----------------------------
Independent investigations in two different laboratories (Lab 2 and Lab 3) confirmed the uptake of SiO~2~ and polymer NPs into HepG2 cells via CLSM imaging. The SiO~2~ NPs were internalized after 24 h with a perinuclear localization (Fig. [6A and B](#Fig6){ref-type="fig"}). However, the CLSM images of Lab 3 indicate an intracellular accumulation of the NPs as evidenced by the green fluorescence signal (Fig. [6A](#Fig6){ref-type="fig"}), Lab 2 detected beside large perinuclear aggregates, also a high interaction with the cell surface (Fig. [6B](#Fig6){ref-type="fig"}, as probed by the yellow fluorescence caused by colocalization of red (membrane) and green (nanoparticle) fluorescence.Figure 6Association of SiO~2~ NPs with cultured HepG2 cells. (**A**) Confocal laser scanning microscopy images of HepG2 after treatment with 500 µg/mL SiO~2~ NPs for 24, 48 and 72 h. (Lab 3). (**B**) Confocal laser scanning microscopy images of HepG2 after treatment with 500 µg/mL SiO~2~NP for 24, 48 and 72 h. (Lab 2). (**C**) Flow cytometry analysis of the uptake after exposure from the experiment conducted by Lab 2. Blue: nucleus, red: cell membrane, green: NPs (\*\*p \< 0.01; \*\*\*\*p \< 0.0001, compared to untreated controls).
A possible explanation for the discrepant results of the two laboratories could be different gain adjustment of the signal detectors (photomultiplier tubes) as well as different surfaces of the cover slips used for cultivation of cells before analysis by CLSM (a parameter which was not harmonized beforehand). Indeed, SiO~2~ NP were detected as larger agglomerates on polymer (ibidi) coated compared to regular glass cover slips (Fig. [6B](#Fig6){ref-type="fig"}, 24 h), which might affect cellular localization and uptake. A control experiment without cells confirmed the finding that SiO~2~ NPs were more agglomerated on the ibidi than the glass surface (Supplementary Figure [2](#MOESM1){ref-type="media"}).
The polymer NPs also showed a perinuclear internalization pattern inside of HepG2 cells (Fig. [7](#Fig7){ref-type="fig"}). In contrast to the SiO~2~ NPs, the polymer NPs were more adherent to the glass and less to the ibidi surface (Fig. [7A](#Fig7){ref-type="fig"}, 48 h). Possibly, this is caused by a repulsion effect between the polymer NPs and the polymer coated ibidi slides. These results show that the outcome of uptake experiments can also depend on the appropriate combination of slide surfaces and nanoparticle chemistry.Figure 7Internalisation of polymer NPs into cultured HepG2 cells. Confocal laser scanning microscopy images of HepG2 after treatment with 500 µg/mL polymer-NP for 24 and 48 h in (**A**) Lab 3 and in (**B**) Lab 2. (**C**) Flow cytometry analysis of the uptake from the experiment conducted by Lab 2. Blue: nucleus, red: cell membrane, green (\*\*p \< 0.01; \*\*\*p \< 0.001; \*\*\*\*p \< 0.0001 compared to untreated controls).
Flow cytometry data for the SiO~2~ and polymer NPs further confirmed that the NPs were internalized after 24 and 48 h, if one considers the results of each laboratory independently. Nevertheless, differences in the extent of nanoparticle accumulation in cells between labs 2 and 3 were observed (Figs [6](#Fig6){ref-type="fig"} and [7](#Fig7){ref-type="fig"}).
Discussion {#Sec25}
==========
This is the first attempt to cross-validate different assays used for nanoparticle uptake and toxicity studies. All these tests have been employed in the research on nanoparticle toxicity in the literature before, but a direct comparison of their performance in a round robin approach using specified experimental conditions has not been attempted to our knowledge. Here the laboratories involved in the intercomparison study of the DFG Priority Program SPP1313 have agreed on using this array of tests in order to consistently determine their use and differences. Importantly, particular emphasis was devoted to the colloidal stability of the nanoparticle formulations. An accurately defined protocol for the nanoparticle transfer from water into DMEM/FCS led to reproducible colloids for all partner groups within the round robin experiment.
The fact that all NPs were colloidally stable is an important issue for many biological experiments involving NPs^[@CR4]^. In fact, NP size alterations (in most cases due to agglomeration) will influence its effective density, settling rate and ultimately the effective dose delivered to cells with consideration of a given density and viscosity of the suspension medium. Due to the high ionic strength of cell culture media components, colloids, which are not perfectly stabilized against any interparticular interaction, will form agglomerates with distinctly altered diffusion and sedimentation properties compared to the primary particles. This may lead to a situation where the local NP concentration on the cell surface can be markedly higher than the initial nominal concentration simply by an increased contribution of particle sedimentation. The density of such agglomerates and hence, their sedimentation properties, can be distinctly different from those of the primary particles, i.e. loose agglomerates with a distinct amount of interstitial water sediment slower than compact agglomerates of the same material^[@CR23]^. Under these conditions, the factual dose rate to the cell membrane is not necessarily a simple function of the initial NP concentration but is significantly affected by NP stability^[@CR24]^. In case of the SiO~2~ particles (density ≈ 2 g/cm³), sedimentation and diffusion velocities were of the same order of magnitude and the largest agglomerates might have sedimented up to 5--6 times faster than the primary particles. Consequently slight deviations between the expected and the actual NP dose might have occurred. In case of polystyrene particles (density ≈ 1.04 g/cm³) sedimentation is negligible since their density is almost the same like that of water.
When particularly comparing NP cytotoxicity in 2D-cell cultures, one should play particular attention to the utilization of stable NP suspensions, the measuring of the NP size and effective density and the application of sedimentation models to be able to calculate the *actual* dose delivered to cells (for further details see refs [@CR25]--[@CR27]).
In our study, NP dosimetry was not the primary focus, since we aimed at comparing the NP impact on cells among different labs (and not of NP) using identical cell lines and culture medium, identical NP suspensions with verified colloidal stability. Due to the homogeneity of experimental parameters, the NP effective density, fate and transport behavior were expected to be comparable among all the labs involved.
Interestingly, it turned out in the inter-laboratory experiments that small variations of the transfer process for the SiO~2~ NPs can strongly affect the stability and, hence, aggregation state of the NPs in cell culture medium. It took several rounds of careful testing and refining, until NPs from the same batch had the same properties after re-dispersion in the individual labs. Many of the factors that were found to have a distinct impact on the state of the NPs are often not been considered in similar experiments involving single or fewer laboratories. Specifically, it was observed i. e., that the power of the ultrasonic bath, an incomplete immersion of the centrifuge tubes with the NP in the ultrasonic bath, different centrifugal accelerations, and even the geometry of the centrifuge tubes, i.e. the use of small (1--3 mL) conically shaped centrifuge tubes instead of round bottomed or larger conical ones resulted in substantially aggregated samples with hydrodynamic diameters of up to 1000 nm. The impact of sonication parameters such as power output, frequency and sample volume on nanoparticle agglomeration and stability in liquid dispersions has already been shown by other authors^[@CR28]--[@CR30]^. The observed drastic effect of the container shape on the sonication processes has not been reported so far. Utmost care should also be taken whenever mailing NPs between laboratories by standard mail, as the conditions during transport are frequently uncontrolled and not reproducible, if the carrier is not well selected. Low (less than \<0 °C) and high temperatures (\>40 °C) lead to a distinct alteration of the aqueous dispersion of silica nanoparticles. Only the careful elimination of all thermal impact on the samples leads to reliable results by thoroughly standardizing the transport and dispersion procedures. This allowed us to work with high standard samples. In this context it should also be pointed out that such problems can be strongly related to the individual properties of different NPs: the polymer NPs used in this study showed reproducible properties and excellent colloidal stability already in the first round of shipment to the individual groups indicating that they are less prone to be affected by environmental factors such as temperature variations.
From the results presented above we can conclude that the SiO~2~ NPs are not toxic in the tested concentration range. However, there are numerous reports on the potential of SiO~2~ NPs to induce autophagy in liver cells as a result of reactive oxygen species^[@CR31]^. Yet, in most cases where toxicity of amorphous silica is observed cells are exposed to NPs in the absence of fetal calf serum. Meanwhile it is well known that the presence of serum proteins prevents toxicity of silica NPs in epithelial cells^[@CR32],\ [@CR33]^ due to the formation of a protein corona^[@CR21]^ which most likely also explains the absence of adverse effects in our present study.
Interestingly the polymeric NPs showed in the MTS, ATP and LDH tests a trend towards a moderate toxicity, while the 7-AAD and the Alamar-Blue tests did not indicate this (Fig. [4](#Fig4){ref-type="fig"}). This could be due to the different test principles. 7-AAD will determine if the cell is capable of excluding the dye 7-AAD from the inner of each single cell. Therefore it is crucial that the cell is still detectable in the flow cytometer and has not disintegrated into cell debris which is regularly excluded from analysis. Also 7-AAD does include the absolute cell mass as all events on a forward scatter/sideward scatter plot are counted as cells and set as 100%. Of these the cells excluding 7-AAD actively by cell membrane processes are given. 7-AAD can also detect apoptotic and (early) necrotic cells. As stated above very late necrotic events as well as a decrease in proliferation may be missed. MTS and ATP tests in contrary assess the total amount of dehydrogenases or the "total energy level" as ATP. Here doubling or halving cell numbers will result in double or half of the corresponding reading. On the other hand when a cells metabolic capacity is impaired dehydrogenase levels and ATP content may also be decreased. For ATP every energy-consuming event -- like even nanoparticle uptake -- may decrease the amount of ATP^[@CR34]^. Therefore these tests are sensitive to several factors unlike the 7-AAD test.
Interestingly, cell viability data between the AlamarBlue test and MTS were slightly different (as shown in Figs [2](#Fig2){ref-type="fig"} and [4](#Fig4){ref-type="fig"}) even though both are based on the determination of cellular dehydrogenase activity. Yet, different enzymes might contribute to the metabolism of the dyes in both assays, which could account for the differences observed. Data in the literature about the comparison of different vitality assays are controversial. Hamid *et al*. and Benoit *et al*. reported that Alamarblue is slightly more sensitive with a better reproducibility and a cheaper redox indicator than XTT and MTT^[@CR35],\ [@CR36]^. On the other hand, the addition of some compounds to cells like nickel leads to the reduction of Alamarblue so that kinetic experiments may be unreliable unless strictly controlled^[@CR37]^. Overall, the lack of gross toxicity especially in the presence of serum and efficient uptake of carboxylated polystyrene NPs is in agreement with previous reports^[@CR38],\ [@CR39]^. Clearly adsorbed proteins also for polymeric nanoparticles govern uptake^[@CR40]^ and their intracellular fate^[@CR41]^.
The differences in the CLSM images between Lab 2 and 3 were not expected and were at first confusing. Although uptake was detected in both labs (Figs [4](#Fig4){ref-type="fig"} and [6](#Fig6){ref-type="fig"}), clearly Lab 2 detected more NPs on the cell membrane hereby indicating that NPs had agglomerated before they could have been taken up. After intense discussions we determined that both laboratories had adhered to the standardized parameters, but the choice of the slide chamber used for imaging was not determined beforehand as this was not suspected to have any influence. Interestingly, in the case of silica NPs we could show that nanoparticle agglomeration is higher on the ibidi slide while glass slides did not show this effect. Additionally, the differences of nanoparticle uptake detected by CLSM might be attributed to different sensitivity adjustment of detectors between the laboratories. Furthermore, the comparison of these data with flow cytometry, as a quantitative method, showed that flow cytometry was less sensitive at concentrations below 75 µg/mL due to cellular auto-fluorescence. Hence, using flow cytometry as single method could lead to misleading results.
The importance of interlaboratory evaluation of different toxicity assays in various cell lines has recently been appreciated and several studies approached this issue focusing on multiple aspects relevant for nanotoxicology^[@CR42]^. Our findings extend th**o**se efforts highlighting not only the relevance of harmonized procedures to prepare NP dispersions but also pinpointing the importance to use several independent toxicity assays and defined cell culture surfaces for proper particle imaging.
Conclusion {#Sec26}
==========
Our round robin test revealed that the outcome of a seemingly simple cytotoxicity assay might be highly dependent upon the respective protocol used in a given laboratory. Even apparently not relevant deviations, such as the choice of cell growth matrix material, have shown to be meaningful. Another aspect refers to the importance of utilization of clearly defined and characterized NP samples. Especially, the procedure to prepare reproducible and stable NP stock dispersions is influenced by multiple factors such as even the geometry of the tube used for resuspension. Moreover, comparative analyses of multiple *in vitro* screening methods are absolutely essential and reliance on a single assay could provide false positive or negative results. In summary, our findings could help in getting more consistent and conclusive data for hazard assessment of nanomaterials in the near future.
Electronic supplementary material
=================================
{#Sec27}
Supplementary Information
**Electronic supplementary material**
**Supplementary information** accompanies this paper at doi:10.1038/s41598-017-02958-9
**Publisher\'s note:** Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
This study was supported by the German Research Foundation (DFG) within the Priority Program Bio-Nano-Responses (SPP1313). We gratefully acknowledge technical assistance from M. Dröge and S. Burgold.
Nanoparticle synthesis and characterization: D.N., P.S., Q.G., J.G., S.G., L.T., C.G., K.L.; analysis of nanoparticle effects on cells: L.L., S.R., S.D., V.M., C.W.; manuscript preparation: L.L., V.M., K.L., C.W., R.Z., C. G., E.R., I.H.; final approval: all authors.
Competing Interests {#FPar1}
===================
The authors declare that they have no competing interests.
| {
"pile_set_name": "PubMed Central"
} |
Introduction
============
Chronic obstructive pulmonary disease (COPD) is a chronic progressive disease which is characterized by an inexorable decline in respiratory function, exercise capacity, and health status. This chronic clinical course of COPD is punctuated by periods of increased symptoms that are termed "acute exacerbations". Acute exacerbations are significant events in the course of the disease as they accelerate disease progression, impair quality of life, and are the major contributor to morbidity and mortality in COPD. In addition they are the major cause of excess health care costs as they often result in unscheduled health care visits, treatment costs, and hospitalizations. Therefore, preventing exacerbations is a major therapeutic goal that has not been achieved with currently available treatments. The major causes of exacerbations are respiratory infections with both viruses and bacteria. Although a role for infection in both stable COPD and COPD exacerbations has long been established, controversy remains regarding a number of issues including the effect of bacterial infection in stable COPD, the relative contributions of viruses and bacteria in COPD exacerbations, mechanisms of susceptibility to infection, and the role of antibiotics.
Microbial flora in the normal lung and in COPD patients
=======================================================
In order to investigate the role of infection in COPD, an understanding of the normal microbial flora in healthy individuals is required. Most of the surfaces of the upper respiratory tract (including nasal and oral passages, nasopharynx, and oropharynx) are colonized by bacteria in healthy individuals. These organisms constitute the normal flora of the respiratory tract and rarely cause disease. The nose is commonly colonized with predominantly S*taphylococcus epidermidis* and C*ornyebacteria* but S*taphylococcus aureus* can also be present. The nasopharynx is predominantly colonized with non-hemolytic and alpha-hemolytic *Streptococci* and *Neisseria* species with occasional carriage of *Streptococcus pneumonia* and *Haemophilus influenzae*.[@b1-copd-7-555] Determining the prevalence of bacteria in the lower respiratory tract has proved problematic due to the difficulty of obtaining lower respiratory tract samples uncontaminated by upper airway secretions. Bronchoscopy with protected brush specimens is the best method of avoiding upper airway contamination. A report by Rosell et al pooled the results of studies using bronchoscopy in healthy subjects and both stable and exacerbated COPD.[@b2-copd-7-555] In 70 healthy participants, potentially pathogenic microorganisms (PPMs) were cultured in only 6% of subjects and bacterial loads were low. Two subsequent studies not included in this analysis reported that PPMs were present in 10% of subjects undergoing diagnostic bronchoscopy in whom no pathology was found,[@b3-copd-7-555] and in 3% of a group of healthy smokers and ex-smokers.[@b4-copd-7-555] Therefore based on these findings, conventional wisdom has been that the lower respiratory tract in healthy individuals is sterile. However it has been estimated that over 70% of the bacterial species present on mucosal surfaces cannot be cultured using currently available standard culture techniques. Several recent studies have used novel molecular approaches for bacterial detection rather than standard culture techniques and have challenged the idea that the lower respiratory tract is predominantly sterile. These studies will be discussed in further detail in a later section.
In contrast to healthy individuals, bacteria can often be cultured from lower airway secretions in COPD patients. The pooled analysis by Rosell et al reported that PPMs are present at a concentration of 10^2^ cfu/mL or greater in 29% of stable COPD patients[@b2-copd-7-555] and Sethi et al detected bacteria in 34.6% of patients.[@b4-copd-7-555] No bacteria were present in patients with mild disease (GOLD criteria), but 24% of moderate patients, 45% of severe patients, and 33% of very severe patients had bacteria detected. This relationship between lung function impairment and bacteria has been reported in some studies[@b5-copd-7-555] but other studies have found no relationship.[@b3-copd-7-555],[@b4-copd-7-555],[@b6-copd-7-555]--[@b8-copd-7-555] Studies using sputum to detect bacterial infection have generally found higher rates of bacterial isolation than the bronchoscopy studies with bacteria found in up to half of stable COPD patients.[@b9-copd-7-555]--[@b11-copd-7-555] This is likely due to higher rates of contamination with upper respiratory organisms in sputum.
The most common organisms cultured in COPD from both sputum and bronchoscopic samples are consistently *Haemophilus influenza*, *Streptococcus pneumonia*, and *Moraxella catarrhalis*. However the bacterial flora in COPD varies with disease severity as Gram negative organisms such as *Pseudomonas aeruginosa* are more commonly detected in patients with more severe airflow obstruction in both stable COPD[@b12-copd-7-555],[@b13-copd-7-555] and in exacerbations.[@b14-copd-7-555]--[@b16-copd-7-555] Other organisms detected less frequently include *Haemophilus parainfluenzae* and *Staphylococcus aureus* but the significance of these organisms is debated ([Table 1](#t1-copd-7-555){ref-type="table"}).
Respiratory microbiome
======================
Our current understanding of the role of bacteria in the pathogenesis and progression of obstructive airways disease is predominantly based upon classical microbial culture techniques. However, these suffer from a number of limitations, are very labor intensive, and using standard conditions can culture only 30% of bacteria.[@b17-copd-7-555] It is therefore understandable that historically the lungs have been considered sterile, despite their continuity with the upper airways, proximity to the gastrointestinal tract, and continuous exposure to the environment. Over the past decade molecular culture independent techniques, initially developed in the ecology field, have identified bacteria previously not amenable to culture.[@b18-copd-7-555] These techniques when combined with advances in sequencing technologies have produced a powerful tool for investigating the role of bacteria in health and disease and have recently begun to shed more light on the role of bacteria in COPD.
Molecular tools rely on genomic evolutionary relationships between bacteria and use similarities in housekeeping genes, such as the highly conserved 16S rRNA gene, to assign phylogeny. These techniques can not only be used to rapidly identify individual bacterial species but also to build up a picture of the complete microbial community in an environment (the microbiome), offering a more comprehensive analysis than classical culture-based techniques.
Hilty et al employed these techniques to clearly demonstrate the presence of a wide variety of bacteria in the airways of healthy non-smoking subjects, establishing the existence of a respiratory microbiome and challenging the dogma of lung sterility.[@b19-copd-7-555] Their study also suggested the healthy respiratory microbiome differs from that associated with a number of respiratory diseases including COPD.
COPD is linked inherently to smoking and we must therefore first look at the effects of smoking on the respiratory microbiome. Interestingly, three studies have now found no overall difference in total bacterial levels between lower airways samples from smokers, non-smokers, and COPD subjects.[@b19-copd-7-555]--[@b21-copd-7-555] However Charlson et al have demonstrated changes in bacterial communities in the upper respiratory tract of smokers, with significant loss of bacterial diversity when compared to the oro-and naso-pharynx of nonsmokers. [@b22-copd-7-555] Whether this represents outgrowth of an organism from a healthy community leading to loss of diversity remains unclear, but loss of microbial diversity has been associated with an increased incidence of disease in other systems.[@b23-copd-7-555]
Although the number of studies examining the lower airways microbiome in COPD is limited, it is already clear that there is a significant overlap between the bacteria seen in COPD and healthy individuals.[@b19-copd-7-555],[@b20-copd-7-555],[@b24-copd-7-555] This has led some authors to postulate the existence of a core respiratory microbiome made up of *Pseudomonas, Streptococcus, Prevotella, Fusobacteria, Haemophilus*, and *Veillonella* species whose relative abundance differs between health and disease.[@b25-copd-7-555] Certainly, in the initial study by Hilty et al, *Haemophilus* species were most strongly associated with the presence of COPD,[@b19-copd-7-555] while the dominance of *Pseudomonas* species and associated lower diversity of the microbiome was observed in subjects with moderate or severe COPD but not in mild disease.[@b20-copd-7-555] However, a recent study by Sze et al did not observe any reduction in microbial diversity in lung tissue from patients with very severe COPD (GOLD stage IV). Although they do report evidence of a unique bacterial community structure in this group with a significantly greater abundance of Firmicutes compared to smoking and non-smoking controls.[@b21-copd-7-555]
Using explanted lungs, Erb-Downward and colleagues were able to characterize the microbial communities at different locations without the need for bronchoscopy, thereby avoiding any potential upper airways contamination. They demonstrated bacterial communities similar to those seen in bronchoalveolar lavage (BAL), and interestingly demonstrated significant micro-anatomical differences in community composition within the lungs. Highlighting the power of technology, the segmental bronchus of the left upper lobe bronchus in one subject was dominated by *Haemophilus*, while in the distal bronchus of the left upper lobe bronchus, *Stentrophomonas* dominated the community.[@b20-copd-7-555]
Molecular microbiology has currently raised more questions than it has answered in the respiratory field. These techniques have clearly demonstrated the presence of PPMs in the lungs of healthy smokers and COPD subjects and have begun to identify differences between the bacterial communities in these groups and between severe and mild disease. However, it remains unclear what impact this could have on our understanding of the disease pathogenesis and progression. How does the microbiome change before, during, and after an exacerbation? What effect do viral infections have on the bacterial communities? Further studies are clearly needed to answer these questions but with these tools and the evolving field of metatranscriptomics, we have the opportunity to study the role of bacteria in obstructive lung disease more comprehensively than ever before.
Significance of bacterial infection in stable COPD
==================================================
The presence of bacteria in the lower respiratory tract in stable COPD patients is usually termed "colonization" rather than "infection", implying that the bacteria present have no or minimal pathological significance. However, studies have now established that there may be relationships between the presence of bacteria and both airways inflammation and adverse clinical outcomes in COPD patients. Therefore the term "colonization" may be misleading and the presence of bacteria in COPD patients may not be as benign as previously thought.
Increased levels of cellular and soluble inflammatory markers in the airways including neutrophils, CXCL8, leukotriene B~4~ (LTB~4~), myeloperoxidase (MPO), matrix metalloproteinase (MMP), tumor necrosis factor-α (TNF-α), IL-1β, IL-6, IL-12, and neutrophil elastase (NE) have been reported in stable COPD patients with bacterial infection.[@b4-copd-7-555],[@b6-copd-7-555],[@b8-copd-7-555],[@b11-copd-7-555],[@b26-copd-7-555]--[@b29-copd-7-555] In addition two studies have reported a quantitative relationship between bacterial load and sputum inflammatory markers.[@b6-copd-7-555],[@b9-copd-7-555] Bacterial infection is also associated with adverse clinical outcomes with an increased frequency of exacerbations,[@b9-copd-7-555],[@b11-copd-7-555],[@b30-copd-7-555] impaired health status,[@b26-copd-7-555],[@b29-copd-7-555] and systemic inflammation.[@b26-copd-7-555],[@b29-copd-7-555] One study did not find an association between bacterial infection and exacerbation frequency.[@b30-copd-7-555] Therefore these studies would appear to provide evidence that bacterial infection, even in patients who are clinically stable, causes airways and systemic inflammation, increases exacerbations, and impairs health status. However, as these studies were cross-sectional they could not determine the direction of the association between bacterial colonization and the outcomes measured. Bacterial infection may cause increased airways inflammation but reverse causation may be equally plausible, ie, patients with greater airways inflammation may be more susceptible to developing bacterial infection. Parameswaran et al attempted to resolve this issue this by studying the effect of the acquisition of a new strain of *Moraxella catarrhalis* on sputum inflammatory markers.[@b31-copd-7-555] Sputum samples collected after acquisition of *Moraxella catarrhalis* had higher levels of TNF-α, CXCL-8, and NE compared to pre-acquisition samples, supporting the hypothesis that bacteria induce airways inflammation. However, inflammatory markers in subjects who did and did not develop infection were not compared prior to acquisition so the hypothesis that airways inflammation predisposes to bacterial infection is still possible.
This study also highlighted the fact that bacterial colonization/infection is not a static phenomenon. This was further demonstrated by Wilkinson et al who studied 30 stable COPD patients over 1 year and collected sputum at the start and end of this time period.[@b10-copd-7-555] Fifty three percent of subjects had bacteria present in sputum at the start of the study. After 1 year half of these had the same bacterial species but in the other 50% of subjects a different bacterial species was present. There was a significant increase in the mean bacterial load over the year and this correlated with a reduction in the forced expiratory volume in 1 second (FEV~1~), as did acquisition of a new bacterial species. A study of 40 COPD patients over an 8-month period used molecular typing to identify the strain of sputum organisms. Fifty-three percent of patients had PPMs in sputum at both time points but only 15% had the same organism, whereas 38% had acquired a new organism.[@b27-copd-7-555]
Therefore, both the presence and type of bacteria in the airways in stable COPD are not static phenomena and may be affected by factors such as disease progression, exacerbations, and treatments including antibiotics and inhaled corticosteroids. Conversely the presence of bacteria can influence disease progression and exacerbations in COPD patients. The relationships between clinical outcomes and bacteria in COPD are undoubtedly complex and further studies to investigate these should be long-term longitudinal studies using strain typing in addition to culture. If a causal relationship between the presence of bacteria and adverse outcomes in stable COPD patients is conclusively demonstrated, then intervention studies with antibiotics in stable COPD may be warranted.
Mechanisms of susceptibility to bacterial infection in COPD
===========================================================
In healthy individuals, bacteria are constantly inhaled but infection does not develop as a result of sophisticated host defenses. These include mechanical factors such as tight epithelial lining cells and mucociliary clearance, antimicrobial peptides (including pentraxin-3, lysozyme, lactoferrin, defensins, elaf in, secretory leukoprotease inhibitors \[SLPI\], and cathelicidin), local immune responses such as secretory IgA, resident phagocytes such as airway macrophages, and acquired immune responses.[@b32-copd-7-555] The presence of bacterial infection in the lower airways suggests that host pulmonary defenses are impaired and a number of abnormalities in pulmonary immune mechanisms have been reported in COPD. Cigarette smoking has a number of effects on the mechanical barriers in the lung that may favor the development of infection. Smoking reduces ciliary beat frequency, induces squamous metaplasia resulting in reduced numbers of ciliated cells, and increases goblet cells and submucosal mucous glands.[@b33-copd-7-555] This is likely to result in excess production and impaired clearance of mucous from the airways, which favors bacterial growth. Impaired mucociliary clearance[@b34-copd-7-555] and ciliary abnormalities have been reported in smokers and COPD[@b35-copd-7-555] and therefore these may be factors contributing to bacterial infection in COPD. There is also evidence that cigarette smoke damages the epithelial junctions leading to increased epithelial permeability.[@b36-copd-7-555]
Another mechanism that has been extensively investigated is impaired function of airway immune cells such as neutrophils and alveolar macrophages in COPD. These cells play a key role in removing microorganisms present in the airways by phagocytosis and initiating immune responses. Absolute numbers of macrophages and neutrophils in the airways are actually increased in smokers and COPD patients, and therefore, research has focused on whether macrophage function in COPD is impaired. A number of studies have reported that macrophage phagocytosis of microorganisms is impaired in COPD[@b37-copd-7-555]--[@b42-copd-7-555] and a number of mechanisms of impaired function of macrophages in COPD have been investigated. Reduced levels of the toll-like receptor TLR2 have been reported in macrophages[@b43-copd-7-555] and neutrophils[@b44-copd-7-555] from COPD patients. MARCO (macrophage receptor with collagenous structure) is a class A scavenger receptor expressed on macrophages that mediates binding and uptake of Gram-positive and Gram-negative bacteria.[@b45-copd-7-555] Cigarette smoking reduces the expression of MARCO on macrophages resulting in impaired responses to infection in vitro[@b46-copd-7-555] and in vivo in mice.[@b47-copd-7-555] Interest in macrophage function has increased recently with studies that have shown that macrophage function can be restored with pharmaceutical agents,[@b42-copd-7-555],[@b47-copd-7-555]--[@b49-copd-7-555] raising the possibility that improving the phagocytic capacity of macrophages is a potential therapeutic option in COPD.
Antimicrobial peptides (AMPs) are soluble molecules present in the airway surface fluid that constitute an important first line of defense against both bacterial and viral pathogens. AMPS include surfactant proteins (SP), defensins, elafin, cathelicidin, SLPI, and lysozyme. Studies of AMPs in COPD have had varying results with both increased[@b50-copd-7-555] and decreased[@b51-copd-7-555],[@b52-copd-7-555] human β-defensin, increased α-defensin,[@b53-copd-7-555] and increased SLPI and elafin[@b52-copd-7-555] reported. Therefore, it remains unclear whether reduced levels of AMPs contribute to impaired immune responses in COPD.
α1-Antitrypsin (AAT) is a member of the serine protease inhibitor (*SERPIN*) supergene family and AAT deficiency is associated with the development of early-onset COPD. The key function of AAT is the inhibition of proteases such as neutrophil elastase, and excess protease activity is considered the key mechanism underlying the development of COPD in AAT deficiency. New evidence has emerged that, in addition to anti-protease activity, AAT possesses anti-inflammatory, immunomodulatory, and both antibacterial[@b54-copd-7-555] and antiviral properties.[@b55-copd-7-555] Therefore, it is possible that AAT deficiency contributes to susceptibility to infection in COPD. AAT deficient patients experience frequent exacerbations[@b56-copd-7-555] but there is no evidence that infections are more frequent than non-AAT deficient patients. AAT augmentation reduces exacerbations and this will be discussed further in the "Novel therapies" section.
Animal models of bacterial infection in COPD
============================================
Since there are no existing human models of experimental bacterial infection in COPD, there has been considerable interest in the use of animal models to investigate the role of bacterial infection in COPD. Animal models use three main approaches to establishing a COPD phenotype: inhalation of noxious stimuli (most commonly cigarette smoke exposure), instillation of tissue-degrading proteinases such as elastase, and genetic manipulation.[@b57-copd-7-555],[@b58-copd-7-555] Gaschler et al administered cigarette smoke to mice for 8 weeks and subsequently challenged intranasally with nontypeable *H. influenzae* (NTHI). This led to increased pulmonary inflammation, upregulation of inflammatory mediator expression but decreased bacterial burden compared to control mice.[@b59-copd-7-555] Drannik et al similarly showed that cigarette smoke altered the response to infection with *Pseudomonas aeruginosa* with a worsening of clinical score, increased inflammation, and increased expression of pro-inflammatory cytokines.[@b60-copd-7-555] Although cigarette smoke-induced models have the advantage of using the primary disease-causing agent, even with prolonged exposure, only mild pulmonary abnormalities are seen in mice, equivalent to human GOLD stage 1 or 2 disease.[@b58-copd-7-555],[@b61-copd-7-555] Therefore, studies using such models may be of less relevance to the study of bacterial infection which occurs more frequently in patients with more severe disease. Instillation of proteinases such as porcine pancreatic elastase produces more severe emphysematous changes which mimic those seen in advanced disease and therefore other studies have chosen to use these techniques to model bacterial COPD exacerbations.
Pang et al used a mouse model of elastase-induced emphysema to show that pulmonary clearance of NTHI was impaired.[@b62-copd-7-555] They also showed that intercellular adhesion molecule 1 (ICAM-1) was downregulated in the airway epithelium of elastase-treated mice and suggested that this may provide an underlying mechanism for the impaired bacterial clearance observed. However, notably, they did not show any differences in localized lung inflammation between the two groups and therefore, whether delayed bacterial clearance leads to clinically relevant exacerbation of disease in this model is unclear. Ganesan et al used a more complex model system of combined lipopolysaccharide and elastase administration to mimic airway inflammation and emphysema in mice.[@b63-copd-7-555] When challenged with NTHI, elastase/LPS mice showed delayed bacterial clearance with an increase in neutrophilic inflammation and prolonged mucus secretion assessed by mucin gene expression and periodic acid schiff (PAS) staining on lung histology. Furthermore, ex vivo macrophages taken from elastase/LPS mice showed deficient phagocytosis and this was demonstrated to be due to decreased expression of scavenger receptor A. This receptor has previously been shown to be important in bacterial clearance in mice[@b64-copd-7-555] and genetic mutations are also associated with increased risk of developing COPD in humans.[@b65-copd-7-555] However, the direct applicability of these findings to mechanisms leading to susceptibility to bacterial infections in patients with COPD requires further characterization.
Developing animal models of chronic bacterial infection has been difficult due to rapid clearance of bacterial agents by the rodent immune system, typically within 1--2 days of administration.[@b57-copd-7-555] Wang and colleagues used a hamster model of elastase-induced emphysema and administered *H. influenzae* encased in agar beads.[@b66-copd-7-555] This led to prolonged survival and persistence of bacteria for up to 3 weeks in emphysematous airways. Such models may be invaluable in the future to aid the study of bacterial colonization in COPD and its relevance to exacerbations and disease pathogenesis. Furthermore, studies using molecular techniques are also ongoing to assess the bacterial flora in the lower respiratory tract of COPD mouse models. It remains to be seen whether this will correlate with the emerging evidence that bacterial communities are disordered in human patients with COPD.[@b19-copd-7-555],[@b20-copd-7-555]
In summary, animal models have provided some useful insights into the mechanisms of host defense against bacterial infection in COPD. However, COPD is a complex, multisystem disorder that is incompletely understood in humans and therefore, the direct relevance of animal studies to clinical disease remains unclear. Future studies should aim to further evaluate which aspects of human bacterial COPD exacerbations existing models accurately mimic, as well as translating findings from existing animal work into clinical disease studies.
Bacteria and COPD exacerbations
===============================
Perhaps equally controversial to the role of bacteria in the pathogenesis of stable COPD is the contribution of bacteria to COPD exacerbations.[@b67-copd-7-555] Bacteria are often detected in COPD exacerbations but the high isolation rates of bacteria in stable COPD means the presence of bacteria does not prove a causative role. As bacteria are often present in patients who are clinically stable, studies have attempted to determine whether bacterial infection does in fact induce exacerbations and the potential mechanisms underlying this.
New acquisition of bacteria
===========================
Detection rates of bacterial infection in samples collected during exacerbations are high, but as bacteria are often present in stable COPD, this does not prove that the organism is newly acquired and caused the exacerbation. There are surprisingly few studies directly comparing infection rates in stable and exacerbated patients. Two studies from the East London COPD cohort reported infection rates of 48.2% and 43% in sputum in stable patients compared to 69.6% and 76% in exacerbated patients[@b68-copd-7-555],[@b69-copd-7-555] (*P* = 0.009 for the study by Hurst et al,[@b69-copd-7-555] *P-*value not provided for the study by Wilkinson et al[@b68-copd-7-555]). The pooled analysis by Rosell et al detected bacterial infection in 29% of protected brush specimens from stable COPD patients and 54% of exacerbated patients (*P* \< 0.001).^2^ However, Papi et al reported a non-significant higher incidence of bacterial detection in exacerbated patients (54.7% vs 37.5% when stable, *P* = 0.08)[@b70-copd-7-555] and Bafadhel et al detected bacterial infection in 28% of stable patients and 35% of exacerbations (*P*-value not provided).[@b71-copd-7-555] Therefore, not all studies have conclusively demonstrated that infection rates are higher in exacerbated patients.
Sethi et al analyzed changes in bacteria in 81 patients over a 56-month period using both culture and molecular typing methods.[@b72-copd-7-555] Isolation of a bacterial pathogen by culture was associated with a significant increase in the incidence of exacerbations. An exacerbation was present in 23.6% of visits at which pathogens were isolated from sputum compared with 18% of visits at which no pathogens were isolated from sputum (*P* \< 0.001). An analysis of individual bacterial species showed that isolation of *M. catarrhalis* and *S. pneumoniae* was associated with a significant increase in the frequency of exacerbations whereas *H. influenzae, P. aeruginosa,* and Gram-negative bacilli were not. However, as can be seen from these data, the majority of visits at which a pathogen was detected were not associated with an exacerbation. Therefore, acquisition of bacterial infection in a COPD patient does not automatically result in an exacerbation and it is likely that there are both host and pathogen factors that determine the outcome of bacterial infection. Further studies are needed to identify the key factors that determine the outcome of bacterial infection in COPD.
Increased bacterial load
========================
If acquisition of a new organism does not occur in all exacerbations, another possible mechanism of exacerbation is an increase in the load of a pre-existing organism present in stable COPD patients. Two studies from the East London cohort reported a significant increase in bacteria load at exacerbation compared to the stable state.[@b68-copd-7-555],[@b69-copd-7-555] Conversely, Sethi et al reported no significant difference between concentrations in sputum during stable disease and exacerbation for *H. influenzae*, *H. haemolyticus*, and *M. catarrhalis* and actually found an inverse relationship between exacerbation occurrence and bacterial concentrations in sputum for *S. pneumoniae* and *H. parainfluenzae*.[@b73-copd-7-555] When paired comparisons of concentrations of the same strain of pathogens isolated from the same patient during exacerbations and stable visits was performed, only *H. influenzae* was present at a higher concentration in exacerbation samples. Therefore, whether increase in bacterial load is a major mechanism of exacerbation in COPD remains unclear.
New bacterial strain
====================
Detection of bacteria by culture does not distinguish between different strains of the same organism. Acquisition of a new strain of a bacterium has been investigated as another mechanism of exacerbation. In the study by Sethi et al described previously, isolation of a new bacterial strain was associated with a significant increase in the frequency of exacerbation. Thirty-three percent of visits at which new strains were isolated were associated with exacerbations, compared with 15.4% of visits at which no new strains were isolated (*P* \< 0.001).[@b72-copd-7-555] The relative risk of an exacerbation in association with the isolation of a new strain of *H. influenzae* was 1.69 (*P* \< 0.001), whereas there was no association between the isolation of *H. influenzae* by culture and an exacerbation. The relative risk of an exacerbation in association with the isolation of a new strain was 2.96 for *M. catarrhalis* and 1.77 for *S. pneumoniae* so strain analysis strengthened the association seen with culture alone. New strains of *P. aeruginosa* were not associated with exacerbations. Further evaluation of this study reported that bacteria identified as variant *H. influenzae* were subsequently characterized as *Haemophilus haemolyticus*.[@b74-copd-7-555] Re-assessment of the results accounting for the previously misinterpreted *H. haemolyticus* showed that new strain acquisition of *H. influenzae* was associated with a four-fold increase in the incidence of COPD exacerbation. Additional follow-up of these patients subsequently also reported a significant association between acquisition of *P. aeruginosa* and exacerbation contrary to what was previously reported[@b13-copd-7-555] and a study of *M. catarrhalis* determined that acquisition of a new strain is associated with 10% of exacerbations.[@b75-copd-7-555] In another publication from this group, the clinical and inflammatory responses associated with acquisition of a new strain of *H. influenzae*, *S. pneumoniae, M. catarrhalis*, and *P. aeruginosa* were compared with exacerbations in which a pre-existing strain of one of these four species was present, exacerbations in which *S. aureus* or Gram-negative bacilli other than *P. aeruginosa* were isolated and pathogen-negative exacerbations.[@b76-copd-7-555] Of the 150 exacerbations, 26% were new strain, 10% were pre-existing strain, 20% were other pathogen, and 44% were pathogen-negative exacerbations. New strain exacerbations had the largest increase in airway inflammation with significantly greater changes in TNF-α and NE when compared with pathogen-negative and other pathogen exacerbations and serum CRP levels were significantly higher with new strain exacerbations in comparison to each of the other three groups. New strain exacerbations were also associated with greater clinical symptom scores.
Chin et al studied the pro-inflammatory effects of different strains of *H. influenzae* and demonstrated that strains isolated during exacerbations caused more airway neutrophil recruitment in a mouse model of bacterial infection and more inflammatory pathway activation in primary human airway epithelial cells cultures when compared with bacterial strains isolated from stable patients.[@b77-copd-7-555] Therefore, these data suggest that acquisition of a new bacterial strain is likely to be an important mechanism linking bacterial infection and COPD exacerbations, and that the risk of exacerbation differs with different *H. influenzae* strains.
Thus, the role of bacteria in COPD exacerbations is complex and the previously held concept of a newly acquired bacterial infection causing an exacerbation based on studies using sputum culture is likely to prove simplistic. New paradigms have emerged from studies using more sophisticated molecular techniques such as strain typing, and as these become more widely available, our understanding of the role of bacteria is likely to evolve further.
Virus infection in COPD
=======================
Historically, bacterial infections have been considered the predominant infectious causative agents of COPD exacerbations and this is reflected in the widespread use of antibiotics in their treatment. Epidemiological data reporting a greater frequency of exacerbations in the winter months,[@b78-copd-7-555] and frequent coryzal symptoms preceding exacerbations suggest that exacerbations may also be associated with respiratory virus infections.[@b79-copd-7-555] Studies investigating the role of viruses in COPD exacerbations carried out in the 1970s and 1980s detected viruses in only \~10%--20% of exacerbations,[@b80-copd-7-555],[@b81-copd-7-555] casting doubt on the role of virus infection. However, the diagnostic methods used in these studies had low sensitivity, especially for the detection of rhinoviruses, which are the most common cause of viral upper respiratory tract infections. Following the development of PCR-based techniques for the detection of respiratory viruses in clinical samples, the role of viruses in a number of clinical syndromes including COPD exacerbations has been re-evaluated. More recent studies using PCR have detected the presence of a virus in 47%--56% of exacerbations.[@b70-copd-7-555],[@b82-copd-7-555]--[@b85-copd-7-555] In most studies, picornaviruses (predominantly rhinoviruses) were the most frequently detected viruses, followed by influenza, parainfluenza, respiratory syncytial virus (RSV), and adenoviruses.[@b86-copd-7-555] Therefore, these reports suggest that up to half of COPD exacerbations are associated with respiratory virus infection. However, the role of virus infection in COPD exacerbations continues to be debated for a number of reasons and some authorities have questioned whether virus infections can cause exacerbations in their own right or whether they simply predispose to secondary bacterial infection. PCR-based diagnostic techniques are able to detect very small amounts of viral RNA or DNA and therefore do not definitively prove the presence of live virus or prove a causative relationship. As is the case with bacteria, viruses can be detected in stable COPD but this has been studied to a much lesser extent and detection rates have varied from 0%[@b83-copd-7-555],[@b87-copd-7-555] to 19%.[@b82-copd-7-555] Few studies have investigated both viral and bacterial infection in the same exacerbations and detection rates have varied from 6%--25%.[@b70-copd-7-555],[@b85-copd-7-555] However, it is also possible that the role of viral infections in COPD exacerbations has been underestimated, as patients are evaluated at the time of presentation which often occurs considerably later than the onset of exacerbation and viruses may no longer be detectable by this time point. As rapid diagnostic methods and antiviral agents become available, the relationship between virus infections and COPD will no longer be of just academic interest but will have potential therapeutic implications and therefore warrants further study.
Experimental rhinovirus infection in COPD
=========================================
A novel tool for investigating relationships between virus infection and COPD exacerbations is experimental rhinovirus infection. Experimental infection studies have been previously conducted in asthmatics and yielded important insights into the mechanisms linking virus infection to exacerbations in asthma.[@b88-copd-7-555],[@b89-copd-7-555] Our group has recently reported the first experimental rhinovirus infection study in COPD patients.[@b90-copd-7-555] COPD patients and non-obstructed control subjects were infected with rhinovirus followed by sequential measurement of symptoms, lung function, inflammatory markers, and virus load. Rhinovirus infection induced symptomatic colds that were followed by the typical features of a COPD exacerbation, ie, lower respiratory symptoms, increased airflow limitation, and airways inflammation. Virus was detected in airway samples following inoculation but prior to the onset of symptoms and viral clearance was followed by symptom resolution and return of inflammatory markers to baseline levels. Virus load correlated strongly with inflammatory markers and the rhinovirus was grown from airway samples, confirming the presence of live virus. Therefore, this study is the first to directly link respiratory virus infection to symptoms, airflow obstruction, and airways inflammation in COPD patients so providing novel evidence that rhinovirus infection causes COPD exacerbations.
Virus infection and stable COPD
===============================
The majority of studies that have investigated virus infection in both exacerbated and stable patients have detected viruses at a greater frequency during acute exacerbations compared to stable state. However, there is some evidence that RSV is detected more frequently in stable patients with one study reporting a frequency of around 25% in the stable state.[@b79-copd-7-555] A further study found RSV in 30% of sputum samples from stable patients, with detection being related to greater airway inflammation and to a faster decline in lung function.[@b91-copd-7-555] However, other studies have not reported high rates of RSV infection in stable COPD patients.[@b70-copd-7-555],[@b82-copd-7-555],[@b92-copd-7-555] A study comparing virus loads between infants with acute respiratory infections and adult COPD patients found that virus loads were 2000-fold higher in the infants, suggesting low-grade virus infection in COPD.[@b93-copd-7-555] The disparity between these findings is likely to be due to a combination of factors including differing sensitivity of PCR techniques used, differences in severity of COPD patients included, or differences in populations studied.[@b94-copd-7-555] Latent infection by adenovirus has also been proposed to be involved in the pathogenesis of COPD. Lung tissue from COPD patients has been demonstrated to carry more group C adenoviral DNA than matched non-obstructed smokers.[@b95-copd-7-555] Latent adenoviral infection in combination with cigarette smoke exposure in a guinea pig model caused an increase in lung volumes, airspace volume, and reduced surface to volume ratio compared to smoke exposure alone.[@b96-copd-7-555] Additionally, adenovirus detection has been shown to be similar in exacerbated and stable COPD patients.[@b97-copd-7-555] Some authors have postulated that the presence of RSV and adenovirus in stable COPD may contribute to the pathogenesis of the disease as there are some common pathologic features between respiratory viral infection and COPD including a predominance of CD8+ T lymphocytes. However, this remains a largely unproven hypothesis.
Relationships between exacerbation etiology and clinical and inflammatory parameters in COPD
============================================================================================
COPD exacerbations are associated with systemic and pulmonary inflammation and increased levels of inflammatory mediators and cells have been measured in airway samples including tumor necrosis factor-alpha (TNF-α),[@b98-copd-7-555] IL-8,[@b98-copd-7-555],[@b99-copd-7-555] IL-6,[@b100-copd-7-555] leukotriene B4,[@b101-copd-7-555] and neutrophils,[@b70-copd-7-555],[@b99-copd-7-555],[@b102-copd-7-555],[@b103-copd-7-555] lymphocytes,[@b99-copd-7-555],[@b103-copd-7-555] and eosinophils.[@b70-copd-7-555],[@b99-copd-7-555],[@b103-copd-7-555] Few studies have examined the relationships between the clinical and inflammatory parameters and specific pathogens. It has been reported that virus-positive exacerbations are associated with more symptoms, greater falls in lung function,[@b79-copd-7-555],[@b104-copd-7-555],[@b105-copd-7-555] higher sputum levels of IL-6,[@b104-copd-7-555],[@b105-copd-7-555] and IP-10,[@b106-copd-7-555] however, these studies did not perform bacterial culture. Conversely, others have reported that only exacerbations associated with purulent sputum (presumed bacterial) are associated with airways inflammation and greater falls in lung function[@b68-copd-7-555],[@b101-copd-7-555],[@b107-copd-7-555],[@b108-copd-7-555] and that bacterial load correlates with levels of inflammatory markers,[@b108-copd-7-555] but did not perform virological analysis. Studies that directly compared clinical and inflammatory parameters in bacteria and virus-associated exacerbations have had contrasting results. Two studies reported that the presence of bacteria but not viruses was associated with airways inflammatory markers such as neutrophils, IL-8, and TNF-α.[@b69-copd-7-555],[@b109-copd-7-555] Papi et al found no differences in clinical outcomes between bacteria-and virus-associated exacerbations. Both were associated with increased neutrophils and neutrophil elastase but only virus-associated exacerbations had significantly increased sputum eosinophils.[@b70-copd-7-555] However, Bafadhel et al reported bacterial exacerbations were associated with increased sputum total cell count, neutrophils and eosinophils, and CXCL-8, TNF-α, IL-1β, CCL5, CCL17, CXCL11, CCL3, and CCL13, whereas virus-associated exacerbations had greater falls in lung function and increased CCL5, CXCL10, and CXCL11.[@b71-copd-7-555] Sethi et al compared the inflammatory responses to different bacteria species in COPD exacerbations and reported that *H. influenzae* and *M. catarrhalis* were associated with inflammatory markers in sputum whereas the inflammatory profile of *H. parainfluenzae*-associated exacerbations was no different from pathogen-negative exacerbations.[@b108-copd-7-555] Therefore, no clear pattern of clinical features or inflammatory markers has emerged that is specific to either bacterial or viral infection in COPD exacerbations. It is possible that the inflammatory pathways activated are the same irrespective of exacerbation etiology; however, studies are needed with intensive sampling and advanced diagnostic methods for both bacteria and viruses to investigate this further.
Susceptibility to virus infection in COPD
=========================================
Cigarette smoking increases susceptibility to infection with respiratory viruses, but it is not established whether COPD is associated with increased risk of virus infection. In vitro studies have shown that cigarette smoke impairs the release of IFN-β and IFN-α.[@b110-copd-7-555] BAL cells from COPD patients infected ex vivo with rhinovirus, demonstrated deficient induction of IFN-β with similar trends for deficient induction of IFNs-α and -λ, associated with deficiency of the interferon-stimulated gene CXCL10.[@b90-copd-7-555] Similar findings have been reported in a mouse model where persistence of rhinovirus, increased airways inflammation, and deficient induction of IFNs-α, β, and -γ were reported in COPD mice compared to controls.[@b111-copd-7-555] However, in vitro rhinovirus infection of epithelial cells from COPD patients resulted in higher virus load and increased inflammatory mediators, but no differences in interferon production compared to cells from control subjects.[@b112-copd-7-555] Rhinoviruses bind to cells via intercellular adhesion molecule-1 (ICAM-1, major group rhinoviruses) or members of the low-density lipoprotein receptor family (minor group rhinoviruses). ICAM-1 is upregulated on the bronchial epithelium of patients with COPD,[@b112-copd-7-555],[@b113-copd-7-555] and therefore, it is possible that increased ICAM-1 levels may permit greater virus binding and increased viral entry into epithelial cells in COPD patients. Further studies investigating mechanisms of virus infection in COPD are required and this may lead to potential future therapies for virus-induced exacerbations.
Treatment of infection in COPD exacerbations
============================================
The controversy surrounding the role of bacteria in COPD is mirrored in the debate regarding the efficacy of antibiotics in COPD exacerbations. The current GOLD guidelines acknowledge this ongoing controversy and the poor quality of the clinical trials on antibiotics. This is reflected in the fact that there is no Grade A evidence on which the guidelines based their recommendations.[@b114-copd-7-555] However, antibiotic use in COPD exacerbations is widespread with \>80% of patients in secondary care and 50% of patients managed in primary care treated with antibiotics.[@b117-copd-7-555],[@b118-copd-7-555] Therefore, antibiotic use in COPD may be excessive and contribute to antimicrobial resistance.
Antibiotics in COPD exacerbations
=================================
In the GOLD guidelines, the evidence for antibiotic use in COPD exacerbations is classed as Category B which is defined as "few randomized trials exist, they are small in size, they were undertaken in a population that differs from the target population of the recommendation, or the results are somewhat inconsistent." There are numerous studies of antibiotics in COPD exacerbations but the majority of these are head-to-head comparisons of different antibiotics and there are few placebo-controlled trials. The placebo-controlled trials that are available are affected by a number of methodological issues including small numbers of patients, different outcome measures assessed at different time points, inclusion of patients without COPD, varying severities of exacerbation, and different antibiotics. A Cochrane review in 2006 identified 11 placebo-controlled trials of antibiotics in COPD including a total of 917 patients. Antibiotic use was associated with significant reductions in mortality, treatment failure, and sputum purulence, but no improvement in arterial blood gas parameters or peak expiratory flow.[@b119-copd-7-555] However, there was considerable heterogeneity across the trials and a number of issues that cast doubt on how applicable these findings are to modern management of COPD exacerbations. The review included trials carried out over a wide time period (1965--2001) with only four studies post-1990 and as the definition of COPD has changed markedly, no uniform COPD classifications such as the GOLD stages could be determined from the studies. Studies included a mixture of exacerbation severities ranging from patients treated in the community to patients requiring mechanical ventilation, corticosteroid use was poorly documented, a range of different antibiotics was used, and outcome measures such as quality of life measures not recorded. Therefore, it is debatable whether it is valid to extrapolate these findings to modern day practice. A more recent study included hospitalized COPD patients, all of which were treated with oral corticosteroids and randomized to receive doxycycline or placebo.[@b120-copd-7-555] The primary end point was clinical response on day 30 and the study found no difference between groups in this outcome measure. Doxycycline was shown to be superior to placebo in a number of secondary outcomes. On day 10, clinical success was achieved in 80% of patients in the doxycycline group and 69% in the placebo group (OR: 1.9; 95% CI: 1.1--3.2; *P* = 0.03) in the intention-to-treat analysis but this significant difference was lost in the per-protocol population. Open-label antibiotic treatment for lack of efficacy, day 10 symptom scores, and positive bacteriological cultures were higher in the placebo group. However, treatment failure and changes in lung function were not significantly different between groups.
Therefore, despite the widespread use of antibiotics in COPD exacerbations, the evidence for a beneficial effect is not compelling. Studies have attempted to identify factors that can identify patients who will benefit significantly from antibiotics therapy. Anthonisen et al studied 173 COPD patients with 362 exacerbations for 3.5 years and treated with antibiotics or placebo in a randomized, double-blind, crossover fashion.[@b121-copd-7-555] Exacerbations were classified into three groups: type 1 exacerbations comprised worsening dyspnea, increased sputum volume, and purulence, type 2 exacerbations were any two of these symptoms, and type 3 was any one of these symptoms with evidence of fever or an upper respiratory tract infection. Outcome was defined as success, no resolution, or deterioration, all assessed at 21 days. Patients with a type I exacerbation receiving antibiotics had a higher success rate compared to placebo (62.9% vs 43%, *P* = 0.01) but no benefit was seen for patients with type II (70% vs 60%) and type III (74% vs 70%) exacerbations. However in the study by Daniels et al the treatment effect did not differ significantly between the Anthonisen groups.[@b120-copd-7-555] A meta-analysis of ten trials including 1557 patients by Puhan et al reported that the effects of treatment varied according to the severity of exacerbations. Antibiotics did not significantly reduce treatment failure in mild to moderate exacerbations treated as outpatients but did so in severe exacerbations requiring hospital admission.[@b122-copd-7-555] These findings were confirmed by Quon et al and together these studies indicate that the beneficial effect of antibiotics may be restricted to more severe exacerbations.[@b123-copd-7-555] Sputum purulence has also been proposed as a marker of bacterial infection[@b124-copd-7-555]--[@b128-copd-7-555] and there is evidence that withholding antibiotics from patients without purulent sputum is safe,[@b124-copd-7-555],[@b129-copd-7-555] but these studies were not randomized.
The use of procalcitonin (PCT) as a marker of bacterial infection has been studied in a number of clinical syndromes including COPD exacerbations. Procalcitonin levels in the serum are low in the healthy state and become elevated in response to bacterial infections, but this response is attenuated by cytokines released in response to viral infection. Christ-Crain et al evaluated the use of procalcitonin to guide antibiotic therapy in respiratory tract infections including patients with COPD exacerbations.[@b130-copd-7-555] Patients were assigned to either a standard therapy arm where antibiotics were prescribed according to clinical judgment, or a procalcitonin-guided arm in which antibiotics were given only to patients with raised PCT levels. Eighty-seven percent of the COPD patients in the standard therapy group received antibiotics compared to 38% in the procalcitonin group (*P* = 0.0001) and there were no differences in outcomes between the groups. This group then carried out a similar study in COPD exacerbations only and reported that antibiotic prescribing was reduced by 32% in the procalcitonin group compared to the standard therapy group.[@b131-copd-7-555] There were no differences in outcomes between the two study arms and importantly, antibiotic prescribing was not different up to 6 months after the index exacerbation. There was no relationship between PCT levels and sputum purulence, Anthonisen type, or sputum cultures.
Therefore, despite the widespread use of antibiotics in COPD exacerbations, the evidence for this usage is surprisingly weak. Antibiotic use has a number of adverse effects both in terms of direct side effects for patients and effects on antimicrobial resistance. With increasing use of self-management plans for COPD patients[@b132-copd-7-555] and increasing numbers of elderly patients with COPD, the prescribing of antibiotics and the associated adverse effects are likely to increase with current management strategies. The validation of biomarkers and identification of clinical characteristics that identify exacerbations likely to be bacterial are urgently needed to guide rational antibiotic prescribing in COPD.
Antibiotics in stable COPD
==========================
Recognition of the potential role of bacteria in amplifying airways inflammation in stable COPD has led to interest in treating bacterial infection not only in exacerbations but also in stable patients. This approach is not new as trials of prophylactic antibiotics in chronic bronchitis were first carried out in the 1960s. A meta-analysis of nine such trials showed some reduction in exacerbation time and a small but non-significant effect on exacerbation frequency.[@b133-copd-7-555] The lack of clear efficacy and concerns around the development of antibiotic resistance meant that this approach to antibiotic use in COPD was not pursued until more recently. Long-term macrolide antibiotics have been used in a number of chronic respiratory diseases including cystic fibrosis, asthma, and obliterative bronchiolitis. It is unclear whether the clinical benefits of macrolides are due to their antimicrobial effects as anti-inflammatory and immunomodulatory effects have also been described in vitro. In view of their beneficial effect in pulmonary diseases, their effect in COPD has been examined. Seemungal et al assessed the effects of erythromycin administered over 1 year in COPD and reported a significant reduction in exacerbations in the treatment group.[@b134-copd-7-555] No effect on airway or systemic inflammatory markers was seen, suggesting that exacerbations were reduced through an antibacterial mechanism. A trial of the macrolide antibiotic azithromycin 250 mg once daily for 1 year in patients with COPD reported a reduction in exacerbations and improvement in quality of life measures in the treatment arm.[@b135-copd-7-555] However, azithromycin was associated with a slight increase in hearing impairment and a doubling of macrolide resistance in respiratory pathogens isolated from nasopharyngeal swabs. An association between exposure to macrolides and macrolide resistance in pneumococci in COPD has also been reported,[@b136-copd-7-555] and therefore, caution may be required before embarking on long-term antibiotic treatment in large populations of COPD patients.
The risk of development of antibiotic resistance is reduced if either short courses or pulsed courses of antibiotics are used. A short course (5 days) of moxifloxacin in stable COPD patients with bacterial infection demonstrated eradication rates of 75% compared to 30% with placebo (*P* = 0.01). However, infection rates were similar at 8 weeks and there was no difference in exacerbation rates over a 5-month follow-up period indicating that the effect of the antibiotic was short-lived.[@b137-copd-7-555] A randomized controlled trial of the fluoroquinolone antibiotic moxifloxacin to patients with COPD as a pulsed dose for 5 days every 8 weeks reported reduced exacerbations in the per-protocol analysis, but no significant effect was noted on exacerbations in the intention-to-treat analysis.[@b138-copd-7-555] A subgroup analysis of the per-protocol group showed that patients with mucopurulent or purulent sputum at baseline had a 45% reduction in the occurrence of exacerbations. There was no evidence to suggest development of significant bacterial resistance during the trial.
Therefore, the role of antibiotics in stable COPD remains to be clearly defined and the benefits weighed against the potential risks of increased microbial resistance and other adverse effects.
Novel treatments for COPD exacerbations
=======================================
In view of the evidence implicating respiratory viruses in COPD exacerbations, anti-viral agents may have potential as treatment in COPD exacerbations. Drugs have been developed that are active against rhinoviruses including the capsid-binding agent pleconaril.[@b139-copd-7-555] Despite demonstrating clinical benefit in reducing the severity of symptoms, pleconaril was not approved as a treatment for the common cold. Pleconaril has also been investigated as a treatment for asthma exacerbations but the results have not been reported yet.[@b140-copd-7-555] If antiviral agents demonstrate benefit in asthma, then trials in COPD may be warranted. The observation that AAT has antimicrobial activity suggests administration of AAT as a potential novel therapeutic approach for COPD.[@b55-copd-7-555] In AAT-deficient COPD patients, AAT augmentation therapy is associated with reductions in exacerbations[@b141-copd-7-555],[@b142-copd-7-555] and in cystic fibrosis, inhaled AAT reduces *Pseudomonas* ssp. load.[@b54-copd-7-555] However it remains to be determined whether administration of AAT is beneficial in non-AAT deficient COPD.
Conclusion
==========
In conclusion, infection with both bacteria and viruses plays a major role in both stable COPD and acute exacerbations. It is clear that infection can have multiple outcomes including resolution, chronic infection, and exacerbation, and complex interactions between host and pathogen factors are likely to determine the outcome. Previous concepts of "colonization" in stable patients are no longer valid with the evidence that even in stable patients, the presence of bacteria is associated with adverse outcomes. New diagnostic methods that can detect different strains of the same bacterial species and reveal the presence of bacteria not detected by culture methods are likely to further advance our understanding of the role of bacteria in COPD. Finally, the development of anti-viral agents and the validation of biomarkers that can identify specific infectious agents will lead to the use of pathogen-directed therapy in exacerbations.
**Disclosure**
The authors report no conflicts of interest in this work.
######
The Respiratory Microbiome in Health, in stable COPD and exacerbated COPD
Healthy individuals Stable COPD (mild to moderate) Stable COPD (moderate to severe) Exacerbated COPD
-------------------------------------------------------------------------- -------------------------------------------------------------- ------------------------------------------------------------------------------- -------------------------------------------------------------
*Staphylococcus epidermidis*[@b1-copd-7-555] *Haemophilus influenza*[@b12-copd-7-555],[@b13-copd-7-555] *Haemophilus influenza*[@b12-copd-7-555],[@b13-copd-7-555],[@b19-copd-7-555] *Moraxella catarrhalis*[@b69-copd-7-555]
*Cornyebacteria*[@b1-copd-7-555] *Streptococcus pneumonia*[@b12-copd-7-555],[@b13-copd-7-555] *Streptococcus pneumonia*[@b12-copd-7-555],[@b13-copd-7-555] *Streptococcus pneumonia*[@b69-copd-7-555]
*Staphylococcus aureus*[@b1-copd-7-555] *Moraxella catarrhalis*[@b12-copd-7-555],[@b13-copd-7-555] *Moraxella catarrhalis*[@b12-copd-7-555],[@b13-copd-7-555] *Haemophilus influenza*[@b70-copd-7-555]
*Non-hemolytic streptococci*[@b1-copd-7-555] *Haemophilus parainfluenzae*[@b73-copd-7-555] *Pseudomonas aeruginosa*[@b12-copd-7-555],[@b13-copd-7-555],[@b23-copd-7-555] *Pseudomonas aeruginosa*[@b14-copd-7-555],[@b16-copd-7-555]
*Alpha-hemolytic streptococci*[@b1-copd-7-555],[@b19-copd-7-555] *Staphylococcus aureus*[@b70-copd-7-555] *Haemophilus parainfluenzae*[@b73-copd-7-555] *Staphylococcus aureus*[@b70-copd-7-555],[@b71-copd-7-555]
*Neisseria* spp.[@b1-copd-7-555] *Staphylococcus aureus*[@b70-copd-7-555] *Haemophilus parainfluenzae*[@b73-copd-7-555]
*Streptococcus pneumonia* (occasional)[@b1-copd-7-555],[@b19-copd-7-555]
*Haemophilus influenza* (occasional)[@b1-copd-7-555],[@b19-copd-7-555]
*Prevotella* spp.[@b19-copd-7-555]
*Fusobacteria*[@b19-copd-7-555]
*Veillonella*[@b19-copd-7-555]
**Abbreviation:** COPD, cheronic obstructive pulmonary disease.
| {
"pile_set_name": "PubMed Central"
} |
Introduction {#s0001}
============
Clinical research trials sponsored by the pharmaceutical industry have profound impacts on the practice of medicine.[@cit0001] The industry can support a number of trials at all stages of a product's life with significant financial investment assigned to the innovation.[@cit0002] Therefore, there is potential for considerable monetary loss if the results and conclusions of these trials are unfavourable for the sponsor.[@cit0001],[@cit0003] A number of systematic reviews have documented industry sponsorship of drug studies to be associated with findings favourable to the sponsor.[@cit0004]--[@cit0008] More recently, a review written by the Cochrane Collaboration found there to be less correspondence between the results and conclusions of industry-funded studies when compared to non-industry-funded studies.[@cit0001]
In an era of rapidly developing therapies, physicians often rely on the peer-reviewed literature -- especially the abstracts of published studies -- to remain well informed about their respective fields of practice. It is therefore of paramount importance to evaluate outcome reporting bias in study abstracts. This was assessed in the ophthalmic literature by Alasbali and colleagues, who found industry-funded studies on the ocular hypotensive efficacy of topical prostaglandin analogues to be more likely to report proindustry abstract conclusions, which often did not correspond with studies' results.[@cit0009] Our group performed a similar analysis of randomized clinical trials on the efficacy of intravitreal anti-vascular endothelial growth factor (anti-VEGF) therapy for retinal vein occlusion published in high-impact journals and did not find industry-sponsorship to be associated with an increased rate of reporting bias.[@cit0010] The focus on high-impact journals allowed the authors to capture journals that were most likely to be referred to by physicians, and the results of this study were reassuring given the rapid adoption of anti-VEGF therapy for a number of ocular conditions.
Treatment of diabetic macular edema (DME) remains controversial among vitreoretinal specialists.[@cit0011],[@cit0012] Given the increased recognition of the role of inflammation in the development of DME, intravitreal corticosteroid therapy has been shown to provide promising anatomical and visual benefits, especially when compared to laser therapy.[@cit0013]--[@cit0015] Compared to anti-VEGF therapy, intraocular corticosteroids do carry class-specific risks such as cataract progression and ocular hypertension that theoretically could be downplayed in abstract presentation.[@cit0015] The purpose of this study was to examine the relationship between industry funding and the presence of spin in high-impact studies assessing the efficacy and safety of intravitreal corticosteroid therapy for DME.
Methods {#s0002}
=======
This systematic review was conducted in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement.[@cit0016] Approval from an Institutional Review Board was not required for this study as no human subjects were involved, and analysis was based upon information from published literature.
Search Methods {#s0002-s2001}
--------------
Ovid Medline and Ovid Embase were searched from inception through July 16, 2018, for studies investigating the treatment of patients with diabetic macular edema with intravitreal corticosteroid therapy (ie, triamcinolone, fluocinolone, or dexamethasone) used in at least one of their treatment arms. The search strategy ([[Appendix 1](https://www.dovepress.com/get_supplementary_file.php?f=262085.pdf)]{.ul}) was designed to generate studies published in English journals with an IF greater than 2 as per the Clarivate Analytics 2017 Journal Citation Report.[@cit0017] The inclusion of studies published in journals with IF greater than 2 is presumed to reflect studies more likely to be read by clinicians when browsing recent medical literature.
Study Selection {#s0002-s2002}
---------------
Search results were imported into DistillerSR (Evidence Partners; Ottawa, Canada) to manage all identified records. Title and abstract screening was completed to select for studies that were primary RCTs and meta-analyses. Full-text screening was then completed to select for studies that reported on main outcomes of visual acuity, retinal thickness, and/or complications. Studies that published secondary or sub-analyses on previously published RCTs were excluded. [Figure 1](#f0001){ref-type="fig"} outlines the flow of study selection as per the PRISMA statement.[@cit0016]Figure 1Selection of randomized clinical trials and meta-analyses.
Data Collection and Analysis {#s0002-s2003}
----------------------------
Our methods of study evaluation and data extraction have been previously described.[@cit0010] Study quality was assessed with the scoring scale utilized by Alasbali et al (2009) and is outlined in [Table 1](#t0001){ref-type="table"}.[@cit0009] Correspondence between significance of the main outcome measure result and abstract conclusion was assessed by surveying whether the wording of the abstract conclusion matched the statistical analysis of the results as they pertained to the main outcome measure(s). Following independent data extraction, all discrepancies on the standardized data sheet were resolved by unanimous agreement amongst the authors (HN, AK, JS). One author (HN) also collected objective data on the included studies: sample size, source of funding (industry vs non-industry), whether the publication was authored by an industry employee as reported in the manuscript, and whether any of the co-authors had a potential conflict of interest (COI). The presence of a potential author COI was determined by assessing the study's disclosure statement and was defined as any previous relationship between a co-author and the company that manufactured the drug intervention(s) being studied. Corresponding authors of the included studies were contacted if any of the objective information was not evident as part of the published article.Table 1Criteria Utilized for Grading of Study Quality[@cit0009]Quality ScoreCriteria1: Meta-analysis (To assign this level, all of the following criteria must be met).1. The paper reports a comprehensive search for evidence.\
2. The authors avoid bias in selecting articles for inclusion.\
3. The authors assess each article for validity.\
4. The paper reports clear conclusions that are supported by the data and appropriate analysis.1: Large RCT (To assign this level, all of the following criteria must be met).1. Patients were randomly allocated to treatment groups.\
2. Follow-up was at least 80% complete.\
3. Both the patients and the investigators were blind to the treatment the patient received.\
4. Patients were analyzed in the treatment groups to which they were assigned.\
5. The sample size was large enough to detect the outcome of interest.2: RCTRCT or overview that did not meet level 1[^1]
The primary outcome of the present study was the association between funding source and the correspondence between the studies' abstract conclusion and statistical significance of their main outcome, expressed as an odds ratio. Exploratory secondary analyses were also performed to determine any associations between our variables of interest, including study sample size, journal impact factor and the presence of any COI. Statistical analysis included the Fisher exact test for categorical data, and the Mann--Whitney *U*-test/Kruskal--Wallis test or Student's *t*-test/one-way ANOVA for numerical data, as appropriate. A p-value \< 0.05 was considered statistically significant. All data were extracted and stored in Microsoft Excel software (Microsoft; Redmond, WA). Statistical analysis was performed using SPSS (IBM Corp; Armonk, New York; software version 22).
Results {#s0003}
=======
The original search of both databases yielded 10,073 articles, reduced to 7158 following the removal of duplicates. After title and abstract screening, the full texts of 73 articles were completed. Twenty-nine articles were then excluded due to being a secondary analysis (n=17), analysis of a main outcome not relevant to the present study (n=4), or inadequate study design (n=8). Therefore, 44 (41 RCTs and 3 meta-analyses) publications were included in the present analysis ([Figure 1](#f0001){ref-type="fig"}).[@cit0018]--[@cit0061] Of these 44 studies, 36 (82%) were of higher impact factor (IF≥3), and 31 (70%) were assigned a study quality score of 2. Fourteen (32%) received industry funding, five (11%) had an author who was an industry employee, and 17 (39%) had an author(s) with a potential COI. [Table 2](#t0002){ref-type="table"} outlines characteristics of the included studies.Table 2Summary of Full-Text Study AssessmentsArticleImpact FactorStudy QualitySample SizeInterventionsMOM(s)MOM (p\<0.05)?Correspondence?Any COIIndustry AuthorIndustry SponsorSponsorCommentsHIGHER-IMPACT JOURNALS (Impact Factor ≥ 3.0)Heng et al (2016)[@cit0018]3.806280Macular Laser ± IV DEX ImplantBCVANoYesCOI PresentNoYesAllerganSmall sample sizeShah et al (2016)[@cit0019]3.7250IVB vs IV DEX implantBCVA and CSTYesYesCOI PresentNoYesAllerganFailed to mention steroid-related IOP elevation rate, Small sample sizeMaturi et al (2015)[@cit0020]3.7240IVB± IV DEX ImplantBCVA and CSTYesYesCOI PresentNoYesAllerganSmall sample sizeGillies et al (2014)[@cit0021]8.2288IV DEX Implant vs IVBBCVANoYesCOI Not PresentNoYesAllerganFailed to mention steroid-related IOP elevation rate, Small sample sizeCallanan et al (2013)[@cit0022]8.22253Laser ± IV DEX ImplantBCVANoYesCOI PresentYesYesAllergan-Elman et al (2010)[@cit0023]8.22854Laser ± IVR or IVTBCVANoYesCOI PresentNoYesGenentech, Allergan-Ip et al (2008)[@cit0024]8.22840IVT vs Focal/Grid LaserBCVAYesYesCOI PresentNoYesAllergan-Chew et al (2007)[@cit0025]8.22129STT ± Focal LaserBCVA and CSTNoYesCOI PresentNoNoSoley Non-Industry-Googe et al (2011)[@cit0026]3.72345IVR vs IVTBCVA and CSTYesYesCOI PresentNoYesGenentech, Allergan-Campochiaro et al (2012)[@cit0027]8.21953Fluocinolone Vitreous InsertsBCVAYesYesCOI PresentYesYesAlimera Sciences-Pearson et al (2011)[@cit0028]8.21196IV Fluocinolone ImplantBCVAYesYesCOI PresentYesYesBausch & Lomb-Elbendary et al (2011)[@cit0029]3.7232IV Diclofenac vs IVTVA, CMT, IOPNoYesCOI Not PresentNoNoNoneSmall sample sizeCampochiaro et al (2011)[@cit0030]8.21953Low vs High dose Fluocinolone IV ImplantBCVAYesYesCOI PresentYesYesAlimera Sciences-Gillies et al (2011)[@cit0031]8.2184Laser ± IVTBCVAYesYesCOI Not PresentNoNoSoley Non-IndustrySmall sample sizeTakata et al (2010)[@cit0032]3.7224IV vs ST infusion of TriamcinoloneBCVA, CMT, IOPYesYesCOI Not PresentNoNoSoley Non-IndustrySmall sample sizeMirshahi et al (2010)[@cit0033]3.7236PRP and MPC ± IVTBCVA, CMTNoYesNo Disclosure ProvidedNoNoNoneSmall sample sizeGillies et al (2010)[@cit0034]3.38181IVTBCVANoYesCOI Not PresentNoNoSoley Non-IndustrySmall sample sizeSoheilian et al (2009)[@cit0035]8.21150IVB ± IVT vs Macular LaserBCVAYesYesCOI Not PresentNoNoSoley Non-Industry-Maia et al (2009)[@cit0036]5.052244Laser ± IVTBCVA, CMT, TMVYesYesCOI Not PresentNoNoSoley Non-IndustrySmall sample sizeHauser et al (2008)[@cit0037]3.7242IVTBCVA, CMT, IOP, CataractNoYesCOI Not PresentNoNoNoneSmall sample sizeOckrim et al (2008)[@cit0038]3.806288IVT vs LaserBCVANoYesCOI Not PresentNoNoSoley Non-IndustrySmall sample sizePaccola et al (2008)[@cit0039]3.806226IVT vs IVBBCVA and CMTYesYesCOI Not PresentNoNoSoley Non-IndustryFailed to mention steroid-related IOP elevation rate, Small sample sizeSoheilian et al (2007)[@cit0040]3.71103IVB ± IVT vs Macular LaserBCVAYesYesCOI Not PresentNoNoSoley Non-Industry-Lam et al (2007)a[@cit0041]8.22111IVT vs Grid LaserBCVA, CFTNoNoNo Disclosure ProvidedNoNoSoley Non-IndustryLaser alone had similar BCVA and CFT outcomes at final follow up; Failure to mention steroid-related IOP elevation rateLam et al (2007)b[@cit0042]3.806263IVTBCVA, CFT, IOPYesNoCOI Not PresentNoNoSoley Non-IndustryFailed to mention high rates of steroid-related IOP elevations. Small sample sizeAudren et al (2006)[@cit0043]5.052232IVTCMTNoYesNo Disclosure ProvidedNoNoNoneFailed to mention high rates of steroid-related IOP elevations, Small sample sizeBonini-Filho et al (2005)[@cit0044]3.38228STT vs IVTBCVA, CMT, IOP, Lens StatusYesYesCOI Not PresentNoNoSoley Non-IndustrySmall sample sizeSpandau et al (2005)[@cit0045]3.806227IVTBCVA, IOPYesYesCOI Not PresentNoNoNoneSmall sample sizeCardillo et al (2005)[@cit0046]8.2224IVT vs STTBCVA, CMTYesYesCOI Not PresentNoNoSoley Non-IndustrySmall sample sizeTunc et al (2005)[@cit0047]8.2260Focal Laser ± STTBCVAYesYesCOI Not PresentNoNoNoneSmall sample sizeMaturi et al (2018)[@cit0048]5.62129IVR ± IV DEX ImplantBCVANoYesCOI PresentNoYesGenentech, Allergan-Sarao et al (2017)[@cit0049]3.157242PRN vs Single IV DEX ImplantBCVAYesYesCOI PresentNoNoNoneSmall sample sizeIsaac et al (2012)[@cit0050]3.157122IVT vs IVBCFTYesYesNo Disclosure ProvidedNoNoNoneSmall sample sizeKim et al (2008)a[@cit0051]3.7233IVTBCVANoYesNo Disclosure ProvidedNoNoSoley Non-IndustryFailed to mention high rates of steroid-related IOP elevations, Small sample sizeSutter et al (2004)[@cit0052]8.2169IVTBCVAYesNoCOI PresentNoNoSoley Non-IndustryInfectious endophthalmitis developed in one IVT eye, Small sample sizeYilmaz et al (2009)[@cit0053]8.21293IVT vs STTBCVANoYesCOI Not PresentNoNoNoneMeta-AnalysisLOWER-IMPACT JOURNALS (Impact Factor ≥ 2.0 and \< 3.0)Callanan et al (2017)[@cit0054]2.3492363IV DEX implant vs IVRBCVAYesYesCOI PresentYesYesAllergan-Ramu et al (2015)[@cit0055]2.2752100PRN vs fixed dosing IV DEX ImplantBCVANoYesCOI PresentNoYesAllergan-Kriechbaum et al (2014)[@cit0056]2.275230IVB vs IVTBCVA and CSTYesYesCOI PresentNoNoNoneSmall sample sizeZhang et al (2013)[@cit0057]2.2381434IVT vs IVBBCVA and CMTYesYesCOI Not PresentNoNoSoley Non-IndustryMeta-AnalysisDoi et al (2012)[@cit0058]2.349240PPV vs IVTBCVA and CMTYesYesCOI Not PresentNoNoSoley Non-IndustrySmall sample sizeAhmadieh et al (2008)[@cit0059]2.3491115IVB ± IVTCMTNoYesCOI Not PresentNoNoSoley Non-Industry-Qi et al (2012)[@cit0060]2.2381172IVT vs STTBCVA, CMT, IOPNoYesCOI Not PresentNoNoNoneMeta-AnalysisKim et al (2008)b[@cit0061]2.68246STTVA+CMT+DR progressionYesYesCOI Not PresentNoNoNoneSmall sample size[^2]
Correspondence Between Main Outcome Measure and Abstract Conclusion {#s0003-s2001}
-------------------------------------------------------------------
Statistically significant main outcome measures were present in 26 of 44 (59%) of the included studies. There was correspondence between wording of abstract conclusions and study results in 41 of 44 (93%) articles. Reasons for non-correspondence included the failure to mention high rates of steroid-related intraocular pressure (IOP) elevation (n=2)[@cit0041],[@cit0042] and the implication of safety despite a case of endophthalmitis in a small sample (n=1) ([Table 2](#t0002){ref-type="table"}).[@cit0052]
Funding {#s0003-s2002}
-------
Among the 14 studies that received industry funding, two were funded by Alimera Sciences, Inc.,[@cit0027],[@cit0030] eight studies by Allergan, Inc.,[@cit0018]--[@cit0022],[@cit0024],[@cit0054],[@cit0055] three by both Allergan, Inc. and Genentech, Inc.,[@cit0023],[@cit0026],[@cit0048] and one study was funded by Bausch & Lomb, Inc.[@cit0028] Eighteen studies received funding from a nonindustry sponsor.
Comparing Industry-Funded versus Nonindustry-Funded Studies {#s0003-s2003}
-----------------------------------------------------------
A statistically significant main outcome measure was reported in 8 of 14 (57%) industry-funded studies and in 18 of 30 (60%) nonindustry-funded studies (p=1.00, Fisher exact test). Correspondence between abstract conclusions and significance of main outcome was present in 14 of 14 (100%) industry-funded and 27 of 30 (90%) nonindustry-funded studies. The odds ratio of industry funding being associated with noncorrespondence was 0.27 (95% CI: 0.01 to 5.61, p=0.54). Industry-funded studies had significantly greater sample sizes (p=0.01), but similar mean study quality (p=0.50) and journal impact factor (p=0.14) when compared to the nonindustry-funded studies. These data are summarized in [Table 3](#t0003){ref-type="table"}.Table 3Summary of Studies Investigating Steroid Therapy for Diabetic Macular Edema, Based on Funding StatusOutcome StudiedIndustry-Funded (n=14)Nonindustry-Funded (n=30)p-valueCorrespondence of main outcome and conclusions14 (100%)27 (90%)0.54\*Statistically significant (p\<0.05) main outcome8 (57%)18 (60%)1.00\*Sample size, mean ± SD (95% CI)374.6 ± 360.0 (278.4--470.8)82.7 ± 88.3 (66.6--98.8)0.01^†^Study quality, mean ± SD (95% CI)1.79 ± 0.43 (1.68--1.90)1.67 ± 0.48 (1.58--1.76)0.50^‡^Journal impact factor, mean ± SD (95% CI)5.90 ± 2.51 (5.23--6.57)4.68 ± 2.27 (4.27--5.09)0.14^‡^[^3][^4]
Comparing Higher-Impact versus Lower-Impact Publications {#s0003-s2004}
--------------------------------------------------------
When publications were stratified by journal IF into a "high-impact" group (n = 36) with IF ≥ 3 and a "low-impact" group (n=8) with IF\<3, statistically significant main outcome measures were reported in 21 of 36 (58%) high-impact publications and in 5 of 8 (63%) of low-impact publications (p=1.00). Correspondence between abstract conclusions and significance of the main outcome was present in 33 of 36 (92%) high-impact publications and in 8 of 8 (100%) of low-impact publications (p=1.00). When comparing high-impact and low-impact publications there were no significant differences in rates of significant main outcome measures (p=1.00), rates of industry funding (p=1.00), rates of author COI (p=0.76), sample size (p=0.87), or study quality (p=0.68). These data are summarized in [Table 4](#t0004){ref-type="table"}.Table 4Summary of Studies Investigating Steroid Therapy for Diabetic Macular Edema, Based on Impact FactorOutcome StudiedHigh-Impact (n=36)Low-Impact (n=8)p-valueCorrespondence of main outcome and conclusions33 (92%)8 (100%)1.00\*Statistically significant (p \< 0.05) main outcome21 (58%)5 (63%)1.00\*Industry funding12 (33%)2 (25%)1.00\*Any author COI14 (39%)3 (38%)0.76\*Sample size, mean ± SD (95% CI)178.4 ± 270.3 (133.4--223.4)162.5 ± 154.1 (108.0--217.0)0.87^†^Study quality, mean ± SD (95% CI)1.72 ± 0.45 (1.64--1.80)1.63 ± 0.52 (1.45--1.81)0.63^‡^Journal impact factor, mean ± SD (95% CI)5.67 ± 2.21 (5.30--6.04)2.34 ± 0.14 (2.29--2.39)\<0.01^‡^[^5][^6]
Comparing COI-Present versus COI-Absent Publications {#s0003-s2005}
----------------------------------------------------
After compiling responses from corresponding authors with published disclosures, 17 studies had a COI, 22 had no COI and five did not have a disclosure statement. Correspondence between abstract conclusions and significance of main outcome was present in 16 of 17 (94%) studies with a COI, 21 of 22 (96%) studies without a COI and in four of five (80%) studies without a COI disclosure statement (p=0.48). Studies with a COI had a significantly greater mean sample size when compared to studies without a COI or a disclosure statement (p\<0.01). There were no significant differences in rates of significant main outcome measures (p=0.25), study quality (p=0.61), or journal impact factor (p=0.61). These data are summarized in [Table 5](#t0005){ref-type="table"}.Table 5Summary of Studies Investigating Steroid Therapy for Diabetic Macular Edema, Based on Any Author COIOutcome StudiedAuthor COI Present (n=17)No Author COI (n=22)No COI Disclosure (n=5)p-valueCorrespondence of main outcome and conclusions16 (94%)21 (96%)4 (80%)0.48\*Statistically significant (p\<0.05) main outcome11 (65%)14 (64%)1 (20%)0.25\*Sample size, mean ± SD (95% CI)319.2 ± 347.7 (263.7--374.7)93.8 ± 98.8 (45.0--142.6)46.8 ± 36.3 (−55.5--149.1)\<0.01^†^Study quality, mean ± SD (95% CI)1.76 ± 0.44 (1.65--1.87)1.64 ± 0.49 (1.54--1.74)1.80 ±0.45 (1.6--2.0)0.61^‡^Journal impact factor, mean ± SD (95% CI)5.66 ± 2.58 (5.08--6.24)4.68 ± 2.30 (4.17--5.19)4.76 ± 2.05 (3.69--5.83)0.61^‡^[^7][^8]
Discussion {#s0004}
==========
This study aimed to examine whether the presence of industry funding affected the likelihood of biased outcome reporting among studies of intravitreal corticosteroid therapy for DME. Overall, the results of the present study indicated that abstract outcome reporting corresponded with their statistical results for almost all of the included studies, and that funding source was not a predictor for biased reporting. Journal impact factor and the presence of a COI were not predictors of biased outcome reporting. Industry-funded studies and studies with a COI had greater samples sizes but were of similar impact and quality when compared to their counterparts.
The results of this study are reassuring given that biased outcome reporting has been identified in a number of published studies. In 2009, Berwanger et al published the results of a systematic survey of RCT abstract reporting in high-impact general medical journals and found that 29% of studies lacked a definition of the primary outcome and that half of the studies did not report on side effects or harms.[@cit0062] In RCTs of wound treatments, Lockyer et al found that among studies of wound care treatment that did not have a statistically significant result, 71% had some form of biased reporting.[@cit0063] Among RCTs in oncology, Vera-Badillo et al found biased reporting of efficacy outcomes to be common in studies with a negative primary endpoint and that toxicity was underreported.[@cit0064] Recently, biased outcome reported has been shown to be prevalent among high-impact neurology journals.[@cit0065]
Although biased outcome reporting has been identified as a concern in biomedical research, the role of industry sponsorship has been debated in the literature. Recently, a meta-analysis of "spin" in the medical literature found that clinical trials had the greatest variability in the prevalence of spin, with common practices being detracting from statistically nonsignificant results and inappropriately using causal language.[@cit0066] Although the industry sponsorship was hypothesized by the authors to be associated with spin, the results of this meta-analysis were inconclusive.[@cit0066] Published reviews have found that industry funding was not associated with biased reporting among oncology trials,[@cit0067] musculoskeletal studies,[@cit0068] general medical journals,[@cit0069] or in gastrointestinal research.[@cit0070]
Within the ophthalmic literature, Alasbali et al investigated whether funding source was associated with biased abstract conclusions among studies of topical prostaglandins for intraocular pressure lowering.[@cit0009] Their study found 62% of industry-funded articles to have an abstract conclusion that was not consistent with the results of the main outcome measure, while none of the non-industry-funded articles had noncorrespondence. Additionally, while only 24% of the industry-funded studies had a statistically significant main outcome measure, 90% of the industry-funded studies had a proindustry abstract conclusion.[@cit0009] These findings contrast those of the present study quite dramatically and may reflect differences in intervention efficacy (topical prostaglandins vs intravitreal corticosteroids) or differences in methodologies between the studies. Namely, the present study only included RCTs and meta-analyses published in relatively higher-impact journals, which may explain the difference in non-correspondence rates. Recently, our group published a study using a very similar methodology examining the effect of funding source on reporting bias in studies of intravitreal anti-VEGF therapy for retinal vein occlusion.[@cit0010] Similar to the present study, rates of biased abstract reporting were low and were unaffected by funding source, reflecting no differences despite the increased risk profile of intraocular corticosteroids compared to intravitreal anti-VEGF therapy. Finally, the rigorousness of the peer-reviewed process between 2009 and 2019 may partly explain the difference in results. It is interesting to note that all 3 studies with non-correspondence were published prior to 2008.
Although the present study found an overall abstract conclusion and study results correspondence rate of 93%, it is important to note that only the primary outcome was evaluated. The most common primary outcomes among the included studies were visual acuity and retinal thickness. Adverse events, namely intraocular pressure elevation, were rarely reported as a primary outcome and are especially relevant in the context of intravitreal steroid therapy. This review identified seven studies that failed to mention the increased prevalence of steroid-related intraocular pressure elevations in their respective abstracts. If this adverse effect was included as reported as a primary outcome in these studies, the overall non-correspondence rate of the present review would have increased by roughly 11%. This highlights the importance of comprehensive outcome reporting to allow readers to fully understand and appreciate the risks and benefits of therapies they later offer to their patients.
The major limitation of the present study was its highly selective inclusion criteria. Unlike prior studies, the present study only included RCTs and meta-analyses that were published in journals with an impact factor greater than 2. RCTs and meta-analyses are considered to provide the highest level of evidence and are likely preferentially assessed by physicians. Although the present study may have excluded high-quality studies published in journals of lower impact, the authors feel that this analysis captured articles that would more likely be read by physicians when scanning the recent medical literature. Although the present study did not identify differences in the rates of noncorrespondence between the subgroups of journal impact factor, future studies may find it useful to examine biased reporting among studies published in lower-impact journals (impact factor \<2). Nonetheless, it is reassuring to note that among these higher-impact publications of intravitreal corticosteroid therapy for DME, biased abstract reporting overall appears to be uncommon and unrelated to industry sponsorship or authorship, or to journal impact factor.
The data presented in this manuscript were presented in part at the 2019 Annual Meeting of the American Society of Retina Specialists in Chicago, Illinois.
Disclosure {#s0005}
==========
JS is a consultant for Alcon Laboratories, Alimera Science, Oxurion, Regeneron, and Thrombogenics. AEK is a consultant for Allergan, Alimera Sciences, Regeneron, and Valeant and receives grant funding from Bausch Health, Roche/Genentech and Second Sight. The authors report no other conflicts of interest in this work.
[^1]: **Abbreviation:** RCT, randomized controlled trial.
[^2]: **Abbreviations:** MOM, main outcome measure; COI, conflict of interest; IV DEX, intravitreal dexamethasone; IVB, intravitreal bevacizumab; IVR, intravitreal ranibizumab; IVT, intravitreal triamcinolone; STT, subtenon triamcinolone; PRP, panretinal photocoagulation; MPC, macular photocoagulation; PRN, pro re nata; BCVA, best corrected visual acuity; CST, central sub-foveal thickness; CMT, central macular thickness; IOP, intraocular pressure; TMV, total macular volume; CFT, central foveal thickness.
[^3]: **Notes:** \*Fisher exact test; ^†^Student's *t*-test; ^‡^Mann--Whitney *U*-test.
[^4]: **Abbreviations:** 95% CI, 95% confidence interval; SD, standard deviation.
[^5]: **Notes:** \*Fisher exact test; ^†^Student's *t*-test; ^‡^Mann--Whitney *U*-test.
[^6]: **Abbreviations:** 95% CI, 95% confidence interval; SD, standard deviation.
[^7]: **Notes:** \*Fisher exact test; ^†^One-way ANOVA; ^‡^Kruskal--Wallis test.
[^8]: **Abbreviations:** 95% CI, 95% confidence interval; SD, standard deviation.
| {
"pile_set_name": "PubMed Central"
} |
{#sp1 .17}
{#sp2 .18}
{#sp3 .19}
{#sp4 .20}
{#sp5 .21}
{#sp6 .22}
{#sp7 .23}
{#sp8 .24}
{#sp9 .25}
{#sp10 .26}
{#sp11 .27}
{#sp12 .28}
{#sp13 .29}
{#sp14 .30}
{#sp15 .31}
{#sp16 .32}
{#sp17 .33}
{#sp18 .34}
{#sp19 .35}
{#sp20 .36}
{#sp21 .37}
{#sp22 .38}
{#f1 .21}
{#f2 .23}
| {
"pile_set_name": "PubMed Central"
} |
Introduction {#S1}
============
Methicillin-resistant *Staphylococcus aureus* (MRSA) is an opportunistic pathogen that colonizes humans as well as animals ([@B33]). In humans, MRSA causes a wide range of infections including skin and soft tissue infections (SSTIs), and invasive infections such as pneumonia and endocarditis ([@B23]). Since it was first reported in the United Kingdom ([@B31]), MRSA has evolved into three types, including health-care-associated MRSA (HA-MRSA), which was isolated from patients in health-care settings ([@B10]); community-acquired MRSA (CA-MRSA), which was initially isolated from healthy individuals with no previous exposure to health-care facilities ([@B61]); and the new strains that were associated with livestock in the early 2000s ([@B3]; [@B64]), which were designated as livestock-associated MRSA (LA-MRSA). Initially, livestock-associated *S. aureus* isolates causes major problems in agriculture and are the leading cause of bovine mastitis ([@B21]). In addition, MRSA isolates associated with livestock (called LA-MRSA) belonging to specific lineages such as ST398 have been reported to cause infections in animals and animal handlers ([@B3]; [@B64]). In recent years, these LA-MRSA lineages were able to break the species barrier to colonize and cause infections in humans with or without contact with livestock ([@B20]; [@B28]).
Several molecular typing methods including staphylococcal protein A (*spa*) typing, multilocus sequence typing (MLST), pulsed-field gel electrophoresis, staphylococcal cassette chromosome *mec* (SCC*mec*) typing, DNA microarray, and whole genome sequencing (WGS) have been used to characterize and identify *S. aureus* lineages including LA-MRSA. Several LA-MRSA lineages have been identified including CC398 and CC9, which are predominant in Europe and Asia, respectively ([@B35]; [@B11]). Other clones associated with livestock include ST72, ST97, ST5, ST1, and ST433 ([@B35]; [@B11]).
The SCC*mec* genetic element is a mobile genetic element that confers methicillin resistance and resistance to other beta-lactam antibiotics to susceptible strains following its acquisition. The SCC*mec* element is variable in structural organization and the carriage of additional genetic structures such as transposons and insertion sequence elements. The high diversity in its structural organization and composition has formed the basis of SCC*mec* typing of MRSA strains ([@B29]). HA-MRSA isolates carry SCC*mec* types I, II, and III; CA-MRSA isolates carry SCC*mec* types IV, V, and VI; and LA-MRSA can carry any of the SCC*mec* types associated with CA-MRSA or HA-MRSA. For example, CC9 isolated from pigs in Asia were reported to harbor SCC*mec* III, SCC*mec* V, or SCC*mec* IX element ([@B14]; [@B46]). Similarly, CC398 isolates have been reported to carry different SCC*mec* types including SCC*mec* IV, SCC*mec* V, and SCC*mec* IX ([@B62]).
*Staphylococcus aureus* is endowed with multiple virulence factors, such as toxins, enzymes, hemolysins, and leukocidins including Panton--Valentine leukocidin (PVL), that enhance the capacity of the bacterium to cause disease in humans. Genomic studies have revealed that LA-MRSA clones such as CC398 lack or rarely carry specific virulence factors including PVL and toxic shock syndrome toxin (TSST), which are considered major contributors in *S. aureus* infections ([@B30]; [@B49]). PVL is a pore-forming cytotoxin that plays a major role in *S. aureus* infections by targeting leukocytes ([@B38]).
Although HA-MRSA and CA-MRSA have been widely reported in patients attending Kuwait hospitals ([@B7]), LA-MRSA has only recently been detected among MRSA isolates obtained from patients admitted to Kuwait hospitals. This paper reports the molecular characterization of LA-MRSA isolates obtained from patients in Kuwait hospitals in 2016--2017. LA-MRSA isolates selected for this study were identified based on their clonal complex (CC), which was determined by DNA microarray.
Materials and Methods {#S2}
=====================
Ethical Approval {#S2.SS1}
----------------
The study did not require ethical approval, because all the MRSA isolates were obtained as part of routine diagnostic microbiology investigations.
Methicillin-Resistant *Staphylococcus aureus* Strains {#S2.SS2}
-----------------------------------------------------
A total of 4726 MRSA isolates were obtained from clinical samples in 2016 (*N* = 2305) and 2017 (*N* = 2421) in 11 public Kuwait hospitals. MRSA isolates were obtained from clinical samples submitted to the clinical microbiology diagnostic laboratory in the 11 hospitals. The isolates were identified using biochemical tests and tube coagulase at the diagnostic microbiology laboratory. Once it was identified as MRSA in the diagnostic laboratories, the isolates were sent to MRSA Reference Laboratory located in the Department of Microbiology, Faculty of Medicine, Kuwait University, for molecular typing. The information accompanying the submitted MRSA isolates was sample ID, date of isolation, patient location, patient ID, and clinical source. The isolates were subcultured twice on brain--heart infusion agar (BHIA) plates to obtain pure colonies and incubated at 35°C for 18 h. Pure cultures were preserved in beads and stored at −20 and −80°C. They were recovered on BHIA and incubated at 35°C prior to testing.
Antibiotic Susceptibility Testing {#S2.SS3}
---------------------------------
Antibiotic susceptibility testing was performed using the disc diffusion method according to the Clinical Laboratory Standards Institute (CLSI) ([@B12]). The susceptibility testing was performed with 13 antibiotics including benzyl penicillin (10 U), cefoxitin (30 μg), kanamycin (30 μg), mupirocin (200 and 5 μg), gentamicin (10 μg), erythromycin (15 μg), clindamycin (2 μg), chloramphenicol (30 μg), tetracycline (10 μg), trimethoprim (2.5 μg), fusidic acid (10 μg), rifampicin (5 μg), and ciprofloxacin (5 μg). Minimum inhibitory concentration (MIC) for cefoxitin, vancomycin, and teicoplanin were determined with Etest strips (bioMérieux, Marcy l'Etoile, France) according to the manufacturer's instructions. *S. aureus* strains ATCC25923 and ATCC29213 were used as quality control strains for the disc diffusion and MIC determination, respectively. The *D*-test was used to test for inducible resistance to clindamycin. Susceptibility to fusidic acid was interpreted according to the British Society to Antimicrobial Chemotherapy (BSAC) ([@B8]).
DNA Isolation for Amplification {#S2.SS4}
-------------------------------
Three to five identical colonies of an overnight culture were picked using a sterile loop and suspended in a microfuge tube containing 50 μl of lysostaphin (150 μg/ml) and 10 μl of RNase (10 μg/ml) solution. The tube was incubated at 37°C in the heating block (ThermoMixer, Eppendorf, Hamburg, Germany) for 20 min. To each sample, 50 μl of proteinase K (20 mg/ml) and 150 μl of Tris buffer (0.1 M) were added and mixed by pipetting. The tube was then incubated at 60°C in the water bath (VWR Scientific Co., Shellware Lab, United States) for 10 min. The tube was transferred to a heating block at 95°C for 10 min in order to inactivate proteinase K activity. Finally, the tube was centrifuged, and the extracted DNA was stored at 4°C till used for PCR.
*Spa* Typing {#S2.SS5}
------------
Amplification of *spa* gene was performed using synthetic primers published previously ([@B25]) in a total volume of 25 μl. The PCR protocol consisted of an initial denaturation at 94°C for 4 min, followed by 25 cycles of denaturation at 94°C for 1 min, annealing at 56°C for 1 min, and extension for 3 min at 72°C, and a final cycle with a single extension for 5 min at 72°C. Five microliters of the PCR product was analyzed by 1.5% agarose gel electrophoresis to confirm amplification. The PCR product was purified using MicroElute Cycle-Pure Spin kit (Omega Bio-Tek Inc., United States) according to the manufacturer's protocol. The purified DNA was used for sequencing PCR. A total of 10 μl of the sequencing reaction mixture containing 2 μl of big dye terminator mix, 2 μl of 5× sequencing buffer, 3 μl of nuclease-free water, 1 μl of 3.2 pM primer (forward and reverse), and 2 μl of purified DNA were prepared. The sequencing PCR protocol consisted of initial denaturation for 1 min at 94°C, followed by 25 cycles of denaturation for 10 s at 96°C, annealing at 55°C for 5 s, and extension for 4 min at 66°C. Ultra-Sep Dye Terminator Removal kit (Omega Bio-Tek Inc., United States) was used to purify DNA. Purified DNA was sequenced in an automated 3130 × 1 genetic analyzer (Applied Biosystems, United States) in accordance with the manufacturer's protocol. The sequence of *spa* gene was analyzed using the Ridom Staph Type software (Ridom GmbH, Wurzburg, Germany). The software detected the *spa* repeats and assigned the *spa* type for each isolate.
DNA Microarray {#S2.SS6}
--------------
The *S. aureus* Genotyping kit 2.0 (Alere GmbH, Germany) was used for clonal assignment and the detection of genes encoding antibiotic resistance and virulence factors for MRSA isolates representing each *spa* type identified using the protocol provided by the manufacturer ([@B43]).
Results {#S3}
=======
Detection of Livestock-Associated Methicillin-Resistant *Staphylococcus aureus* Isolates {#S3.SS1}
----------------------------------------------------------------------------------------
A total of 4,726 MRSA isolates collected from patients in 11 hospitals in 2016--2017 were investigated by *spa* typing. From this, 2,823 isolates selected on the basis of *spa* types were analyzed by DNA microarray to assign the isolates into CCs. The 2,823 isolates included isolates from all clinical samples in all hospitals with the same *spa* types. The DNA microarray results identified 202 (7.1%) of the 2,823 isolates as LA-MRSA isolates. The isolates were defined as LA-MRSA solely on the basis of molecular rather than epidemiological definition. The LA-MRSA isolates that belonged to CC96 (31 isolates), CC97 (169 isolates), and CC398 (2 isolates) were identified as LA-MRSA. The CC96 isolates were obtained in 2016 (*N* = 21) and 2017 (*N* = 10). Eighty-three and 86 CC97 isolates were isolated in 2016 and 2017, respectively. CC398 isolates were only detected in 2016. Of the non-LA-MRSA isolates, the dominant CCs were CC5 (*N* = 796), CC22 (*N* = 397), CC8 (*N* = 304), CC1 (*N* = 239), CC6 (*N* = 223), CC30 (*N* = 179), CC80 (*N* = 178), and CC88 (*N* = 88). This report focuses on the characteristics of the LA-MRSA isolates.
The characteristics of the LA-MRSA are summarized in [Table 1](#T1){ref-type="table"}. The results for each isolate is presented in the [Supplementary Table S1](#TS1){ref-type="supplementary-material"}. The LA-MRSA isolates were obtained from clinical samples of patients treated in Kuwait hospitals. The samples were collected from 137 inpatients and 50 outpatients. The locations of 17 patients were not specified. The isolates were obtained from nasal swab (47; 23.2%), skin and soft tissue (47; 23.2%), groin (13; 6.4%), high vaginal swab (HVS) (13; 6.4%), blood (12; 5.9%), sputum (12; 5.9%), throat (10; 4.9%), eye (5; 2.4%), trachea (4; 1.9%), urine (4; 1.9%), axilla (2; 0.9%), ear (2; 0.9%), and fluid (1; 0.5%). The source of 30 (14.8%) isolates was unspecified.
######
Genotypic characterization of LA-MRSA isolates.
**Clonal complex (CC)** **Specimen** **LA-MRSA strain** **Number of isolates** ***Spa* type**
------------------------- ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- --------------------------------------------------------- ------------------------ ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
CC398 Nasal CC398-MRSA-IV 1 t899 (1)
Swab CC398-MRSA-V (PVL^+^) 1 t034 (1)
CC96 Nasal (11), sputum (1), throat (2), groin (2), wound (4), skin (6), blood (1), unspecified (3) CC96-MRSA-IV 30 t1028 (2), t11822 (13), t1198 (1), t1234 (1), t14838 (2), t203 (1), t4955 (4), t8154 (2), t8731 (1), t9867 (2), ND (1)
Pus CC96-MRSA-IV (PVL^+^), Central Asian caMRSA/WA MRSA-119 1 t8731 (1)
CC97 Groin CC97-MRSA- (*mec* V/V~T~+*fus*+*ccrAB*2) 1 t2297 (1)
Unspecified CC97-MRSA-(*mec* VI+*fus*) 2 t359 (2)
Wound (4), nasal (2), pus (1), urine (1), swab (1), trachea (1), unspecified CC97-MRSA-IV, WA MRSA-54/63 11 t267 (2), t359 (5), t521 (1), t1234 (1), t693 (1), ND (1)
Wound (1), throat (1), HVS (2), groin (2), urine (1), nasal (2), ear (2), unspecified (3) CC97-MRSA-V 14 t1234 (5), t203 (1), t267 (4), t359 (1), t4955 (1), t2297 (1), t693 (1)
Unspecified (17), HVS (10), fluid (1), wound (17), nasal (31), skin (5), swab (2), groin (8), pus (8), blood (11), axilla (2), throat (7), eye (5), sputum (11), trachea (3), urine (2) CC97-MRSA-V (*fusC*^+^) 139 t1814 (2), t1965 (1), t267 (67), t2734 (1), t359 (30), t376 (2), t521 (4), t527 (1), t189 (4), t16486 (1), t2802 (1), t2297 (10), t044 (1), t15069 (1), t16606 (2), t17281 (1), t17282 (1), t17330 (1), t2770 (1), t701 (1), t9638 (1), ND (5)
HVS CC97-MSSA (1) 1 t16903 (1)
ND, not determined; LA-MRSA, livestock-associated methicillin-resistant
Staphylococcus aureus
; PVL, Panton--Valentine leukocidin; HVS, high vaginal swab.
The CC398 belonged to two genotypes including PVL-negative CC398-MRSA-IV/t899 and PVL-positive CC398-MRSA-V/t034.
Thirty CC96 isolates belonged to CC96-MRSA-IV, and one isolate belonged to CC96-MRSA-IV (PVL^+^), also known as the Central Asian caMRSA/WA MRSA-119.
Seven different subtypes were identified among the CC97 isolates. These included CC97-MRSA-V (*fusC*^+^) (139 isolates), CC97-MRSA-V (14 isolates), CC97-MRSA-IV WA MRSA-54/63 (11 isolates), CC97-MRSA-(*mec* VI+*fus*) (2 isolates), CC97-MRSA-(*mec* V/V~T~+*fus*+*ccrAB*2) (1 isolate), CC97-MRSA-(V+*fus*) (1 isolate), and CC97-MSSA (1 isolate).
The LA-MRSA isolates belonged to 35 *spa* types. CC398 isolates consisted of two *spa* types, t899 and t034. Ten *spa* types were identified among CC96 isolates including t11822 (*N* = 13), t4955 (*N* = 4), t1028 (*N* = 2), t8154 (*N* = 2), t8731 (*N* = 2), t9867 (*N* = 2), t14838 (*N* = 2), t1234 (*N* = 1), t1198 (*N* = 1), and t203 (*N* = 1). *Spa* type was not determined (ND) for one isolate.
Twenty-six *spa* types were identified among CC97 isolates, with t267 (*N* = 67), t359 (*N* = 39), and t2297 (*N* = 12) as the common *spa* types in this lineage. The other *spa* types were detected less frequently. *Spa* types t203, t1234, and t4955 were identified in both CC96 and CC97 isolates ([Table 1](#T1){ref-type="table"}).
Antibiotic Susceptibility Testing and Antibiotic Resistance Genes {#S3.SS2}
-----------------------------------------------------------------
All LA-MRSA isolates were tested for susceptibility to antimicrobial agents. The results for the disk susceptibility testing and the genetic resistance determinants are summarized in [Table 2](#T2){ref-type="table"}. Antibiotic susceptibility testing showed that all LA-MRSA isolates were resistant to cefoxitin and were positive for *mecA*. A total of 187 (92.5%) isolates were resistant to penicillin mediated by *blaZ.* Sixteen CC97 isolates were phenotypically resistant to penicillin by disc diffusion method but lacked *blaZ* ([@B27]).
######
Phenotypic and genotypic characteristics of LA-MRSA isolates.
**Locus** **CC96 (*N* = 31)** **CC97 (*N* = 169)** **CC398 (*N* = 2)** **Total (*N* = 202)**
--------------------------------------------- --------------------- ---------------------- --------------------- -----------------------
**Virulence factors**
*Sea* 30 30
*Sec* 20 22
*Sed* 2 2
*Sel* 22 22
*sek* 2 2
*seq* 2 2
*tst* 1 1
PVL 1 1 2
PVL (P83) 4 4
*chp-scn-sak* (Type B) 30 2 32
*scn-sak* (Type E) 1 196 197
*can* 31 2 33
**Antibiogram/antibiotic resistance genes**
**Penicillin resistance**
Penicillin (phenotypic) 28 (90.3%) 157 (92.8%) 1 (50%) 188 (92.1%)
*blaZ* 31 141 1 175
**Methicillin-resistance**
Cefoxitin (phenotypic) 31 (100%) 169 (100%) 2 (100%) 204 (100%)
*mecA* 31 169 2 204
**MLS-B resistance**
Erythromycin (phenotypic) 26 (83.8%) 12 (7.1%) 2 (100%) 42 (20.5%)
Clindamycin (phenotypic) 26 (83.8%) 7 (4.1%) 2 (100%) 35 (17.1%)
*erm*(A) 1 1
*erm*(C) 26 4 1 31
*msr*(A) 3 5
Aminoglycoside resistance
Gentamicin (phenotypic) 1 (3.2%) 120 (71.0%) 121 (59.3%)
Kanamycin (phenotypic) 1 (3.2%) 119 (70.4%) 120 (58.8%)
*aacA-aphD* 118 118
**Trimethoprim resistance**
Trimethoprim (phenotypic) 4 (2.3%) 1 (50%) 5 (2.4%)
*dfrS1* 2 1 3
**Fusidic resistance**
Fusidic acid (phenotypic) 2 (6.4%) 138 (81.6%) 140 (68.6%)
*fusC* 142 142
*fusB* 1 1 2
**Tetracycline resistance**
Tetracycline (phenotypic) 30 (17.7%) 2 (100%) 32 (15.6%)
*tet*(K) 27 27
*tet*(M) 1 1
**Chloramphenicol resistance**
Chloramphenicol (phenotypic) 1 (0.59%) 1 (0.49%)
*cat* 2 2
**Streptogramin A resistance**
*vga*(A) 3 1 4
**Fosfomycin resistance**
*fosB* 2 4
LA-MRSA, livestock-associated methicillin-resistant
Staphylococcus aureus
; PVL, Panton--Valentine leukocidin.
Gentamicin resistance and kanamycin resistance were detected in 120 CC97 isolates, with only 118 isolates positive for *aacA--aphD*. One CC96 and two CC97 isolates were phenotypically resistant to gentamicin and kanamycin but lacked any of the aminoglycoside resistance genes in the DNA microarray panel ([Table 2](#T2){ref-type="table"}).
Macrolide--lincosamide--streptogramin-B (MLS-B) resistance was detected in 20.7% of the LA-MRSA. All CC96 isolates resistant to erythromycin and clindamycin carried *erm*(C). MLS-B resistance in the CC398 isolates was mediated by *erm*(A) and *erm*(C). Four CC97 erythromycin- and clindamycin-resistant isolates carried *erm*(C), three isolates were resistant only to erythromycin carried *msr*(A), and five isolates phenotypically resistant to erythromycin and clindamycin lacked any of the MLS-B resistance genes in the microarray panels.
Fusidic acid resistance genes *fusB* and *fusC* were identified in CC96 and CC97 isolates. Two CC96 isolates were phenotypically resistant to fusidic acid, but only one isolate carried *fusB*. *fusB* was also detected in one isolate belonging to CC97, whereas 142 isolates carried *fusC* ([Table 2](#T2){ref-type="table"}).
Tetracycline resistance was detected in 30 CC97 isolates with 27 isolates carrying *tet*(K). Tetracycline resistance gene, *tet*(M), was found in one CC398 isolate ([Table 2](#T2){ref-type="table"}). Trimethoprim resistance gene *dfrS1* was detected in one CC398 isolate and in two of the four phenotypically resistant CC97 isolates. Chloramphenicol resistance mediated by *cat* was detected in a single CC97 isolate ([Table 2](#T2){ref-type="table"}).
*vga*(A), which confers resistance to streptogramin A compound (not tested phenotypically in this study), was found in three CC97 isolates and one isolate belonging to CC398. Fosfomycin was not tested in this study. However, two isolates belonging to CC97 carried *fosB*, which confers fosfomycin resistance ([Table 2](#T2){ref-type="table"}).
Virulence Encoding Genes {#S3.SS3}
------------------------
DNA microarray demonstrated that all LA-MRSA isolates carried genes for virulence factors including genes for adhesions, accessory gene regulators (*agr*), capsular polysaccharides (*cap*), and enzymes but varied in their carriage of genes for exotoxins ([Table 2](#T2){ref-type="table"}). All isolates were positive for *agr* and *cap* but differed in the types of *agr* and *cap* alleles. CC398 and CC97 carried *agr* 1 and *cap* 5, whereas CC96 isolates carried *agr* III and *cap* 8.
Of the 204 LA-MRSA isolates, 80 (39.6%) isolates carried enterotoxins. Two CC97 isolates carried *sed*, *sek*, and *seq*, whereas CC96 isolates variably carried *sec*, *sea*, and *sel* ([Table 2](#T2){ref-type="table"}). No enterotoxins were detected among CC398 isolates ([Table 2](#T2){ref-type="table"}).
Gene for TSST-1, *tst*, was detected in one CC97 isolate. PVL was only found in two CC96 isolates.
All isolates carried microbial surface components recognizing adhesive matrix molecules (MSCRAMMs). However, they varied in the carriage of collagen adhesion (*cna*). The *cna* gene was detected in CC96 and CC398 isolates but not in CC97 isolates ([Table 2](#T2){ref-type="table"}).
The immune evasion cluster (IEC) genes (*scn*, *chp*, and *sak*) were identified in all LA-MRSA isolates. CC398 carried *scn*, *chp*, and *sak* encoding genes (IEC type B). All CC97 and one CC96 isolates lacked *chp* and carried *scn* and *sak* (IEC type E).
Discussion {#S4}
==========
Livestock-associated methicillin-resistant *S. aureus* was initially isolated from livestock and later in isolates from humans who were in contact with livestock ([@B32]). Subsequently, LA-MRSA isolated from individuals with no contact with livestock was reported in different places including Italy ([@B47]), Spain ([@B36]), Australia ([@B42]), and Saudi Arabia ([@B44]). In this study, we characterized LA-MRSA obtained from human patients in Kuwait hospitals. The result of the study revealed that LA-MRSA constituted 7.1% of MRSA isolated from patients in hospitals in Kuwait in 2016--2017. The low prevalence of LA-MRSA reported in this study is similar to results reported in patients in Europe including the United Kingdom ([@B26]); Luxembourg, Poland, and Norway ([@B32]); and Asian countries such as China, Taiwan, Japan, and Malaysia ([@B11]). Human colonization with LA-MRSA is more common in areas with high density of livestock ([@B16]; [@B32]). Nevertheless, colonization with LA-MRSA isolates was also reported in people with no contact with livestock. A study published in 2013 in the United States reported CC398 MSSA as the dominant strain among detainees in the Dallas County Jail ([@B17]), which showed the transmission of these isolates in the absence of an animal source.
The LA-MRSA isolates belonged to three CCs, including CC96, CC97, and CC398, in this study. CC97 was found in 169 isolates, making CC97 the dominant LA-MRSA clone in Kuwait hospitals in 2016--2017. Prior to this report, CC97 isolates were reported to cause an outbreak in a neonatal intensive care unit (ICU) in a Kuwait hospital in 2007 ([@B60]; [@B59]) and was detected in four isolates in another hospital in 2010 ([@B7]). The increase in the prevalence of CC97 in recent years in Kuwait may suggest an increased transmission among patients in hospitals. CC97 has also been sporadically reported in patients in Saudi Arabia ([@B44]), Spain ([@B36]; [@B51]), and Australia ([@B42]). Although CC97 is rarely reported in humans ([@B24]; [@B53]), it is commonly isolated from cattle, and it is considered one of the most common causes of bovine mastitis ([@B57]; [@B1]; [@B58]) and one of the most common in the Italian pig industry ([@B4]; [@B19]). The increase in the prevalence of CC97 among human patients observed in this study is of concern, as it highlights the increasing transmission of LA-MRSA among human patients. The CC97 isolates consisted of 26 *spa* types and six genotypes, revealing the diversity of the isolates. CC97-MRSA-V (fusC^+^)-t267 was the dominant strain carrying few enterotoxin genes (*sed*, *sek*, and *seq*) and bovine PVL (P83), which is similar to the bovine CC97 isolates reported in Italy ([@B19]). *Spa* type t267 was also reported in isolates obtained from bovine in Portugal ([@B13]), Switzerland ([@B6]), and Brazil ([@B50]). The other *spa* types associated with CC97 isolates in this study (t359, t521, t2297) have also been reported in patients as well as animals in other studies ([@B2]; [@B19]), making them successful zoonotic subtypes.
Few reports have described the virulence profiles of CC97 isolates. A study in South Africa ([@B54]) showed that CC97 isolated from bovine and humans carried few enterotoxins genes including *sec* and *sel*. Similarly, the CC97 isolates in this study harbored *sec* and *sel*, suggesting that *sec* and *sel* may be common features of CC97 MRSA isolates.
CC97 isolates in this study were multiresistant to antibiotics, including resistance to gentamicin, kanamycin, erythromycin, clindamycin, and tetracycline. Similarly, multiresistant strains of CC97 were isolated from human patients in Saudi Arabia ([@B44]) and in bovine in Spain ([@B22]). In contrast, non-multiresistant isolates of CC97 were obtained from dairy milk in China ([@B65]). We observed differences between the antibiotic susceptibility patterns and the carriage of antibiotic resistance genes in CC97 isolates obtained in this study. Penicillin resistance was detected in 157 of the CC97 isolates, but only 141 carried *blaZ*, which could be due to a lack of signal to *blaZ*, *blaI*, and *blaR*. A similar observation was reported by [@B67] in two CC398-t034 isolates in which *blaZ* could not be detected by DNA microarray, although they were resistant to penicillin and the resistance was confirmed by the detection of penicillinase with nitrocefin. A similar pattern was also observed in isolates resistant to erythromycin, gentamicin, kanamycin, trimethoprim, and tetracycline, in which the corresponding resistance genes could not be detected by the arrays. This could be due to the presence of other resistance mechanisms that are not in the DNA microarray panel or due to intrinsic resistance in these isolates. Intrinsic resistance was documented in *S. aureus* with penicillin-binding protein 2a (PBP2a), which renders the effectiveness of the beta-lactam antibiotics.
CC96 was the second most common LA-MRSA clone detected in this study. Isolates from this lineage were first detected in Kuwait in 2016 (21 isolates) and then in 2017 (10 isolates). CC96 MRSA isolates are rare in humans with only single isolates reported previously from Russia ([@B40]) and Saudi Arabia ([@B55]; [@B39]). However, ST96-MSSA is a common pathogen of rabbits, where it causes different infections ([@B40]; [@B63]; [@B41]; [@B45]). An isolate of ST96-MRSA belonging to *spa* type t1190 was isolated from a rabbit meat sample that could not be characterized as either CA-MRSA or HA-MRSA ([@B18]), suggesting that a previously ST96-MSSA/t1190 had acquired the *mecA* determinant. Although the ST96-MRSA in this study belongs to different spa types, we argue that they are probably related to the rabbit ST96 lineage. Furthermore, the CC96-MRSA in this study belonged to agr III and cap8 and were resistant to erythromycin mediated by *ermC* similar to ST96-MSSA isolates isolated from rabbits ([@B41]). The lack of information on the epidemiology and the genetic characteristics of ST96-MRSA in the literature warrants further studies to describe the origin, prevalence, and molecular characteristics of the emerging CC96-MRSA.
Since its discovery in the Netherlands in the early 2000s, CC398 has become the most common LA-MRSA clone circulating in Europe ([@B9]). This is the first report of CC398 in human patients in Kuwait hospitals and as far as we know in the Gulf Cooperation Council (GCC) countries. CC398 is the most prevalent lineage in pigs ([@B9]; [@B19]), but it has also been reported in horses, poultry, cattle, and companion animals ([@B9]). The CC398-MRSA was classified into two strains carrying SCC*mec* IV and V each and belonged to *spa* types t899 and t034, respectively. The CC398-IV-t899 isolate was PVL negative, similar to isolates obtained from animals in United Kingdom ([@B5]) and human patients in Spain ([@B37]), Italy ([@B47]), and Denmark ([@B34]). MRSA belonging to spa type t899 has been described as a hybrid LA-MRSA of CC9 and CC398 ([@B34]; [@B56]).
Apparently, CC398-MRSA of animal origin usually lacks enterotoxins and PVL ([@B19]), whereas the early-branching East Asian strain is positive for PVL ([@B68]). In this study, one CC398-MRSA-V isolate was positive for PVL, as has also been reported in human patients in China ([@B68]), Sweden ([@B66]), Finland ([@B52]), and New Zealand ([@B67]), suggesting that our strain may belong to the human branch of CC398. The lack of enterotoxins in these two CC398-MRSA isolates is consistent with previous results suggesting that absence of enterotoxins may be a characteristic feature of isolates from this lineage ([@B37]; [@B19]).
CC398 strains of animal origin usually lack the IEC genes that facilitate the colonization and invasion of MRSA in human hosts ([@B49]). The two CC398 isolates detected in this study carried the IEC genes *scn*, *chp*, and *sak* (IEC type B). Similarly, recent studies by [@B15] and [@B48] also reported CC398 isolates of human origin carrying the IEC genes. The presence of IEC genes in these isolates may explain the ability of these strains to jump from livestock and successfully adapt and colonize human beings. One of the CC398 isolates belonged to *spa* type t899.
The major limitation of this study is the lack of information on the patients' travel history or contact with animals. Therefore, it is difficult to determine if these LA-MRSA strains were acquired by contact with livestock or by household members who are in contact with livestock. Nevertheless, the detection of these clones in human patients is significant because it shows their expansion beyond the usual livestock hosts, which may pose new problems for infection control.
Conclusion {#S5}
==========
In conclusion, this study described the characteristics of LA-MRSA strain belonging to CC96, CC97, and CC398 in patients in Kuwait hospitals. Genotyping showed that our isolates are diverse and belonged to different lineages including CC398, which are prevalent in Europe. This is the first report of CC398 LA-MRSA in Kuwait. The study also revealed that most of the isolates belonging to CC97 expressed resistance to multiple classes of antibiotics. The CC97 and CC398 MRSA isolates shared characteristics similar to those obtained from bovine and human patients in Europe. These observations suggest that LA-MRSA isolates were introduced to Kuwait via different routes. Further surveillance studies are required to monitor future transmission patterns of these isolates. Although LA-MRSA may have the same virulence potential and causes similar infections as *S. aureus*, identifying these isolates can inform on their origin, which will help in controlling the spread of MRSA in the clinical settings.
Data Availability Statement {#S6}
===========================
All datasets generated for this study are included in the article/[Supplementary Material](#TS1){ref-type="supplementary-material"}.
Author Contributions {#S7}
====================
SB, BM, BN, TV, and ST carried out the laboratory work. SB performed the data analysis. EU performed the experimental design. SB and EU carried out the manuscript writing and editing. All authors read and approved the final manuscript.
Conflict of Interest {#conf1}
====================
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest. The reviewer SM declared a past co-authorship, with several of the authors, SB and EU, to the handling editor.
The authors are grateful to the technical staff in the MRS Reference Laboratory located in Microbiology Department, Faculty of Medicine, for their technical assistance.
Supplementary Material {#S9}
======================
The Supplementary Material for this article can be found online at: <https://www.frontiersin.org/articles/10.3389/fmicb.2019.02912/full#supplementary-material>
######
Click here for additional data file.
[^1]: Edited by: Ziad Daoud, University of Balamand, Lebanon
[^2]: Reviewed by: Phil Giffard, Charles Darwin University, Australia; Stefan Monecke, Leibniz Institute of Photonic Technology (IPHT), Germany
[^3]: This article was submitted to Antimicrobials, Resistance and Chemotherapy, a section of the journal Frontiers in Microbiology
| {
"pile_set_name": "PubMed Central"
} |
All relevant data are within the manuscript and its Supporting Information files.
Introduction {#sec001}
============
Climate change can strongly influence agriculture with temperature extremes, cold temperatures being a significant cause of damage in limiting crop yield. Although rice (*Oryza sativa* L.), is one of the world\'s most important crops consumed as a major part of the diet \[[@pone.0218019.ref001]\], it is sensitive to cold compared to the temperate crops such as wheat and barley, due to its origin and adaptation for cultivation in tropical and subtropical regions of the world.
The sensitivity and symptoms of plants to cold stress vary with the growth stage. Rice, exposed to cold stress at the vegetative stage, shows symptoms like yellowing of leaves, lower stature, and decreased tillering \[[@pone.0218019.ref002]\]. Other symptoms include damage to the photosynthetic machinery, more specifically the ultrastructure of chloroplasts, altering the light-harvesting chlorophyll antenna complexes \[[@pone.0218019.ref003]\] and/or modifying thylakoid structures \[[@pone.0218019.ref004]\], and an overall reduction in photosynthetic processes by cold temperatures thereby leading to a deficit in plant energy resources.
Under cold stress, reactive oxygen species (ROS) accumulation is induced \[[@pone.0218019.ref005]\] and can cause severe damage to various cellular components such as altering the membrane lipid composition due to excess accumulation of malondialdehyde (MDA), and an increase in antioxidants that can scavenge ROS and protect rice plants against oxidative damage \[[@pone.0218019.ref006]\].
In this research report we present results on the physiological responses of a sample of rice *O*. *sativa* sub-species *japonica* genotypes to cold stress, estimated through quantification of photosynthetic parameters, ROS mediated damage, accumulation of antioxidants and osmolytes, that distinguish the sensitive and tolerance rice phenomes. In addition, we present here the analyses of several stress-responsive genes *OsBURB-16*, *OsGH3-2*, *OsSFR6*, *ZFP245*, *OsACA6*, *Ctb1*, *OsSAP1*, *OsTPP1* and *OsSRO1a* ([S1 Table](#pone.0218019.s001){ref-type="supplementary-material"}) that can potentially contribute towards the observed mechanisms of tolerance.
Results and discussion {#sec002}
======================
Rice is a major global food crop and a model crop for cereals. To understand the basis of acclimation stability under cold, we used a set of rice genotypes contrasting in their tolerance towards cold and evaluated the photosynthetic, biochemical, gene and protein expression response parameters. The results we describe here suggest the presence of complex mechanisms that involve the interaction of many biochemical and physiological pathways along with hormonal cross-talk contributing to cold tolerance.
The analysis of cold-stress-responsive gene expression at the transcript and protein level, along with the phenotypic response, provides an understanding of cold stress tolerance mechanisms in the multiple plant systems. This information is needed since there is little information available on the signaling pathways responsible for low-temperature acclimation and the differential expression of genes at the transcript and proteins level that provide a crucial role in chilling stress signaling \[[@pone.0218019.ref007],[@pone.0218019.ref008]\].
The response of photosynthetic parameters to cold stress {#sec003}
--------------------------------------------------------
The first effects of cold stress on plants are observed in photosynthesis \[[@pone.0218019.ref009]\], which was found to be highly affected in the rice genotypes studied ([Fig 1A](#pone.0218019.g001){ref-type="fig"}). However, Secano do Brazil and Cypress display \> 80% reduction compared to Nipponbare and M202. The reduced air and leaf temperature usually reduce the evaporative demand \[[@pone.0218019.ref010],[@pone.0218019.ref011]\], observed as reduced transpiration in all genotypes studied ([Fig 1B](#pone.0218019.g001){ref-type="fig"}). However, the tolerant genotypes displayed the highest water use efficiency in comparison to sensitive ones under cold stress ([Fig 1C](#pone.0218019.g001){ref-type="fig"}). Under well-watered conditions all the genotypes showed very similar levels but on exposure to stress there was a significant reduction by about 75%.
{#pone.0218019.g001}
The chilling stress caused direct effects on stomata, provoking two potential causes of stomatal closure. In Secano do Brazil and Cypress, the direct inhibition of mesophyll photosynthesis ([Fig 1A](#pone.0218019.g001){ref-type="fig"}) caused a rise in *c*~*i*~ ([Fig 1D](#pone.0218019.g001){ref-type="fig"}), with an associated stomatal closure ([Fig 1E](#pone.0218019.g001){ref-type="fig"}) \[[@pone.0218019.ref012],[@pone.0218019.ref013]\]. In Nipponbare and M202, the stomata were the primary target of the chilling stress and their closure led to a reduction in *c*~*i*~, prompting a decline in photosynthesis \[[@pone.0218019.ref010],[@pone.0218019.ref014]\]. Photo-inhibition can be one of the primary causes of reduction in photosynthesis after cooling \[[@pone.0218019.ref015]--[@pone.0218019.ref016]\] and is characterized by a reduction in Quantum Efficiency of PSII \[[@pone.0218019.ref017]\]. A significant reduction was seen in Secano do Brazil and Cypress under stress, but for the same conditions, the tolerant genotypes did not suffer photoinhibition ([Fig 1F](#pone.0218019.g001){ref-type="fig"}). The fluorescence parameter Fv'/Fm' is regarded as a suitable assay for plant tolerance and sensitivity to cold \[[@pone.0218019.ref018]\], due to inherent tolerance or physiological acclimation.
Biochemical parameters affected by cold-stress {#sec004}
----------------------------------------------
When plants are exposed to low-temperature stress, Chlorophyll biosynthesis is affected ([Fig 2A](#pone.0218019.g002){ref-type="fig"}), and the results of the experiment show that Nipponbare and M202 do not sense the stress like the sensitive genotypes Secano do Brazil and Cypress, which show a significant impact on the biosynthesis of Chlorophyll ([Fig 2A](#pone.0218019.g002){ref-type="fig"}). According to \[[@pone.0218019.ref017]\], this is because the impact on Chlorophyll biosynthesis is due to down-regulation of gene expression and protein abundance of several enzymes involved in tetrapyrrole metabolisms described in other studies \[[@pone.0218019.ref019],[@pone.0218019.ref020],[@pone.0218019.ref021]\].
{#pone.0218019.g002}
The anthocyanins, which are induced by environmental stresses \[[@pone.0218019.ref022]\], have a role in modifying the quantity and quality of captured light \[[@pone.0218019.ref023]\], by protecting from the effects of UV‐B \[[@pone.0218019.ref024]\], and scavenging of reactive oxygen intermediates under stress. However, our data showed a significant reduction in M202 by about 50% and by about 75% in Secano do Brazil, while Cypress showed no difference between the control and stressed plants ([Fig 3A](#pone.0218019.g003){ref-type="fig"}). These results do not support the hypothesis that anthocyanins are needed to serve an auxiliary photo-protective role in leaves, because this variation is independent of the concentrations of chlorophyll (Figs [2A](#pone.0218019.g002){ref-type="fig"} and [3A](#pone.0218019.g003){ref-type="fig"}). However, the anthocyanins are found predominantly associated with leaf mesophyll \[[@pone.0218019.ref025]\], a location that is unsuitable for screening out UV-B but ideal for the scavenging of oxygen radicals produced by chloroplasts. The phenolic content showed an increase in all genotypes, with the highest content in the sensitive genotypes Secano do Brazil and Cypress ([Fig 3B](#pone.0218019.g003){ref-type="fig"}). This increase may help in restricting the penetration of UV-B into the inner tissues of the plant \[[@pone.0218019.ref026]\], and contribute to their antioxidant ability, which inhibits lipid peroxidation by trapping the lipid alkoxyl radicals \[[@pone.0218019.ref027],[@pone.0218019.ref028]\].
{#pone.0218019.g003}
Proline and soluble sugars are also known to protect rice from damage due to cold stress \[[@pone.0218019.ref029]\]. Proline content was found to be enhanced by cold stress in all genotypes ([Fig 4A](#pone.0218019.g004){ref-type="fig"}), with the tolerant lines showing a higher synthesis in control and stress conditions, with results similar to that obtained earlier \[[@pone.0218019.ref030]\]. Proline is involved in the removal of stress-related excess H^+^, maintains oxidative respiration at optimal cytosolic pH \[[@pone.0218019.ref031]\], acts as a reservoir of carbon and nitrogen, and increases protein water-binding ability through its hydrophobic interactions with the surface residues of proteins \[[@pone.0218019.ref032]\].
{#pone.0218019.g004}
Soluble sugars like glucose and sucrose can accumulate in plants under stress, and act as osmoprotectants against freezing/dehydration damage as described previously \[[@pone.0218019.ref033]\]. The tolerant genotypes (Nipponbare and M202) showed no significant difference in glucose content, although the sensitive genotypes showed a decrease ([Fig 4B](#pone.0218019.g004){ref-type="fig"}). In contrast, the sucrose content showed a high increase in the tolerant genotypes compared to the control, while for the sensitive genotypes a significant drop was observed ([Fig 4C](#pone.0218019.g004){ref-type="fig"}). Such an increase in sucrose has been documented previously \[[@pone.0218019.ref034]\], suggesting that an increase under low temperature can be a useful marker for cold tolerance in rice.
Many plant subcellular locations such as the cell wall, plasma membrane, mitochondria, and chloroplast, at the site of thylakoid electron transport; and the nucleus are intrinsically responsible for producing ROS in response to stress \[[@pone.0218019.ref035],[@pone.0218019.ref036],[@pone.0218019.ref037]\]. To avoid disastrous damage to protein and lipid components, plants have numerous antioxidant systems. Among these, ROS (e.g. H~2~O~2~) can cause cellular and tissue damage by degradation of polyunsaturated lipids to form MDA, which is a reactive aldehyde causing toxic cellular stress \[[@pone.0218019.ref038]\]. Nipponbare, despite showing the highest H~2~O~2~ production under stress ([Fig 2B](#pone.0218019.g002){ref-type="fig"}), showed the lowest damage on the plasma membrane ([Fig 2C](#pone.0218019.g002){ref-type="fig"}); whereas the tolerant M202 under stress, with no significant H~2~O~2~ induction, showed damage with MDA induction. In contrast, the sensitive genotypes had high production of H~2~O~2~ and MDA, causing a great impact on the cell membranes, as shown previously \[[@pone.0218019.ref039]\].
The antioxidant enzymes SOD, POD and CAT, present in tolerant genotypes can compete against ROS formation \[[@pone.0218019.ref040]\], providing antioxidant activity to offer protection from oxidative stress damage \[[@pone.0218019.ref041]\].These enzymatic mechanisms, found among plants challenged to cold and other abiotic stresses, function in ROS scavenging by the reduction of superoxide radicals into H~2~O~2~ as analyzed here for the SOD ([Fig 3C](#pone.0218019.g003){ref-type="fig"}), CAT ([Fig 3D](#pone.0218019.g003){ref-type="fig"}) and POD ([Fig 3E](#pone.0218019.g003){ref-type="fig"}) expression activities that catalyze H~2~O~2~ into H~2~O and protect the plant cells from H~2~O~2~ accumulation \[[@pone.0218019.ref042]\]. The bulk of H~2~O~2~ from SOD catalysis remains biologically toxic. SOD activity observed under stress ([Fig 3C](#pone.0218019.g003){ref-type="fig"}) in cold tolerant genotypes, can limit plant damage from ROS, and is lower in cold-sensitive genotypes. H~2~O~2~ production by SOD enzymes can also function in oxidative stress signaling, to play the role of a secondary messenger and protect reactions leading to induced CAT and POD activity in plants \[[@pone.0218019.ref043]\]. Biochemical analysis of cold-sensitive genotypes revealed that the lower increase of CAT and POD activity could reduce the efficiency of the plant cells to scavenge damaging free radicals. The analyses suggest that most of the plant's response in increasing antioxidant activity has an important role towards cold stress tolerance. The high stability and increased rate of CAT and POD activity are known to confer cold-induced oxidative stress tolerance \[[@pone.0218019.ref044],[@pone.0218019.ref045]\].
DPPH application was used to measure free radicals as a measure of stress tolerance \[[@pone.0218019.ref046]\]. The increase in DPPH radical scavenging ([Fig 3F](#pone.0218019.g003){ref-type="fig"}), observed in Nipponbare and M202, also appears to be correlated with the degree of plant stress tolerance \[[@pone.0218019.ref047],[@pone.0218019.ref048]\]. Cold tolerance has been quantified by measuring the reduction in growth rate and cell membrane stability \[[@pone.0218019.ref049]\]. The DPPH assay reveals a higher antioxidant capacity in the tolerant genotypes ([Fig 3F](#pone.0218019.g003){ref-type="fig"}), supporting the relationship between antioxidant capacity and cold tolerance. The observed relationship between CAT and POD activity, and cell membrane stability, supports the importance of sustaining an optimum antioxidant content under stress for the expression of cold tolerance. Cold-tolerance in plants is generally associated with a higher antioxidant capacity that is induced under stress, compared to the response of cold-sensitive plants \[[@pone.0218019.ref050]\].
Gene expression responses to cold-stress treatment {#sec005}
--------------------------------------------------
To support our studies on phenotypic and biochemical responses to cold stress of tolerant and sensitive rice genotypes, a bibliographic review was conducted and identified several stress-responsive genes under low-temperature conditions. From this literature, we selected nine cold-regulated genes, and their expression behavior was characterized in the four different genotypes at different time points after stress (3h, 6h, 24h, and 48h) in the vegetative plant growth stages ([Fig 5](#pone.0218019.g005){ref-type="fig"}). Cold acclimation can involve alterations in gene expression and changes in the levels of particular proteins following cold treatment \[[@pone.0218019.ref051]\].
{ref-type="supplementary-material"}) conferring stress tolerance to low temperatures in rice.\
Stress tolerance related genes are shown in a time-course of 03--48 h after stress initiation. Data are results from three biological replicates and are expressed as the relative quantification (RQ) ratio of fold change of stress treatment to control.](pone.0218019.g005){#pone.0218019.g005}
The plant cell wall, which can be considered a layer to safeguard against abiotic stress such as cold \[[@pone.0218019.ref036]\], is a complex structure inserted in a physiologically active pectin matrix, cross-linked with structural proteins and, depending on the tissue/organ, with lignin \[[@pone.0218019.ref052],[@pone.0218019.ref053]\]. The *OsBURP16* gene, encodes a putative precursor of PG1β, a subunit that regulates the activity of polygalacturonase (PG), an enzyme which hydrolyzes pectin, and changes the composition of the plant cell wall \[[@pone.0218019.ref054]\]. Our data showed that the expression of *OsBURP16* was induced maximally in sensitive genotypes (Secano do Brazil and Cypress) and had an increased transcript level with time, while Nipponbare and M202 had little variability in the same period ([Fig 5](#pone.0218019.g005){ref-type="fig"}). Therefore, the *OsBURP16* gene can be considered as an indicator of cold stress sensitivity \[[@pone.0218019.ref054]\]. Unlike *OsBURP16*, *OsACA6* exhibited an increased level of expression in tolerant genotypes and Secano do Brazil (sensitive), however, Nipponbare had a higher expression. Located in the plasma membrane are pumps or antiporters, which act on the cold response pathway involving the influx of Ca^2+^ from the apoplast to cytosol \[[@pone.0218019.ref055]\].
Under cold stress, plants exhibit a decrease in H~2~O~2~ and MDA content, with an increase in membrane stability, CAT, SOD and APX expression \[[@pone.0218019.ref035]\], and proline content \[[@pone.0218019.ref029]\]. These changes have been shown to indicate a positive response of cold tolerance in transgenic tobacco \[[@pone.0218019.ref056]\].
The *OsGH3-2* and *Ctb1* genes show early and maximal induction after 3 h of stress and decrease with time in the tolerant Nipponbare and M202, while in Cypress and Secano do Brazil there is an increase in expression with time ([Fig 5](#pone.0218019.g005){ref-type="fig"}). The overexpression of *OsGH3-2* \[[@pone.0218019.ref057]\], has been shown to decrease drought resistance and stomatal closure, as well as increase water loss and improvement of cold and oxidative stress tolerance in rice at the vegetative stage. On the other hand, \[[@pone.0218019.ref058]\] report that in association with CAT, *Ctb1* participates directly in the regulatory pathway of small-RNAs and promotes cold-tolerance at the reproductive stage.
*OsSRO1a* and *OsSAP1* were quickly induced after 3 h, reaching a maximum at 48 h in Nipponbare and M202 ([Fig 5](#pone.0218019.g005){ref-type="fig"}). For the sensitive genotypes, we found the same pattern, although the expression was lower. SRO are proteins involved in ADP-ribose conjugation, DNA repair, apoptosis, transcription, and chromatin remodeling \[[@pone.0218019.ref059]\]. They possess a C-terminal RCD1-SROTAF4 domain and interact with *AP2/EREBP* and transcription factors *OsDREB2A* \[[@pone.0218019.ref060]\]. In rice, *OsSRO1* has a role in drought and oxidative stress tolerance, stomatal closure and H~2~O~2~ accumulation \[[@pone.0218019.ref061]\]. *SAPs* can interact with proteins via their zinc-finger domains \[[@pone.0218019.ref062]\], such as *OsSAP1* with cytoplasmic kinase *OsRLCK253* \[[@pone.0218019.ref063]\]. *OsSAP1* can regulate the stress responses by either modulating the expression of genes or by interaction with other proteins \[[@pone.0218019.ref064]\]. Altogether, our results suggest that this interaction may have relevance in stress physiology and cold-acclimation.
Trehalose-6-phosphate phosphatase (TPP) is a sugar storage metabolic regulator and acts in protection against abiotic stress \[[@pone.0218019.ref065],[@pone.0218019.ref066]\]. In rice, overexpression has shown increased tolerance to abiotic stresses \[[@pone.0218019.ref067],[@pone.0218019.ref068]\]. Our data showed that the *OsTPP1*gene is induced after 24 h and increases more at 48 h in the tolerant genotypes, while Cypress and Secano do Brazil show a slight induction at 3 h that goes down completely after 48h ([Fig 5](#pone.0218019.g005){ref-type="fig"}). These results agree with other studies \[[@pone.0218019.ref069]\], that show overexpression of *OsTPP1*is enhanced in salt and cold tolerance of rice. Moreover, they also showed that in plants with high expression of *OsBURP16* there is a decrease in trehalose content.
Among the genes studied for expression under cold stress, *OsZFP245* and *OsSFR6* showed the highest induction in the tolerant genotypes, with the peak induction at 48 h. In Cypress, *OsSFR6* showed increased expression unto 24 h. However, in the sensitive Secano do Brazil both genes (*OsZFP245* and *OsSFR6*) were induced early with maximal expression at 3 h, and sensitive Cypress induced *OsSFR6* early with the highest expression at 24h ([Fig 5](#pone.0218019.g005){ref-type="fig"}). Located in the nucleus, *OsSFR6* acts to induce Cold-On Regulated (COR) genes via transcription factors CRT⁄DREs 2A and CBFs 1--3 \[[@pone.0218019.ref070]\], and *OsZFP245*, a zinc finger protein gene with the role of increasing proline content and antioxidant enzymes \[[@pone.0218019.ref071],[@pone.0218019.ref072]\].
A model for low-temperature tolerance mechanisms ([Fig 6](#pone.0218019.g006){ref-type="fig"}) was developed by integrating information on gene expression response to cold in the genotypes, to improve our understanding of how stress is perceived by cells and how the regulatory cascade of signals act to promote tolerance to suboptimal temperatures.
{ref-type="supplementary-material"}) leading to low temperature tolerance, with an increase in concentration due to the presence of ROS. Abbreviations shown indicate the changes in components affected. **Pm**: Plasma membrane; **Cw**: Cell wall; **Ch**: Chloroplast; **N**: Nucleus; **Grey arrow**: Calcium efflux **Black arrows**: Cold perception; **X**: Degradation of pectin caused by increased polygalacturonase induced by increased expression of *OsBURP16*; **Red arrow**: Induction; **Blue arrow**: Induction of cold tolerance; **Between brackets**: Increase in concentration due to ROS. **Dotted**: Association of *Ctb1* and CAT for miRNA induction. **Upward arrows**: Increased concentration due to induction of ROS; **X**: Association of *Ctb1* and CAT for miRNA induction. The regulatory cascade of perception and induction of damage in response to low temperatures, and response of genes to the stress treatment, as well as the induction of biochemical responses leading to tolerance to low temperatures, leading from an increased concentration of ROS.](pone.0218019.g006){#pone.0218019.g006}
Stress is perceived by cells through molecular signals, which in turn, induce a concomitant expression of multiple genes (cited in the text). They act in the perception and induction of damage caused by suboptimal temperatures generating responses that may, for example, increase the activity of antioxidant and osmoprotectant enzymes, and result in greater tolerance to stress.
Differential protein expression under cold-stress {#sec006}
-------------------------------------------------
To improve our understanding on the response of rice plants to cold stress, we used a comparative proteomics approach to study the effect of cold stress on rice genotypes differing in tolerance and observed differential protein expression between tolerant and sensitive genotypes ([S1 Table](#pone.0218019.s001){ref-type="supplementary-material"}).
The leucine-rich repeats protein kinases (LRR-RLKs) play an important role in regulating plant responses to abiotic stress \[[@pone.0218019.ref073]\], supported by our data that shows tolerant genotypes have higher expression compared to the sensitive. These results were also observed earlier \[[@pone.0218019.ref074]\], which supports that *GsLRPK* increases kinase activity in the presence of cold stress and increases expression of low-temperature- responsive genes, resulting in an enhancement in the tolerance to cold stress. Other kinases like the Fructokinase-2, Phosphoribulokinase, and Nucleoside diphosphate kinase 1 genes are also regulated by environmental stresses and show the same expression behavior in tolerant genotypes.
GF14 shows significant homology with protein kinase-dependent regulatory proteins \[[@pone.0218019.ref075]\]. In rice, the 14-3-3 proteins GF14b, GF14c, GF14e and GF14f, interact with target proteins that are involved in stress response \[[@pone.0218019.ref076]\]. GF14-a showed low expression in sensitive genotypes and high expression for GF14-f, suggesting that they may be involved in mechanisms of tolerance and/or acclimatization to adverse environmental conditions.
The bHLH proteins are a group of transcription factors that carry out key roles in phytochrome signal transduction, cell fate determination and stomatal differentiation \[[@pone.0218019.ref077]\]. These transcription factors are stress-inducible under drought, cold or high-salinity and are brassinosteroid-responsive\[[@pone.0218019.ref077],[@pone.0218019.ref078]\], and this inductive behavior was also observed in the cold tolerant M202 and Nipponbare genotypes.
In plants, methylglyoxal a by-product of glycolysis, is toxic and causes damage to cells, and high cellular concentrations are generated from unfavorable environmental conditions \[[@pone.0218019.ref079],[@pone.0218019.ref080]\]. Glyoxalase proteins are very important for limiting methylglyoxal levels, and for this, the plants have a glyoxalase system with multiple isoforms of both GLYI and GLYII proteins\[[@pone.0218019.ref081],[@pone.0218019.ref082],[@pone.0218019.ref083]\]. Due to its role in methylglyoxal detoxification, over-expression of glyoxalase system in plants confers significant tolerance against adverse environmental conditions \[[@pone.0218019.ref079],[@pone.0218019.ref084],[@pone.0218019.ref085],[@pone.0218019.ref086],[@pone.0218019.ref087],[@pone.0218019.ref088]\]. Evidence of increased tolerance was demonstrated by elevated protein expression (Putative glyoxalase I) in the tolerant genotypes studied.
Lipid transfer proteins (LTPs) play an important role in abiotic stress tolerance \[[@pone.0218019.ref089]\], and can facilitate the inter-membrane exchange and transfer of various amphiphilic molecules including phospholipids, glycolipids, steroids, acyl-CoAs, and fatty acids \[[@pone.0218019.ref090]\]. Transcript levels of LTPs increased in response to drought \[[@pone.0218019.ref091]\], salt \[[@pone.0218019.ref092]\] and cold \[[@pone.0218019.ref093]\] and the increase of non-specific lipid-transfer protein 1 expression in tolerant genotypes give more support to its role of increasing tolerance to abiotic stress.
Conclusions {#sec007}
===========
This study supports that the genotypes Nipponbare and M202 have tolerance to low temperatures with the evidence of physiological responses, such as photosynthesis showing lower reduction, better efficient use of water without suffering photoinhibition, or reduction in the Quantum Efficiency of PSII. The biochemical profile showed that for the same genotypes, chlorophyll biosynthesis was not affected. Among the anthocyanins, a significant decrease in their content was observed, which identified pigments associated with the leaf mesophyll that act directly in the elimination of oxygen radicals produced by the chloroplasts. Accumulation of proline, glucose, and sucrose was also observed, these being osmoprotectants against freezing and dehydration damage. Antioxidants in the same tolerant genotypes, despite showing high production of H~2~O~2~ under stress, did not cause a high impact on the plasma membrane or the high activity of the antioxidant enzymes. SOD, CAT, POD and DPPH enzymes play an important role in stress tolerance. Differential expression of genes and proteins: the genes *OsGH3-2*, *OsSRO1a*, *OsZFP245* and *OsTPP1*; and the LRR-RLKs, BHLH, GLYI, and LTP1 proteins, showed a clear difference in expression between tolerant and sensitive, thus suggesting that these genes are good candidates for identification of low-temperature tolerant genotypes in rice that are capable of maintaining growth, development, and production at the desired agronomic levels. Finally, based on our studies, a schematic representative model of cold tolerance in rice ([Fig 6](#pone.0218019.g006){ref-type="fig"}) is proposed outlining mechanisms of action of the genes analyzed with differential responses in resistant genotypes, with the objective of improving our understanding of the operation of tolerance to low temperatures. To summarize the results, our analysis shows for the first time the role of different antioxidants and osmolytes in modulating the physiological responses contributing to tolerance. In addition, this report also identifies markers for screening of cold tolerance in multiple rice genotypes, along with few putative protein markers identified from LCMS/MS analysis.
Materials and methods {#sec008}
=====================
Plant growth conditions and cold stress treatment {#sec009}
-------------------------------------------------
Seeds of the genotypes temperate *japonica* Nipponbare and M202 (tolerant), and the tropical *japonica* Cypress and Secano do Brazil (sensitive) from the USDA mini-core collection \[[@pone.0218019.ref094]\] were sterilized, immersed in deionized water and germinated in the dark for five days. After germination, seedlings of the same size were transplanted into 500 ml capacity plastic pots filled with commercial substrate (Redi-earth) with known water retention capacity and weight. All the individual genotypes in the pots were grown in a Conviron growth chamber set at26/22° C ±1°C day/night temperature and light intensity of 600 μmolm^-2^s^-1^, with day/night cycle of 14/10h, and kept in trays filled with water simulating flooded conditions with periodic commercial fertilization using Miracle-Gro (Scotts Miracle-Gro Products) for the entire period of the experiment.
Fifteen days after transplanting, the uniformly growing plants were divided into two groups with 20 pots each, five pots per genotype, each containing one plant. One group was kept under ideal conditions (28° C) and served as control, while another set of trays was used for low treatment temperature (10° C) as described \[[@pone.0218019.ref095]\] with adaptations. For both treatments, the plants were kept under the conditions described for a period of 72h. For this experiment, the experimental design was a complete randomized design and the data were submitted to analysis of variance (ANOVA) and Student\'s t-test (1%), using the SAS 9.3 statistical program (SAS Institute, Cary, NC).
For the analysis of gene expression, 2 cm leaf tissue fragments were collected at four times 3h, 6h, 24h, and 48h, after low-temperature initiation, while the samples for the biochemical and proteomic analyses were collected only at 72 hours. The photosynthesis and photochemical efficiency of PSII were evaluated using the second fully expanded leaf at 72h using a portable photosynthesis meter (LI-6400XT; LI-COR) at a CO~2~ concentration of 370 μmol mol^-1^light intensity of 1,000 μmol m^-2^ s^-1^ and 55% to 60% relative humidity.
Biochemical analysis {#sec010}
--------------------
Chlorophyll content was estimated following the method of \[[@pone.0218019.ref096]\], using absorbance for chlorophyll a at 663 nm and chlorophyll b at 645 nm, that was measured with a UV-Vis spectrophotometer. The hydrogen peroxide content was determined as described \[[@pone.0218019.ref097]\], with absorbance at 390 nm. Malondialdehyde (MDA) Buffer solution was made with 0.07% NaH~2~PO~42~H~2~O and 1.6% Na~2~HPO~4~, 1% H~2~O~2~, and 20% trichloroacetic acid containing 0.5% thiobarbituric acid. The absorbance of the supernatant was read at 532nm and MDA concentration was calculated using the MDA extinction coefficient of 155 mM^-1^cm^-1^ \[[@pone.0218019.ref098]\]. Anthocyanin was quantified as described \[[@pone.0218019.ref099]\], the absorbance measured at 525 nm and the anthocyanin concentration calculated in mg/gm of fresh weight using the millimolar extinction coefficient of 31.6. The content of phenolics in the extract was determined according to \[[@pone.0218019.ref100]\] with some modifications, with absorbance measured at 760 nm.
For the determination of Superoxide dismutase (SOD), Catalase (CAT) and total Peroxidases (POD), the total proteins were extracted as described \[[@pone.0218019.ref101]\]. Total Superoxide dismutase activity, the basis of its ability to inhibit the photochemical reduction of nitroblue tetrazolium (NBT) \[[@pone.0218019.ref102]\], was assayed as described \[[@pone.0218019.ref103]\] with some modifications, the absorbance of the reaction mixture measured at 560 nm. Catalase enzyme activity was assayed as described \[[@pone.0218019.ref104]\], with the decrease in H~2~O~2~ assayed by a decrease in optical density at 240 nm, and the activity calculated using the extinction coefficient of 40 mm^-1^ for H~2~O~2~. The peroxidase (POD) activity was determined using the method of \[[@pone.0218019.ref105]\].
The antioxidant activity of the extracts, based on the scavenging activity of the stable 1,1-diphenyl-2-picrylhydrazyl (DPPH) free radical, was determined following \[[@pone.0218019.ref106]\]. The absorbance at 517 nm was used to calculate \[(A~0~-A~1~)/A~0~\] x 100 where A~0~ is the absorbance of the control, and A~1~ is the extract. Glucose was estimated using 3,5-dinitrosalicylic acid (DNS) according to the method described \[[@pone.0218019.ref107]\]. The absorbance was recorded at 570 nm where glucose served as the standard. The sucrose was estimated using the method described by \[[@pone.0218019.ref108]\], and the absorbance was recorded at 620 nm. Free proline content was determined according to the procedure of \[[@pone.0218019.ref109]\], and the red color intensity was measured at 520 nm.
Analysis of Gene expression {#sec011}
---------------------------
Total RNA was extracted, using Trizol reagent (Invitrogen), from rice genotypes and cDNA was synthesized using 2 μg total RNA treated with DNAse using GoScript reverse transcription system (Promega). RT-qPCR reactions were performed using GoTaq qPCR Master Mix (Promega) with Ubiquitin as an internal reference gene \[[@pone.0218019.ref110]\]in a 96-CFX thermocycler(Bio-Rad).The temperature increase (0.5° C 10 s^-1^) from 55°C to 95°C was used for the analysis of the melting curve. Non-transcribed RNA was also run as a negative control. For qPCR analysis the primers were selected based on literature searches on their role in cold tolerance and sequences derived from accession numbers obtained from rice genome databases were used for primer designing using IDT primer designing tool. The primers used are listed in [S2 Table](#pone.0218019.s002){ref-type="supplementary-material"}. The relative expression differences for each of the samples in individual experiments was determined by normalizing the Ct value for each gene in relation to Ubiquitin Ct value and the relative fold change was calculated using the equation 2^-ΔΔCt^ \[[@pone.0218019.ref111]\]. The expression analysis was performed with three biological replicates and two technical replicates. Total RNA isolated from three different leaves collected from three different plants under control and treatment are treated as biological replicates while an aliquot of same sample for each sample was duplicated serving as technical replicates. Data were used in the analysis of variance (ANOVA) and Student\'s t-test (1%), using the statistical program SAS 9.3 (SAS Institute, Cary, NC).
Analysis of protein expression {#sec012}
------------------------------
Total protein of four rice genotypes used in this work was extracted using the Protein Isolation Buffer and methodology as described by \[[@pone.0218019.ref095]\]. The protein concentration in each sample was determined by the Bradford assay \[[@pone.0218019.ref112]\] using bovine albumin as the standard (Fraction V, Sigma). Total protein samples were loaded onto SDS-PAGE-Gel, with samples of 90 μg of protein. Spots of interest, showing differences were excised from the gel and digested using the protocol described by \[[@pone.0218019.ref113]\]. All MALDI-MS and MS/MS analyses were performed using Ultraflex II MALDI-TOF/TOF mass spectrometer (Bruker Daltonik, Bremen, Germany). All LC-MS/MS was performed using Bruker Amazon-SL quadrupole ion trap mass spectrometer with a captive spray ionization source. The resulting LC-MS/MS spectra were analyzed by Skyline-daily 3.6.9 software and shown in [S1 Table](#pone.0218019.s001){ref-type="supplementary-material"} \[[@pone.0218019.ref114]\].
Supporting information {#sec013}
======================
###### Genes with accession numbers, forward and reverse primer sequences, and efficiency.
(PDF)
######
Click here for additional data file.
###### Summary of proteins identified by MS/MS (MALDI TOF/TOF) differentially expressed between the tolerant (Nipponbare and M202) and sensitive (Secano do Brazil and Cypress) genotypes in response to cold stress.
(PDF)
######
Click here for additional data file.
###### Quantification of biochemical changes under cold stress.
(XLSX)
######
Click here for additional data file.
###### Quantification of whole plant gas-exchange physiological changes under cold stress.
(XLSX)
######
Click here for additional data file.
###### Quantitative PCR results of stress responsive differentially expressed genes.
(XLSX)
######
Click here for additional data file.
[^1]: **Competing Interests:**The authors have declared that no competing interests exist.
| {
"pile_set_name": "PubMed Central"
} |
Background {#Sec1}
==========
In situ cholangiocarcinoma, i.e., epithelial carcinoma without submucosal invasion, is asymptomatic and very difficult to detect by multidetector-row computed tomography (MDCT) or direct cholangiography. Thus, this type of carcinoma is rarely resected with a preoperative definitive diagnosis, especially, nonpapillary flat type in situ carcinoma, which is extremely rare.
Here, we present a case of flat type in situ perihilar cholangiocarcinoma that was incidentally accompanied by hilar bile duct stricture after open cholecystectomy. Preoperative diagnostic imaging studies did not identify the lesion, and only aspiration bile cytology showed a positive result.
Case presentation {#Sec2}
=================
A 70-year old man was diagnosed with gallbladder cancer (Fig. [1](#Fig1){ref-type="fig"}a) and received open cholecystectomy with lymphadenectomy of the hepatoduodenal ligament at a local hospital. The final pathology showed that the tumor was a moderately differentiated adenocarcinoma with an invasion depth of the mucosal layer. No lymph node metastases were found, and all of the surgical margins were negative (Fig. [1](#Fig1){ref-type="fig"}b). The patient was discharged from the hospital, but 3 months later, magnetic resonance imaging (MRI) showed hilar bile duct stricture with intrahepatic biliary dilatation (Fig. [2](#Fig2){ref-type="fig"}a), probably due to delayed bile duct ischemia caused by lymphadenectomy. Then, biliary drainage was performed with endoscopic biliary stenting (Fig. [2](#Fig2){ref-type="fig"}b). Three months later, the patient experienced cholangitis with septic shock, and percutaneous transhepatic biliary drainage (PTBD) into the left intrahepatic bile duct was performed (Fig. [2](#Fig2){ref-type="fig"}c). Unexpectedly, the aspiration bile cytology from the PTBD catheter showed malignant cells (Fig. [2](#Fig2){ref-type="fig"}d). Percutaneous transhepatic cholangioscopy (PTCS) was performed via the sinus tract of PTBD, but the examination failed to detect any malignant lesions in the biliary tree. The patient was referred to our clinic for possible surgical treatment. Fig. 1Images of gallbladder cancer. **a** Computed tomography revealing a papillary tumor in the gallbladder. **b** Resected specimen. The yellow arrows indicate a papillary tumor Fig. 2Images of the bile stricture. **a** MRCP showing a hilar bile duct stricture (yellow arrowheads). **b** Endoscopic biliary stent placed in the right posterior inferior bile duct (B6). **c** Percutaneous transhepatic biliary drainage into the left lateral anterior bile duct (B3). **d** Aspiration bile cytology from the PTBD showing malignant cells. MRCP, magnetic resonance cholangiopancreatography; CBD, common bile duct; PTBD, percutaneous transhepatic biliary drainage
After admission, the patient's cholangiograms were re-evaluated (Fig. [3](#Fig3){ref-type="fig"}a, b). The right posterior inferior bile duct (B6) was the infraportal type and joined the common hepatic duct. The hilar bile duct stricture was 3 cm in length. Intraductal ultrasonography (IDUS) did not detect any malignant cells in the bile duct around the hilar stricture, and no cancer cells were found in the endoscopic biopsy specimen. MDCT demonstrated a left portal vein obstruction, probably due to the inadvertent puncture of PTBD performed at the local hospital. Overall, no findings that indicated cholangiocarcinoma were observed. However, we determined that surgery was needed to treat this complicated biliary stricture. Left hepatectomy with caudate lobectomy and extrahepatic bile duct resection was performed. Severe adhesion around the hepatoduodenal ligament resulted in a difficult surgery. Intraoperative frozen section examination revealed no malignancy of the resected common bile duct stump and biliary stricture. The stumps of the resected B6 and the right hepatic duct were reconstructed by Roux-en-Y cholangiojejunostomy (Fig. [3](#Fig3){ref-type="fig"}c). The operative time was 591 min, and blood loss was 1745 ml. Fig. 3Cholangiogram, preoperative schema, and operative picture. **a** Cholangiogram showing a hilar bile duct stricture (orange arrow). **b** Preoperative scheme. The planned surgery was left hepatectomy with caudate lobectomy. **c** Left hepatectomy with en bloc resection of the caudate lobe and extrahepatic bile duct was performed as scheduled. The right hepatic duct (RHD) and the right posterior inferior bile duct (B6a) were reconstructed separately
The postoperative course was uneventful, and the patient was discharged on postoperative day 12. In the final pathology, no malignant cells were observed in the hilar stricture, but flat type in situ carcinoma was found at the confluence of the right and left hepatic ducts, which was distant from the stricture (Fig. [4](#Fig4){ref-type="fig"}). Histologically, the carcinoma was a well-differentiated adenocarcinoma and 1 cm in diameter. No lymph node metastases were found, and all of the surgical margins were negative. There were no BilIN lesions in the background bile duct. The patient is still alive and in good health 54 months after the hepatectomy. Fig. 4Macroscopic and microscopic findings of specimens. **a** Resected specimen. **b** Flat type in situ carcinoma located near the hepatic confluence (the yellow dotted circle). The carcinoma is distant from the bile duct stricture. **c**, **d** Microscopic findings of in situ carcinoma
Discussion {#Sec3}
==========
When arterial blood flow to the bile duct is restricted, a bile duct stricture will develop \[[@CR1], [@CR2]\]. Several authors have reported delayed bile duct strictures due to ischemia after excision of the tissue surrounding the biliary tree \[[@CR3]--[@CR5]\]. Ishizuka et al. \[[@CR3]\] reported two cases of delayed ischemic biliary stricture after radical lymphadenectomy in the hepatoduodenal ligament with skeletonization of the extrahepatic bile ducts for malignant diseases: in both cases, histologic examination of the subsequently resected biliary strictures revealed evidence of ischemia. Skeletonization of the extrahepatic bile duct may induce ischemia then delayed stricture formation. Lymphadenectomy of the hepatoduodenal ligament is routinely performed in advanced gallbladder carcinoma. However, when preserving the extrahepatic bile duct, this surgical procedure may induce bile duct stricture, and the present case is a case of delayed ischemic stricture. For prevention of this complication, assessment of arterial perfusion in the bile duct wall using indocyanine green near-infrared imaging may be a promising way \[[@CR5]\].
Flat type precursor lesions are called biliary intraepithelial neoplasias and are classified into three grades \[[@CR6]\]. Especially with severe atypia, the lesions are identical to in situ carcinoma. This kind of epithelial carcinoma is often detected as a lesion accompanied by invasive carcinoma \[[@CR7]\], while in situ carcinoma alone is rarely detected. According to our previous study \[[@CR8]\], only 3 cases of in situ carcinoma were found in 545 consecutive resections of perihilar cholangiocarcinoma. Thus, the incidence of in situ perihilar cholangiocarcinoma was 0.55%.
In this case, aspiration bile cytology alone showed a positive result. We performed aspiration cytology from the PTBD catheter five times, which was positive three times. Tsuchiya et al. \[[@CR9]\] reported that the positivity of aspiration bile cytology increased when repeatedly performed but reached a plateau after five examinations. Aspiration cytology is easy to perform and repeatable \[[@CR10]\]; thus, it should be used, especially in cases of negative biopsy results. However, aspiration cytology never indicates location of the lesion, which is a major limitation of this examination.
Conclusions {#Sec4}
===========
We experienced a rare case of flat type in situ perihilar cholangiocarcinoma that was incidentally accompanied by a benign bile duct stricture. When the tumor is undetectable but aspiration cytology is positive, in situ cholangiocarcinoma may exist; thus, surgery should be carefully considered.
CBD
: Common bile duct
ENBD
: Endoscopic nasobiliary drainage
IDUS
: Intraductal ultrasonography
MDCT
: Multidetector-row computed tomography
PTBD
: Percutaneous transhepatic biliary drainage
PTCS
: Percutaneous transhepatic cholangioscopy
**Publisher's Note**
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
We would like to thank American Journal Experts ([www.aje.com](http://www.aje.com)) for the English language editing.
MN, NW, MN, and KS performed the surgery and perioperative management on the patient, and MN drafted the manuscript. All authors read and approved the final manuscript.
The authors declare that they received no financial support pertaining to this case report.
The authors declare that all data in this article are available within this published article.
Not applicable.
Written informed consent was obtained from the patient for the publication of this case report and any accompanying images.
The authors declare that they have no competing interests.
| {
"pile_set_name": "PubMed Central"
} |
Introduction
============
Endovascular treatment (EVT) for iliofemoral artery lesions has become common in the last decade due to technical developments related to self-expandable nitinol stents, and EVT is now an alternative first-line treatment with long-term outcomes that are not inferior to those after surgery.[@b1-mder-8-467],[@b2-mder-8-467] The high procedural success rate and reduced invasiveness has encouraged extension of indications for endovascular revascularization, but with little focus on hemodynamic significance. The current guidelines for peripheral artery disease indicate that EVT should be used with optimal drug treatment and supervised exercise therapy when clinical findings suggest a reasonable likelihood of symptomatic improvement. These guidelines also recommend pressure measurements across lesions at rest and under induced hyperemia in cases in which the hemodynamic significance of the target lesion is unclear.[@b3-mder-8-467],[@b4-mder-8-467] Various parameters have been proposed to determine hemodynamic significance, including mean pressure gradient (MPG), peak-to-peak PG, and ratio of mean pressure (also referred to as the peripheral fractional flow reserve \[pFFR\]) with or without vasodilators.[@b5-mder-8-467]--[@b9-mder-8-467] However, the validity of these parameters for treatment planning is unclear. Therefore, the purpose of this study was to verify whether trans-stenotic absolute PG and pFFR measured using a 0.014-inch pressure wire under hyperemia can be used to establish hemodynamic significance in iliofemoral angiographically intermediate stenosis.
Methods
=======
Study design and patient population
-----------------------------------
Pressure measurements were performed for 37 consecutive patients (40 lesions) with claudication and critical limb ischemia who had iliofemoral atherosclerotic artery disease for which EVT was planned according to comprehensive consideration of symptoms, quantitative vessel angiography results or duplex ultrasound (DUS) study in advance between April and August 2012 at Kokura Memorial Hospital. Patients who were planning to undergo EVT for TransAtlantic Inter-Society Consensus (TASC) II class C or D lesions, infrapopliteal artery lesions, chronic total occlusion, and surgical bypass grafts were excluded. Based on these criteria, the subjects were 22 patients (23 lesions) with intermediate stenosis, defined as \<75% stenosis on quantitative vessel angiography, identified retrospectively from our database. We did not defer the EVT from the results of pressure measurements. Kokura Memorial Hospital review board/ethics committee approved the study protocol. Written informed consent was obtained from each patient.
Pressure measurements
---------------------
Pressure measurements were performed using a 0.014-inch pressure wire (300 cm PrimeWire Prestage^®^; Volcano Corporation, San Diego, CA, USA) following angiography and just before starting EVT. The pressure wire was calibrated before the procedure, equalized at the end of the guiding system, advanced over the lesion, and positioned in a distal healthy area. The mean trans-stenotic absolute PG (baseline MPG) and mean pressure ratio (baseline MPR) were then measured. Next, 250 µg of isosorbide dinitrate (ISDN) was injected via the guiding system to induce a hyperemic state. Hyperemic MPG and MPR were then measured, usually approximately 1 minute after flushing the guiding system with saline. MPG was defined as the difference between the mean pressure in the distal healthy area across the lesion from the pressure wire (mean distal pressure) and the mean central aorta pressure from the guiding system (mean aorta pressure). MPR was calculated as mean distal pressure/mean aorta pressure. Hyperemic MPR was considered equivalent to pFFR.
DUS study
---------
The peak systolic velocity ratio (PSVR) on duplex (DUS) was used to define the hemodynamic severity of lesions. Tests were performed by certified sonographers within 30 days before pressure measurements. PSVR was calculated by dividing the velocity measured at the point of maximum stenosis by the velocity in the closest adjacent healthy vessel segment.
Outcomes
--------
Comparisons of baseline MPG to hyperemic MPG, and baseline MPR to hyperemic MPR (pFFR) were performed to assess responses of these parameters to vasodilators. Correlations of PSVR on DUS with hyperemic MPG and pFFR were examined. Optimal cut-off values were determined for hyperemic MPG and pFFR as indicators of hemodynamically significant stenosis (PSVR \>2.5 on DUS).
Statistical analysis
--------------------
Continuous variables are presented as means with standard deviation and categorical variables as numbers with percentages. Continuous variables were compared by Student's *t*-test and correlations between continuous variables were examined using simple linear regression. A receiver operating characteristic analysis was used to determine the optimal cut-off values for hyperemic MPG and pFFR. *P*\<0.05 was regarded as significant. All analyses were performed using JMP^®^ version 10.0 (SAS Institute Inc., Cary, NC, USA).
Results
=======
The characteristics of the patients are shown in [Table 1](#t1-mder-8-467){ref-type="table"} and the characteristics of lesions are shown in [Table 2](#t2-mder-8-467){ref-type="table"}. The patients had a mean age of 72.2±6.8 years (range: 61--86 years) and 19 were male. Almost half of the patients had coronary artery disease (CAD) and three were receiving hemodialysis. All patients had intermittent claudication, except for four with critical limb ischemia. The mean % stenosis diameter was 57.3%±13.2%, and 91% of the lesions were classified as TASC II A.
MPG increased significantly after injection of ISDN (from 9.0±5.7 to 16.3±6.2 mmHg; *P*\<0.05, [Figure 1](#f1-mder-8-467){ref-type="fig"}). Baseline MPG was not significantly correlated with PSVR (*R*=0.33; *P*=0.12); however, hyperemic MPG was significantly correlated with PSVR (*R*=0.52; *P*\<0.05, [Figure 2](#f2-mder-8-467){ref-type="fig"}).
MPR decreased significantly after injection of ISDN (from 0.92±0.06 to 0.81±0.07; *P*\<0.05, [Figure 3](#f3-mder-8-467){ref-type="fig"}). Baseline MPR was not significantly correlated with PSVR (*R*=−0.32; *P*=0.16); however, hyperemic MPR (pFFR) was significantly correlated with PSVR (R=−0.50; *P*\<0.05, [Figure 4](#f4-mder-8-467){ref-type="fig"}).
The optimal cut-off value for pFFR for hemodynamically significant stenosis (defined as PSVR \>2.5) was 0.85 (area under the curve \[AUC\] 0.72; sensitivity 94%; specificity 50%, *P*\<0.05, [Figure 5](#f5-mder-8-467){ref-type="fig"}). Hyperemic MPG was not a significant predictor of hemodynamically significant stenosis (AUC 0.48; sensitivity 71%; specificity: 87%, *P*=0.95).
Discussion
==========
The major findings in this study of iliofemoral angiographically intermediate stenosis were that hemodynamic significance can be established more accurately under hyperemia induced by ISDN, compared to a resting state; that hyperemic MPG and hyperemic MPR (pFFR) are both significantly correlated with PSVR on DUS; and that the optimal cut-off value for pFFR as an indicator of hemodynamic significance (PSVR \>2.5) was 0.85.
In CAD, common physiological findings are included in planning a revascularization strategy for non-severe stenosis. We routinely employ stress tests for coronary blood flow through exercise or pharmacologic stimulation, and coronary pressure measurements are commonly used for assessment of the ischemic potential of a stenosis in our daily practice. Recent large clinical trials have established fractional flow reserve (FFR) measured using a pressure wire as a standard diagnostic tool in patients with non-severe stenosis.[@b10-mder-8-467],[@b11-mder-8-467] In peripheral artery disease, hyperemic translesional MPG and renal FFR are important as physiological parameters in endovascular revascularization for renal artery stenosis (RAS). Current American College of Cardiology/American Heart Association guidelines recommend pressure measurements for understanding the functional significance in RAS.[@b12-mder-8-467],[@b13-mder-8-467]
Unlike in CAD or RAS, pressure measurements are not commonly used for iliofemoral disease with intermediate stenosis although there might be high frequencies of visual-functional mismatch between moderate morphologic stenosis (\>50% in angiography) and pressure ratio.[@b14-mder-8-467] A residual MPG \<10 mmHg after intervention has been suggested as an indicator of hemodynamic significance in iliac arteries.[@b4-mder-8-467] However, this criterion is based on findings obtained using 5 Fr catheters, and large diameter catheters always overestimate the PG compared with a pressure wire.[@b5-mder-8-467] In addition, there are no data on the clinical value of using vasodilators and pFFR measurements for iliofemoral disease with intermediate stenosis. In contrast, DUS is frequently used for planning of iliofemoral artery revascularization (both endovascular and surgical) in daily practice. The quality of this examination depends on the experience of sonographers and data interpretation can be time-consuming; but the method is noninvasive, and provides information on lesion locations and stenosis in two dimensions, and on hemodynamic severity based on PSVR.[@b15-mder-8-467],[@b16-mder-8-467]
The results of the current study revealed the importance of inducing a hyperemic state in evaluation of iliofemoral artery circulation. The findings support our hypothesis that pressure measurements for iliofemoral angiographic mild to moderate stenosis can be used in place of DUS to evaluate hemodynamic significance. Thus, measurement of hyperemic MPG and pFFR using a pressure wire can be used to assess the hemodynamic severity of a lesion (which was not evaluated in DUS before EVT) in a catheter laboratory and to perform intervention with the pressure wire across the lesion based on pFFR reaching the threshold described above. We note that modern pressure wires have improved durability and torque, and thus EVT using a pressure wire is not stressful for operators, except for treating complex lesions such as those in the TASC II C/D class. Hyperemic MPG also had a significant correlation with PSVR on DUS, but we were unable to establish a threshold MPG as an indicator of hemodynamic significance in the current study. This parameter is an absolute value that depends strongly on blood pressure. Therefore, there is likely to be a wide range of hyperemic MPG values in iliofemoral arteries that may or may not reflect hemodynamic significance ([Figure 2](#f2-mder-8-467){ref-type="fig"}). Thus, we conclude that hyperemic MPR (pFFR) is a better measure of hemodynamic significance.
Limitations
===========
There were several limitations in the study. First, the study population was identified retrospectively and the sample size was small. Second, correlations between hyperemic pressure measurements and PSVR on DUS were significant, but only moderate. Third, some cases had diffuse long lesions (especially for superficial femoral lesions); thus, some of the lesions on which pressure measurements were performed in our catheter laboratory did not strictly correspond to the positions of lesions evaluated on DUS by sonography in the vascular laboratory. This inadequate correspondence might be one of the reasons behind the poor specificity in pFFR measurement (only 50%) in the present study. Finally, a dose of 250 µg of ISDN was used to induce hyperemia, but a different dose of ISDN or another vasodilator (nitroglycerine, papaverine, adenosine) may alter the results.[@b17-mder-8-467] In the future, a validation study examining the type and dose of vasodilator would be needed.
Conclusion
==========
Induction of a hyperemic state is required for pressure measurements in iliofemoral artery circulation. Hyperemic MPR (pFFR) measured using a 0.014-inch pressure wire is reliable for prediction of hemodynamic significance in iliofemoral angiographically mild to moderate stenosis. The optimal cut-off value of pFFR for hemodynamic significance was 0.85.
The authors thank Hiroyoshi Yokoi, MD (Fukuoka Sanno Hospital, Cardiovascular Medicine Center) and Professor Akira Yamashina (Tokyo Medical University, Cardiology) for providing critical review of the manuscript.
**Disclosure**
The authors have no conflicts of interest to disclose.
{#f1-mder-8-467}
{#f2-mder-8-467}
{#f3-mder-8-467}
{#f4-mder-8-467}
{#f5-mder-8-467}
######
Patient characteristics for pressure measurements (N=22)
Mean ± SD or Number (%)
--------------------------- -------------------------
Age, years 72.2±6.8
Sex, male 19 (86)
Left lower extremity 14 (64)
Hypertension 19 (86)
Dyslipidemia 13 (59)
Diabetes mellitus 14 (64)
Current smoking 6 (27)
Previously smoked 12 (55)
Coronary artery disease 13 (59)
Ejection fraction (%) 59±13
Cerebrovascular disease 6 (27)
Hemodialysis 3 (14)
Rutherford classification
0 1 (5)
1 1 (5)
2 6 (27)
3 11 (50)
4 3 (14)
5 1 (5)
6 0
**Abbreviation:** SD, standard deviation.
######
Lesion characteristics (N=23)
Mean ± SD or Number (%)
-------------------------------- -------------------------
Minimum lesion length (mm) 2.1±1.3
Reference vessel diameter (mm) 5.2±2.9
Diameter stenosis (%) 57.3±13.2
Lesion length (mm) 32.0±19.0
Lesion location
Iliac 8 (35)
Femoropopliteal 15 (65)
Restenosis 8 (35)
Below the knee runoff
0/3 0
1/3 6 (26)
2/3 7 (10)
3/3 10 (43)
TASC II classification
A 21 (91)
B 2 (9)
**Abbreviations:** SD, standard deviation; TASC, TransAtlantic Inter-Society Consensus.
| {
"pile_set_name": "PubMed Central"
} |
Plants from the genus *Oenanthe* (Umbelliferae family) are among the most poisonous species of the European flora.([@ref1]) They bear a general resemblance to parsnip and carrot and are still a cause of often fatal human poisonings.([@ref1]) Sardinian *Oenanthe* species have a special ethnopharmacological relevance, being considered the most likely candidate for the (in)famous sardonic herb, a neurotoxic plant used in pre-Roman Sardinia for the ritual killing of elderly people.([@ref2]) According to the ancient historians, elderly people unable to support themselves were intoxicated with the sardonic herb and then killed by dropping from a high rock or by beating to death.([@ref2]) The facial muscular contraction induced by the sardonic herb mimicked a smile, and the expression *risus sardonicus* (sardonic smile) to indicate a sinister smile is well documented in the Latin and Greek literature and in most modern European languages. It even found its way into the mainstream medical lingo as the hallmark of lockjaw (*trismus*), the spasm of the muscles of mastication.([@ref3])
The poisonous constituents of *Oenanthe* species are a series of polyacetylenic alcohols exemplified by oenanthotoxin (**1a**).([@ref4]) Surprisingly, the absolute configuration at the stereogenic carbon (C-14) of oenanthotoxin is still unknown, and little information on the molecular details of its neurotoxic activity has been reported. As part of a study on poisonous Sardinian plants,([@ref5]) we have investigated *O. fistulosa* L. (tubular water-dropwort), a species that, despite its very broad distribution in Europe, had so far been overlooked in terms of phytochemical studies. Large amounts of the diacetylene epoxydiol **2** were obtained from the seeds, while the roots gave, along with the widespread acetylene falcarindiol (**3**), oenanthotoxin (**1a**) and dihydroenanthotoxin (**1b**). The absolute configuration of **1a** and **1b** was clarified by application of the modified Mosher method, while the activity of oenanthotoxin was profiled against GABA receptors, a target involved in the action of many convulsivant toxins of plant origin.([@ref6])
An acetone extract from the underground parts of *O. fistulosa* was fractionated by gravity column chromatography to afford large amounts (ca. 0.80% of dried plant material) of (+)-(3*R*, 8*S*)-falcarindiol (**3**)([@ref7]) and a mixture of oenanthotoxin (**1a**) and its dihydroderivative (**1b**) (0.070%).([@ref4]) Compounds **1a** and **1b** were easily separated by preparative HPLC and identified by comparison with authentic samples of (+)-oenanthotoxin and (+)-dihydrooenanthotoxin obtained, in over 10-fold higher yield, from the roots of *O. crocata*, a more common species in Sardinia (see Experimental Section). A complete assignment of the ^1^H NMR resonances of both **1a** and **1b** is reported in the Experimental Section.
Oenanthotoxin (**1a**) is isomeric with cicutoxin (**4**) from water hemlock (*Cicuta virosa* L.), with the only difference between these C-17 acetylenic toxins being the position of a single double bond, located next to the secondary oxymethine in cicutoxin and adjacent to the primary hydroxyl in oenanthotoxin.([@cit4b]) The absolute configuration at C-14 of cicutoxin was established as *R* by application of a CD exciton chirality method on its *p*-methoxybenzoate, taking advantage of the allylic nature of the stereogenic center, and was confirmed by application of the Mosher method.([@ref8]) Since the stereogenic center of oenanthotoxin is not allylic, the Mosher method([@ref9]) was used to assign its absolute configuration.
To achieve this aim, two aliquots of oenanthotoxin (**1a**) were dissolved in dry pyridine and allowed to react overnight with DMAP and (*R*)- or (*S*)-MTPA chloride, respectively, affording the (*S*)- and (*R*)-MTPA diesters **5a** and **5b**, respectively. Analysis of the Δδ(*S−R*) values of the protons neighboring the oxygenated methine according to the Mosher model([@ref9]) (Figure [1](#fig1){ref-type="fig"}) allowed the assignment of the *R* configuration at C-14 of **1a**. The same methodology was applied to dihydrooenanthotoxin (**1b**), which, by analysis of the Δδ(*S−R*) values between its (*S*)- and (*R*)-MTPA diesters **6a** and **6b** (Figure [1](#fig1){ref-type="fig"}), turned out to have the same *R* configuration at C-14. In both cases, concomitant MTPA esterification at C-1 did not exert any influence on the anisotropic effect of the phenyl group of the MTPA moiety bound to the C-14, located 13 carbons away from the primary hydroxyl.
{#fig1}
The seeds of *O. fistulosa* were totally devoid of oenanthotoxin-like polyacetylenes and contained rather large amounts of the diacetylenic epoxydiol **2**. The structure elucidation of this compound (C~17~H~24~O~3~, HRMS) was helped greatly by comparison with falcarindiol (**3**),([@ref7]) since the two compounds showed similar spectroscopic features.([@ref7]) Thus, **2** differed from **3** only in the epoxidation of the disubstituted double bond, as evident from the replacement of the two mutually coupled olefin protons of **3** with two oxymethine signals \[δ 3.06 (dd, *J* = 7.1, 4.0 Hz, H-9) and 2.80 (dt, *J* = 5.9, 4.0 Hz, H-10\]. Corresponding changes were observed in the ^13^ C NMR spectrum (upfield shift of the C-9 and C-10 methines, now resonating at δ 57.9 and 57.4, respectively). Full assignment of ^1^H and ^13^C NMR resonances of **2** was achieved through analysis of 2D NMR (COSY, HSQC, and gHMBC) data, which confirmed that **2** is an epoxidized derivative of falcarindiol. To further confirm this assignment and establish the configuration of the epoxide ring, the reaction of (+)-(3*R*,8*S*)-falcarindiol (**3**)([@ref7]) with peracids was investigated. Enines can be chemoselectively epoxidized, and treatment of falcarindiol with *meta*-chloroperoxybenzoic acid (MCPBA) under controlled conditions (monitoring of the conversion by ^1^H NMR analysis, see Experimental Section) gave the epoxide **2** in satisfactory (66%) yield and as the only reaction product. This was identical in terms of both spectroscopic properties and optical rotation to the natural product isolated from *O. fistulosa*. Prolonged reaction times or the addition of NaHCO~3~ to buffer the acidity of the medium led to extensive degradation of the reaction product. The epoxidation of allylic alcohols is a *syn* addition, and the peroxy acid is directed by hydrogen bonding to the allylic hydroxyl, which, in the case of secondary alcohols, reacts in the conformation that minimizes allylic strain (A^(1,3)^-strain) between the alkyl group bound to the oxymethine and the distal substituent of the adjacent double bond (C-7 and C-11, respectively, in the case of the epoxidation of **3**).([@ref10]) Taking into account that the stereogenic double bond of falcarindiol has a *cis* configuration, these considerations lead to a 9*S*,10*R* configuration for the epoxide ring of **2**.([@ref11]) A compound with the planar structure of **2** and spectroscopic features compatible with those of **2** was described from *Panax quinquefolium*.([@ref12]) However, the optical rotation was different \[+ 107 (MeOH) for the epoxydiol from *O. fistulosa* and +87 (MeOH) for that from *P. quinquefolium*\], and in the absence of an authentic sample, the identity of the two compounds could not be established.
Polyacetylene neurotoxins are believed to act like biological analogues of the sesquiterpene lactone picrotoxin, binding to the convulsant site of GABA~A~ receptors and causing death by inducing convulsions and respiratory paralysis.([@ref13]) In addition, oenanthotoxin has also been reported to dramatically affect cationic (calcium and sodium) currents in excitable membranes.([@ref14]) We found that both oenanthotoxin and dihydrooenanthotoxin could potently block the GABAergic responses in neuronal cell cultures, with an EC~50~ value in the low micromolar range for both compounds. Currents elicited by exogenous GABA were recorded in the whole-cell configuration by rapid application of 3 μM GABA at the membrane voltage of −40 mV, and a typical GABA-evoked response, recorded in these conditions, is presented in Figure [2](#fig2){ref-type="fig"} (left). The effect of **1a** and **1b** on current amplitude was assessed by dividing the current amplitude measured in the presence of this drug by the amplitude value in control conditions measured from the same cell. Oenanthotoxin (**1a**) clearly inhibited the GABAergic responses in a dose-dependent manner (Figure [2](#fig2){ref-type="fig"}A). A tentative fitting of the logistic equation (*y* = 1/(1 + (*c*/Ec~50~)*h*)) to the amplitude dose-dependence for oenanthotoxin shown in Figure [2](#fig2){ref-type="fig"}B yielded a EC~50~ value of 1.39 μM (*h* = 0.59). Dihydrooenanthotoxin (**1b**) exhibited a similar inhibitory activity on currents elicited by 3 μM GABA, but was slightly more potent (EC~50~ = 0.835 μM, *h* = 0.58). The effect of both test compounds was reversible, although at least 10 min of wash was required to restore the control responses. Compound **1b** can be viewed both as 2,3-dihydrooenanthotoxin and as 12,13-dihydrocicutoxin (cf. **1a** and **4**), showing that, in polyacetylenic toxins, the presence of allylic double bonds is redundant for activity and that the conjugated system required for the activity is shorter than previously postulated,([@ref13]) encompassing four, and not five, unsaturations.
{#fig2}
Unlike other plant toxins, the convulsant polyacetylenes from water dropwort and related plant do not evoke unpleasant taste (bitter) or chemesthetic (burning) sensations, and the roots of *O. crocata*, an exceedingly poisonous plant, have a paradoxical sweetish and pleasant taste and odor.([@ref15]) The large concentration of falcarindiol, a bitter compound,([@ref16]) and the lower contents of polyacetylene toxins in *O. fistulosa* when compared to *O. crocata* could presumably underlie the observation that the former species has not yet been associated with human or animal poisoning. The name *Oenanthe* signifies "wine flower", because the plant produces a state of stupefaction similar to drunkenness.^[@ref1],[@ref2]^ This, as well as locked jaws (*risus sardonicus*), has been documented in human poisoning from *O. crocata*,^[@ref1],[@ref2]^ and there is little doubt that *herba sardonica* of the ancient medical literature should be identified with *O. crocata*,([@ref2]) a plant that, within the Mediterranean area, is common only in Sardinia.([@ref2]) The results of our investigation provide a further rationale for this identification, proposing a molecular mechanism for the *risus sardonicus* described by the ancient authors.
Neurotoxins have emerged as powerful neurochemical tools,^[@ref17],[@ref18]^ and the scarce attention given so far to the C-17 neurotoxic acetylenes can best be explained by their chemical instability, especially in the highly purified form required for biological studies. Thus, oenanthotoxin (**1a**) and dihydrooenanthotoxin (**1b**), just like many other polyacetylenes, easily turned into highly colored and insoluble polymeric materials upon storage, even at low temperature (−18 °C). To overcome this problem, we have developed a flash chromatography protocol to separate oenanthotoxin (**1a**) and dihydroenanthotoxin (**1b**) from a crude extract of *O. crocata* and found that these compounds, while being highly unstable in the solid state, can be conveniently stored for at least 6 months in benzene or DMSO solution at 4 °C (see Experimental Section). This might foster interest in this fascinating class of compounds, with its unique capacity of dramatically perturbing calcium and sodium ionic currents in excitable membranes([@ref14]) and specifically interacting with ligand-gated ion channels like the GABA~A~ receptor.
Experimental Section {#sec2}
====================
General Experimental Procedures {#sec2.1}
-------------------------------
Optical rotations (CHCl~3~) were measured at 589 nm on a Perkin-Elmer 192 polarimeter equipped with a sodium lamp (λ = 589 nm) and a 10 cm microcell. ^1^H (500 MHz) and ^13^C (125 MHz) NMR spectra were measured on a Varian INOVA spectrometer. Chemical shifts were referenced to the residual solvent signal (CDCl~3~: δ~H~ 7.26, δ~C~ 77.0). Homonuclear ^1^H connectivities were determined by the COSY experiment. One-bond heteronuclear ^1^H−^13^C connectivities were determined with the HSQC experiment. Two- and three-bond ^1^H−^13^C connectivities were determined by HMBC experiments optimized for a ^2,3^*J* of 7 Hz. Through-space ^1^H connectivities were evidenced using a ROESY experiment with a mixing time of 500 ms. Two- and three-bond ^1^H−^13^C connectivities were determined by gradient 2D HMBC experiments optimized for a ^2,3^*J* of 9 Hz. Low- and high-resolution EIMS spectra (70 eV) were performed on a VG Prospec (Fisons) mass spectrometer. ESIMS spectra were performed on a LCQ Finnigan MAT mass spectrometer. Silica gel 60 (70−230 mesh) was used for gravity column chromatography. Reactions were monitored by TLC on Merck 60 F~254~ (0.25 mm) plates, which were visualized by UV inspection and/or staining with 5% H~2~SO~4~ in ethanol and heating. Organic phases were dried with Na~2~SO~4~ before evaporation.
Plant Material {#sec2.2}
--------------
*Oenanthe fistulosa* was collected near San Basilio (CA, Sardinia, Italy) on September 14, 2007. The plant was identified by M.B., and a voucher specimen (171007) is held at the University of Cagliari. *O. crocata* was collected near Bonacardo (Abbasanta, NU, Italy) in May 8, 2007, and a voucher specimen (OCR080506BO) is held at the University of Cagliari.
Extraction and Isolation {#sec2.3}
------------------------
\(A\) Powdered, dried underground parts of *O. fistulosa* (608 g) were extracted with acetone (2 × 2 L) at room temperature, affording 7.94 g of a dark oil, which was fractionated by gravity column chromatography on silica gel (*n*-hexane−EtOAc, 9:1 to 6:4, as solvent gradient system). Fractions eluted with *n*-hexane−EtOAc (9:1) were further purified by HPLC (*n*-hexane−EtOAc, 85:15) to yield falcarindiol (**3**, 344 mg). Fractions eluted with *n*-hexane−EtOAc (6:4) were further purified by HPLC (*n*-hexane−EtOAc, 65:35) to yield oenathotoxin (**1a**, 68 mg) and dihydrooenanthotoxin (**1b**, 67 mg). (B) Powdered, dried seeds of *O. fistulosa* (379 g) were extracted with acetone (2 × 2 L) at rt, affording 21.3 g of a yellowish oil, which was fractionated by gravity column chromatography on silica gel (*n*-hexane−EtOAc, 9:1 to 6:4, as solvent gradient system). Fractions eluted with *n*-hexane−EtOAc (7:3) were further purified by HPLC (*n*-hexane−EtOAc, 55:45) to yield epoxyfalcarindiol (**2**, 432 mg). (C) Dried underground parts of *O. crocata* (95 g) were extracted with acetone (2 × 1 L) to afford 3.1 g of a brownish oil, which was purified with the Biotage SP1 system with a silica gel (10 g) FLASH 12+M column and a *n*-hexane−EtOAc gradient of increasing polarities (flow rate 9 mL/min, UV detector 316 nm; fractionation mode: volume 9 mL/fraction). Fractions 29−31 and 34−36 (*n*-hexane−EtOAc, 6:4) afforded about 88 mg of oenanthotoxin (**1a**) and 89 mg of dihydrooenathotoxin (**1b**), respectively. These compounds were obtained as colorless foams and could not be stored in this form. However, solutions in hydrocarbon solvents (benzene, toluene) or DMSO could be stored for at least 6 months at 4 °C without any apparent (^1^H NMR analysis) degradation or development of a reddish color.
Oenanthotoxin (**1a**): {#sec2.4}
-----------------------
amorphous foam; \[α\]~D~^25^ +34 (*c* 0.5, CHCl~3~); ^1^H NMR (500 MHz, CDCl~3~) δ 6.71 (1H, dd, *J* = 15.5, 11.0 Hz, H-9), 6.39 (1H, dt, *J* = 15.0, 5.0 Hz, H-2); 6.12 (1H, dd, *J* = 15.0, 11.0 Hz, H-10), 5.87 (1H, overlapped H-3), 5.86 (1H, overlapped, H-11); 5.57 (1H, d, *J* = 15.5 Hz, H-8), 4.25 (2H, d, *J* = 5.0 Hz H~2~-1), 3.60 (1H, m, H-14), 2.30 (1H, m, H-12a), 2.20 (1H, m, H-12b), 1.78 (1H, m, H-13a), 1.50 (1H, overlapped, H-13b), 1.49 (1H, overlapped, H-15a), 1.40 (1H, m, H-15b), 1.30 (2H, m, H~2~-16), 0.92 (3H, d, *J* = 7.3, Hz H~3~-17); ESIMS (positive ion) *m*/*z* 281 \[M + Na\]^+^.
Dihydrooenanthotoxin (**1b**): {#sec2.5}
------------------------------
amorphous foam; ^1^H NMR (500 MHz, CDCl~3~) δ 6.65 (1H, dd, *J* = 15.5, 10.6 Hz, H-9), 6.12 (1H, dd, *J* = 15.0, 11.0 Hz, H-10), 5.84 (1H, dt, *J* = 15.0, 5.0 Hz, H-11), 5.50 (1H, d, *J* = 15.5 Hz, H-8), 3.77 (2H, t, *J* = 5.0 Hz, H~2~-1), 3.60 (1H, m, H-14), 2.45 (2H, t, *J* = 5.0 Hz, H~2~-3), 2.30 (1H, m, H-12a), 2.21 (1H, m, H-12b), 1.79 (2H, t, *J* = 5.0 Hz, H~2~-2), 1.75 (1H, m, H-13a), 1.50 (1H, overlapped, H-13b), 1.49 (1H, overlapped, H-15a), 1.40 (1H, m, H-13b), 1.30 (2H, m, H~2~-16), 0.93 (3H, d, *J* = 7.3 Hz, H~3~-17); ESIMS (positive ion): *m*/*z* 283 \[M + Na\]^+^.
9-Epoxyfalcarindiol (**2**): {#sec2.6}
----------------------------
colorless oil, \[α\]~D~^25^ +107 (*c* 0.9, MeOH); IR (liquid film) ν~max~ 3400, 3120, 2080, 1451, 1293, 1238, 1164, 983 cm^−1^; ^1^H NMR (500 MHz, C~6~D~6~) δ 5.78 (1H, ddd, *J* = 17.0, 10.2, 5.45 Hz, H-2), 5.37 (1H, brd, *J* = 17.0 Hz, H-1a), 5.03 (1H, brd, *J* = 10.3 Hz, H-1b), 4.67 (1H, d, *J* = 5.4 Hz, H-3), 4.20 (1H, d, *J* = 7.1 Hz, H-8), 3.06 (1H, dd, *J* = 7.1, 4.0 Hz, H-9), 2.80 (1H, dt, *J* = 5.9, 4.0 Hz, H-10), 1.53 (2H, brm, H-11a,b), 1.45−1.30 (10H, m, H-12a,b + H-13a,b + H-14a,b + H-15a,b + H-16a,b), 1.01 (3H, t, *J* = 7.1 Hz, H-17); ^13^C NMR (125 MHz, CDCl~3~) δ 136.1 (d, C-2), 116.2 (t, C-1), 79.3 (s, C-7), 79.1 (s, C-4), 70.3 (s, C-5), 70.1 (s, C-6), 63.2 (d, C-3), 61.0 (d, C-8), 57.9 (d, C-9), 57.4 (d, C-10), 31.9 (t, C-15), 27.6 (t, C-11), 29.4, 29.3 (t, C-13 and C-14) 26.6 (t, C-12), 22.8 (t, C-16), 14.1 (q, C-17). HREIMS *m*/*z* 276.1737 \[M\]^+^ (calcd for C~17~H~24~O~3~, 276.1725).
MTPA Esters of Oenanthotoxin and Dihydrooenanthotoxin {#sec2.7}
-----------------------------------------------------
Oenanthotoxin (**1a**, 2.5 mg) was dissolved in 0.5 mL of dry pyridine and treated with (−)-*R*-MTPA chloride (20 μL) and *N*,*N*-dimethylaminopyridine (DMAP, a spatula tip), then maintained at room temperature under stirring overnight. After removal of the solvent, the reaction mixture was purified by HPLC on a SI60 column (eluent *n*-hexane−EtOAc, 9:1), affording the (*S*)-MTPA diester **1c** in a pure state (3.5 mg, 53% yield). Using (+)-*S*-MTPA chloride, the same procedure afforded the (*R*)-MTPA diester **1d** in the same yield. This procedure was repeated for dihydrooenanthotoxin (**1b**, 2.0 mg) to obtain the (*S*)-MTPA diester **1e** and (*R*)-MTPA diester **1f** in the same yield.
Oenanthotoxin-1-*O*-14-*O*-(S)-MTPA diester (**5a**): {#sec2.8}
-----------------------------------------------------
amorphous solid; ^1^H NMR (500 MHz, CDCl~3~) δ 7.45 and 7.35 (MTPA phenyl protons), 6.66 (H-9, dd, *J* = 15.5, 10.6 Hz), 6.28 (H-2, dt, *J* = 15.0, 5.0 Hz), 6.09 (H-10, dd, *J* = 15.0, 10.6 Hz), 5.84 (H-3, overlapped), 5.82 (H-11, overlapped), 5.57 (H-8, d, *J* = 15.5 Hz), 5.10 (H-14, m), 4.86 (H~2~-1, d, *J* = 5.0 Hz), 3.55 (MTPA OCH~3~, s), 2.16 (H-12a, overlapped), 2.12 (H-12b, overlapped), 1.81 (H-13a, m), 1.73 (H-15a, m), 1.65 (H-13b, m), 1.58 (H-15b, m), 1.29 (H~2~-16, m), 0.88 (H~3~-17, d, *J* = 7.3 Hz); FABMS (glycerol matrix, positive ions) *m*/*z* 691 \[M + H\]^+^.
Oenanthotoxin-1-*O*-14-*O*-(R)-MTPA diester (**5b**): {#sec2.9}
-----------------------------------------------------
amorphous solid; ^1^H NMR (500 MHz, CDCl~3~) δ 7.55 and 7.53 (MTPA phenyl protons), 6.69 (H-9, dd, *J* = 15.5, 10.6 Hz), 6.26 (H-2, dt, *J* = 15.0, 5.0 Hz), 6.16 (H-10, dd, *J* = 15.0, 10.6 Hz), 5.91 (H-11, overlapped), 5.86 (H-3, overlapped), 5.57 (H-8, d, *J* = 15.5 Hz), 5.10 (H-14, m), 4.86 (H~2~-1, d, *J* = 5.0 Hz), 3.58 (MTPA OCH~3~, s), 2.35 (H-12a, m), 2.23 (H-12b, m), 1.98 (H-13a, m), 1.68 (H-13b, m), 1.61 (H-15a, m), 1.57 (H-15b, m), 1.28 (H~2~-16, m), 0.88 (H~3~-17, d, *J* = 7.3 Hz); FABMS (glycerol matrix, positive ions) *m*/*z* 691 \[M + H\]^+^.
Dihydrooenanthotoxin-1-*O*-14-*O*-(S)-MTPA diester (**6a**): {#sec2.10}
------------------------------------------------------------
amorphous solid; ^1^H NMR (500 MHz, CDCl~3~) δ 7.45 and 7.35 (MTPA phenyl protons), 6.64 (H-9, dd, *J* = 15.5, 10.6 Hz), 6.09 (H-10, dd, *J* = 15.0, 10.6 Hz), 5.83 (H-11, dt, *J* = 15.0, 5.0 Hz), 5.45 (H-8, d, *J* = 15.5 Hz), 5.20 (H-14, m), 4.39 (H~2~-1, m), 3.55 (MTPA OCH~3~, s), 2.36 (H~2~-3, t, *J* = 6.5 Hz), 2.25 (H-12a, overlapped), 2.21 (H-12b, overlapped), 1.91 (H~2~-2, m), 1.75 (H-15a, m), 1.71 (H-13a, m), 1.60 (H-15b, m), 1.56 (H-13b, m), 1.30 (H~2~-16, m), 0.90 (H~3~-17, d, *J* = 7.3 Hz); FABMS (glycerol matrix, positive ions) *m*/*z* 693 \[M + H\]^+^.
Dihydrooenanthotoxin-1-*O*-14-*O*-(R)-MTPA diester (**6b**): {#sec2.11}
------------------------------------------------------------
amorphous solid; ^1^H NMR (500 MHz, CDCl~3~) δ 7.55 and 7.53 (MTPA phenyl protons), 6.70 (H-9, dd, *J* = 15.5, 10.6 Hz), 6.11 (H-10, dd, *J* = 15.0, 10.6 Hz), 5.85 (H-11, dt, *J* = 15.0, 5.0 Hz), 5.48 (H-8, d, *J* = 15.5 Hz), 5.20 (H-14, m), 4.39 (H~2~-1, m), 3.58 (MTPA OCH~3~, s), 2.36 (H~2~-3, overlapped), 2.35 (H-12a, overlapped), 2.28 (H-12b, overlapped), 1.91 (H~2~-2, m), 1.90 (H-13a, m), 1.68 (H-15a, overlapped), 1.67 (H-13b, overlapped), 1.54 (H-15b, m), 1.27 (H~2~-16, m), 0.90 (H~3~-17, d, *J* = 7.3 Hz); FABMS (glycerol matrix, positive ions) *m*/*z* 693 \[M + H\]^+^.
Epoxidation of Falcarindiol {#sec2.12}
---------------------------
To a solution of falcarindiol (85 mg, 0.36 mmol) in dry CH~2~Cl~2~ (3 mL) was added 85% *m*-chloroperbenzoic acid (MCPBA, 73 mg, 0.36 mmol, 1 molar equiv). The course of the reaction was followed by ^1^H NMR analysis of a CDCl~3~ solution containing an equimolecular mixture of falcarindiol and MCPBA. After stirring 2 h at rt, the reaction was worked up by dilution with CH~2~Cl~2~ and the addition of silica gel (1 g) impregnated with a 5% water solution of Na~2~S~2~O~3~. After stirring a few minutes, the slurry was filtered and the filtrate was dried (Na~2~SO~4~) and evaporated, to afford 60 mg (66%) of 9-epoxyfalcarindiol (**2**), spectroscopically (^1^H and ^13^C NMR) identical to the natural product and having \[α\]~D~^25^ +101 (*c* 1.1, MeOH).
Cell Culture for Electrophysiological Recordings {#sec2.13}
------------------------------------------------
Neuronal cell culture was prepared as previously described in detail by Andjus et al.([@ref19]) Briefly, P1−P3-day-old Wistar rats were decapitated. This procedure is in accordance with the regulation of the Polish Animal Welfare Act. Hippocampi were removed, manually sliced, treated with trypsin, mechanically dissociated, centrifuged twice at 40 g, plated in Petri dishes, and cultured. Experiments were performed on cells between 10 and 17 days in culture.
Electrophysiological Recordings {#sec2.14}
-------------------------------
Currents were recorded in the whole-cell configuration of the patch-clamp technique using the Axopatch 200B amplifier (Molecular Instruments, Sunnyvale, CA) at a holding potential (*V*~h~) of −40 mV. The intrapipet solution contained (in mM) CsCl 137, CaCl~2~ 1, MgCl~2~ 2, 1,2-bis(2-aminophenoxy)ethane-*N*,*N*,*N′*-tetraacetic acid (BAPTA) 11, ATP 2, and HEPES 10 (pH 7.2 with CsOH). The composition of the standard external solution was (in mM) NaCl 137, KCl 5, CaCl~2~ 2, MgCl~2~ 1, glucose 20, and HEPES 10 (pH 7.2 with NaOH). For the whole-cell recordings, patch pipets had a resistance of 2.5−4.0 MΩ when filled with internal solution. The whole-cell recordings were included in the statistics when the access resistance was below 10 MΩ. The cells in which series resistance showed a clear tendency to increase during recordings were not considered in the analysis. Recordings in the whole-cell mode were started at least 3 min after establishing the whole-cell mode. This time was sufficient to stabilize the recording conditions. The agonist was applied using the RSC-200 multibarrel rapid perfusion system (Bio-Logic, Grenoble, France). With this system, in the whole-cell configuration, the solution exchange occurred within 30−100 ms. Before each control recording, cells were washed with normal external solution for at least 3 min. In studies aiming at assessment of the impact of considered test compounds \[oneantotoxin (**1a**), dihydrooneantotoxin (**1b**)\] on GABA-elicited responses, these compounds were present both in the washing solution (for at least 3 min before agonist application) and in the agonist-containing saline. For acquisition and data analysis, pClamp 9.2 software was used (Molecular Devices Corporation). To avoid the data scatter due to cell-to-cell variability, the effect of studied drugs was assessed by calculating the relative amplitude values with respect to the controls recorded from the same cell. For the analysis of currents, recorded in the whole-cell configuration, the current signals were low-pass filtered at 3 kHz with a Butterworth filter and sampled at 10 kHz using the analog-to-digital converter Digidata 1322A (Molecular Device Corporation) and stored on a computer hard disk. Data are expressed as mean ± SEM. A paired Student *t* test was used to assess the significance of differences between considered data sets. All experiments were performed at room temperature, 22−24 °C.
P.W., K.S., and J.W.M. were supported by a Wellcome Trust International Senior Research Fellowship in Biomedical Science (grant no. 070231/Z/03/Z).
[^1]: Università del Piemonte Orientale.
[^2]: Consorzio per lo Studio dei Metaboliti Secondari.
[^3]: Università di Milano.
[^4]: Università di Napoli Federico II.
[^5]: Wroclaw Medical University.
| {
"pile_set_name": "PubMed Central"
} |
Hintergrund {#Sec1}
===========
Alexander Langmuir definierte 1963 die „Surveillance von Krankheiten" als „eine kontinuierliche Beobachtung von Verteilung, Trends und Inzidenz durch systematische Sammlung, Konsolidierung und Evaluation von Berichten über Morbidität, Mortalität und anderer relevanter Daten". Diese Definition schloss eine Weiterleitung der Daten an alle ein, die diese Informationen benötigen \[[@CR1]\]. Fünf Jahre später wurde auf der 21. Weltgesundheitsversammlung bezüglich der nationalen und globalen Surveillance übertragbarer Krankheiten zum Aspekt der Datenverbreitung hinzugefügt, dass die Daten zu Einrichtungen gelangen sollten, die Public-Health-Maßnahmen initiieren können \[[@CR2]\]. Eine erweiterte Definition der Centers for Disease Control and Prevention (CDC) von 1986 beinhaltet die Verbindung der Surveillance-Kette zur Nutzung der Daten für Prävention und Kontrolle. Hierzu gehört die Kapazität, Daten zu sammeln, zu analysieren sowie die Weitergabe von Daten an Public-Health-Programme \[[@CR3]\].
Die aktive Form der epidemiologischen Surveillance beinhaltet, dass die Daten von der Institution, die die Surveillance durchführt, zum Beispiel per E-Mail oder Telefon angefordert werden. Bei der passiven Form eines epidemiologischen Surveillance-Systems werden Fälle von im Gesundheitswesen tätigen oder anderen autorisierten Personen gemeldet, ohne dass aktiv nachgefragt wurde. Die das Surveillance-System koordinierende Instanz erwartet Meldungen, um diese dann auszuwerten. Ein Beispiel hierfür ist der Meldeweg meldepflichtiger Infektionskrankheiten gemäß Infektionsschutzgesetz (IfSG); der Arzt meldet eine Infektionskrankheit (§ 6 IfSG), das Labor einen Erregernachweis an das zuständige Gesundheitsamt (§ 7 IfSG). Dort werden die Informationen des Patienten zusammengeführt. Sind bestimmte Kriterien erfüllt, übermittelt das Gesundheitsamt die Meldungen an die zuständige obere Landesgesundheitsbehörde; in der Folge werden die Daten an das RKI weitergeleitet (§ 7 Abs. 3 IfSG). Ein weiteres Beispiel ist die Meldepflicht im Rahmen der Internationalen Gesundheitsvorschriften (IGV). Die IGV bilden das völkerrechtliche Fundament zur internationalen Bekämpfung von Infektionskrankheiten; neben verbindlichen Verpflichtungen enthalten sie eine Reihe maßnahmenorientierter Empfehlungen der Weltgesundheitsorganisation (WHO). Dazu gehören Empfehlungen, die nach Feststellung einer gesundheitlichen Notlage von internationaler Tragweite (GNIT) oder bei Vorliegen einer bestimmten Gefahr von der WHO ausgesprochen werden können. Liegt ein solches Ereignis vor, so ist dieses immer gemeinsam mit den bereits durchgeführten Maßnahmen innerhalb von 24 Stunden nach der Bewertung an die WHO zu melden (Artikel 6 IGV) \[[@CR4]\]. Im Rahmen der IGV gibt es auch eine aktive Komponente; sie besteht aus einer vorgeschriebenen Rückmeldung des Vertragsstaates auf WHO-Anfragen für die Bestätigung (oder Widerlegung) von Ereignissen, die der WHO weder durch Meldung noch durch Konsultation bekannt sind (Artikel 9.1 und 10 IGV) \[[@CR5]\].
Indikatorbasierte Surveillance-Systeme beruhen auf einer strukturierten Datensammlung durch Routine-Surveillance, zum Beispiel Meldesysteme. Die ereignisbasierte (event-based) Surveillance hingegen bezieht unstrukturierte Informationen aus unterschiedlichen Quellen \[[@CR6], [@CR7], [@CR8]\]. Zum Beispiel können webbasierte Informationsquellen zu einem frühen Erkennen von Ausbrüchen und einer Sensibilisierung der Öffentlichkeit beitragen \[[@CR9]\]. Im November 2002 erschien in einer chinesischen Zeitung ein Artikel, der ungewöhnlich hohe Fallzahlen an Patienten mit akuten respiratorischen Erkrankungen in Rettungsstellen von Krankenhäusern in der Provinz Guangdong beschrieb. Nach diesem frühen Hinweis auf SARS (Severe Acute Respiratory Syndrome) berichteten chinesische Medien in der Folge von respiratorischen Erkrankungen bei Krankenhauspersonal. Diese Ereignisse wurden vom Global Public Health Intelligence Network (GPHIN), einem von der Public Health Agency of Canada betriebenen webbasierten Frühwarnsystem, registriert \[[@CR10], [@CR11], [@CR12]\]. Parallel wurde online im ProMED-Mail-System über den Ausbruch diskutiert, bevor die chinesische Regierung offiziell darüber berichtete \[[@CR13]\]. Die WHO erhielt den ersten offiziellen Bericht aus China am 11. Februar 2003 \[[@CR14]\].
In Deutschland existiert, ebenso wie in vielen anderen EU-Ländern, eine etablierte indikatorbasierte Surveillance. Um adäquat auf neue Bedrohungen reagieren zu können, kann sich der Öffentliche Gesundheitsdienst (ÖGD) nicht mehr ausschließlich auf traditionelle Surveillance-Systeme verlassen \[[@CR15], [@CR16]\]. Am 13. Januar 2009 wurde von der Abteilung für Infektionsepidemiologie am Robert Koch-Institut (RKI) eine wöchentliche telefonische epidemiologische Bund-Länder-Lagekonferenz (EpiLag) eingeführt, an der überwiegend Vertreter der oberen Landesgesundheitsbehörden und des RKI, darüber hinaus gelegentlich Referenten für Infektionsschutz der obersten Landesgesundheitsbehörden und Vertreter der Bundeswehr teilnehmen. Ziel war es, für den ÖGD auf Länder- und Bundesebene eine Plattform zur Diskussion aktueller infektionsepidemiologischer Ereignisse zu schaffen und unabhängig von der Meldepflicht den Austausch über relevante Ereignisse zu verbessern. Hierzu zählen neben Ereignissen mit potenziell überregionaler infektionsepidemiologischer Relevanz auch Ereignisse, die ungewöhnlich sind oder bei denen Diskussionsbedarf besteht. Diese Herangehensweise entspricht einer ereignisbasierten Surveillance-Komponente. Die Kombination der indikatorbasierten Komponente aus der klassischen Surveillance mit einer ereignisbasierten Surveillance entspricht dem Konzept der „Epidemic Intelligence" (EI). Beiden Surveillance-Komponenten soll Beachtung geschenkt werden, da Signale von beiden zu Public-Health-Maßnahmen führen können \[[@CR2]\].
Ziel des vorliegenden Beitrages ist es, die Organisation, die Durchführung und den Inhalt der wöchentlichen EpiLag zu beschreiben und ihren infektionsepidemiologischen Nutzen für den ÖGD und damit für die Bevölkerung darzustellen.
Die Telefonkonferenz {#Sec2}
====================
Inhalt {#Sec3}
------
Die EpiLag bietet den Teilnehmern einen strukturierten, wöchentlichen und damit regelmäßigen Austausch über aktuelle infektionsepidemiologische Ereignisse. Alle Teilnehmer haben die Möglichkeit, über infektionsepidemiologisch relevante oder ungewöhnliche Ereignisse auf nationaler oder internationaler Ebene zu berichten und sie zu diskutieren. Ereignisse können auch in mehreren Telefonkonferenzen diskutiert werden.
Organisation und Durchführung {#Sec4}
-----------------------------
Die Organisation und Durchführung der EpiLag erfolgt durch die Abteilung für Infektionsepidemiologie am RKI ([Abb. 1](#Fig1){ref-type="fig"}). Um für die Konferenz relevante und aktuelle Informationen zu erhalten, werden innerhalb des RKI vor jeder EpiLag die für bestimmte Infektionskrankheiten zuständigen Fachgebiete, die Pressestelle und die Redaktion des Epidemiologischen Bulletin sowie die infektionsepidemiologische Rufbereitschaft ([Abb. 1](#Fig1){ref-type="fig"}) zu wochenaktuellen, nationalen und internationalen infektionsepidemiologischen Ereignissen und Informationen befragt.
{#Fig1}
Weitere Informationen werden dem Weekly Threat Report des Europäischen Zentrums für die Prävention und Kontrolle von Krankheiten (ECDC) sowie der IGV Event Information Page der WHO entnommen. Weitere „Internet-based tools" (IBT), die als Quellen für die EpiLag dienen, sind das Early Warning and Response System (EWRS) und ProMED ([Tab. 1](#Tab1){ref-type="table"}).
+----------------------------------------------------------------------------------------------------------------------+------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+
| ProMED-mail (Program for Monitoring Emerging Diseases) | ProMED ist ein globales, Internet-basiertes elektronisches Berichtsystem zur Entdeckung, Überwachung und zum schnellen Informationsaustausch zu Ausbrüchen von Infektionskrankheiten und anderer akuter Gefährdungen durch Giftstoffe, die die Gesundheit von Menschen, Tieren und Nahrungspflanzen beeinflussen könnten. ProMED wurde 1994 von der „International Society of Infectious Disease" gegründet. Als Informationsquellen dienen unter anderem Medienberichte, offizielle Berichte, Online-Zusammenfassungen und lokale Beobachtungen. Der Zugang zu ProMED ist öffentlich |
| | |
| http://www.promedmail.org | |
+----------------------------------------------------------------------------------------------------------------------+------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+
| EWRS (Early Warning and Response System) | Das EWRS für Infektionskrankheiten ist ein internetbasiertes Netzwerk der Europäischen Kommission (EC), der nationalen Gesundheitsbehörden der Mitgliedsstaaten, die für die Durchführung von Maßnahmen zur Kontrolle von Infektionskrankheiten zuständig sind, sowie des Europäischen Zentrums für die Prävention und Kontrolle von Krankheiten (ECDC). EWRS wurde von der EC mit dem Ziel gegründet, der EU eine schnelle und effektive Reaktion auf Ereignisse, Notfallsituationen und Bedrohungen in Verbindung mit Infektionskrankheiten zu ermöglichen (17). Der Zugang zum EWRS gliedert sich in einen öffentlichen sowie einen geschlossenen Teil, der nur zugänglich ist für Public Health-Institutionen der EU-Mitgliedsländer, die offiziell von ihrer Regierung ihres jeweiligen Landes und der EC zu Mitgliedern des EWRS erklärt wurden |
| | |
| https://ewrs.ecdc.europa.eu/ | |
+----------------------------------------------------------------------------------------------------------------------+------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+
| CDTR (Communicable Disease Threats Report) | Der CDTR des ECDC dient als Werkzeug für Europäische Epidemiologen, die auf nationaler Ebene im Bereich Surveillance für die Epidemic Intelligence zuständig sind. Der wöchentlich erscheinende Bericht bezieht seine Informationen aus unterschiedlichen Quellen, wie zum Beispiel Medien (z. B. TV und „web-news"), offizielle Websites internationaler Organisationen, Bulletins, andere webbasierte Frühwarn- und Meldesysteme, z. B. ProMed, das Global Public Health Intelligence Network (GPHIN) und andere. Relevant sind Infektionskrankheiten mit potenzieller Bedrohung für die EU. Der Zugang zu CDTR-Berichten muss von den EU-Mitgliedsstaaten autorisiert werden. Auch Institute oder Einrichtungen wie zum Beispiel die WHO-Euro, CDC Atlanta, die Public Health Agency Canada und andere Public Health Partner, mit denen das ECDC ein MoU (Memorandum of Understanding) vereinbart hat, haben Zugang |
| | |
| http://www.ecdc.europa.eu/en/activities/epidemicintelligence/Pages/EpidemicIntelligence_ThreatCommunicationCDTR.aspx | |
+----------------------------------------------------------------------------------------------------------------------+------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+
| Meldungen gemäß IGV (Internationale Gesundheitsvorschriften) | Die IGV regeln die internationale Zusammenarbeit bei der Kontrolle von Infektionskrankheiten und sind auf alle Ereignisse anwendbar (natürlich oder beabsichtigt), die eine Gefahr für die öffentliche Gesundheit darstellen können (biologisch, chemisch oder durch ionisierende Strahlen). Im Rahmen der IGV existiert ein internetbasiertes Netzwerk zu Meldungen von Infektionskrankheiten an die WHO durch die nationalen „IGV-Anlaufstellen" von 193 Mitgliedsstaaten. In Deutschland fungiert das Lagezentrum des Bundesministeriums des Innern (BMI) als nationale IGV-Anlaufstelle. In allen Fällen, die Infektionskrankheiten betreffen, werden die Bearbeitung der Meldung sowie ggf. notwendige weitere Maßnahmen vom RKI koordiniert |
| | |
| http://www.who.int/ihr/en/ | |
| | |
| http://www.rki.de/cln_048/nn_205760/DE/Content/Infekt/IGV/igv\_\_node.html?\_\_nn | |
+----------------------------------------------------------------------------------------------------------------------+------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+
Die Vorbereitung auf die EpiLag und die Erstellung des Protokolls beanspruchen pro Woche zwischen fünf und neun Stunden. Die EpiLag selbst dauert circa 30 Minuten. Der gesamte zeitliche Aufwand für das RKI liegt also zwischen fünf und zehn Stunden. Für die wöchentliche Vor- und Nachbereitung der EpiLag ist je ein im Turnus wechselnder Mitarbeiter der Abteilung für Infektionsepidemiologie verantwortlich. Bei der Konferenz selbst sind vom RKI zumindest der für das Sammeln und Zusammenstellen der Informationen verantwortliche Mitarbeiter und ein leitender Moderator anwesend. Die Teilnehmer der EpiLag wählen sich in die Telefon-Konferenz ein.
Methode {#Sec5}
=======
Auswertung der Protokolle {#Sec6}
-------------------------
Die wöchentlichen EpiLag-Protokolle der ersten drei Quartale 2009 wurden hinsichtlich Beteiligung, Häufigkeit diskutierter infektionsepidemiologischer Ereignisse und ihrer Informationsquellen ausgewertet.
Evaluation der Konferenz {#Sec7}
------------------------
Um die Qualität der EpiLag zu evaluieren, führte das RKI zwischen dem 23.6. und dem 1.7.2009 einen Online-Survey (http://www.surveymonkey.com) durch. Es wurden 23 Fragen zu den Themen Logistik (zehn Fragen), Inhalt (fünf Fragen) und Qualität (acht Fragen) der Telefonkonferenz gestellt. Die überwiegende Zahl der Fragen konnte mit den vorgegebenen Antwortkategorien „stimme voll zu/stimme zu/neutral/stimme nicht zu/stimme gar nicht zu", „sehr gut/gut/weniger gut/unzufrieden", „ja/nein" oder einem freien Textfeld beantwortet werden.
Ergebnisse {#Sec8}
==========
Teilnahme {#Sec9}
---------
Seit Einführung der EpiLag am 13.1.2009 lag die durchschnittliche Beteiligung der oberen Landesgesundheitsbehörden in den ersten drei Quartalen 2009 bei 78%. An den wöchentlichen Telefonkonferenzen nahmen in unregelmäßigen Abständen auch Vertreter der Bundeswehr sowie einzelne Referenten für Infektionsschutz der Bundesländer als Vertreter der obersten Landesgesundheitsbehörden teil.
In der EpiLag diskutierte Ereignisse {#Sec10}
------------------------------------
In der EpiLag wurde bisher über das Auftreten von insgesamt 46 verschiedenen Infektionskrankheiten (darunter auch Häufungen, bei denen kein Erreger diagnostiziert worden war) diskutiert. Die mit Abstand häufigsten infektionsepidemiologischen Ereignisse, von denen berichtet wurde, waren Masern und Meningokokken-Meningitis ([Tab. 2](#Tab2){ref-type="table"}). Internationale Ereignisse wurden zum Großteil vom RKI berichtet, nationale Ereignisse überwiegend von den Vertretern der zuständigen Landesgesundheitsbehörden. Als national wurden auch die infektionsepidemiologischen Ereignisse eingestuft, bei denen aus dem Ausland importierte Erreger die Ursache waren. Über Mumps, hämorrhagisches Fieber und CJK (Creutzfeldt-Jakob-Krankheit) wurde überwiegend international berichtet; Infektionskrankheiten wie Pest, aviäre Influenza, saisonale Influenza oder Poliomyelitis wurden im Rahmen von Updates über die globale epidemiologische Situation behandelt.
+-----------------------------------+-----------------------+----------+-----------+------------------------------------+----------+---+---+
| Krankheit | Anzahl der Ereignisse | Ereignis | Krankheit | Anzahl der Ereignisse | Ereignis | | |
+===================================+=======================+==========+===========+====================================+==========+===+===+
| Masern | 24 | 16 | 8 | Listeriose | 2 | 1 | 1 |
+-----------------------------------+-----------------------+----------+-----------+------------------------------------+----------+---+---+
| Meningitis | 16 | 14 | 2 | Tollwut | 2 | 2 | 0 |
+-----------------------------------+-----------------------+----------+-----------+------------------------------------+----------+---+---+
| Pandemische Influenza (H1N1) 2009 | 11 | 10 | 1 | Typhus | 2 | 2 | 0 |
+-----------------------------------+-----------------------+----------+-----------+------------------------------------+----------+---+---+
| Hämorrhagisches Fieber | 11 | 3 | 8 | Akute respiratorische Infektion | 1 | 0 | 1 |
+-----------------------------------+-----------------------+----------+-----------+------------------------------------+----------+---+---+
| Poliomyelitis | 9 | 0 | 9 | Anthrax | 1 | 0 | 1 |
+-----------------------------------+-----------------------+----------+-----------+------------------------------------+----------+---+---+
| Salmonellose | 9 | 7 | 2 | Aviäre Influenza (globales update) | 1 | 0 | 1 |
+-----------------------------------+-----------------------+----------+-----------+------------------------------------+----------+---+---+
| Norovirus-Erkrankung | 8 | 7 | 1 | Cholera | 1 | 0 | 1 |
+-----------------------------------+-----------------------+----------+-----------+------------------------------------+----------+---+---+
| Saisonale Influenza | 8 | 6 | 2 | Dengue-Fieber | 1 | 0 | 1 |
+-----------------------------------+-----------------------+----------+-----------+------------------------------------+----------+---+---+
| Mumps | 7 | 2 | 5 | Dengue-Fieber (globales update) | 1 | 0 | 1 |
+-----------------------------------+-----------------------+----------+-----------+------------------------------------+----------+---+---+
| Kuhpocken | 6 | 6 | 0 | Diphtherie | 1 | 1 | 0 |
+-----------------------------------+-----------------------+----------+-----------+------------------------------------+----------+---+---+
| Legionellose | 6 | 2 | 4 | Hand-Fuß-Mund-Krankheit | 1 | 1 | 0 |
+-----------------------------------+-----------------------+----------+-----------+------------------------------------+----------+---+---+
| Tuberkulose | 6 | 4 | 2 | HIV/AIDS | 1 | 1 | 0 |
+-----------------------------------+-----------------------+----------+-----------+------------------------------------+----------+---+---+
| Botulismus | 5 | 3 | 2 | HPV Infektion | 1 | 0 | 1 |
+-----------------------------------+-----------------------+----------+-----------+------------------------------------+----------+---+---+
| Hepatitis A | 5 | 3 | 2 | Milzbrand | 1 | 0 | 1 |
+-----------------------------------+-----------------------+----------+-----------+------------------------------------+----------+---+---+
| Pertussis | 4 | 3 | 1 | MRSA-Infektion | 1 | 1 | 0 |
+-----------------------------------+-----------------------+----------+-----------+------------------------------------+----------+---+---+
| Pest | 4 | 0 | 4 | Poliomyelitis (globales update) | 1 | 0 | 1 |
+-----------------------------------+-----------------------+----------+-----------+------------------------------------+----------+---+---+
| West-Nil-Fieber | 4 | 0 | 4 | Q-Fieber | 1 | 1 | 0 |
+-----------------------------------+-----------------------+----------+-----------+------------------------------------+----------+---+---+
| Chikungunya-Fieber | 3 | 1 | 2 | Ringelröteln | 1 | 1 | 0 |
+-----------------------------------+-----------------------+----------+-----------+------------------------------------+----------+---+---+
| FSME | 3 | 1 | 2 | Rotavirus-Erkrankung | 1 | 1 | 0 |
+-----------------------------------+-----------------------+----------+-----------+------------------------------------+----------+---+---+
| Gastrointestinale Erkrankungen | 3 | 3 | 0 | Saisonale Influenza | 1 | 1 | 0 |
| | | | | | | | |
| | | | | Sommer Surveillance (update) | | | |
+-----------------------------------+-----------------------+----------+-----------+------------------------------------+----------+---+---+
| Nicht diagnostizierte Erkrankung | 3 | 2 | 1 | Saisonale Influenza | 1 | 1 | 0 |
| | | | | | | | |
| | | | | Winter Surveillance (update) | | | |
+-----------------------------------+-----------------------+----------+-----------+------------------------------------+----------+---+---+
| Lassa-Fieber | 3 | 2 | 1 | Skabies | 1 | 1 | 0 |
+-----------------------------------+-----------------------+----------+-----------+------------------------------------+----------+---+---+
| Shigellose | 3 | 2 | 1 | Tularämie | 1 | 0 | 0 |
+-----------------------------------+-----------------------+----------+-----------+------------------------------------+----------+---+---+
| Creutzfeldt-Jakob-Krankheit | 2 | 0 | 2 | Windpocken | 1 | 1 | 0 |
+-----------------------------------+-----------------------+----------+-----------+------------------------------------+----------+---+---+
| Myositis, epidemische | 2 | 2 | 0 | Zeckenfieber | 1 | 0 | 1 |
+-----------------------------------+-----------------------+----------+-----------+------------------------------------+----------+---+---+
*N* National, *I* International, *FSME* Frühsommer-Meningoenzephalitis, *HPV* Humanes Papillomavirus, *MRSA* Methicillin-resistenter *Staphylococcus aureus*.
Infektionsepidemiologisch relevante Ereignisse wurden teilweise über mehrere aufeinanderfolgende Wochen in der EpiLag thematisiert. Dazu zählte zum Beispiel ein schwer zu kontrollierender Masern-Ausbruch in Hamburg, der über 22 Kalenderwochen (KW) diskutiert wurde und auch zu Fällen in Niedersachsen, Nordrhein-Westfalen und Baden-Württemberg führte. In Hamburg wurden 216 Fälle gemeldet. Der letzte Fall erkrankte am 17.6.2009. Ein Hepatitis-A-Ausbruch in Lettland wurde über einen Zeitraum von 10 KW in der Telekonferenz behandelt. Über Chikungunya-Ausbrüche im Indischen Ozean, Südostasien und Ostasien wurde über einen Zeitraum von acht KW berichtet. In diesem Rahmen wurde in der 23. und 24. KW ein Chikungunya-Ausbruch in Thailand besprochen: Fünf der acht bis zum 4. Juni 2009 in Deutschland übermittelten Fälle hatten sich in Thailand, Malaysia oder Singapur, die drei anderen Fälle auf den Philippinen, Sri Lanka und in Indien infiziert. In der 36. und 37. KW wurde vom ersten Chikungunya-Ausbruch auf der Insel La Réunion nach dem Ausbruch von 2005/2006 berichtet; im Sommer 2007 kam es in Italien zur ersten lokalen Übertragung des Chikungunya-Virus in Europa \[[@CR18]\].
Als Beispiel für eine nicht-meldepflichtige Erkrankung, die in der EpiLag diskutiert wurde und damit einen raschen und regelmäßigen Austausch zwischen den zuständigen Landesgesundheitsbehörden ermöglichte, zählt ein Kuhpocken-Ausbruch mit Erkrankungsfällen in Frankreich und mehreren Bundesländern Deutschlands. Kleintiernager sind das natürliche Reservoir der Kuhpockenviren und können von infizierten Tieren auf den Menschen übertragen werden. In der 4. und 5. KW 2009 wurden in der EpiLag erstmals Kuhpocken-Fälle in Bayern und Nordrhein-Westfalen (NRW) diskutiert, danach wurde auch von Fällen in Frankreich berichtet. Die Fälle in Frankreich und NRW konnten auf Heimtierratten („Schmuseratten") zurückgeführt werden, die von einem Händler aus Tschechien exportiert wurden. Jedoch ergab sich bei diesem Händler und seinen Zulieferern kein Hinweis auf eine Kuhpocken-Infektion der Tiere. Die tschechischen Behörden wurden über das EWRS informiert. Die Kuhpocken-Fälle in Bayern konnten auf einen bayerischen Rattenzüchter zurückgeführt werden, bei dem die zuständige Veterinärbehörde 17 von 31 Heimtierratten positiv testete.
In der 29. KW kam es erneut zu einem Kuhpockenausbruch, der durch tiefgefrorene Futterratten übertragen wurde, die aus einem polnischen Zuchtbetrieb stammten. Futterratten sind zur Fütterung anderer Tiere bestimmt. Baden-Württemberg konnte als erstbetroffenes Bundesland mithilfe der EpiLag andere Bundesländer zeitnah benachrichtigen: Das RKI erhielt aus Baden-Württemberg eine Empfängerliste der Futterratten des Händlers, leitete sie an die betroffenen Bundesländer weiter, die wiederum den Gesundheitsämtern die Lieferadressen mitteilen konnten.
Am 1.7.2009 wurde in Deutschland die Meldepflicht für MRSA eingeführt. Die EpiLag behandelte dieses Thema intermittierend über insgesamt sieben KW: Inhalte waren die MRSA-Falldefinition, die Art der Meldung, die provisorische Erfassung, die technischen Vorgehensweisen sowie die Übermittlung der Fälle mit SurvNet, einer Software, die der Eingabe von Fallmeldungen entsprechend IfSG sowie der Weiterleitung (Übermittlung) der erfassten Daten von der Peripherie (Gesundheitsämter) über die zuständigen Landesgesundheitsbehörden an das RKI dient. Es wurde deutlich, dass die EpiLag zur Klärung technischer Fragen sowie zur adäquaten Umsetzung neuer Anforderungen beitragen kann.
Besonders bei internationalen Ereignissen, die über mehrere Wochen diskutiert wurden, zeigt sich, dass in der EpiLag unterschiedliche Informationsquellen genutzt wurden ([Tab. 3](#Tab3){ref-type="table"}).
ProMed ECDC Threat Report EWRS IGV Bundesländer RKI
------------------------- -------- -------------------- ------ ----- -------------- -----
Masernausbruch D (HH) X
Hepatitis A Lettland X X
Chikungunya Fieber D, B X X X
Meldepflicht MRSA D X
Masernausbruch D (NRW) X
Kuhpocken D X X X
Salmonellose USA X X X
*D* Deutschland, *HH* Hamburg, *B* Belgien, *NRW* Nordrhein-Westfalen, *ProMed* Program for Monitoring Emerging Diseases, *ECDC* European Centre for Disease Prevention and Control, *EWRS* Early Warning and Response System, *IGV* Internationale Gesundheitsvorschriften, *RKI* Robert Koch-Institut.
Ergebnisse der Evaluation {#Sec11}
-------------------------
Beim Online-Survey antworteten 22 Teilnehmer. Alle 16 oberen Landesgesundheitsbehörden waren vertreten, und sämtliche Befragten hatten eine ihr Bundesland repräsentierende Funktion.
Mit der Logistik der wöchentlichen EpiLag waren die Teilnehmer generell zufrieden. Über 90% der Befragten gaben an, die Koordination der Telefonkonferenzen durch das RKI sei vollständig und ausreichend, und es gebe ausreichend Zeit, Fragen zu stellen sowie über relevante Vorkommnisse in der Region zu berichten. Fast 71% hielten die Häufigkeit der Telefonkonferenzen für angemessen. Alle Teilnehmer waren der Ansicht, dass die Protokolle zur Telefonkonferenz ausreichend zeitnah erstellt werden. Über 90% gaben an, die Protokolle seien in einem geeigneten Format (PDF), und über 80% begrüßten die Möglichkeit, dem RKI Feedback zu den Protokollen geben zu können. Insgesamt 85% der Teilnehmer gaben an, dass die Anmerkungen der Länder während der Telefonkonferenzen im Protokoll exakt wiedergegeben worden seien. Die Informationen zur aktuellen epidemiologischen Lage wurden von 91% der Teilnehmer als ausreichend angesehen. 85% der Befragten waren der Ansicht, dass Informationen über nationale Vorkommnisse für ihre tägliche Arbeit relevant seien, mit Blick auf die Informationen über internationale Ereignisse waren 71% dieser Ansicht. Alle Teilnehmer nutzen die Informationen der EpiLag, um sich über infektionsepidemiologische Ereignisse in anderen Bundesländern bzw. die nationale Situation in Deutschland, und 88%, um sich über die internationale Situation zu informieren. 95% der Teilnehmer betrachten die Telefonkonferenz als eine Informationsquelle, und 85% nutzen die Informationen der Konferenz für interne Zwecke.
Die allgemeine Qualität der EpiLag hinsichtlich Durchführung, Inhalt und Format der Protokolle sowie die Qualität der länderspezifischen Informationen wurde von allen Teilnehmern mit entweder „gut" oder „sehr gut" bewertet. Die Wichtigkeit der internationalen Ereignisse bezeichneten 85% der Teilnehmer zumindest als „hoch". Zum Zeitpunkt der Evaluation wurden die Informationen aus dem Lagebericht von vier zuständigen Landesgesundheitsbehörden an die zugeordneten Gesundheitsämter und von fünf zuständigen Landesgesundheitsbehörden an ihre Landesministerien weitergeleitet. Das Protokoll wird aber auch als Diskussionsgrundlage oder als Hilfe zur Bearbeitung aktueller infektionsepidemiologischer Probleme genutzt.
Die Kommentare der Teilnehmer hinsichtlich weiterer Vorschläge für die EpiLag bezogen sich überwiegend auf die seit Ende April 2009 in Deutschland auftretende pandemische Influenza (H1N1) 2009, die bis Ende Juni 2009 in der Telefonkonferenz nicht behandelt wurde. Die Besprechung des Themas in der EpiLag wurde von den Teilnehmern für wichtig gehalten.
Diskussion {#Sec12}
==========
Vor Einrichtung der EpiLag gab es zwischen Bund und Ländern keinen strukturierten, häufigen und regelmäßigen Austausch über aktuelle infektionsepidemiologische Ereignisse. Von Beginn an wurde die EpiLag vom RKI im Rahmen des „Epidemic Intelligence"-Ansatzes (EI) konzipiert. Die konstant hohe Beteiligung der oberen Landesgesundheitsbehörden lässt darauf schließen, dass eine Lücke geschlossen und das EI-Konzept von den Teilnehmern angenommen wurde. Die Evaluation zur allgemeinen Qualität, Organisation und Relevanz der Konferenz ergab hohe Zustimmungsraten. Die Weiterleitung des epidemiologischen EpiLag-Protokolles an die Gesundheitsämter und Landesministerien durch einige Teilnehmer ist ein weiterer Beleg für die Qualität und Relevanz der Konferenz und Protokolle.
Berichte über nationale Ereignisse wurden bei der Evaluation zu 85% als relevant eingestuft. Die Relevanz internationaler Ereignisse wurde von 85% der Teilnehmer zumindest als „hoch" bezeichnet; 88% nutzen die EpiLag auch dazu, sich über die internationale Situation zu informieren, 71% betrachten die internationalen Informationen als relevant für ihre tägliche Arbeit. Die Aussagen lassen den Schluss zu, dass das Informationsbedürfnis über internationale Ereignisse höher ist als der praktische Nutzen für die tägliche Arbeit.
Die pandemische Influenza (H1N1) 2009 war zunächst in der EpiLag ausgespart worden, um inhaltliche Dopplungen und Überschneidungen mit anderen Konferenzen zu vermeiden. In der Evaluation wurde deutlich, dass die Vertreter der oberen Landesgesundheitsbehörden es als wichtig betrachteten, nicht auf dieses Thema zu verzichten. In der Folge wurde die pandemische Influenza (H1N1) 2009 wieder in die EpiLag aufgenommen. Die EpiLag soll jedoch nicht als Ersatzforum für eine regelmäßige Abstimmung der Managementstrategien zwischen RKI und den Referenten für Infektionsschutz der einzelnen Bundesländer dienen.
Durch die EpiLag haben die zuständigen Landesgesundheitsbehörden die Möglichkeit, vom RKI über infektionsepidemiologisch relevante internationale Informationen aus dem EWRS und dem CDTR des ECDC informiert zu werden -- nationale Anlaufstelle für dieses System ist das RKI. Das RKI hat wiederum durch die EpiLag die Möglichkeit, relevante Informationen an die zuständigen Landesgesundheitsbehörden weiterzugeben sowie Informationen von ihnen zu erhalten.
Die EpiLag erleichtert den horizontalen Informationsaustausch zwischen den Bundesländern, wie sich vor allem am Beispiel der Kuhpockenausbrüche durch Schmuse- und Futterratten zeigte.
Generell bietet eine Telefonkonferenz den Vorteil, dass die teilweise zeit- und kostenaufwendige An- und Abfahrt wegfällt und somit die zeitlichen und finanziellen Ressourcen aller Teilnehmer effektiver genutzt werden können. Durch standardisierte Arbeitsweise konnte seit Einführung der EpiLag der wöchentliche zeitliche Aufwand bei der Vorbereitung reduziert werden.
Schlussfolgerung {#Sec13}
================
Das Epidemic Intelligence (EI)-Konzept der EpiLag stellt den Teilnehmern Meldungen über Ereignisse aus der indikatorbasierten Surveillance sowie aus der eventbasierten Surveillance aus unterschiedlichen Quellen zur Verfügung. Informationen aus beiden Komponenten können mithilfe der EpiLag beschleunigt weitergegeben und evaluiert werden. Somit können auch Public-Health-Maßnahmen rascher eingeleitet werden. Mit Einführung der EpiLag wurde eine Informationslücke geschlossen.
Die Implementierung eines nationalen EI-Surveillance-Systems sollte spezifisch auf Monitoring und Evaluation von Infektionskrankheiten, die die Gesundheit der Bevölkerung bedrohen, zugeschnitten sein \[[@CR19]\]. Da sich die EpiLag als infektionsepidemiologische Plattform zwischen nationaler Ebene (RKI) und den zuständigen Landesgesundheitsbehörden bewährt hat, könnten auch zwischen den zuständigen Landesgesundheitsbehörden und lokalen Gesundheitsämtern Telefonkonferenzen eingerichtet werden, auf denen die Teilnehmer regelmäßig, strukturiert und zeitnah infektionsepidemiologische Ereignisse diskutieren.
Danksagung {#d29e1389}
==========
Wir danken allen Vertretern der oberen Landesgesundheitsbehörden, den Referenten für Infektionsschutz der Bundesländer sowie der Bundeswehr für die aktive Teilnahme an der EpiLag-Telekonferenz. Unser besonderer Dank gilt Ramona Keuchel, Inge Mücke, Denise Neugebauer, Mona Poorbiazar, Annicka Reuß, Irene Schöneberg, Birgitta Schweickert und Maria Wadl vom RKI für ihren Beitrag zu diesem Manuskript.
Interessenkonflikt {#d29e1394}
==================
Der korrespondierende Autor gibt an, dass kein Interessenkonflikt besteht.
| {
"pile_set_name": "PubMed Central"
} |
These third party data are available from NIH GRASP. The authors did not have any special access privileges and interested researchers can access the data at <https://grasp.nhlbi.nih.gov/FullResults.aspx> (Trait(s): Ankle injury).
Introduction {#sec001}
============
Ankle sprains and strains are the most common musculoskeletal injuries in athletes, especially indoor or court sports \[[@pone.0185355.ref001]--[@pone.0185355.ref003]\]. Low ankle sprains occur with injury to ankle ligaments, most commonly the lateral ligament complex. Ankle strains describe pathologic stretching or tearing of muscle or tendon. Other ankle joint derangements, such as instability, may occur with insufficiency of the soft tissue restraints of the ankle or malalignment. These injuries (collectively referred to as ankle injuries) often occur in athletic activity with axial loading of an inverted, plantar-flexed foot as the most common mechanism. Ankle sprains are more common in women than men, and in children or adolescents compared to adults \[[@pone.0185355.ref002],[@pone.0185355.ref003]\].
Little is known about genetic factors that affect risk for ankle injury. One study reported an association of the R577X mutation in the *actinin 3* (*ACTN3*) gene with acute ankle sprain from a study involving 142 cases of injury (p = .011)\[[@pone.0185355.ref004]\]. *ACTN3* encodes alpha-actinin skeletal muscle isoform 3, which is an actin-binding protein expressed in all skeletal muscles. In principle, genetic studies such as this have the potential to identify contributing factors to ankle injuries and to provide diagnostic markers informing individuals about their personal risk for injury.
In order to identify genetic factors that may provide insight about ankle injuries, we screened the entire genome for loci associated with these injuries. We identified individuals who had suffered an ankle injury from a cohort of 99,342 patients in the Research Program on Genes, Environment, and Health (RPGEH) of the Kaiser Permanente, Northern California (KPNC) health plan. A gene association analysis revealed one locus on chromosome 21 associated with ankle injury with genome-wide significance, and a second locus on chromosome 9 with an association just below genome-wide significance. We re-tested the *ACTN3* SNP for association with ankle injury in our cohort but did not see a significant association.
Methods {#sec002}
=======
A genome-wide association screen was performed for ankle injury using data from the genotyped Genetic Epidemiology Research on Adult Health and Aging (GERA) cohort of the Research Program in Genes, Environment and Health (RPGEH). The data generation and data analysis pipeline have been previously described \[[@pone.0185355.ref005]\]. A complete description of the cohort and study design can be found in dbGaP (Study Accession: phs000674.v1.p1).
Our analysis cohort (n = 99,342) includes 57,606 females, 41,670 males, and 66 individuals of uncertain sex ([Table 1](#pone.0185355.t001){ref-type="table"}). Sex was determined previously based on heterozygosity of the X chromosome (dbGAP Study Accession: phs000674.v1.p1). Moreover, our analysis cohort is ethnically diverse, including 83,264 European-White (EUR); 8,560 Latino (LAT) and 7,518 East Asian (EAS) individuals based on ancestry principle components.
10.1371/journal.pone.0185355.t001
###### Demographic factors of the GERA study population used in genome-wide association analyses of ankle injury.
{#pone.0185355.t001g}
Cases[^a^](#t001fn001){ref-type="table-fn"} Controls Overall
------------------------------------------------------------ --------------------------------------------- ---------------- --------------
**Subjects** (%) 1,696 (1.7%) 97,646 (98.3%) 99,342
**Sex** (%)[^**b**^](#t001fn002){ref-type="table-fn"}
Female 1,049 (1.8%) 56,557 (98.2%) 57,606
Male 645 (1.5%) 41,025 (98.5%) 41,670
Undetermined 1 (1.5%) 65 (98.5%) 66
**Ancestry** (%)[^**c**^](#t001fn003){ref-type="table-fn"}
European 1,421 (1.7%) 81,843 (98.3%) 83,264
Latin American 154 (1.8%) 8,406 (98.2%) 8,560
East Asian 121 (1.6%) 7,397 (98.4%) 7,518
**Age**[^**d**^](#t001fn004){ref-type="table-fn"} 60.1 (±13.1) 62.9 (±13.7) 62.8 (±13.7)
^a^ Cases with ankle injury as defined by individuals with one or more qualifying ICD-9, ICD-10 or CPT-4 code in their EHR. For details, see [Methods](#sec002){ref-type="sec"}.
^b^ Sex/gender as determined by an individual's genetic data, reported as the number and percentage of total.
^c^ Race/ethnic groups as determined by PCA on an individual's genetic data from the GERA cohort. Reported as the number and percentage of total for each respective group.
^d^ Age at subject enrollment in the GERA cohort, reported as mean age with standard deviation.
Participants were genotyped at over 650,000 SNPs \[[@pone.0185355.ref006]\]. Genotypes were then imputed using standard procedures with a cutoff of R^2^ \> 0.3 \[[@pone.0185355.ref007]--[@pone.0185355.ref009]\]. The quality of the imputed data was previously validated in Jorgenson et al., 2015 \[[@pone.0185355.ref010]\].
Determination of genetic ancestry was performed by principal component analysis (PCA), as previously described \[[@pone.0185355.ref005],[@pone.0185355.ref011]\]. These ancestry principal components were used in the GWAS to adjust for genetic ancestry.
Phenotype definition {#sec003}
--------------------
Ankle injury cases were identified in the GERA cohort based on clinical diagnoses and surgical procedures captured in the KPNC electronic health record system. The electronic health record includes reported injuries over the entire lifetime of the patients, including those that occurred prior to enrollment in KPNC as well as those that occurred after the genotyping analysis was performed, if reported by the patient and recorded by the physician. International Classification of Disease, Ninth Revision (ICD-9), International Classification of Disease, Tenth Revision (ICD-10) and Common Procedure Terminology, Fourth Edition (CPT-4) codes, were used to identify cases of ankle injury ([Table 2](#pone.0185355.t002){ref-type="table"}). [Table 2](#pone.0185355.t002){ref-type="table"} includes codes for: 1) ankle sprain, 2) ankle strain, 3) surgical repair for disrupted ankle ligament and 4) joint derangement of the ankle or foot. Ankle sprains and strains are not differentiated in the ICD-9 codes, while they are in the ICD-10 codes. Joint derangement includes joint instability.
10.1371/journal.pone.0185355.t002
###### Ankle injury phenotypes classified by ICD and/or CPT codes.
{#pone.0185355.t002g}
ICD-9 code Code Description[^a^](#t002fn001){ref-type="table-fn"} N
------------- ------------------------------------------------------------------------ -----
845.01 Sprain/Strain Deltoid Ligament 38
845.02 Sprain/Strain Calcaneofibular Ligament 21
845.03 Sprain/Strain Tibiofibular Ligament, Distal 34
845.09 Other Sprains/Strains Of Ankle 789
718.87 Other Joint Derangement, Not Elsewhere Classified, Ankle And Foot 444
ICD-10 code
S93.401A Sprain Of Unspecified Ligament Of Right Ankle, Initial Encounter 190
S93.401D Sprain Of Unspecified Ligament Of Right Ankle, Subsequent Encounter 44
S93.401S Sprain Of Unspecified Ligament Of Right Ankle, Sequela 1
S93.402A Sprain Of Unspecified Ligament Of Left Ankle, Initial Encounter 165
S93.402D Sprain Of Unspecified Ligament Of Left Ankle, Subsequent Encounter 45
S93.402S Sprain Of Unspecified Ligament Of Left Ankle, Sequela 1
S93.409A Sprain Of Unspecified Ligament Of Unspecified Ankle, Initial Encounter 3
S93.411A Sprain Of Calcaneofibular Ligament Of Right Ankle, Initial Encounter 2
S93.412A Sprain Of Calcaneofibular Ligament Of Left Ankle, Initial Encounter 2
S93.421D Sprain Of Deltoid Ligament Of Right Ankle, Subsequent Encounter 1
S93.422A Sprain Of Deltoid Ligament Of Left Ankle, Initial Encounter 1
S93.422D Sprain Of Deltoid Ligament Of Left Ankle, Subsequent Encounter 2
S93.422S Sprain Of Deltoid Ligament Of Left Ankle, Sequela 1
S93.431A Sprain Of Tibiofibular Ligament Of Right Ankle, Initial Encounter 1
S93.432D Sprain Of Tibiofibular Ligament Of Left Ankle, Subsequent Encounter 1
S93.432S Sprain Of Tibiofibular Ligament Of Left Ankle, Sequela 1
S93.492A Sprain Of Other Ligament Of Left Ankle, Initial Encounter 4
S93.492D Sprain Of Other Ligament Of Left Ankle, Subsequent Encounter 1
S93.492S Sprain Of Other Ligament Of Left Ankle, Sequela 1
CPT
27695 Repair, Primary, Disrupted Ligament, Ankle; Collateral 15
^a^ International Statistical Classification of Diseases and Related Health Problems (ICD-9 or ICD-10) and Current Procedural Terminology (CPT-4) codes extracted from KPNC electronic health records of GERA cohort subjects.
Genome-wide association and meta-analysis {#sec004}
-----------------------------------------
Genome-wide association analyses of the GERA cohort genomic data were conducted as previously described \[[@pone.0185355.ref005],[@pone.0185355.ref011]\]. SNP associations were tested with ankle injury with a logistic regression model using allele counts for typed and imputed SNPs in an additive genetic model for each of the race/ethnic populations. The model was adjusted for genetic sex, age at enrollment into the RPGEH cohort, race/ethnicity using principal components, and variations in genotyping protocol \[[@pone.0185355.ref005],[@pone.0185355.ref011]\]. We used 10 principal components for European (EUR), 6 for Latin American (LAT) and 3 for East Asian (EAS). The final number of SNPs that were analyzed was 8,795,348 for EUR; 9,153,118 for LAT and 8,055,053 for EAS. To account for inflation due to population stratification, the genomic control parameter (λ~gc~) was calculated: EUR (1.008), LAT (1.008), EAS (1.048). Subsequently, p-values were adjusted for genomic control in each population. Results from each population were combined by inverse-variance, fixed-effects meta-analysis as previously described \[[@pone.0185355.ref005],[@pone.0185355.ref011]\]. SNPs that did not contain data for EUR were removed, as EUR comprises more than 80% of the cohort. The final number of SNPs that was analyzed in the fixed-effects meta-analysis was 8,183,964. Power calculations were made using the software at <http://csg.sph.umich.edu/abecasis/cats/gas_power_calculator/index.html> \[[@pone.0185355.ref012]\].
We examined the level of heterogeneity as previously described \[[@pone.0185355.ref005],[@pone.0185355.ref011]\].
To perform the sensitivity analysis, the total set of 1696 cases was split into a subset that are known to involve an ankle injury (1275 cases) and another subset (421 cases) that are either ankle or foot derangements as they were identified solely by ICD-9 code 718.87 (Other Joint Derangement, Not Elsewhere Classified, Ankle And Foot). Logistic regression was used to calculate the association of chr21:47156779:D and rs13286037 with each subset of cases. Because the number of cases in the subgroups was smaller than in the total, only six principal components were used for the EUR ancestry group. P-values were calculated from a fixed-effects meta-analysis.
Further bioinformatics investigation of the top genome-wide significant loci from the meta-analysis was conducted as previously described \[[@pone.0185355.ref005],[@pone.0185355.ref011]\].
Summary statistics for all SNPs from the fixed-effects meta-analysis are available at NIH GRASP: <https://grasp.nhlbi.nih.gov/FullResults.aspx>.
Ethical considerations {#sec005}
----------------------
This study analyzed stored data from RPGEH subjects who consented to genomic testing and use of their genomic data, as well as health data from the KPNC electronic health record, for future research studies. The health and genotype data for the subjects were de-identified. All study procedures were approved by the Institutional Review Board of the Kaiser Permanente Research Institute.
Results {#sec006}
=======
Study population and genotype information {#sec007}
-----------------------------------------
We performed a logistic regression for DNA variants associated with ankle injuries. Ankle injuries (which refers to ankle sprains, strains and other derangements) were identified by ICD-9, ICD-10 and CPT codes ([Table 1](#pone.0185355.t001){ref-type="table"}). There were 1,696 cases and 97,646 controls in the GERA cohort. Overall, the period prevalence of ankle injury was 1.7%. Participation in sports was not included in the electronic health record, and hence we were not able to determine the incidence rate for the subset of the population who were athletes. Men showed a lower incidence of ankle injury than women that was statistically significant (p = 1.3x10^-3^; OR = .85; 95% CI = 0.77--0.94), consistent with previous results \[[@pone.0185355.ref002],[@pone.0185355.ref003]\]. There was a small difference in the age of enrollment between the cases and controls (1.8 yrs) that was statistically significant (p = 1.5x10^-8^)([Table 1](#pone.0185355.t001){ref-type="table"}). One possibility is that this might be caused by an ascertainment bias where some elderly patients that enrolled in the RPGEH program might be systematically missed as cases if they incurred the ankle injury when they were young, before electronic records were in common practice.
Genome-wide study for association with ankle injury {#sec008}
---------------------------------------------------
The RPGEH cohort, genotyping data, methodological approach and logical flow presented here overlap those used in previous work by the same authors on MCL injury, shoulder dislocation, plantar fasciitis, ACL injury and Achilles tendon injury \[[@pone.0185355.ref005],[@pone.0185355.ref011],[@pone.0185355.ref013],[@pone.0185355.ref014]\]. However, the analyses presented here present new results and concepts on the genetic basis for ankle injury, which has not previously been analyzed at the genome-wide level.
We compared the observed p-values to the distribution of p-values expected by chance in a Q-Q plot ([Fig 1](#pone.0185355.g001){ref-type="fig"}). The black dots deviate from the red line for the lowest observed p-values in the upper right hand corner, indicating that the observed association signals are significantly different than the signals that would be expected by chance.
{#pone.0185355.g001}
The p-value for every SNP from the meta-analysis is shown in a Manhattan plot in [Fig 2](#pone.0185355.g002){ref-type="fig"}. Using p = 5x10^-8^ as a cut-off for genome-wide significance, chr21:47156779:D on chromosome 21 showed a genome-wide significant association with ankle injury (p = 3.8x10^-8^, [Table 3](#pone.0185355.t003){ref-type="table"}). rs13286037 on chromosome 9 showed an association that was nearly genome-wide significant (5.1x10^-8^)([Table 3](#pone.0185355.t003){ref-type="table"}). For both chr21:47156779:D and rs13286037, the minor allele frequency was below 5%, meaning that there were a relatively small number of individuals carrying either one or two copies of the risk allele. Fisher's exact test (i.e. linear discriminant analysis) is an alternative way to analyze the association of a DNA variant with ankle injury for small sample sizes. We repeated the analysis for association with ankle injury using Fisher's exact test for these two DNA variants, and obtained p-values that were similar although slightly less strong than the values using logistic regression. Specifically, the p-values using Fisher's exact test for chr21:47156779:D and rs13286037 were 4.8x10^-8^ and 9.5x10^-8^, respectively.
{#pone.0185355.g002}
10.1371/journal.pone.0185355.t003
###### Genome-wide association analyses for ankle injury.
{#pone.0185355.t003g}
Variant Gene(s) EA[^a^](#t003fn001){ref-type="table-fn"} EAF[^b^](#t003fn002){ref-type="table-fn"} P-value[^c^](#t003fn003){ref-type="table-fn"} OR (95% CI)[^d^](#t003fn004){ref-type="table-fn"}
-------------------------------------------------------- ----------- ------------------------------------------ ------------------------------------------- ----------------------------------------------- ---------------------------------------------------
chr21:47156779:D[^e^](#t003fn005){ref-type="table-fn"} *COL18A1* A .011 3.8x10^-8^ 1.99 (1.75--2.23)
*SLC19A1*
*PCBP3*
rs13286037[^e^](#t003fn005){ref-type="table-fn"} *NFIB* A .025 5.1x10^-8^ 1.63 (1.46--1.80)
^a^ Effect allele (minor allele).
^b^ Effect allele frequency in the control population.
^c^ P-value from fixed-effects meta-analysis.
^d^ Adjusted allelic odds ratio with 95% confidence interval.
^e^ Genotypes for these SNPs were imputed: chr21:47156779:D (R^2^ = 0.73) and rs13286037 (R^2^ = 0.94).
Neither chr21:47156779:D nor rs13286037 were directly genotyped on the Affymetrix chips, but rather their genotype data was imputed ([Table 3](#pone.0185355.t003){ref-type="table"}; [S1 Table](#pone.0185355.s001){ref-type="supplementary-material"}). For chr21:47156779:D, the R^2^ value was 0.73, indicating that the genotype was only partially accurate using imputation and thus care should be taken until true genotype data can be obtained. Inaccuracies in the genotype of chr21:47156779:D caused by imputation would not be expected to lead to a spurious association between chr21:47156779:D and ankle injury. Rather one might expect that the noise introduced by inaccurate imputation would weaken the association between the true genotype of chr21:47156779:D and ankle injury. For rs13286037, the R^2^ value was 0.94 indicating that the imputed genotype is fairly accurate.
Of the 1696 cases of ankle injury, 421 were identified based solely on ICD-9 code 718.87, which pertains to instability or hypermobility of joints in either the ankle or foot. For this code, most of the injuries involve the ankle but some may involve foot derangements instead. Because some of the diagnoses with this code may have involved a foot rather than an ankle derangement, we wanted to know whether the association of chr21:47156779:D and rs13286037 was similar between the subset of cases known to affect the ankle (1275 cases) versus the subset identified by ICD-9 718.87 (421 cases). We split the cases into the two sub-groups, then repeated the logistic regression and meta-analysis for each sub-group ([Table 4](#pone.0185355.t004){ref-type="table"}). The results indicate that both chr21:47156779:D and rs13286037 show an association with both the ankle sprain/strain and the ankle/foot derangement phenotypes. The odds ratios are similar for the two phenotypes. As expected, the p-values become weaker as the number of cases drops in each sub-group. These results indicate that the association of the top two SNPs with the ankle sprain/strain and the ankle/foot derangement phenotypes are qualitatively similar.
10.1371/journal.pone.0185355.t004
###### Sensitivity analysis for association with ankle sprains/strains and ankle/foot derangements.
{#pone.0185355.t004g}
chr21:47156779:D Cases P value [^a^](#t004fn001){ref-type="table-fn"} OR (95% CI) [^b^](#t004fn002){ref-type="table-fn"}
--------------------------------------------------------------- ------- ------------------------------------------------ ----------------------------------------------------
All cases 1696 3.8x10^-8^ 1.99 (1.75--2.23)
Ankle sprains/strains[^c^](#t004fn003){ref-type="table-fn"} 1275 1.2x10^-5^ 1.89 (1.60--2.18)
Ankle/foot derangements[^d^](#t004fn004){ref-type="table-fn"} 421 5.2x10^-4^ 2.26 (1.80--2.72)
rs13286037
All cases 1696 5.1x10^-8^ 1.63 (1.46--1.80)
Ankle sprains/strains[^c^](#t004fn003){ref-type="table-fn"} 1275 6.0x10^-7^ 1.67 (1.47--1.87)
Ankle/foot derangements[^d^](#t004fn004){ref-type="table-fn"} 421 1.6x10^-2^ 1.54 (1.19--1.89)
^a^ P-value from fixed-effects meta-analysis.
^b^ Adjusted allelic odds ratio with 95% confidence interval.
^c^Cases known to an ankle sprain or strain.
^d^Cases identified only by ICD-9 code 718.87 involving either an ankle or a foot derangement.
Two loci associated with ankle injuries {#sec009}
---------------------------------------
For chr21:47156779:D, individuals that carried one copy of the risk allele (genotype A/AG) had a 1.86-fold increased risk of ankle injury compared to individuals with no risk alleles (genotype AG/AG)([Table 5](#pone.0185355.t005){ref-type="table"}). For rs13286037, individuals carrying one copy of the risk allele (A/T) had a 1.58-fold higher risk for ankle injury compared to individuals with no risk alleles (T/T)([Table 5](#pone.0185355.t005){ref-type="table"}). For both genetic variants, the risk was even higher in people carrying two copies of the risk allele, but there were too few such individuals for this result to be statistically significant ([Table 5](#pone.0185355.t005){ref-type="table"}).
10.1371/journal.pone.0185355.t005
###### Genotype distributions for chr21:47156779:D and rs13286037.
{#pone.0185355.t005g}
chr21:47156779:D A/A A/AG AG/AG
------------------------------------------------------------------------------- -------------- ------- --------
Cases 1 67 1,585
Controls 8 2,091 92,988
Overall 9 2,158 94,573
Risk for ankle injury 0.111 .0310 .0167
Relative risk for ankle injury[^a^](#t005fn001){ref-type="table-fn"} (95% CI) 6.65 1.86 1.00
(0.84--52.36) (1.44--2.37)
rs13286037 A/A A/T T/T
Cases 3 131 1,612
Controls 83 4,845 95,394
Overall 86 4,976 97,006
Risk for ankle injury .0361 .0263 .0166
Relative risk for ankle injury[^a^](#t005fn001){ref-type="table-fn"} (95% CI) 2.17 1.58 1.00
(0.69--6.89) (1.35--1.90)
^a^ Risk relative to individuals homozygous for the protective allele (95% CI).
The GWAS results were analyzed to determine whether the association with ankle injury for either chr21:47156779:D or rs13286037 was stronger in some ancestry groups than in others, a phenomenon known as heterogeneity \[[@pone.0185355.ref015]\]. [Table 6](#pone.0185355.t006){ref-type="table"} shows the p-values and odds ratios for these two SNPs for each ancestry group. The logistic regression analysis did not converge on a p-value for the EAS ancestry for either chr21:47156779:D or rs13286037 due to limited number of cases. As expected, the smallest p-value for both SNPs was observed for the European population, since 82% of the cohort was European. For the LAT ancestry group, the p-values were 0.94 for chr21:47156779:D and 0.08 for rs13286037. The odds ratios for each race were in the same direction and of similar magnitude. Using I^2^ and Cochran's Q to assess heterogeneity, we saw no evidence of significant heterogeneity for rs13286037 between the EUR and LAT ancestry groups ([Table 6](#pone.0185355.t006){ref-type="table"}). For chr21:47156779:D, the I^2^ estimate was 42%, suggesting that there might be heterogeneity between the EUR and LAT ancestry groups. However, the 95% confidence interval for I^2^ was 0--90, indicating that the presence and extent of heterogeneity is not certain.
10.1371/journal.pone.0185355.t006
###### Association statistics for chr21:47156779:D and rs13286037 with ankle injury in individual ancestry groups.
{#pone.0185355.t006g}
Race SNP EA[^a^](#t006fn001){ref-type="table-fn"} P-value[^b^](#t006fn002){ref-type="table-fn"} OR I^2^ [^d^](#t006fn004){ref-type="table-fn"} Q [^e^](#t006fn005){ref-type="table-fn"}
-------------- ------------------ ------------------------------------------ ----------------------------------------------- ------ --------------------------------------------- ------------------------------------------
EUR chr21:47156779:D A 1.57x10^-8^ 2.07 42 0.19
(1.61--2.66) (0--90)
LAT chr21:47156779:D A 0.94 1.04
(0.38--2.81)
EAS chr21:47156779:D A ND[^f^](#t006fn006){ref-type="table-fn"} ND
EUR rs13286037 A 2.62x10^-7^ 1.61 0 0.70
(1.34--1.93) (0--90)
LAT rs13286037 A 0.08 1.85
(.93--3.68)
EAS rs13286037 A ND ND
^a^Effect allele.
^b^P value adjusted for lambda genomic inflation factor from fixed-effect meta-analysis.
^c^ Allelic odds ratio (95% confidence interval).
^d^ Percentage of variability between ancestry groups that is due to heterogeneity (95% confidence interval).
^e^ Cochran's Q, p-value that the association is different between ancestry groups.
^f^No data.
chr21:47156779:D is located in the intergenic region between the protein-coding genes *COL18A1*, *SLC19A1* and *PCBP3* on chromosome 21 ([Fig 3](#pone.0185355.g003){ref-type="fig"}). *COL18A1* encodes the alpha chain of type XVIII collagen, which is a structural component of tendons and ligaments \[[@pone.0185355.ref016]\]. *SLC19A1* encodes Solute Carrier Family 19, which transports folate into cells \[[@pone.0185355.ref017]\]. *PCBP3* encodes Poly(rC)-Binding Protein 3, which binds poly(C) stretches in RNA \[[@pone.0185355.ref018]\]. rs13286037 is located in an intron of *NFIB*, which encodes a transcriptional repressor protein ([Fig 4](#pone.0185355.g004){ref-type="fig"})\[[@pone.0185355.ref019]\].
{#pone.0185355.g003}
{#pone.0185355.g004}
We searched for a mechanism whereby these SNPs or a linked SNP might affect the activity of nearby genes to account for their effects on ankle injury. Using R^2^\>0.7 as a threshold, there are three other SNPs (rs76694187, rs138382277 and rs118069956) that are in the same linkage disequilibrium block as chr21:47156779:D spanning about 71 kb on chromosome 21 ([S1 Table](#pone.0185355.s001){ref-type="supplementary-material"}). Any one of these four SNPs might be responsible for affecting risk for ankle injury, with the others showing an association simply due to linkage.
None of the four SNPs in the linkage block on chromosome 21 (chr21:47156779:D, rs76694187, rs138382277 and rs118069956) are in a coding region ([Fig 3](#pone.0185355.g003){ref-type="fig"}). ChIP seq experiments from ENCODE indicate that rs118069956 is located within a binding site for the transcription factors GATA2 and REST, raising the possibility that variation at rs118069956 might alter binding of these transcription factors and thereby influence expression of nearby genes \[[@pone.0185355.ref020]\]. rs118069956 is also located in a DNAse I hypersensitive region, which is a region of open chromatin often caused by binding of transcription factors \[[@pone.0185355.ref020]\]. However, gene expression experiments have not yet been able to show that allelic variation in rs118069956 leads to changes in expression of nearby genes \[[@pone.0185355.ref021]\]. A second SNP in this region (rs138382277) might be responsible for variation in expression of a nearby long non-coding RNA. The Genotype-Tissue Exchange (GTeX) Portal has shown that rs138382277 is an expression quantitative trait locus (eQTL) for *LINC00205* (Long Intergenic Non-Coding RNA); specifically, the minor allele of rs138382277 is associated with lower expression of *LINC00205* and increased risk for ankle injury. *LINC00205* is located about 350 kb from rs138382277with no known function, although this type of RNA has been found to alter chromatin structure and affect levels of gene expression \[[@pone.0185355.ref022]\].
For the locus on chromosome 9, rs35128680 is located 8.8 kb away from the sentinel SNP rs13286037 with R^2^ = 0.90, indicating that the genotypes at these two SNPs are highly correlated ([Fig 4](#pone.0185355.g004){ref-type="fig"}). Data from the ENCODE projects show that rs35128680 is located within the central portion of the binding sites for three transcription factors (SMARCC1, TRIM28, MAX) and is also in a DNAse I hypersensitive site \[[@pone.0185355.ref020]\]. These results suggest that rs35128680 might affect binding of one or more of these transcriptions factors, thereby changing expression of a nearby gene and affecting risk for ankle injury. However, expression data from the GTeX consortium has not yet shown that rs35128680 is an eQTL for *NFIB* or any other closely-linked gene, possibly because the relevant cell or tissue type for ankle injury has not been tested.
Re-testing *ACTN3* R577X for association with ankle injury {#sec010}
----------------------------------------------------------
Shang et al. have reported that the R577X mutation in *ACTN3* (rs1815739) showed an association with acute ankle sprains \[[@pone.0185355.ref004]\]. We looked up the values for this SNP in our cohort and found that it did not show any signal of association (p = 0.90; OR = 1.00; 95% CI = 0.93--1.08).
Discussion {#sec011}
==========
Ankle injuries, including sprains, strains, and other derangements, are common in athletes \[[@pone.0185355.ref002],[@pone.0185355.ref003]\]. While ankle injuries are usually caused by an acute eversion or inversion, certain populations may be at greater risk of injury following such insults. Several risk factors have been described, however the genotype of the athlete may also impact an individual's risk for ankle injury, as well as the severity of injury and rate of recovery \[[@pone.0185355.ref002],[@pone.0185355.ref003]\]. Such individuals may ameliorate their risk with preventative training, tailored conditioning and appropriate footwear \[[@pone.0185355.ref023]--[@pone.0185355.ref027]\].
Genetic markers for ankle injury {#sec012}
--------------------------------
This study provides new information about the genetic factors associated for ankle injury. We demonstrated the first evidence for genetic factors affecting ankle injury with genome-wide significance, with large-scale genotype and phenotype data from the RPGEH with 99,342 individuals including 1,696 ankle injuries. Power calculations indicate that a cohort of this size would have about a 90% chance of detecting a SNP with an association to ankle injury at genome-wide significance (assuming genotype relative risk of 1.7, minor allele frequency of 5%).
An indel (chr21:47156779:D) showed an association with ankle injury that was genome-wide significant (3.8x10^-8^), and rs13286037 showed an association that was nearly genome-wide significant (5.1x10^-8^). A previous study found that most genetic associations with moderate p-values (defined as p≤10^−7^ and p≥5x10^-8^) were validated in subsequent studies \[[@pone.0185355.ref028]\]. These results should be replicated in an independent population in future studies. For chr21:47156779:D, we note that the imputed genotype was inferred with only 73% accuracy, indicating that the association with ankle injury should be viewed with some caution.
Potential genetic mechanisms for ankle injury {#sec013}
---------------------------------------------
The chr21:47156779:D locus on chromosome 21 contains four linked variants that are located in the intergenic region between *COL18A1*, *SLC19A1* and *PCBP3* ([Fig 3](#pone.0185355.g003){ref-type="fig"}). *COL18A1* stands out as it encodes a collagen protein that might have a structural role in tendons or ligaments in the ankle \[[@pone.0185355.ref016]\]. chr21:47156779:D is an indel that has not been queried by either the ENCODE or GTeX project about whether it affects gene expression. rs118069956 is located 25 kb from chr21:47156779:D, and is situated in the binding region for the transcription factors GATA2 and REST \[[@pone.0185355.ref020]\]. A different SNP in this locus, rs138382277, is located 41 kb from chr21:47156779:D and is an eQTL for *LINC00205*, a long non- coding RNA situated 350 kb away. In summary, the association between genetic variation at this locus on chromosome 21 and ankle injury might involve changes in expression of a collagen gene or one of the two other nearby genes, or it might involve changes in expression of *LINC00205* located 350 kb away.
The locus on chromosome 9 contains two SNPs that show a moderate association with ankle injury ([S1 Table](#pone.0185355.s001){ref-type="supplementary-material"}). Both SNPs lie within an intron of a transcription factor gene *NFIB* ([Fig 4](#pone.0185355.g004){ref-type="fig"}). rs35128680 may affect expression of nearby genes as it is located within the binding regions of three transcription factors (SMARCC1, TRIM28, MAX)\[[@pone.0185355.ref020]\].
Predictive power of genetic testing for ankle injury {#sec014}
----------------------------------------------------
Individuals who have one copy of the risk allele for chr21:47156779:D (A) or rs13286037 (A) have an increased risk of 1.86-fold or 1.58-fold compared to individuals lacking a risk allele in our cohort, respectively. For the general population, a 58% or 86% increased relative risk for ankle injury may not warrant a change in lifestyle. For elite athletes participating in a jumping sport, however, this level of risk may warrant attention with regard to training regimen, because the consequences of injury can be substantial.
We were not able to replicate the association of the R577X mutation in *ACTN3* (rs1815739) with ankle injuries \[[@pone.0185355.ref004]\]. Power calculations show that our chance for replicating this result from a cohort of 1696 cases at p≤0.05 was 95% (assuming genotype relative risk≥1.2). One explanation for the lack of validation is that the previous study looked at cases of acute ankle sprain in young, Chinese male soldiers. The difference in injury phenotype (ankle sprains/strains/derangements vs. acute ankle sprains), race (mostly European vs. Han Chinese) or population (general population in the Bay Area vs. soldiers) could account for the difference in the findings. Nevertheless, evidence from many other studies suggests that candidate gene associations need to be independently replicated, otherwise their credibility is low \[[@pone.0185355.ref029],[@pone.0185355.ref030]\].
Limitations and future directions {#sec015}
---------------------------------
As noted with previous analyses of this cohort, there are several limitations to this type of study \[[@pone.0185355.ref005],[@pone.0185355.ref011]\]. First, the phenotypes were defined from codes contained in the electronic health records, which may be inaccurate. 877 of the 1696 cases (52%) were identified based on four ICD-9 codes: 845.1, 845.2, 845.3 and 845.9. These four ICD-9 codes do not distinguish between ankle sprains and ankle strains. Furthermore, ICD-9 code 718.87 identified 421 cases of either ankle or foot derangements not elsewhere classified. Thus, some of the 718.87 diagnoses may have involved the foot rather than the ankle, or bony or paralytic derangements rather than sprains and strains.
Second, the ankle is composed of different ligaments, including the deltoid, lateral and tibio-fibular ligaments. It is unclear whether or not these ligaments have different aetiologies or underlying mechanisms for injury. The specific ligaments affected in each injury are usually not indicated by the ICD-9, ICD-10 or CPT codes in the electronic health record.
Third, the electronic health records do not distinguish between acute and chronic ankle injuries, which might have different aetiologies. Hence, the electronic health records do not provide knowledge about the nature or specific site of the ankle injury. The genetic association results presented here may derive from any one, or all, of the types of ankle injuries. Further investigation is warranted to study the differences in underlying genetic influences between specific ankle ligaments or chronic versus acute trauma on ankle injuries.
Fourth, the cohort included people regardless of whether or not they participated in a sport. We cannot document whether the statistical association of chr21:47156779:D and rs13286037 with ankle injury was derived predominantly from the subset of the population that were active in one or more sports.
Fifth, the number of individuals of Latin-American and East Asian ethnicity was relatively small (8,560 and 7,518, respectively). The association results for these ancestry groups are weaker than those from the European ancestry group, as would be expected due to smaller sample size. Heterogeneity analysis did not show a significant difference in the effect of either chr21:47156779:D or rs13286037 between the Latin-American or the East Asian ancestry group compared to the European ancestry group.
Sixth, the genotypes of chr21:47156779:D and rs13286037 were not directly measured but rather deduced by imputation. Care should be taken until the genetic association results can be replicated using direct genotyping of these loci in an independent cohort.
In the future, it will be important to replicate these gene association results with ankle injury in independent cohorts. Additional studies are warranted to begin to illuminate the underlying biological mechanism for the association of variation near *COL18A1*, *SLC19A1* and *PCBP3* on chromosome 21 and *NFIB* on chromosome 9 with ankle injury. It will also be interesting to perform the analysis on populations of athletes competing in sports with high rates of ankle injury, such as basketball or soccer. The results from these studies may reveal whether certain genetic polymorphisms such as chr21:47156779:D or rs13286037 could be used as diagnostic markers to help predict which athletes harbor a higher risk for ankle injury. Preventative measures could then be taken to alleviate that risk, thereby reducing the overall incidence of injury.
Supporting information {#sec016}
======================
###### Linkage disequilibrium blocks on chromosomes 9 and 21.
(XLS)
######
Click here for additional data file.
The authors thank the Kaiser Permanente Northern California RPGEH team for access to data and assistance in data management. This work was supported by grants from the NIH (5RO1AG025941). Data for this study were provided by the Kaiser Permanente Research Program on Genes, Environment and Health (RPGEH). Participant enrollment, survey and sample collection for the RPGEH were supported by grants from the Robert Wood Johnson Foundation, the Ellison Medical Foundation, the Wayne and Gladys Valley Foundation, and Kaiser Permanente. Development of genotypic data was supported by grant RC2 AG036607 from the National Institutes of Health. Information about data access can be obtained at: <http://www.ncbi.nlm.nih.gov/projects/gap/cgibin/study.cgi?study_id=phs000674.v1.p1> and <https://researchbank.kaiserpermanente.org/for-researchers/>.
[^1]: **Competing Interests:**The authors have declared that no competing interests exist.
| {
"pile_set_name": "PubMed Central"
} |
Jin C, Li Z, Zheng X, et al. Development and validation of T‐ARMS‐PCR to detect CYP2C19\*17 allele. J Clin Lab Anal. 2020;34:e23005 10.1002/jcla.23005
1. INTRODUCTION {#jcla23005-sec-0005}
===============
P450 cytochromes (CYPs) are a superfamily of enzymes with an absorbance peak of 450 nm. They play critical roles in xenobiotic detoxification, drug metabolism, and natural product biosynthesis.[1](#jcla23005-bib-0001){ref-type="ref"} CYP2C19 is a member of P450 family of proteins and is involved in activating/inactivating reactions of a wide spectrum of drugs, such as antidepressants, antiplatelets, antihypertensives, anticancers, antiulcers, and anticoagulant drugs.[2](#jcla23005-bib-0002){ref-type="ref"} The human *CYP2C19* gene is located on chromosome 10q24.1‐q24.3, and 35 alleles have been identified. CYP2C*19\*17* is one of the single nucleotide polymorphisms (SNPs) in the *CYP2C19* gene and the only SNP that possesses increased function of CYP2C19 (<https://www.pharmvar.org/gene/CYP2C19>). It is characterized by a −806 C \> T change (rs12248560) in the promoter region, and its carriers have an ultrarapid metabolization rate of substrates.[3](#jcla23005-bib-0003){ref-type="ref"} The mechanism of *CYP2C19\*17* affecting metabolization was shown to be an increase in *CYP2C19* expression level. It was shown that the heterozygotes and homozygotes of *CYP2C19\*17* had 1.8‐fold and 2.9‐fold increased *CYP2C19* mRNA levels over *CYP2C19\*17* noncarriers in the liver, respectively.[4](#jcla23005-bib-0004){ref-type="ref"} A study showed that a specific nuclear protein bound to the element carrying −806T, but not −806C, which consequently caused the increased transcriptional activity of the *CYP2C19* gene.[5](#jcla23005-bib-0005){ref-type="ref"} The impact of *CYP2C19\*17* in the clinic is mainly shown in drug pharmacokinetics. One of the most studied drugs is clopidogrel. It is an antithrombus drug that works by preventing platelet aggregation. Clopidogrel is now used as the standard dual antiplatelet treatment with aspirin after coronary operations. Clopidogrel is taken as a prodrug and processed by CYP2C19 to obtain the active thiol metabolite. The active metabolite can inhibit platelet activation and aggregation through binding of the ADP P2Y~12~ receptor on platelet membrane and inhibiting the activation of the downstream glycoprotein IIb/IIIa complex.[6](#jcla23005-bib-0006){ref-type="ref"} Studies have shown that T allele carriers are significantly associated with an enhanced response to clopidogrel[7](#jcla23005-bib-0007){ref-type="ref"}, [8](#jcla23005-bib-0008){ref-type="ref"} and an increased bleeding risk.[7](#jcla23005-bib-0007){ref-type="ref"}, [9](#jcla23005-bib-0009){ref-type="ref"} Omeprazole, the most studied proton pump inhibitor (PPI), is also a substrate of CYP2C19. Studies have shown that patients with the *CYP2C19\*17* allele are likely to have therapeutic failure after omeprazole treatment.[5](#jcla23005-bib-0005){ref-type="ref"}, [10](#jcla23005-bib-0010){ref-type="ref"} Among children with refractory gastroesophageal reflux disease (GERD) after PPI therapy, the *CYP2C19\*17* allele was associated with antireflux surgery (ARS).[11](#jcla23005-bib-0011){ref-type="ref"} Voriconazole, a systematic antifungal drug, is also processed by CYP2C19. *CYP2C19\*17* has been shown as an important factor for voriconazole therapy failure.[12](#jcla23005-bib-0012){ref-type="ref"}
Currently, *CYP2C19\*17* can be detected by polymerase chain reaction‐restriction fragment length polymorphism (PCR‐RFLP),[4](#jcla23005-bib-0004){ref-type="ref"}, [10](#jcla23005-bib-0010){ref-type="ref"}, [12](#jcla23005-bib-0012){ref-type="ref"} TaqMan assay,[7](#jcla23005-bib-0007){ref-type="ref"}, [8](#jcla23005-bib-0008){ref-type="ref"}, [11](#jcla23005-bib-0011){ref-type="ref"} allele‐specific PCR (ASP‐PCR),[13](#jcla23005-bib-0013){ref-type="ref"} high‐resolution melting (HRM),[14](#jcla23005-bib-0014){ref-type="ref"} pyrosequencing,[10](#jcla23005-bib-0010){ref-type="ref"}, [15](#jcla23005-bib-0015){ref-type="ref"} MassARRAY,[15](#jcla23005-bib-0015){ref-type="ref"} and DNA direct sequencing.[5](#jcla23005-bib-0005){ref-type="ref"} All these methods have their respective merits and shortcomings. For example, ASP‐PCR and PCR‐RFLP do not require expensive equipment, and they are less expensive than other methods. However, ASP‐PCR requires duplex PCRs, and PCR‐RFLP involves three steps: amplification, restriction enzyme digestion, and gel electrophoresis. The TaqMan assay can detect SNPs conveniently, but it requires expensive probes and equipment. HRM is a rapid, convenient, high‐throughput, and inexpensive method for SNP genotyping, but the accuracy will be affected when another SNP is present in the same amplicon, and the equipment used in HRM is expensive. pyrosequencing, MassARRAY, and DNA sequencing are all accurate methods for detecting SNPs, and DNA sequencing is acknowledged as the gold standard for DNA sequence analysis. But all these methods require expensive equipment. Compared with the abovementioned assays, tetra‐primer ARMS‐PCR (T‐ARMS‐PCR) can detect SNPs by four primers using only one PCR and do not need restriction enzymes, probes, and expensive equipment.[16](#jcla23005-bib-0016){ref-type="ref"}
Among T‐ARMS‐PCR, PCR‐RFLP, quantitative PCR, and direct sequencing, it was shown that T‐ARMS‐PCR obtains the most favorable cost‐benefit ratio and is reliable, simple, and fast. [17](#jcla23005-bib-0017){ref-type="ref"} In this study, we introduced a simple T‐ARMS‐PCR assay to detect *CYP2C19\*17*.
2. MATERIALS AND METHODS {#jcla23005-sec-0006}
========================
2.1. DNA extraction {#jcla23005-sec-0007}
-------------------
The samples have been described in our previous study.[18](#jcla23005-bib-0018){ref-type="ref"} Oral swab samples were provided by 93 students at Nanchang University, and a salting‐out method was used to extract genomic DNA. After DNA quantification by a spectrophotometer (Nanodrop2000, Thermo Fisher Scientific), each sample was diluted to 10 ng/μL and kept at −20°C. The study was approved by the Medical Ethics Committee of the Second Affiliated Hospital of Nanchang University, and all the participants signed informed consent.
2.2. Gene synthesis {#jcla23005-sec-0008}
-------------------
A DNA fragment containing *CYP2C19\*17* (−1072 to −611, 462 bp) was synthesized and inserted into the Smal cloning site of the pUC57 plasmid. According to the DNA quantification results, the plasmid sample was diluted to 0.1 ng/μL for PCR amplification.
2.3. T‐ARMS‐PCR {#jcla23005-sec-0009}
---------------
Primers FO, RI, and RO in Table [1](#jcla23005-tbl-0001){ref-type="table"} were designed by PRIMER1 online software,[15](#jcla23005-bib-0015){ref-type="ref"} and primer FI was from Scott et al[13](#jcla23005-bib-0013){ref-type="ref"} T‐ARMS‐PCR was performed in a total volume of 10 μL, including 5 μL of 2 × Taq master mix, 0.4 μM of primer FO, 0.4 μM of primer RI, 0.8 μM of primer FI, 0.4 μM of primer RO, 1 μL of DNA template (including genomic DNA and plasmid DNA), and 2 μL of ddH~2~O. The PCR conditions started with the initial denaturation at 94℃ for 3 minutes, 35 cycles of 94℃ for 20 seconds, 56℃ for 20 seconds, and 72℃ for 20 seconds, followed by a final extension at 72℃ for 5 minutes. The PCR products were visualized by 3% agarose gel electrophoresis at 150 V for 30 minutes.
######
Primers used in T‐ARMS‐PCR of *CYP2C19\*17*
Primer Sequence[a](#jcla23005-note-0002){ref-type="fn"} (5′‐3′) Product size
-------------------------------------------- ---------------------------------------------------------- -----------------
FO GAGATCAGCTCTTCCTTCAGTTACAC Common 462 bp
RO CACCTTTACCATTTAACCCCCTAAAAA
FI[b](#jcla23005-note-0003){ref-type="fn"} TTTTTCAAATTTGTGTCTTCTGTTCTCAAA***T***T T allele 227 bp
RI GCGCATTATCTCTTACATCAGAG***C***TG C allele 292 bp
Abbreviations: FI, forward primer; FO, forward outer primer; RI, reverse inner primer; RO, reverse outer primer.
Specificity‐enhancing mismatches are shown in bold italics.
Primer FI was from Scott et al.[13](#jcla23005-bib-0013){ref-type="ref"}
John Wiley & Sons, Ltd
2.4. DNA sequencing validation {#jcla23005-sec-0010}
------------------------------
The results of T‐ARMS‐PCR were validated by DNA sequencing. Six samples showing the CT genotype or CC genotype by T‐ARMS‐PCR were amplified and sent to a company (Sangon) for sequencing. The PCR (10 μL) contained 5 μL of 2 × Taq master mix, 0.4 μM of primer FO, 0.4 μM of primer RO, 0.5 μL of DNA template, and 3.7 μL of ddH~2~O. The thermal conditions were the same as T‐ARMS‐PCR.
3. RESULTS {#jcla23005-sec-0011}
==========
A representative T‐ARMS‐PCR electrophoretogram of *CYP2C19\*17* is shown in Figure [1](#jcla23005-fig-0001){ref-type="fig"}. The CT genotype showed three bands: 227 bp, 292 bp, and 462 bp. The CC genotype showed two bands: 292 bp and 462 bp. The artificially synthesized plasmid containing the TT genotype showed two bands: 227 bp and 462 bp. Among the 93 samples, 91 samples (97.85%) were the CC genotype, and two samples (2.15%) were the CT genotype. The allele frequencies were 98.92% and 1.08% for the C allele and T allele, respectively. The distribution of genotypes was in Hardy‐Weinberg equilibrium (*χ* ^2^ = 0.0110; *P* = .9165).
{#jcla23005-fig-0001}
The results of DNA sequencing are shown in Figure [2](#jcla23005-fig-0002){ref-type="fig"}. The two samples showing three bands by T‐ARMS‐PCR were the CT genotype, and the other four samples showing 292 bp and 462 bp by T‐ARMS‐PCR were the CC genotype. The results of T‐ARMS‐PCR were completely concordant with the DNA sequencing results.
{#jcla23005-fig-0002}
4. DISCUSSION {#jcla23005-sec-0012}
=============
In addition to the important role of predicting the clinical outcome of drugs metabolized by CYP2C19, *CYP2C19\*17* is also associated with a decreased risk of some diseases, such as breast cancer and Behcet\'s disease.[19](#jcla23005-bib-0019){ref-type="ref"}, [20](#jcla23005-bib-0020){ref-type="ref"} Thus, it is necessary to establish a simple and reliable method to detect this polymorphism.
Because of its reliability, simplicity, and low cost, T‐ARMS‐PCR has been widely used to detect SNPs and mutations in recent years.[21](#jcla23005-bib-0021){ref-type="ref"}, [22](#jcla23005-bib-0022){ref-type="ref"}, [23](#jcla23005-bib-0023){ref-type="ref"}, [24](#jcla23005-bib-0024){ref-type="ref"}, [25](#jcla23005-bib-0025){ref-type="ref"}, [26](#jcla23005-bib-0026){ref-type="ref"}, [27](#jcla23005-bib-0027){ref-type="ref"} Using the four T‐ARMS‐PCR primers designed by PRIMER1 software, we found that the inner primer of the T allele did not have specificity (data not shown) because all the tested samples had the corresponding band of the T allele, but the T allele was rare in Chinese subjects.[28](#jcla23005-bib-0028){ref-type="ref"} After searching the literature, we found that Scott et al developed an ASP‐PCR method to detect the *CYP2C19\*17* allele.[13](#jcla23005-bib-0013){ref-type="ref"} We validated the specificity of their T allele primer, and then, the primer was used as our T allele primer (primer FI). We also did not know whether the C allele primer (primer RI) designed by PRIMER1 software was specific because all the tested samples had the corresponding band of the C allele. To prove the specificity of the C allele primer, we synthesized a DNA fragment of the TT genotype. The position of the synthesized DNA fragment in the *CYP2C19* gene was the same as the PCR product of the two outer primers. We did not obtain a PCR product when the synthesized DNA fragment was amplified by primer FO and primer RI, proving that the two primers had the specificity to amplify the C allele. Then, the four primers were added in a single tube at different ratios to obtain the optimal result. As shown in Figure [1](#jcla23005-fig-0001){ref-type="fig"}, the 227 bp and 292 bp specific fragments of the CT genotype had similar luminance, which suggested that the PCR system and conditions were appropriately optimized. We found that the common band of the CC genotype was absent when we used a specific production lot of 2 × Taq master mix, but the 292 bp specific fragment was not affected. This is consistent with the results of Medrano et al and our previous studies.[28](#jcla23005-bib-0028){ref-type="ref"}, [29](#jcla23005-bib-0029){ref-type="ref"} In this study, no TT genotype sample was found in the 93 samples. The genotype and allele frequencies of *CYP2C19\*17* in our study were very similar to those of two Chinese populations.[15](#jcla23005-bib-0015){ref-type="ref"}, [30](#jcla23005-bib-0030){ref-type="ref"} The results of T‐ARMS‐PCR and DNA sequencing were completely consistent, which further confirmed the reliability of the T‐ARMS‐PCR for *CYP2C19\*17* genotyping.
Compared with PCR‐RFLP, T‐ARMS‐PCR does not need a restriction enzyme, so the cost of T‐ARMS‐PCR is lower than that of PCR‐RFLP. T‐ARMS‐PCR is also cheaper than low‐cost ASP‐PCR. T‐ARMS‐PCR runs PCR in a single tube, while ASP‐PCR requires two reaction tubes, so ASP‐PCR needs more reagents. Compared with the TaqMan assay, HRM, pyrosequencing, MassARRAY, and DNA direct sequencing, T‐ARMS‐PCR does not require special equipment, so it is particularly suitable for use in common laboratories without expensive equipment.
In conclusion, T‐ARMS‐PCR is a reliable, simple, and low‐cost method for detecting *CYP2C19\*17*. It offers an effective screening method for personal medication on *CYP2C19\*17*.
CONFLICT OF INTEREST {#jcla23005-sec-0013}
====================
The authors declare that there are no conflicts of interest.
We thank the students who provided the DNA samples for this study. This work was supported by the Science Research Training Program of Nanchang University (Grant No. 2017268).
| {
"pile_set_name": "PubMed Central"
} |
1. Introduction {#sec1}
===============
Charcot-Marie-Tooth (CMT), also known as hereditary motor and sensory neuropathy, is classified into two major categories: type 1 and type 2 based on the value of motor median nerve conduction velocity (NCV) \[[@B1]\]. CMT1, also known as the demyelinating form of CMT, is clinically characterized by slow NCVs due to myelin sheath abnormalities. CMT2 is as an axonal degeneration and is characterized by the reduction of the amplitudes of motor and sensory nerve action potentials with relatively normal conduction velocities \[[@B2]\]. The most common subtype of CMT2 is CMT2A, an autosomal dominant axonal degeneration of motor and sensory nerves caused by mutations in the mitochondrial mitofusin-2 (*MFN2*) gene (MIM 608507) \[[@B3]\]. Mutations in the *MFN2* gene are the most common cause for CMT2 \[[@B2]--[@B6]\]. *MFN2* is involved in mitochondrial fusion and the maintenance of mitochondrial morphology \[[@B7]--[@B9]\]. Numerous studies have reported mutations in the *MFN2* gene; the majority are point mutations \[[@B10]\]. We have identified the single base change c.2113G \> A changing valine to isoleucine (V705I) of the *MFN2* gene in a patient from a multigenerational Australian family (CMT105) with CMT2 and pyramidal signs. The V705I variant of *MFN2* has been previously reported as a disease causing mutation in single individuals with CMT2 \[[@B11], [@B12]\]. To determine if this variant causes CMT2 in the family, we performed segregation analysis and tested the variant in a cohort of ethnically matched controls of English descent.
2. Methods {#sec2}
==========
2.1. Subjects {#sec2.1}
-------------
Thirty-two individuals from a large, multigeneration Australian CMT2 family (CMT105) were tested ([Figure 1](#fig1){ref-type="fig"}). Sixteen individuals were clinically affected. Informed consent was obtained from all participants according to protocols approved by the Sydney Local Health District Human Ethics Committee. Genomic DNA was extracted from peripheral blood using standard methods by the Molecular Medicine Laboratory, Concord Hospital, Australia.
2.2. High Resolution Melt (HRM) Analysis of *MFN2* {#sec2.2}
--------------------------------------------------
The coding regions of the *MFN2* gene were screened in the index patient IV-6 ([Figure 1](#fig1){ref-type="fig"}). The reference sequence for *MFN2* (NM_014874) was obtained from the National Centre for Biotechnology Information (<http://www.ncbi.nlm.nih.gov/>). Primer3 was used to design primers to amplify the coding exons and intron-exon boundaries of the *MFN2* gene. Primer sequences are available on request. DNA samples were amplified in a total volume of 10 *μ*L containing 10 ng genomic DNA, 1X HRM Master Mix (Idaho Technology), 4 pmol of each primer, and 1X PCR Enhancer (Invitrogen). A PCR Enhancer was not required to amplify exons 8, 9, 10, 13, and 16. DNA samples were amplified using the Eppendorf Mastercycler pro Thermal Cycler. PCR conditions were as follows: initial denaturation of 95°C for 15 min followed by 35 cycles of 95°C for 30 s, appropriate annealing temperature for 30 s, and 68°C for 40 s, with a final extension of 68°C for 5 min. Amplicons were scanned by high resolution melt (HRM) analysis on a 96-well Light Scanner (Idaho Technology). Melt curve profiles were analysed using Light Scanner Call-IT 2.0 software (Version 2.0.0.1331) as described previously \[[@B13]\]. Patient samples showing differential melt curve were sequenced using BigDye Terminator Cycle Sequencing at the Australian Cancer Research Foundation Facility, Garvan Institute of Medical Research, Australia. A sequence analysis was performed using SeqMan II version 5.03 (DNASTAR Inc.).
2.3. Segregation Analysis {#sec2.3}
-------------------------
Exon 18 was screened through additional family members by HRM analysis. Thirty-two individuals from the family were tested for segregation and sixteen individuals were clinically affected with CMT2.
2.4. Control Testing {#sec2.4}
--------------------
One hundred and fourteen healthy chromosomes from 57 ethnically matched unrelated controls (English descent) were analysed for the presence of the V705I variant. Controls were nonconsanguineous spouses of other research subjects who had no signs of neuropathy. A PCR product containing V705I region was cut by *Aat* II restriction enzyme, which recognises GACGT. This site was erased by the V705I variant. A PCR amplicon was digested in a total volume of 20 *μ*L according to the conditions recommended by the manufacture. The products were separated on 2% of agarose gel.
3. Results {#sec3}
==========
3.1. Clinical Findings {#sec3.1}
----------------------
Detailed clinical features of this large Australian family with CMT2 and pyramidal signs have been previously published \[[@B14], [@B15]\]. The family pedigree ([Figure 1](#fig1){ref-type="fig"}) shows autosomal dominant inheritance with the disease phenotype segregating in five generations. Affected individuals in the family showed a reduction of compound amplitudes of motor and sensory nerve action potentials. Nerve conduction velocities were within the normal range. The disease was slowly progressive. Two individuals (IV-6 and III-2) were wheelchair bound with severe CMT2. Electrophysiological findings of these two patients are shown ([Table 1](#tab1){ref-type="table"}).
3.2. High Resolution Melt (HRM) Analysis {#sec3.2}
----------------------------------------
The *MFN2* gene was screened by HRM analysis in the index patient (IV-6) from CMT105 and a differential melt curve for exon 18 was observed when compared with the melt curves of control individuals ([Figure 2](#fig2){ref-type="fig"}). Dideoxy sequence analysis identified the following nucleotide transition c.2113G \> A (V705I). We tested for segregation of the variant in the family. A unique differential melt curve corresponded to individual IV-6 was obtained while the melt curves of other affected individuals tightly grouped together with nonaffected family members and controls ([Figure 2](#fig2){ref-type="fig"}). This demonstrated that the V705I variant was not present in other affected family members and, therefore, the V705I was not segregating with the disease phenotype. To confirm the HRM findings, the index patient IV-6, two additional affected individuals, and a healthy individual from the family were sequenced. The analysis showed that the heterozygous V705I variant was present only in the index patient IV-6 while it was absent in other two affected individuals and the normal individual ([Figure 3](#fig3){ref-type="fig"}). No homozygous V705I variant was identified. We examined the frequency of the V705I in 57 healthy and ethnically matched controls of English descent and identified the heterozygous variant in five control samples (5/57) with an allele frequency of 4.4%.
4. Discussion {#sec4}
=============
The *MFN2* V705I variant was found in an index patient (IV-6) from a large kindred with CMT2 and pyramidal signs. No other known or novel CMT mutations were found in the family. The V705I variant did not segregate with the disease in the family. The mother (III-20) and brother (IV-7) of the index patient are clinically affected but the variant was absent in both cases, indicating that the V705I variant was probably inherited from the father who was said to be unaffected and, unfortunately, was not available for testing. Furthermore, another individual (III-2) had similar clinical severity to the index patient ([Table 1](#tab1){ref-type="table"}) and did not carry the V705I variant. This suggests that the V705I variant has no role in modifying the disease phenotype.
To test the frequency of this variant in the population, we screened 57 unrelated controls (114 chromosomes) and detected the heterozygous V705I variant in 4.4% of these subjects. Our data, therefore, suggests that the variant, in its heterozygous state, is not a disease-causing mutation as previously reported \[[@B11]\].
Compound heterozygous mutations in *MFN2* have been identified in early onset CMT2 where *MFN2* mutations are not pathogenic unless coinherited with another *MFN2* mutation \[[@B16]--[@B18]\]. However, the *MFN2* gene was excluded in our family by linkage analysis \[[@B15]\]. Until a homozygous V705I variant is identified, the pathogenicity of the homozygous state will remain unknown and will require further studies if identified.
The *MFN2* variant, V705I, was first reported as a pathogenic mutation in 2006 \[[@B11]\] in a proband clinically diagnosed with CMT2. This change leads to the exchange of the nonpolar valine to the nonpolar isoleucine at position 705 (c.2113G \> A) (Figures [2](#fig2){ref-type="fig"} and [3](#fig3){ref-type="fig"}). In the initial report, the variant was not detected in 212 ethnically matched control chromosomes from Norway. DNA samples from family members of the proband in the study were not available and therefore segregation of the variant has not been tested. The V705I variant has also been reported in two probands from unrelated Norwegian patients with CMT2 \[[@B12]\]. However, segregation of the variant in the families was not reported.
Our thirty-two family members allowed further validation studies of the variant. The current dbSNP (Build 137, Jun 2012) database (<http://www.ncbi.nlm.nih.gov/projects/SNP/>) showed 690 SNP variants in the *MFN2* gene. The Inherited Peripheral Neuropathies Mutation database (<http://www.molgen.ua.ac.be/CMTMutations/Home/Default.cfm>) shows about 46 pathogenic mutations (6.3%). Therefore, any newly found variant is more likely to be a polymorphism than pathogenic mutation.
Proof that a DNA variant is a disease-causing mutation requires rigorous validation when reporting novel sequence alterations \[[@B19]--[@B21]\]. This approach has been shown to be necessary in other diseases. The P1148A substitution in *FBN1* gene was initially thought to be a pathogenic mutation causing Marfan syndrome in the majority of patients of Asian descent \[[@B22]\]. This variant was subsequently described as a common polymorphism in Asian populations \[[@B23], [@B24]\].
We have now shown that the heterozygous *MFN2* V705I variant does not segregate with the disease proving that it is not a disease-causing mutation in this family. The identification of V705I in 5/57 ethnically matched normal controls establishes that in the heterozygote state this is not pathogenic. Our observation is supported by a recent report in dbSNP 135 (rs142271930), where the V705I variant was found in normal controls with minor allele frequency of 0.4% indicating that it is a rare variant in that population, whereas the frequency in the Australian population may be higher. Furthermore, our findings emphasise the importance of segregation studies and the use of many healthy controls from a variety of ethnic groups when describing novel, potentially pathogenic mutations.
5. Conclusion {#sec5}
=============
We have shown that the heterozygous *MFN2* V705I variant is a polymorphism and not a disease-causing mutation in our family. We have also previously excluded *MFN2* by linkage analysis. As this was previously reported as a disease-causing mutation, our study highlights the importance of variant validation by segregation studies and genotyping in ethnically matched controls. Further studies are needed to investigate the pathogenicity of homozygous *MFN2* V705I variant if identified.
Conflict of Interests
=====================
All authors have no financial relation with the commercial identity mentioned in the paper. There is no conflict of interests for any of the authors.
{#fig1}
{#fig2}
{#fig3}
######
Clinical motor and sensory electrophysiology study performed in the index patient IV-6 and individual III-2.
Patient Age Motor Sensory
--------- ------ ------- --------- ----- ---- --- ---- --- ----
IV-6 41 y 1.5 49 1.2 59 0 NO 0 NO
III-2 66 y --- 56 --- 58 0 0 0 0
MAP: motor action potential (mV), CV: conduction velocity (m/s), SAP: sensory action potential (*μ*V), NO: not obtainable, and (---): not done. Normal motor values were as follows: median MAP \> 4.2 mV, median CV \> 49 m/s, ulnar MAP \> 5.6 mV, and ulnar CV \> 47 m/s. Normal sensory values were as follows: median SAP \> 9.0 *μ*V, median sensory CV \> 56 m/s, ulnar SAP \> 8 *μ*V, and ulnar sensory CV \> 55 m/s.
[^1]: Academic Editor: Eng King Tan
| {
"pile_set_name": "PubMed Central"
} |
Introduction {#sec1_1}
============
Laser retinal photocoagulation is the gold standard treatment for diabetic retinopathy. This treatment partially destroys the outer retina, reducing oxygen demand and the production of inflammatory cytokines. This increases vitreous oxygenation and induces vasoconstriction. Recently, pattern scan laser photocoagulation (PASCAL) has been introduced, enabling short-pulse-duration retinal photocoagulation in many locations simultaneously \[[@B1], [@B2], [@B3], [@B4], [@B5]\]. However, a previous report on approximately 1,300 cases that underwent PASCAL showed that complications, although rare, were varied, including 17 cases of retinal bleeding (1.3%), 2 cases of choroidal detachment (0.15%), and 1 case of exudative retinal detachment (ERD) (0.07%) \[[@B2]\]. Though ERD was the rarest complication, it can severely impair visual function, and its characteristics and clinical course remain unclear. Here, we describe 3 cases in which bullous ERD developed after PASCAL in diabetic retinopathy. There have not been any other published reports on such an unusual case series.
Case Report {#sec1_2}
===========
The first patient (case 1; Table [1](#T1){ref-type="table"}; Fig. [1](#F1){ref-type="fig"}, [2](#F2){ref-type="fig"}) was a 43-year-old man with untreated type 2 diabetes (HbA1c: 11%). He was introduced to our department 11 months before this report and underwent intravitreal bevacizumab in both eyes, followed by pan-retinal laser photocoagulation. At the time of this report, the patient underwent additional treatment with 1,836 PASCAL laser shots for a remaining nonperfusion area (NPA) in his left eye (wave length: 532 nm, diameter: 200 μm, duration: 20 ms, power: 400∼450 mW). Three days after the treatment, ERD occurred and decimal visual acuity (VA) decreased from 0.4 to counting fingers. Unfortunately, though the ERD resolved spontaneously 1 month later, VA remained at 0.03.
The second patient (case 2; Table [1](#T1){ref-type="table"}; Fig. [1](#F1){ref-type="fig"}, [2](#F2){ref-type="fig"}) was a 37-year-old woman with untreated type 2 diabetes (HbA1c: 14%). She was introduced to our department 19 months before this report and underwent pan-retinal laser photocoagulation in both eyes. Her right eye had previously shown mild posterior synechia, and she had undergone cataract surgery and subsequent vitreous surgery with endolaser photocoagulation for vitreous hemorrhage 11 months before this report. At the time of this report, she underwent additional treatment with 1,056 PASCAL laser shots for a remaining NPA in her right eye (wave length: 532 nm, diameter: 200 μm, duration: 20 ms, power: 400∼450 mW). Two days after the treatment, ERD occurred and decimal VA decreased from 0.6 to 0.04. Steroid treatment was considered but was not administered because of a quick (within 3 days) spontaneous reduction of the retinal detachment. Fortunately, the ERD resolved spontaneously 2 months later, and VA recovered to 0.8.
The final patient (case 3; Table [1](#T1){ref-type="table"}; Fig. [1](#F1){ref-type="fig"}, [2](#F2){ref-type="fig"}) was a 71-year-old man with untreated type 2 diabetes (HbA1c: 7.4%). He was introduced to our department 4 months before this report, presenting with vitreous hemorrhage in his right eye. The right eye had undergone vitreous surgery 2 months previously. At the time of this study, the patient underwent additional treatment with 3,521 PASCAL laser shots for a remaining NPA in his right eye (wave length: 561 nm, diameter: 200 μm, duration: 200 ms, power: 270∼400 mW). Three days after the treatment, ERD occurred and decimal VA decreased from 1.0 to 0.02. Fortunately, the ERD resolved spontaneously 3 weeks later, and VA recovered to 0.6.
All patients presented without any signs of inflammation, although mild posterior synechia was present in case 2 on the first visit. Furthermore, no patients had any history of uveitis or rheumatoid disorders.
Discussion {#sec1_3}
==========
Though the reason why ERD occurred after photocoagulation remains unclear, ERD is generally thought to be caused by disruption of the blood-ocular barrier due to intraocular inflammation and endothelial capillary damage in the retina. Thus, blood-retinal barrier breakdown due to retinal pigment cell impairment after retinal PASCAL treatment is a possible cause of bullous ERD in this situation. ERD after photocoagulation is very rare, and, thus, its clinical features have not been elucidated \[[@B2]\]. This report is, therefore, the first to confirm, to some extent, details of the clinical course of ERD after photocoagulation and background characteristics related to the condition. Though ERD has been reported to occur in young male patients with poor glycemic control \[[@B2]\], the current report shows that ERD can occur regardless of gender, age, or glycemic control. Case 3 had relatively good glycemic control (HbA1c: 7.4%) and underwent more than 3,000 PASCAL shots before ERD occurred. Thus, ERD may occur even after a moderate number of laser shots in patients with well-controlled diabetes. The clinical course of the present case series had a number of interesting aspects: ERD resolved spontaneously in all 3 cases; case 2 had good visual recovery despite having the longest duration of ERD (2.5 months); and cases 2 and 3 had good visual recovery under postoperative avitreous status before PASCAL. Case 1, with poor visual recovery, had no history of vitrectomy before PASCAL, suggesting that the vitreous gel may act as a reservoir of photocoagulation-induced inflammatory cytokines that cause irreversible retinal cell damage. Thus, this case series highlights 2 key points: first, ERD can occur regardless of gender, age, glycemic control, or vitreous status and despite a moderate number of laser shots, even with PASCAL; second, ERD in nonvitrectomized eyes may cause irreversible visual loss, even if the ERD resolves within 1 month.
Statement of Ethics {#sec1_4}
===================
Informed consent for the treatment was obtained from the patients. The procedures conformed to the tenets of the Declaration of Helsinki, and the study was approved by the institutional review board of Tohoku University Graduate School of Medicine.
Disclosure Statement {#sec1_5}
====================
None of the authors have any potential conflicts of interest to disclose.
Funding Sources {#sec1_6}
===============
This paper was supported in part by a JST grant from JSPS KAKENHI Grants-in-Aid for Scientific Research (C) (H.K.: 26462629). The funders had no role in the design or conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; or the decision to submit the manuscript for publication.
Author Contributions {#sec1_7}
====================
Data collection, management, and interpretation of the data (S.N., H.K., N.A., and T.N.), drafting of the manuscript (H.K.), and review or approval of the manuscript (H.K. and T.N.). The principal investigators, Dr. Shohei Nishikawa and Dr. Naoko Aizawa, had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.
{#F1}
{#F2}
######
Characteristics of eyes with exudative retinal detachment after retinal pattern scan laser photocoagulation in diabetic retinopathy
Case Age, years Sex Diagnosis HbA1c, % Vitrectomy history Decimal BCVA FT, µm PC spots, *n* Extent of ERD, optic disc area Duration of ERD, days
------ ------------ ----- ----------- ---------- -------------------- -------------- -------- --------------- -------------------------------- ----------------------- ------- ---------- ----
1 43 m PDR 11.0 − 0.4 CF 0.03 207 183 1,836 3 31
2 37 f PDR 14.0 \+ 0.6 0.04 0.8 222 178 1,056 total RD 73
3 71 m PDR 7.4 \+ 1.0 0.02 0.6 292 263 3,521 6 21
BCVA, best-corrected visual acuity; FT, foveal thickness; ERD, exudative retinal detachment; PC, photocoagulation; PDR, proliferative diabetic retinopathy; CF, counting fingers; RD, retinal detachment.
| {
"pile_set_name": "PubMed Central"
} |
The SHRSP represents a suitable animal model for the investigation of the etiopathogenetic basis of hypertensive target organ damage.^[@bib1]^ Feeding SHRSP with JD accelerates both renal and cerebrovascular damage occurrence^[@bib2],\ [@bib3]^ with renal damage preceding stroke.^[@bib2],\ [@bib4],\ [@bib5]^ The gene encoding UCP2 maps nearby the lod score peak of *STR1*/stroke QTL identified on rat chromosome 1 in the SHRSP.^[@bib3]^ UCP2 is a inner mitochondrial membrane protein that exerts an antioxidant effect in various tissues by regulating fatty acid oxidation, mitochondrial biogenesis, substrate utilization and ROS elimination,^[@bib6]^ and is regulated by PPAR*α*.^[@bib7]^ The latter, a member of nuclear receptor family of ligand-activated transcription factors, is known to regulate lipid and energy metabolism through the uncoupling proteins;^[@bib7]^ it also exerts anti-inflammatory and antioxidant effects in many cell types, including cardiovascular cells.^[@bib8]^
UCP2 downregulation associates with increased oxidative stress, atherosclerosis, vascular damage and shorter lifespan in mice.^[@bib9],\ [@bib10],\ [@bib11],\ [@bib12]^ UCP2 overexpression significantly prevented ROS production in endothelial cells and preserved endothelial function by reducing ROS levels.^[@bib13],\ [@bib14]^ Consistently with its ability to decrease endogenous mitochondrial ROS production and to maintain normal mitochondrial membrane potential and ATP levels, a neuroprotective effect of UCP2 has been previously described both *in vitro* and *in vivo*.^[@bib15],\ [@bib16],\ [@bib17],\ [@bib18],\ [@bib19],\ [@bib20],\ [@bib21]^ We previously reported an age-related spontaneous decrease of UCP2 gene and protein expression only in the brain of SHRSP, preceding spontaneous stroke occurrence at 1 year of age.^[@bib22]^
Of interest, we have shown that, in association with increased renal injury, JD significantly downregulates UCP2 gene and protein expression in the kidneys of SHRSP, but not of SHRSR.^[@bib23]^ Consistent findings were obtained in the kidneys of SHRSR/SHRSP-derived stroke congenic lines, depending on the genetic configuration of the transferred *UCP2*.^[@bib24]^ *In vitro*, *UCP2* silencing in renal mesangial cells led to increased inflammation, oxidative stress and cell mortality.^[@bib23]^ Exposure of primary renal proximal tubular epithelial cells isolated from SHRSP to high-NaCl medium led to UCP2 downregulation and reduced viability, which was rescued by recombinant UCP2.^[@bib24]^ Moreover, the PPAR*α*-mediated upregulation of UCP2 gene and protein expression by BO sprouts juice, administered along with JD, completely prevented renal damage occurrence in SHRSP.^[@bib25]^ As expected, the selective inhibition of PPAR*α* reduced the beneficial effects of BO on the renal injury of this strain.^[@bib25]^
Notably, fenofibrate, a compound that exerts renal and neuroprotection in various experimental settings through its impact on several antioxidant enzymes,^[@bib26],\ [@bib27],\ [@bib28]^ and that is also known to stimulate PPAR*α* and UCP2 expression,^[@bib29]^ promoted protection from target organ damage in SHRSP.^[@bib30]^
Based on the above-mentioned observations, the aims of the present study were: (1) to assess for the first time the modulation of *UCP2* in the brain of high-salt-fed SHRSP *versus* SHRSR, as well as in two SHRSR/SHRSP-*STR1*/QTL stroke congenic lines; (2) to explore the impact of PPAR*α* and *UCP2* expression modulation by BO and fenofibrate on the stroke susceptibility of high-salt-fed SHRSP; and (3) to explain part of the mechanisms underlying brain *UCP2* downregulation upon JD in the stroke-prone strain.
Results
=======
Impact of 4 weeks JD feeding on brain *UCP2* expression and related inflammatory and oxidative stress parameters in the four rat lines
--------------------------------------------------------------------------------------------------------------------------------------
Four weeks of JD feeding induced a significant UCP2 gene and protein expression downregulation only in the SHRSP brain ([Figures 1a--c](#fig1){ref-type="fig"}), as previously reported in the kidneys.^[@bib23],\ [@bib24],\ [@bib25]^ The (SHRsp.SHRsr-(D1Rat134-Mt1pa)) congenic line, derived from the SHRSP parental strain and carrying the SHRSR/*STR1* chromosomal fragment, did not downregulate *UCP2* under JD, differently from the SHRSP strain of origin ([Figures 1d and e](#fig1){ref-type="fig"}). Vice versa, the (SHRsr.SHRsp-(D1Rat134-Mt1pa)) congenic line, derived from the SHRSR parental strain and carrying the SHRSP/*STR1* chromosomal fragment, significantly downregulated *UCP2* under JD, differently from the SHRSR strain of origin ([Figures 1f and g](#fig1){ref-type="fig"}). These results confirmed the key role of *UCP2* configuration (SP or SR) for the response to high-salt diet.
[Figure 2](#fig2){ref-type="fig"} shows the NF-*κ*B protein expression level, a marker of inflammation, and the carbonylated protein level, a marker of oxidative stress, in the brains of the parental lines (SHRSR: panels a and b; SHRSP: panels c and d). Both inflammatory and oxidative stress markers were significantly increased only in the brains of JD-fed SHRSP. [Figure 3](#fig3){ref-type="fig"} shows the same parameters in the two *STR1*/QTL stroke congenic lines, the one derived from the SHRSP (panels a and b), and the one derived from the SHRSR (panels c and d). Both inflammatory and oxidative stress markers were decreased in the SHRSP-derived stroke congenic line, carrying the SHRSR/*STR1* chromosomal fragment (panels a and b), whereas these markers increased significantly in the SHRSR-derived stroke congenic line carrying the SHRSP/*STR1* chromosomal fragment (panels c and d). These results confirmed that, whenever *UCP2* expression was downregulated, such as in JD-fed SHRSP and JD-fed (SHRsr.SHRsp-(D1Rat134-Mt1pa)), the degree of inflammation and of oxidative stress increased. Vice versa, no increase of these processes was detected in the brains of both JD-fed SHRSR and JD-fed (SHRsp.SHRsr-(D1Rat134-Mt1pa)), both carrying higher levels of brain *UCP2* expression (as compared with the other two lines).
Impact of fenofibrate administration on brain *UCP2* expression and on stroke occurrence in JD-fed SHRSP
--------------------------------------------------------------------------------------------------------
[Figure 4](#fig4){ref-type="fig"} shows the impact of JD plus fenofibrate *versus* JD alone on brain UCP2 gene and protein expression, and on NF-*κ*B and oxidized total protein levels at the end of 4 weeks of treatment. Fenofibrate could restore UCP2 level (panels a and b) and decrease levels of both NF-*κ*B and oxidative stress (panels c--e).
[Figure 5](#fig5){ref-type="fig"} shows the results of the stroke survival study performed with a long-term fenofibrate administration (3 months) in JD-fed SHRSP. The impact on UCP2 gene and protein expression levels in brains of JD plus fenofibrate treated SHRSP, as compared with animals receiving JD only and JD plus vehicle, is shown in the panels a and b of the [Figure 5](#fig5){ref-type="fig"}. Also at the end of 3 months of treatment, the parallel administration of JD and fenofibrate restored UCP2 level (panels a and b), and decreased levels of both NF-*κ*B and oxidative stress despite the long-term treatment with JD (panels c--e). Importantly, fenofibrate fully protected animals from stroke occurrence over 3 months of follow-up (panel f). In contrast, occurrence of stroke events reached 100% by the seventh week of JD in both JD and JD plus vehicle treated rats, consistently with previous evidence.^[@bib3],\ [@bib5]^ SBP and BW values upon fenofibrate administration are reported in the [Supplementary Table S1](#sup1){ref-type="supplementary-material"}.
Impact of BO administration on brain *UCP2* expression and on stroke occurrence in JD-fed SHRSP
-----------------------------------------------------------------------------------------------
[Supplementary Figure S1](#sup1){ref-type="supplementary-material"} shows the impact of JD plus BO *versus* JD alone on brain UCP2 gene and protein expression, NF-*κ*B and oxidized total protein levels at the end of 4 weeks of treatment. BO restored UCP2 level (panels a and b) and decreased levels of both NF-*κ*B and oxidative stress (panels c--e).
[Figure 6](#fig6){ref-type="fig"} shows UCP2 gene and protein expression levels, at different experimental times during the stroke survival study, in brains of JD plus BO treated rats, as compared with animals receiving JD only and JD plus BO plus PPAR*α* inhibitor (panels a and b). As observed at the end of 4 weeks of the combined treatment, the concomitant administration of JD and BO restored UCP2 level, decreased levels of both NF-*κ*B and oxidative stress despite JD ([Figures 6c and d](#fig6){ref-type="fig"} and [Supplementary Figure S2](#sup1){ref-type="supplementary-material"}), and led to a significant delay of stroke occurrence ([Figure 6e](#fig6){ref-type="fig"}). In fact, 40% of rats survived until the 11th week of treatment. The PPAR*α* inhibitor significantly counteracted the stimulatory effect of the BO juice on *UCP2* expression, therefore leading to 100% stroke occurrence by the eighth week of treatment ([Figures 6a--e](#fig6){ref-type="fig"}), consistently with previous findings.^[@bib25]^ The SBP and BW values upon these treatments are reported in the [Supplementary Table S1](#sup1){ref-type="supplementary-material"}.
Analysis of *UCP2*-targeted microRNAs upon JD *versus* RD in brains of SHRSR and SHRSP
--------------------------------------------------------------------------------------
Out of the compared UCP2-targeted microRNAs in the brains of the SHRSR and SHRSP strains upon the two diets, we detected a remarkable differential expression, very consistent with the parallel differential *UCP2* expression, for the rno-microRNA-503. In fact, this miR was remarkably upregulated (\>2 folds) in the brain of JD-fed SHRSP whereas it was significantly downregulated in the brain of JD-fed SHRSR as compared with RD ([Figure 7a](#fig7){ref-type="fig"}). No other miR showed a significant modulation in relation to the observed *UCP2* expression changes. Based on the results of the microRNAs screening, we further explored the modulation of the microRNA-503 expression in our experimental groups. We discovered that SHRSP receiving either fenofibrate or BO along with JD showed a significant reduction of brain miR-503 expression level ([Figures 7b and c](#fig7){ref-type="fig"}). The expected interference by PPAR*α* inhibitor was observed in SHRSP receiving JD and BO ([Figure 7c](#fig7){ref-type="fig"}). Furthermore, we observed a significant downregulation of brain miR-503 level in the JD-fed SHRSP-derived congenic line containing the SHRSR/*STR1* fragment ([Figure 7d](#fig7){ref-type="fig"}), whereas the SHRSR-derived congenic line, containing the SHRSP/*STR1* segment, showed a significant upregulation of miR-503 upon JD ([Figure 7e](#fig7){ref-type="fig"}). Therefore, the data obtained in the two *STR1*/QTL stroke congenic lines reinforced the evidence obtained in the parental lines of origin.
Impact of microRNA-503 overexpression on viability of HUVECs
------------------------------------------------------------
The *in vitro* overexpression of hsa-miR-503 in HUVECs showed a marked *UCP2* suppression with a linear dose--response ([Figures 8a and b](#fig8){ref-type="fig"}). Importantly, at a miR-503 concentration able to turn off *UCP2* expression by 90%, a significant increase of cell mortality and a significant decrease of cell viability were observed ([Figure 8c](#fig8){ref-type="fig"}). The impact on cell viability was comparable to that obtained upon direct *UCP2* silencing in HUVECs ([Figure 8d](#fig8){ref-type="fig"}).
Discussion
==========
Our study demonstrates that UCP2 gene and protein expression levels are significantly downregulated by Japanese style dietary feeding in brains of SHRSP but not in brains of its related control strain, the SHRSR. This phenomenon was associated with increased inflammation and oxidative stress. Accordingly, a SHRSR-derived stroke congenic line, carrying a fragment of the SHRSP-*STR1*/QTL (containing *UCP2*), showed brain *UCP2* downregulation under JD feeding associated with increased inflammation and oxidative stress. Vice versa, brain *UCP2* expression did not decrease, and both inflammation and oxidative stress were reduced upon JD in the reciprocal congenic line. Consistently, the administration of JD plus fenofibrate, known to stimulate *UCP2* expression,^[@bib25]^ restored brain UCP2 levels, reduced oxidative stress and fully protected from stroke occurrence the high-salt fed SHRSP. The administration of BO sprouts extract, another stimulator of *UCP2* expression,^[@bib25]^ was also able to delay significantly stroke occurrence in JD-fed SHRSP. The PPAR*α* inhibitor antagonized the beneficial effect of BO, confirming previous evidence obtained in the kidneys.^[@bib25]^ The greater efficacy of fenofibrate *versus* BO on stroke protection may be explained by the additional molecular and pharmacological properties of the drug.
Of note, NF-*κ*B, which was characterized in the current study mainly for its important role in inflammation, is a ubiquitous transcription factor that, due to its wide range of gene targets, plays several other functions in mammalian cells, particularly in the nervous system.^[@bib31]^
Based on our results, UCP2 appears to play an important role in the high-salt diet-dependent increased susceptibility to cerebrovascular events, as well as it does for the increased susceptibility to kidney damage of SHRSP.^[@bib23],\ [@bib24],\ [@bib25]^ A common molecular mechanism, dependent on *UCP2* suppression, may underlie the vascular damage observed in different organs of high-salt-fed SHRSP.
Uncoupling the proton flux through UCP2 is a critical pathway in the regulation of senescence.^[@bib8],\ [@bib32],\ [@bib33]^ The involvement of UCP2 in vascular diseases is known, being demonstrated in several pathological contexts,^[@bib9],\ [@bib10],\ [@bib11],\ [@bib12],\ [@bib13]^ and it represents the consequence of its key role in the clearance of ROS within the mitochondria. A neuroprotective effect of UCP2 has been previously associated with its role on oxidative stress.^[@bib16],\ [@bib17],\ [@bib18],\ [@bib19],\ [@bib20],\ [@bib21]^ An association of UCP2 with stroke has been described in other experimental settings.^[@bib34],\ [@bib35]^
A major strength of our data relies on the evidence obtained in the reciprocal SHRSR/SHRSP-*STR1/*QTL stroke congenic lines. The stroke phenotype of these lines clearly depends on the genomic configuration of the inserted chromosomal segment belonging to *STR1*, with evidence that the chromosomal fragment carrying *UCP2* significantly interferes with stroke occurrence.^[@bib5]^ Herein, we report that the introgression of the stroke-prone *STR1* chromosomal segment (carrying *UCP2)* within the stroke-resistant genomic background led to a suppression of *UCP2* expression in contrast to the upregulation of the SHRSR strain. The opposite phenomenon was observed in the reciprocal congenic line, supporting the role of the stroke-prone *UCP2* configuration to obtain downregulation in response to high-salt diet. We recently reported similar findings on UCP2 gene and protein expression with regard to renal damage in the same stroke congenic lines.^[@bib24]^
A fundamental demonstration of the role of *UCP2* in stroke predisposition of SHRSP was provided by the significant protective impact of BO and fenofibrate administration, both stimulator of *UCP2* expression,^[@bib25],\ [@bib26]^ toward stroke occurrence despite JD feeding. Of note, our data on fenofibrate as a protective agent toward stroke confirm and extend previous findings obtained in the same animal model.^[@bib30]^
No evidence of *UCP2* mutations between the two strains was obtained in our previous study.^[@bib23]^ In the attempt to unravel, at least in part, some of the mechanisms underlying brain *UCP2* downregulation upon JD in the stroke-prone strain, we searched for mechanisms involved in the translational regulation of *UCP2*. As a result, we found that the *UCP2* expression modulation upon JD in the brains of SHRSP and SHRSR was related to the microRNA-503. The latter turned out to be significantly upregulated in high-salt-fed SHRSP, whereas it was significantly decreased in the SHRSR upon the same diet. Consistent results were obtained in the two congenic lines, further supporting the role of this microRNA in *UCP2* downregulation. Moreover, treatment with both fenofibrate and BO counteracted the increase of brain microRNA-503 level and the suppression of *UCP2* expression in JD-fed SHRSP. Importantly, both treatments, by their ability to restore regular levels of both microRNA-503 and UCP2, significantly protected from stroke occurrence the JD-fed SHRSP. Finally, miR-503 overexpression *in vitro* abolished *UCP2* expression and caused a high degree of cell mortality, consistently with what observed upon direct *UCP2* silencing. Our results strongly suggest that miR-503 is a modulator of brain *UCP2* expression in high-salt-fed SHRSP and also in SHRSR. Of note, miR-503 exerts multiple actions. It is reported as an antiproliferative and antiangiogenetic factor,^[@bib36]^ and a cell cycle regulator; it is involved in cell adhesion, migration and angiogenesis processes.^[@bib37]^ Its upregulation in diabetes strongly associates with vascular damage.^[@bib36],\ [@bib38]^ On the other hand, a decrease of miRNA-503 upon losartan treatment is associated with an improvement of diabetic nephropathy in an animal model of spontaneous type 2 diabetes.^[@bib39]^ Herein, we report the first evidence that an increase of miR-503 associates with high-salt induced stroke occurrence, through its ability to modulate brain *UCP2* expression, in an animal model of spontaneous hypertension and stroke and that, in turn, miR-503 can be decreased by both pharmacological and nutraceutical approaches to obtain protection from stroke. Further studies will address the interaction between PPAR*α* and miR-503 in the *UCP2* regulation. It will be also interesting to characterize the potential contribution of miRNA-503 in the prevention and/or amelioration of hypertensive target organ damage with the available therapeutic antihypertensive strategies.
Our novel findings support the role of microRNAs in the end-organ damage promotion during hypertension.^[@bib40]^ We are also aware, as a limitation of the current study, that the role of other still unknown *UCP2*-targeted brain microRNAs remains to be determined in our experimental conditions.
In summary, we demonstrate that *UCP2* expression downregulation by high-salt diet associates with increased stroke predisposition whereas *UCP2* upregulation, by both nutraceutical and pharmacological agents, associates with a significant stroke protection in high-salt-fed SHRSP. Our data strengthen the role of UCP2 as a suitable therapeutic target for stroke. Notably, the microRNA-503 behaves as a key determinant of the dietary-dependent regulation of *UCP2* expression in the brain of SHRSP. Thus, the microRNA-503 has a significant potential in unraveling the mechanisms underlying stroke pathogenesis and may reveal a promising therapeutic agent for this disease.
Materials and methods
=====================
Effects of 4 weeks JD feeding on brain UCP2 gene and protein expression in SHRSP, SHRSR and in the SHRSR/SHRSP-*STR1*/QTL stroke congenic lines
-----------------------------------------------------------------------------------------------------------------------------------------------
The following rat lines were used for this study: SHRSP, SHRSR, (SHRsp.SHRsr-(D1Rat134-Mt1pa)), (SHRsr.SHRsp-(D1Rat134-Mt1pa)). The latter two lines are congenic lines carrying the (D1Rat134-Mt1pa) chromosomal segment of *STR1* (containing *UCP2*) in the SHRSR configuration within the SHRSP genomic background (SHRsp.SHRsr-(D1Rat134-Mt1pa)) and, viceversa, in the SHRSP configuration within the SHRSR genomic background (SHRsr.SHRsp-(D1Rat134-Mt1pa)). The preparation of the congenic lines has been previously reported.^[@bib5]^ By following our standardized experimental protocol, six-week-old male rats of both parental and congenic lines were fed with either RD or JD for 4 weeks (parental lines, *n*=7 for each line at each treatment; congenic lines, *n*=6 for each line at each treatment). At the end of 4 weeks of the dietary treatment, after SBP and BW measurement, animals were killed by cervical dislocation and brains were removed for molecular analyses. Tissue total RNA and proteins were extracted by following previously described procedures.^[@bib22],\ [@bib23],\ [@bib24],\ [@bib25]^ The UCP2 gene and protein expression levels, the NF-*κ*B protein expression level and carbonylated total proteins level were assessed by previously described procedures.^[@bib22],\ [@bib23],\ [@bib24],\ [@bib25]^
Impact of fenofibrate administration on brain UCP2 gene and protein modulation and on stroke occurrence in JD-fed SHRSP
-----------------------------------------------------------------------------------------------------------------------
In order to fulfill the first aim, male SHRSP received, starting at 6 weeks of age, JD alone (*n*=4), JD plus fenofibrate (150 mg/kg/die, *n*=4) or vehicle (1% CMC, *n*=4) administered via gavage for 4 weeks. In order to analyze the impact of fenofibrate on stroke survival, 6-week-old SHRSP received JD alone (*n*=6), JD plus fenofibrate (*n*=6) or JD plus CMC (*n*=6). They were monitored for SBP, BW and stroke occurrence up to 3 months of the dietary plus fenofibrate treatment.
Brains of JD plus fenofibrate treated SHRSP, taken at the end of both 1 month and 3 months of treatment (*n*=4 and 6, respectively), were analyzed for UCP2 gene and protein expression levels, NF-*κ*B protein expression and oxidative stress levels. These molecular analyses were compared with those obtained in brains of 4-week JD-fed SHRSP from the above experimental setup and of 4-week JD plus CMC treated SHRSP.
Impact of BO juice administration on brain UCP2 gene and protein modulation and on stroke occurrence in JD-fed SHRSP
--------------------------------------------------------------------------------------------------------------------
In order to fulfill the first aim, male SHRSP received, starting at 6 weeks of age, JD alone (*n*=4) or JD plus BO sprouts extract (*n*=4) for 4 weeks by following previously reported procedures.^[@bib25]^ In order to explore the impact of BO on stroke survival, male SHRSP received JD alone (*n*=6), JD plus BO (*n*=10), JD plus BO plus PPAR*α* inhibitor (*n*=6) as previously reported.^[@bib25]^ SBP, BW and stroke occurrence were monitored up to 3 months of the dietary plus BO juice treatment. Brains of SHRSP, taken either at the end of 4 weeks or at the time of stroke occurrence (between the 8th and 12th week of treatment upon BO; between the 7th and 8th week of treatment upon BO plus PPARα inhibitor), were analyzed for UCP2 gene and protein expression levels, NF-*κ*B protein expression and oxidative stress levels. These molecular analyses were compared with those obtained in brains of 4-week JD-fed SHRSP from the above experimental setup.
All rats used for the experiments described in the paragraphs 1--3 were maintained at the animal facility of the Neuromed Institution in strict compliance with the guidelines set forth by the American Physiological Society. Animal protocols were approved by the Institutional Animal Care and Use Committee of the Neuromed Institution. Climate was controlled, and temperature was set at 22 °C. Diurnal 12-h cycles were kept automatically. Animals were housed two or three per cage with free access to RD (containing 22% protein, 2.7 mg/g Na^+^, 7.4 mg/g K^+^, 0.05 mg/g methionine) and tap water, unless stated otherwise. The JD contained 17.5% protein, 3.7 mg/g K^+^ and 0.03 mg/g methionine (Lab. Piccioni, Milan, Italy), and 1% NaCl was added to the drinking water.
Analysis of *UCP2*-targeted microRNAs expression upon JD *versus* RD in brains of SHRSR and SHRSP
-------------------------------------------------------------------------------------------------
Based on the very limited knowledge of the rat *UCP2* brain modulation by targeted microRNAs, we selected conserved predicted UCP2-targeted microRNAs with all miRSVR scores by searching the [www.microrna.org](http://www.microrna.org) public database. The following miRNAs were considered in this study: Rno-microRNA-1, let-7a, let-7b, let-7c, let-7d, let-7i, 16, 24, 27a, 34a, 138, 206, 214, 218, 298, 497, 503. The RT-PCR for each microRNA was performed in triplicate in brain extracts of both parental strains upon the two diets by specific gene expression Taqman assays (Life Technologies). Based on the evidence of a significant microRNA-503 upregulation in the brain of JD-fed SHRSP as opposed to a significant downregulation in the brain of SHRSR (see Results section), the modulation of this miRNA was verified in JD-fed SHRSP upon fenofibrate, vehicle, BO, BO plus PPARα inhibitor administration, as well as in the brains of the two SHRSR/SHRSP-*STR1*/QTL stroke congenic lines (by analyzing the same rats used in the above described experimental groups).
*In vitro* hsa-microRNA-503 overexpression in HUVECs
----------------------------------------------------
In order to verify directly the impact of microRNA-503 on *UCP2* expression levels, we performed a dose--response experiment *in vitro*. For this purpose, HUVECs (Lonza, Cambrex, Belgium) were seeded in 60-mm well plates (2 × 10^5^ cells/well) and cultured in endothelial growth medium-2 (EGM-2; Lonza) to reach a 70--80% confluence. Then, serial concentrations of 12.5, 25, 50, 100, 200 and 400 nM of hsa-microRNA-503 mimic (Mission microRNA; Sigma-Aldrich (Milan, Italy)) were incubated in OPTIMEM reduced serum medium with a nucleic acid transferring agent (lipofectamine RNAiMAX reagent (Invitrogen, Milan, Italy)) in a final volume of 2 ml/well each for 20 min. Five hours later the complex containing medium was replaced with EGM-2 medium supplemented with 10% fetal bovine serum. Cells transfected with RNAiMax lipofectamine complex and mission miRNA negative control (Sigma-Aldrich) were used as control. Twenty-four hours after transfection cells were extracted for total RNA, by the RNazol procedure,^[@bib23]^ and used for the evaluation of both miR-503 and *UCP2* expression levels by RT-PCR. A specific gene expression Taqman assay (Lifetech, Waltham, MA, USA) was used to assess miR-503 levels, as reported above. The RT-PCR of *UCP2* was performed as reported above. Finally, we assessed the impact of miR-503 overexpression at 100 nM concentration (corresponding to 90% reduction of *UCP2* expression) on cell apoptosis, necrosis and viability, as assessed by FACS. The results of FACS were compared with those obtained by performing *UCP2* silencing with a specific siRNA in the same cell line (by following previously reported procedures^[@bib23]^).
Statistical analysis
--------------------
All values are shown as means±S.E.M. Statistical analysis of SBP, BW, RT-PCR, WB densitometric values and FACS values was performed by one-way ANOVA followed by Bonferroni *post hoc* test. Comparisons between two groups were performed using Student\'s *t-*test followed by *post hoc* Mann--Whitney test. Survivor function in rats monitored over JD feeding alone, compared with JD plus the different treatments, was estimated by the life-table method. Log-rank and Wilcoxon statistics were used for testing equality of survivor functions.
Statistical significance was stated at the *P*\<0.05 level. GraphPad Prism (Ver 5.01 GraphPad Software, Inc., La Jolla, CA, USA) statistical software was used for the statistical analysis.
This work was supported by a grant from the Italian Ministry of Health and from the 5‰ grant.
[Supplementary Information](#sup1){ref-type="supplementary-material"} accompanies this paper on Cell Death and Disease website (http://www.nature.com/cddis)
Edited by A Verkhratsky
The authors declare no conflict of interest.
Supplementary Material {#sup1}
======================
######
Click here for additional data file.
{#fig1}
{#fig2}
{ref-type="fig"}b, for the WB of carbonylated total proteins. Bar graphs on the right side represent corresponding densitometric analysis. \*\*\**P*\<0.0001 for JD *versus* RD. (**c**) WB of NF-*κ*B and (**d**) of carbonylated total proteins in the SHRSR-derived stroke congenic line upon RD or JD and corresponding densitometric analysis shown on the right side. See legend of [Figure 2](#fig2){ref-type="fig"}b for the WB of carbonylated total proteins. \*\*\**P*\<0.0001 for JD *versus* RD. (**d**) \*\**P*\<0.001 for JD *versus* RD](cddis2017278f3){#fig3}
{#fig4}
{ref-type="fig"}b, for the WB of carbonylated total proteins. \*\*\**P*\<0.0001 for JD *versus* RD fed SHRSP; \*\**P*\<0.001 for JD plus fenofibrate *versus* JD plus CMC fed SHRSP. (**f**) Stroke survival rate in the JD, JD plus vehicle and JD plus fenofibrate treated SHRSP. The comparison of JD plus fenofibrate treated SHRSP *versus* both JD and JD plus vehicle treated SHRSP was significant, *P*\<0.001](cddis2017278f5){#fig5}
{ref-type="fig"}b, for the WB of carbonylated total proteins. (**e**) Stroke survival rate in the JD, JD plus BO, JD plus BO plus PPAR*α* inhibitor treated SHRSP. The comparison of JD plus BO treated SHRSP *versus* both JD and JD plus BO plus PPAR*α* inhibitor treated SHRSP was significant, *P*\<0.001 and *P*\<0.001, respectively](cddis2017278f6){#fig6}
{ref-type="fig"}, [2](#fig2){ref-type="fig"}, [3](#fig3){ref-type="fig"}, [4](#fig4){ref-type="fig"}, [5](#fig5){ref-type="fig"}, [6](#fig6){ref-type="fig"}. For number of animals see the previous figures](cddis2017278f7){#fig7}
{#fig8}
| {
"pile_set_name": "PubMed Central"
} |
{#sp1 .142}
{#sp2 .143}
| {
"pile_set_name": "PubMed Central"
} |
Introduction {#Sec1}
============
New and more efficient ways to make use the internal energy converted into heat in electronic devices are of crucial importance for a sustainable future^[@CR1],[@CR2]^. In this field, the engineering of magnetostrictive films onto flexible substrates is an exciting area of research with potential to the development of energy-efficient technologies, covering the impacts of the external stress on the quasi-static and dynamic magnetic properties^[@CR3]--[@CR6]^. Specifically, the interplay of magnetostrictive and temperature effects provides us a playground to the investigation of fundamental anisotropic magnetothermoelectric effects, for instance, the spin polarization dependence of the thermoelectric energy^[@CR7]--[@CR10]^.
In magnetic materials, the thermoelectric voltage, defined as the direct conversion of the temperature gradient to electrical voltage, is investigated through thermomagnetic phenomena, such as the spin Seebeck effect (SSE) and the anomalous Nernst effect (ANE)^[@CR11],[@CR12]^. Both of them essentially consist in the application of a magnetic field and a temperature gradient, thus generating an electric field $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$\overrightarrow{E}$$\end{document}$. Specifically, for the anomalous Nernst effect in nanostructured magnetic materials, an electric field $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\overrightarrow{E}}_{ANE}$$\end{document}$ is induced by the interplay of the magnetization $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$\overrightarrow{m}$$\end{document}$ and the temperature gradient ∇*T*, being a result of $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\overrightarrow{E}}_{ANE}\propto \overrightarrow{m}\times \nabla T$$\end{document}$. Thus, given that the corresponding thermoelectric voltage *V*~*ANE*~ in magnetic materials is observable even with small thermal gradients at room temperature, ANE becomes a key effect to optimally control and employ the internal energy dissipated in electronic devices. In the near past, ANE has been extensively investigated, disclosing that *V*~*ANE*~ is strongly dependent on the crystallographic orientation in magnetic materials^[@CR13]^, type of material (for instance, Fe and Ni have ANE coefficient with opposite signs), thickness of the layers in the film geometry^[@CR12],[@CR14]^, and magnetic anisotropy^[@CR15]^. However, despite the recent advances in this field, there still are several aspects related to the anomalous Nernst effect in nanostructured ferromagnetic materials that are not yet fully understood or explored. Among the different investigated ferromagnets, materials with high-spin polarization are the main elected to constitute spintronics devices. In this context, CoFeB alloys in the film geometry arise as one of the most promising candidates for this kind of application due to their magnetic properties, such as high permeability, high saturation magnetization, low coercive field, and well-defined magnetic anisotropy^[@CR16]--[@CR19]^. These features place the CoFeB alloys suitable to the production of, for instance, magnetic tunnel junctions with high magnetoresistance^[@CR20]--[@CR23]^ and spin-transfer torque^[@CR23]^, as well as to the investigation of a sort of phenomena, as spin pumping^[@CR24],[@CR25]^, spin Hall effect^[@CR26]^, and inverse spin Seebeck effect^[@CR27]^. It is known that the magnetic properties of CoFeB films are strongly dependent on the thickness^[@CR28]--[@CR30]^. Films thinner than 5 nm usually exhibit strong perpendicular magnetic anisotropy (PMA)^[@CR28]^, while the thicker ones commonly show in-plane magnetic anisotropy^[@CR29],[@CR30]^. Moreover, due to its magnetostrictive properties, thick-CoFeB films have been grown onto flexible substrates, thus becoming essential systems to investigate the influence of the stress on the effective magnetic anisotropy^[@CR16],[@CR17]^, as well as appearing as promising materials for application in flexible magnetic devices^[@CR31]^, disposable electronics, smart cards, light-emitting diodes, wearable electronics and a broad range of sensors^[@CR18],[@CR32]--[@CR35]^.
To pave the way, looking for effective integration between the thermoelectric voltage experiments and flexible spintronic devices, the anomalous Nernst effect emerges as a powerful tool to explore thermoelectric effects in samples with magnetostrictive properties. In this work, we perform a theoretical and experimental investigation of the magnetic properties and anomalous Nernst effect in a flexible magnetostrictive film with induced uniaxial magnetic anisotropy and under external stress. Specifically, from the theoretical approach, we calculate the magnetization behavior and the thermoelectric voltage response; experimentally, we verify modifications of the effective magnetic anisotropy in stressed samples through thermoelectric voltage measurements, and explore the anomalous Nernst effect as an efficient way to study flexible magnetostrictive film by modifying both, the magnetic field and external stress. We find quantitative agreement between experiment and numerical calculations, thus elucidating the magnetic properties and thermoelectric voltage response. Besides, we provide evidence to confirm the validity of the theoretical approach to describe the magnetic properties and anomalous Nernst effect in ferromagnetic magnetostrictive films having uniaxial magnetic anisotropy and submitted to external stress.
Results {#Sec2}
=======
Theoretical approach {#Sec3}
--------------------
Here, we focus on a ferromagnetic magnetostrictive film with uniaxial magnetic anisotropy, which is submitted to external stress; and we model it as a planar system, as illustrated in Fig. [1](#Fig1){ref-type="fig"}.Figure 1Schematic configuration of our theoretical system--a ferromagnetic magnetostrictive film with uniaxial magnetic anisotropy, which is submitted to external stress. (**a**) Anomalous Nernst effect in a film. In our numerical calculation, while the magnetization $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$\overrightarrow{m}$$\end{document}$ lies in the film plane, the temperature gradient ∇*T* is normal to the plane. The electric field associated with the anomalous Nernst effect is a result of $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\overrightarrow{E}}_{ANE}\propto \overrightarrow{m}\times \nabla T$$\end{document}$, and thus the thermoelectric voltage *V*~*ANE*~ is proportional to a component of $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\overrightarrow{E}}_{ANE}$$\end{document}$, given that its detection is performed with electrical contacts at the ends of the main axis of the film. (**b**) Definitions of the vectors and angles employed in the numerical calculations of magnetization and thermoelectric voltage. Here, we consider the magnetization vector $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$\overrightarrow{m}$$\end{document}$, whose orientation for each given magnetic field value is set by $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\theta }_{m}$$\end{document}$ and $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\varphi }_{m}$$\end{document}$, the equilibrium angles with respect to the *z* and *x* axes (this latter is also represented by the dashed line above the film), respectively. In particular, due to the film geometry and system thickness, here $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\theta }_{m}$$\end{document}$ is constant and equal to 90°. The magnetic field is kept in the film plane, i.e. $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\theta }_{H}={90}^{\circ }$$\end{document}$, although its orientation can be modified by varying *φ*~*H*~ from 0° up to 360°, having as reference the dashed line indicated in the illustration. The unit vector $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\hat{u}}_{k}$$\end{document}$, not shown here, is defined by $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\theta }_{k}$$\end{document}$ and $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\varphi }_{k}$$\end{document}$, and indicates the direction of the uniaxial magnetic anisotropy induced during deposition; $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\hat{u}}_{\sigma }$$\end{document}$ is given by $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\theta }_{\sigma }={90}^{\circ }$$\end{document}$ and $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\phi }_{\sigma }={90}^{\circ }$$\end{document}$, and describes the direction of the magnetoelastic anisotropy induced by the stress applied to the sample during the experiment; and $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\hat{u}}_{eff}$$\end{document}$, not shown here too, is described by $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\theta }_{{u}_{eff}}$$\end{document}$ and $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\phi }_{{u}_{eff}}$$\end{document}$, and represents the orientation of the effective magnetic anisotropy, a result of the competition between the contributions of the induced uniaxial and magnetoelastic anisotropies. The unit vector normal to the film plane, also not shown, is along the *z* direction, i.e. $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$\hat{n}=\hat{k}$$\end{document}$. Finally, for the thermoelectric voltage calculation, the temperature gradient ∇*T* is normal to the film plane. Further, the *V*~*ANE*~ detection is performed with electrical contacts at the ends of the main axis of the film.
In the anomalous Nernst effect in magnetic materials, an electric field $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\overrightarrow{E}}_{ANE}$$\end{document}$ is induced by the interplay of the magnetization $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$\overrightarrow{m}$$\end{document}$ of the sample and a temperature gradient ∇*T*; and the relationship between these quantities can be expressed as$$\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\overrightarrow{E}}_{ANE}=-\,{S}_{N}(\hat{m}\times \nabla T),$$\end{document}$$where $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${S}_{N}={\lambda }_{N}{\mu }_{\circ }{m}_{s}$$\end{document}$; *λ*~*N*~ is the anomalous Nernst coefficient, *μ*~o~ is the vacuum magnetic permeability, and *m*~*s*~ is the saturation magnetization of the ferromagnetic alloy, which is oriented along to the unit vector $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$\hat{m}\mathrm{.}$$\end{document}$
In our theoretical approach, we consider a typical ANE experiment in a film, in which the temperature gradient is normal to the film plane, while the magnetization lies in the plane, as depicted in Fig. [1(a)](#Fig1){ref-type="fig"}. The corresponding thermoelectric voltage *V*~*ANE*~, detected by electrical contacts at the ends of the main axis of the film, is thus given by$$\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${V}_{ANE}=-\,{\int }_{0}^{L}{\overrightarrow{E}}_{ANE}\cdot d\overrightarrow{l},$$\end{document}$$where the integration limits are set by the distance between the electrical contacts, which in our case is *L*. As a result, the measured *V*~*ANE*~ is proportional to the $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\overrightarrow{E}}_{ANE}$$\end{document}$ component along to the direction defined by the contacts, as we can see in Fig. [1(a,b)](#Fig1){ref-type="fig"}.
For films under an out-of-plane temperature gradient, *S*~*N*~ can be experimentally estimated through$$\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${S}_{N}=({V}_{ANE}^{{S}_{max}}\,{t}_{f}))/(L\,\Delta {T}_{f}),$$\end{document}$$where ***t***~*f*~ is the film thickness, *L* is the own distance between the probe electrical contacts in the experiment, $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$\Delta {T}_{f}$$\end{document}$ is the temperature variation across the film (see Methods for details on the $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$\Delta {T}_{f}$$\end{document}$ estimation and its relation with the temperature variation $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$\Delta T$$\end{document}$ measured experimentally across the sample), and $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${V}_{ANE}^{{S}_{max}}$$\end{document}$ is a very particular maximum *V*~*ANE*~ value that we set from experiment. Specifically, this latter is obtained when the sample is magnetically saturated, with $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$\overrightarrow{m}$$\end{document}$ in the film plane and transverse to the detection direction defined by the electrical contacts; this configuration yields $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\overrightarrow{E}}_{ANE}$$\end{document}$ parallel to the direction of the voltage detection, having its highest magnitude. After all, our theoretical approach provides a normalized *V*~*ANE*~ response, which is rescaled using an experimental *S*~*N*~ value (see Methods for details on the *S*~*N*~ estimation) for comparison.
The theoretical system and the definitions of the vectors considered to perform the numerical calculations are depicted in Fig. [1(b)](#Fig1){ref-type="fig"}. Looking at the thermoelectric voltage at a constant in-plane magnetic field that is high enough to keep the film magnetically saturated, the magnetization follows the orientation of the magnetic field, i.e. $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\phi }_{m}={\phi }_{H}$$\end{document}$. As consequence, the angular dependence of *V*~*ANE*~ with the magnetic field is considerable simplified,$$\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${V}_{ANE}={S}_{N}|\nabla T|L\,\cos \,{\phi }_{m}\mathrm{.}$$\end{document}$$
On the other hand, it is worth remarking that the amplitude and direction of the magnetization may be changed by the application of a magnetic field $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$\overrightarrow{H}$$\end{document}$ and/or external stress *σ*, thus modifying $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\overrightarrow{E}}_{ANE}$$\end{document}$ and, consequently, *V*~*ANE*~. At the non-saturated magnetic regime, the equilibrium angle *φ*~*m*~ of the magnetization is a result of the competition between the applied magnetic field and the magnetic anisotropies of the sample, leading to a complex dependence of the *V*~*ANE*~ with the magnetic field. To investigate the magnetic properties and thermoelectric voltage response of magnetostrictive films with induced uniaxial magnetic anisotropy, we employ a modified Stoner-Wohlfarth model^[@CR36]^. Here, we consider the magnetic free energy density as$$\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$\xi =-\,\overrightarrow{m}\cdot \overrightarrow{H}+4\pi {m}_{s}^{2}{(\hat{m}\cdot \hat{n})}^{2}-\frac{{H}_{k}}{2{m}_{s}}{(\hat{m}\cdot {\hat{u}}_{k})}^{2}-\frac{3}{2}{\lambda }_{s}\sigma {(\hat{m}\cdot {\hat{u}}_{\sigma })}^{2},$$\end{document}$$where $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$\overrightarrow{m}$$\end{document}$ is the magnetization vector of the ferromagnetic layer, $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$\hat{m}$$\end{document}$ is its corresponding versor, *m*~*s*~ is the saturation magnetization, $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${H}_{k}\mathrm{=2}{K}_{u}/{m}_{s}$$\end{document}$ is the anisotropy field, where *K*~*u*~ is the induced uniaxial magnetic anisotropy constant, $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\lambda }_{s}$$\end{document}$ is the saturation magnetostriction constant, *σ* is the applied stress, and $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$\hat{n}$$\end{document}$ is the versor normal to the film plane. The first term of the magnetic free energy density indicates the Zeeman interaction, the second one is the demagnetizing energy density, and the third term is associated to the induced uniaxial magnetic anisotropy oriented along $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\hat{u}}_{k}$$\end{document}$. Finally, the fourth term describes the magnetoelastic energy density for an elastically medium with isotropic magnetostriction. This latter term relates the saturation magnetostriction constant $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\lambda }_{s}$$\end{document}$ and the stress *σ* applied to the system, and gives rise to the magnetoelastic anisotropy contribution along $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\hat{u}}_{\sigma }$$\end{document}$. In particular, the product between $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\lambda }_{s}$$\end{document}$ and *σ* modifies the effective magnetic anisotropy of the sample. While for $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\lambda }_{s}\sigma \mathrm{ > 0}$$\end{document}$ a magnetoelastic anisotropy axis is induced along the same direction of the applied stress, for $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\lambda }_{s}\sigma \mathrm{ < 0}$$\end{document}$, this magnetoelastic anisotropy axis is oriented perpendicularly to the direction of the stress^[@CR36]--[@CR38]^.
From the appropriate magnetic free energy density, a routine for the energy minimization determines the equilibrium angles $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\theta }_{m}$$\end{document}$ and $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\phi }_{m}$$\end{document}$ of the magnetization at a given magnetic field $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$\overrightarrow{H}$$\end{document}$, and we calculate the normalized magnetization $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$m/{m}_{s}$$\end{document}$ and the thermoelectric voltage *V*~*ANE*~ curves.
First of all, to confirm the validity of our theoretical approach, we consider a system consisting of a film with uniaxial magnetic anisotropy and without stress, i.e. $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$\sigma \mathrm{=0}$$\end{document}$ MPa. For the numerical calculations, we take into account the following system parameters: $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${m}_{s}=625$$\end{document}$ emu/cm^3^ and $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\lambda }_{s}=+\,30.0\times {10}^{-6}$$\end{document}$, which are characteristic values of the Co~40~Fe~40~B~20~ alloy^[@CR19],[@CR39],[@CR40]^, $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${H}_{k}\mathrm{=42}$$\end{document}$ Oe, $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\theta }_{k}{\mathrm{=90}}^{\circ }$$\end{document}$, $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\phi }_{k}{\mathrm{=5}}^{\circ }$$\end{document}$ and film thickness of $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${t}_{f}\mathrm{=300}$$\end{document}$ nm. The magnetic field is in the film plane, i.e. $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\theta }_{H}{\mathrm{=90}}^{\circ }$$\end{document}$, and $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\phi }_{H}$$\end{document}$ is varied from 0° up to 360°. For the *V*~*ANE*~ calculation, we assume $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${S}_{N}=1.31\times {10}^{-5}$$\end{document}$ V/K and the $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$\Delta T$$\end{document}$ varies between 0 and 30 K. To illustrate the results obtained with our routine, Fig. [2](#Fig2){ref-type="fig"} presents the magnetization and thermoelectric voltage calculations for this system.Figure 2Numerical calculations of the magnetic response and thermoelectric voltage for a film with uniaxial magnetic anisotropy. For the calculations, we consider the following parameters: $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${m}_{s}=625$$\end{document}$ emu/cm^3^, $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\lambda }_{s}=+\,30.0\times {10}^{-6}$$\end{document}$, $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${H}_{k}=42$$\end{document}$ Oe, $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\theta }_{k}={90}^{\circ }$$\end{document}$, $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\phi }_{k}={5}^{\circ }$$\end{document}$, $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${t}_{f}=300$$\end{document}$ nm, $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$\sigma \mathrm{=0}$$\end{document}$ MPa and $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${S}_{N}=1.31\times {10}^{-5}$$\end{document}$ V/K. For the thermoelectric voltage calculation, the temperature gradient ∇*T* is normal to the film plane. Further, the *V*~*ANE*~ detection is performed with electrical contacts at the ends of the main axis of the film. Notice that this direction is almost transverse to $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\hat{u}}_{k}$$\end{document}$, which defines the induced uniaxial magnetic **a**nisotropy. (**a**) Normalized magnetization curves for distinct *φ*~*H*~ values. (**b**) *V*~*ANE*~ response, with $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$\Delta T=27$$\end{document}$ K, as a function of the magnetic field for the very same *φ*~*H*~ values. (**c**) *V*~*ANE*~, at $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$H=\pm \,300$$\end{document}$ Oe and with $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$\Delta T=27$$\end{document}$ K, as a function of *φ*~*H*~. At this field value, our system is magnetically saturated. (**d**) Similar plot for the *V*~*ANE*~ at $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$H=+\,300$$\end{document}$ Oe for distinct Δ*T* values. In the inset, the dependence of the $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${V}_{ANE}^{Smax}$$\end{document}$ value with Δ*T* is shown.
Figure [2(a)](#Fig2){ref-type="fig"}, in which we show the magnetization calculation for distinct $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\phi }_{H}$$\end{document}$ values, uncovers the expected dependence of the magnetization curves with the orientation between the easy magnetization axis and magnetic field, disclosing all the traditional features of uniaxial magnetic systems^[@CR41]^. Given that $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\varphi }_{k}{\mathrm{=5}}^{\circ }$$\end{document}$, the curve for $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\phi }_{H}={0}^{\circ }$$\end{document}$ reveals the signatures of a typical easy magnetization axis. Specifically, the magnetization loop presents normalized remanent magnetization ∼1 and coercive field of around 35 Oe, this latter close to *H*~*k*~. For the magnetization curve at $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\phi }_{H}={90}^{\circ }$$\end{document}$ in turn, the normalized remanent magnetization reduces to ∼0.08 and the coercive field to around 2 Oe, values compatible with a hard magnetization axis. Remarkably, the small deviation of $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\phi }_{k}$$\end{document}$ yields an hysteretic behavior for $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\phi }_{H}={90}^{\circ }$$\end{document}$, giving rise to a typical hard axis behavior with a non-zero coercive field.
Regarding the thermoelectric voltage calculation, Fig. [2(b)](#Fig2){ref-type="fig"} shows the *V*~*ANE*~ response, with $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$\Delta T=27$$\end{document}$ K, as a function of the magnetic field for distinct $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\phi }_{H}$$\end{document}$ values. Notice the quite-interesting evolution in the shape of the curves as the magnitude and orientation of the field are altered. It is noteworthy that the *V*~*ANE*~ detection is performed with electrical contacts at the ends of the main axis of the film, and here this direction is almost transverse to $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\hat{u}}_{k}$$\end{document}$, which defines the orientation of the induced uniaxial magnetic anisotropy. At small $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\phi }_{H}$$\end{document}$ values, the *V*~*ANE*~ curves seem to mirror the magnetization loops. This feature is a signature that the magnetization is kept close to the easy magnetization axis, thus favoring the alignment between $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\overrightarrow{E}}_{ANE}$$\end{document}$ and the detection direction defined by the electrical contacts. Therefore, *V*~*ANE*~ is directly proportional to the magnitude of the magnetization. On the other hand, as the $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\phi }_{H}$$\end{document}$ value increases and the field is not nearby the easy magnetization axis, the *V*~*ANE*~ curves lose the square shape. This fact is a consequence of the competition between two energy terms, the first associated with the Zeeman interaction and the second one related to the induced uniaxial magnetic anisotropy. At the low magnetic field range, the magnetization remains close to the easy magnetization axis, leading to high *V*~*ANE*~ values. However, as the magnetic field increases, the Zeeman energy term dominates, and the magnetization rotates out from the easy axis, following the field. As a result, the $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\overrightarrow{E}}_{ANE}$$\end{document}$ component along the detection direction defined by the electrical contacts decreases and *V*~*ANE*~ is drastically reduced.
Figure [2(c)](#Fig2){ref-type="fig"} presents the *V*~*ANE*~ values, at $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$H=\pm 300$$\end{document}$ Oe and with $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$\varDelta T\mathrm{=27}$$\end{document}$ K, as a function of $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\phi }_{H}$$\end{document}$. At these field values, our system is magnetically saturated, in a sense that the magnetization follows the orientation of the magnetic field. As expected, the curves draw a clear dependence of *V*~*ANE*~ with the sign of the magnetic field and primarily with $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\phi }_{H}$$\end{document}$, this latter evidenced by a well-defined cosine shape, as described in Eq. ([4](#Equ4){ref-type=""}).
Further, Fig. [2(d)](#Fig2){ref-type="fig"} shows the *V*~*ANE*~ behavior as a function of $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\phi }_{H}$$\end{document}$ at $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$H=+300$$\end{document}$ Oe for selected $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$\varDelta T$$\end{document}$ values. In this case, although the amplitude of the curves is altered, the angular dependence is kept constant, also corroborating Eq. ([4](#Equ4){ref-type=""}). At this magnetic saturation state, the maximum *V*~*ANE*~ value is found for $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\phi }_{H}{\mathrm{=0}}^{\circ }$$\end{document}$, and corresponds to $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${V}_{ANE}^{Smax}$$\end{document}$, according to our setup illustrated in Fig. [1(a,b)](#Fig1){ref-type="fig"}. Moreover, a linear dependence of $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${V}_{ANE}^{Smax}$$\end{document}$ with $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$\varDelta T$$\end{document}$ is obtained, as we can see in the inset of Fig. [2(d)](#Fig2){ref-type="fig"}.
After all, the numerical calculations obtained with our theoretical approach provide us an overview of the ANE for a nanostructured system consisting of a film with induced uniaxial magnetic anisotropy. From now, to go beyond, the main challenge to the description of different systems resides in the appropriate writing of the magnetic free energy density.
Comparison with the experiment {#Sec4}
------------------------------
The previous calculations have qualitatively described the main features of the magnetic behavior and thermoelectric voltage response in a film with uniaxial magnetic anisotropy and without stress.
From now on, we investigate the quasi-static magnetic properties and anomalous Nernst effect in a Co~40~Fe~40~B~20~ (from now on called CoFeB) film with thickness of 300 nm grown onto a flexible (Kapton®) substrate (see Methods for details on the film deposition and experiments). In this context, we perform experiments for the sample with and without external stress.
Figure [3](#Fig3){ref-type="fig"} shows experimental results of the magnetic response and the thermoelectric voltage for the CoFeB film without stress. Figure [3(a)](#Fig3){ref-type="fig"}, in which we present the magnetization response for the film grown onto a flexible substrate as a function of the magnetic field for distinct $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\phi }_{H}$$\end{document}$ values, discloses curves that assign the characteristic behavior of a system with uniaxial magnetic anisotropy close to the direction perpendicular to the main axis of the sample. While coercive field of ∼42 Oe and normalized remanent magnetization of ∼1 are found for $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\phi }_{H}={0}^{\circ }$$\end{document}$, respective values of ∼30 Oe and ∼0.5 are measured at $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\varphi }_{H}{\mathrm{=90}}^{\circ }$$\end{document}$. These values verified for $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\phi }_{H}={90}^{\circ }$$\end{document}$ are indicators of magnetic anisotropy dispersion, which is primarily related to the internal stress stored in the film during the growth.Figure 3Experimental results of the magnetic response and thermoelectric voltage for the CoFeB film without stress. (**a**) Normalized magnetization curves at selected *φ*~*H*~ values for the CoFeB film. (**b**) Similar plot for the thermoelectric voltage *V*~*ANE*~ measurements performed with $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$\Delta T=27$$\end{document}$ K.
Regarding the thermoelectric voltage results, Fig. [3(b)](#Fig3){ref-type="fig"} shows the *V*~*ANE*~ response for the film grown onto a flexible substrate, with $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$\Delta T\mathrm{=27}$$\end{document}$ K, as a function of the magnetic field for distinct $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\phi }_{H}$$\end{document}$ values. The experimental *V*~*ANE*~ curves are also in concordance with the numerical calculations presented in Fig. [2(b)](#Fig2){ref-type="fig"}, reflecting all features of systems with uniaxial magnetic anisotropy and without external stress. Specifically, we find here the evolution in the shape of the curves as the magnitude and orientation of the field are altered. Moreover, at high magnetic field values, when the film is magnetically saturated, we verify the angular dependence of *V*~*ANE*~, having $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${V}_{ANE}^{Smax}\approx 20\,\mu $$\end{document}$V, and the characteristic reduction of its value to zero as $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\phi }_{H}$$\end{document}$ increases to 90°. However, at the low field range, it is noteworthy that a discrepancy between theory and experiment may be found. In particular, as $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\phi }_{H}$$\end{document}$ is raised to 90°, *V*~*ANE*~ does not reach the expected maximum of 20 *μ*V found at $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\phi }_{H}={0}^{\circ }$$\end{document}$, a value also evidenced numerically in Fig. [2(b)](#Fig2){ref-type="fig"}. This divergence is primarily associated with the magnetization process at low magnetic fields and to the existence of magnetic domains in the film, a fact that is not taken into account in our microspin modified Stoner-Wohlfarth theoretical approach. Indeed, the magnitude of the magnetization in the experiment is not kept constant at the low-field levels, leading to a reduction of the *V*~*ANE*~ value.
From the experimental results found for the flexible film, shown in Fig. [3](#Fig3){ref-type="fig"} and compared with the calculated ones presented in Fig. [2](#Fig2){ref-type="fig"}, we corroborate the angular dependence of the magnetization and thermoelectric voltage curves in films with uniaxial magnetic anisotropy. Going beyond, from now on, we focus our efforts on the magnetic response and thermoelectric voltage in a magnetostrictive system under stress. As aforementioned, CoFeB alloy has high saturation magnetostriction constant, reaching $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\lambda }_{s}\approx +\,30.0\times {10}^{-6}$$\end{document}$^[@CR19],[@CR40]^. Therefore, we are able to manipulate here the magnetic properties, i.e. the effective magnetic anisotropy, of our magnetostrictive CoFeB film grown onto a flexible substrate through the application of stress.
Figure [4](#Fig4){ref-type="fig"} shows experimental results of the normalized magnetization curves and normalized *V*~*ANE*~ response for the flexible CoFeB film under distinct stress values. The measurements are performed at $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\phi }_{H}={0}^{\circ }$$\end{document}$, with the stress along the main axis of the sample (see Methods for details on the stress application). For the positive stress values employed here, $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\lambda }_{s}\sigma > 0$$\end{document}$, and it yields a magnetoelastic anisotropy oriented along the main axis of the sample, i.e. $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\theta }_{\sigma }={90}^{\circ }$$\end{document}$ and $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\varphi }_{\sigma }{=90}^{\circ }$$\end{document}$.Figure 4Experimental results of the magnetic response and thermoelectric voltage for the flexible CoFeB film under stress. (**a**) Normalized magnetization curves and (**b**) normalized *V*~*ANE*~ response for the flexible CoFeB film under distinct stress values. The measurements are performed with $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\phi }_{H}={0}^{\circ }$$\end{document}$, and bending the sample along its main axis. Moreover, we set $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$\Delta T=27$$\end{document}$ K for the *V*~*ANE*~ acquisition. In the inset, we present the schematic configuration of the sample in the ANE experiment, uncovering the easy magnetization axis (EA) induced during deposition, the orientation of the stress $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$\sigma $$\end{document}$ in the bent film, as well as the directions of the magnetic field, temperature gradient and *V*~*ANE*~ detection.
From the magnetization response shown in Fig. [4(a)](#Fig4){ref-type="fig"}, we uncover the modifications of the effective magnetic anisotropy with the stress level. Specifically, while the effective anisotropy is well-described by $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\phi }_{{k}_{eff}}$$\end{document}$ roughly close to $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\phi }_{k}$$\end{document}$ for $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$\sigma =0$$\end{document}$, the changes in the coercive field, remanent magnetization, as well as the own shape of the magnetization curves as the stress level increases are straight consequences of the raise of the magnetoelastic anisotropy contribution to the effective magnetic anisotropy. This raise leads to changes in the orientation $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\hat{u}}_{{k}_{eff}}$$\end{document}$ of the effective magnetic anisotropy with the increase of the stress value, thus modifying the whole magnetic behavior and magnetization curves. For $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$\sigma =700$$\end{document}$ MPa, the square curve with smaller coercive field reveals an intermediate magnetic behavior. It arise from an effective magnetic anisotropy that is still roughly close to the direction of the magnetic field at $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\phi }_{H}={0}^{\circ }$$\end{document}$, but has a significant component along the main axis of the sample due to the magnetoelastic contribution.
However, for $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$\sigma =1400$$\end{document}$ MPa, the stress is high enough to set the effective magnetic anisotropy axis close to $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\hat{u}}_{\sigma }$$\end{document}$, i.e. $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\hat{u}}_{eff}\approx {90}^{\circ }$$\end{document}$. Consequently, as expected, the magnetization curve at $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\phi }_{H}={0}^{\circ }$$\end{document}$ has fingerprints of a hard magnetization axis, with non-zero values for coercive field and normalized remanent magnetization due to anisotropy dispersion. Therefore, the whole magnetic behavior of the flexible magnetostrictive film under stress is a result of the competition between the induced uniaxial magnetic anisotropy and the magnetoelastic anisotropy contribution. As a consequence, this competition allows us to tailor the effective magnetic anisotropy and, consequently, the anomalous Nernst effect. Remarkably, as we can confirm from Fig. [4(b)](#Fig4){ref-type="fig"}, modifications in the *V*~*ANE*~ response are already visible at $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\phi }_{H}={0}^{\circ }$$\end{document}$, when just the stress is altered in an experimental manner.
The most striking experimental and theoretical findings here are shown in Fig. [5](#Fig5){ref-type="fig"}, which discloses the evolution of the *V*~*ANE*~ response with the $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\phi }_{H}$$\end{document}$ and stress *σ* for the CoFeB film grown onto a flexible substrate. The numerical calculations for a film with uniaxial magnetic anisotropy and under stress are performed considering parameters similar to those previously employed in Fig. [2](#Fig2){ref-type="fig"}, except for the $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\phi }_{k}$$\end{document}$ and *S*~*N*~ values. In this case, $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${m}_{s}\mathrm{=625}$$\end{document}$ emu/cm^3^ and $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\lambda }_{s}=+\,30.0\times {10}^{-6}$$\end{document}$, $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${H}_{k}=42$$\end{document}$ Oe, $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\theta }_{k}={90}^{\circ }$$\end{document}$, $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\phi }_{k}={30}^{\circ }$$\end{document}$ and $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${t}_{f}=300$$\end{document}$ nm. The orientation of $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\hat{u}}_{k}$$\end{document}$, set by $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\varphi }_{k}$$\end{document}$, is changed in order to mimic the magnetic properties of the flexible film. Further, we consider $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${S}_{N}\mathrm{=1.31}\times {10}^{-5}$$\end{document}$, $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$1.14\times {10}^{-5}$$\end{document}$ and $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$1.09\times {10}^{-5}$$\end{document}$ V/K (see Methods for details on the *S*~*N*~ estimation) for the calculations with $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$\sigma =0$$\end{document}$, 700, and 1400 MPa, respectively. In particular, the decrease of *S*~*N*~ with $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$\sigma $$\end{document}$ is straightly verified through the reduction of the $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${V}_{ANE}^{Smax}$$\end{document}$; and we associate it to modifications in the energy relaxation time close of the Fermi level, which can be altered by the stress application^[@CR42]^. Then, notice the striking agreement between our experiments and numerical calculations, including three important features: the own shape of the *V*~*ANE*~ curves, the $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\phi }_{H}$$\end{document}$ dependence of the evolution in the shape of the curves, and amplitude of the *V*~*ANE*~ response at high magnetic fields, i.e. at the magnetic saturation state.Figure 5Evolution of the thermoelectric voltage with the magnetic field and stress. (**a**) Experimental results and numerical calculations of the *V*~*ANE*~ response as a function of the magnetic field, for selected *φ*~*H*~ values, for the flexible CoFeB film without stress. Below, illustration showing the orientations of $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\overrightarrow{E}}_{ANE}$$\end{document}$, $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$\overrightarrow{H}$$\end{document}$, $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$\overrightarrow{m}$$\end{document}$ and $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\hat{u}}_{eff}$$\end{document}$. The dashed lines represent the direction in which $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$\overrightarrow{m}$$\end{document}$ and $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\overrightarrow{E}}_{ANE}$$\end{document}$ rotates with decreasing the magnetic field. Similar plots for the flexible CoFeB film submitted to stress values of (**b**) $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$\sigma =700$$\end{document}$ MPa and (**c**) $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$\sigma =1400$$\end{document}$ MPa. We set $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$\Delta T=27$$\end{document}$ K for the *V*~*ANE*~ acquisition. For the numerical calculations, we consider the following parameters: $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${m}_{s}=625$$\end{document}$ emu/cm^3^, $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\lambda }_{s}=+\,30.0\times {10}^{-6}$$\end{document}$, $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${H}_{k}=42$$\end{document}$ Oe, $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\theta }_{k}={90}^{\circ }$$\end{document}$, $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\phi }_{k}={30}^{\circ }$$\end{document}$, and $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${t}_{f}=300$$\end{document}$ nm, with $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${S}_{N}=1.31\times {10}^{-5}$$\end{document}$ V/K in (**a**), $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${S}_{N}=1.14\times {10}^{-5}$$\end{document}$ V/K in (**b**), and $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${S}_{N}=1.09\times {10}^{-5}$$\end{document}$ V/K in (**c**). See the movie in Supplementary Information.
From Fig. [5(a)](#Fig5){ref-type="fig"}, the *V*~*ANE*~ curves for the film without stress exhibit all the features found for a film with uniaxial magnetic anisotropy, as expected; hence, their interpretation is the very same to that previously reported for Fig. [2(b)](#Fig2){ref-type="fig"} and 3(a.II,b.II). With respect to the shape of the curves and its evolution with the orientation of the magnetic field, it is worth observing that the *V*~*ANE*~ response pattern changes with $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\phi }_{H}$$\end{document}$ crossing over through $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\phi }_{{k}_{eff}}$$\end{document}$. Specifically, keeping in mind that in this case the effective magnetic anisotropy orientation $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\hat{u}}_{{k}_{eff}}$$\end{document}$ is the own uniaxial magnetic anisotropy direction $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\hat{u}}_{k}$$\end{document}$ with $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\phi }_{k}\approx {30}^{\circ }$$\end{document}$, the $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${V}_{ANE}$$\end{document}$ curves for $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\phi }_{H} < {\phi }_{{k}_{eff}}$$\end{document}$ have shape mirroring the corresponding magnetization loops presented in Fig. [4(a)](#Fig4){ref-type="fig"}. At small $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\phi }_{H}$$\end{document}$ values, represented in Fig. [5](#Fig5){ref-type="fig"}(a) by the curve for $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\phi }_{H}={0}^{\circ }$$\end{document}$, the magnetization is kept close to the magnetic field direction and/or to the easy magnetization axis. It favours the alignment between $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\overrightarrow{E}}_{ANE}$$\end{document}$ and the detection direction defined by the electrical contacts, and therefore *V*~*ANE*~ is directly proportional to the magnitude of the magnetization. The *V*~*ANE*~ response for $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\phi }_{H} > {\phi }_{{k}_{eff}}$$\end{document}$ in turn, depicted by $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\phi }_{H}={40}^{\circ }$$\end{document}$ and 70°, discloses curves with completely different signatures, which lose the square shape. For these cases, at high magnetic field values, the magnetization is aligned with the field and, consequently, *V*~*ANE*~ is drastically reduced; at the low magnetic field range, the magnetization $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$\overrightarrow{m}$$\end{document}$ rotates and remains close to the easy magnetization axis $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\hat{u}}_{eff}$$\end{document}$, as we can see in the schematic representation in Fig. [5(a.III)](#Fig5){ref-type="fig"}, leading to higher *V*~*ANE*~ values.
For the film under $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$\sigma \approx 700$$\end{document}$ MPa, Fig. [5(b)](#Fig5){ref-type="fig"} shows curves with a slightly different profile. In this case, the effective magnetic anisotropy is described by an angle $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\phi }_{{k}_{eff}}$$\end{document}$ that lies between 40° and 70°, in a sense that the change in the *V*~*ANE*~ response pattern takes place within this angular range. Hence, observe that the *V*~*ANE*~ responses at $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\phi }_{H}={0}^{\circ }$$\end{document}$ and 40° have the features of a magnetization curve, just presenting a decrease in the amplitude of the *V*~*ANE*~ signal at the magnetic saturation state, as expected due to the thermoelectric voltage configuration in the experiment. At $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\phi }_{H}={70}^{\circ }$$\end{document}$, the shape of the *V*~*ANE*~ curves considerably, suggesting the competition between the Zeeman interaction and the effective magnetic anisotropy. In this case, at high fields, the magnetization follows the magnetic field, leading to a decrease of *V*~*ANE*~. However, as the field decreases, the Zeeman interaction is reduced and the magnetization $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$\overrightarrow{m}$$\end{document}$ turns to $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\hat{u}}_{eff}$$\end{document}$. As a consequence, there is the increase in the $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\overrightarrow{E}}_{ANE}$$\end{document}$ component along the *V*~*ANE*~ detection direction, as we can see in the schematic representation in Fig. [5(b.III)](#Fig5){ref-type="fig"}. Obviously, as aforementioned, the discrepancy between experiment and theory at this low-field values is primarily associated with the magnetization process and to the existence of magnetic domains in the film, which is not taken into account in our theoretical approach, as already discussed.
At last, for the film under $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$\sigma \approx 1400$$\end{document}$ MPa shown in Fig. [5(c)](#Fig5){ref-type="fig"}, the effective magnetic anisotropy lies in an angle $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\phi }_{{k}_{eff}} > {70}^{\circ }$$\end{document}$, specifically close to $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\phi }_{{k}_{eff}}\approx {90}^{\circ }$$\end{document}$. The presented *V*~*ANE*~ curves have the very same features for all the selected $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\phi }_{H}$$\end{document}$ values. At $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\phi }_{H}={0}^{\circ }$$\end{document}$, the *V*~*ANE*~ curve mirrors the corresponding magnetization loop shown in Fig. [4(a)](#Fig4){ref-type="fig"}. As $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\phi }_{H}$$\end{document}$ increases, we find the decrease in the amplitude of the *V*~*ANE*~ signal at the magnetic saturation state, as expected. Once the magnetic field decreases, irrespective of $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\phi }_{H}$$\end{document}$, the magnetization $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$\overrightarrow{m}$$\end{document}$ rotates to the orientation $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\hat{u}}_{eff}$$\end{document}$ of the effective magnetic anisotropy. Given that $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\overrightarrow{E}}_{ANE}$$\end{document}$ tends to be almost transverse to the *V*~*ANE*~ detection direction, as we can confirm through the schematic representation in Fig. [5(c.III)](#Fig5){ref-type="fig"}, a decrease in the *V*~*ANE*~ value is also found at low field values and, as a consequence, the evolution in the shape of the *V*~*ANE*~ curves is not observed here. In particular, for $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\phi }_{H}={90}^{\circ }$$\end{document}$, $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${V}_{ANE}\approx 0$$\end{document}$ for the whole magnetic field range.
Finally, looking at the angular dependence of the *V*~*ANE*~ signal amplitude at a given magnetic field, Fig. [6](#Fig6){ref-type="fig"} shows the thermoelectric voltage as a function of $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\phi }_{H}$$\end{document}$ for the CoFeB film under selected stress levels. In Fig. [6(a)](#Fig6){ref-type="fig"}, the sample is magnetically saturated, irrespective on the stress level. As a consequence, the *V*~*ANE*~ response is precisely the one described by Eq. ([4](#Equ4){ref-type=""}), with a well-defined cosine shape. Another important test of consistency of our approach is also given by the *V*~*ANE*~ behavior at non-saturated states. The curves are a result of the competition between the uniaxial magnetic anisotropy and the magnetoelastic anisotropy, leading to considerable changes in the shape and amplitude of the *V*~*ANE*~ response. Notice the striking quantitative agreement between experiment and theory in Fig. [6(b)](#Fig6){ref-type="fig"}.Figure 6Angular dependence of the thermoelectric voltage. (**a**) Experimental results and numerical calculations for the *V*~*ANE*~, at $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$H=+\,300$$\end{document}$ Oe and with $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$\varDelta T=27$$\end{document}$ K, as a function of *φ*~*H*~ for the flexible CoFeB film under selected stress levels. At this field value, the sample is magnetically saturated. (**b**) Similar plot for the *V*~*ANE*~ at $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$H=+40$$\end{document}$ Oe, where the sample is at a non-saturated state. For the numerical calculations, we consider the very same parameters emloyed in Fig. [5](#Fig5){ref-type="fig"}, i.e. $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${m}_{s}=625$$\end{document}$ emu/cm^3^, $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\lambda }_{s}=+30.0\times {10}^{-6}$$\end{document}$, $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${H}_{k}=42$$\end{document}$ Oe, $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\theta }_{k}={90}^{\circ }$$\end{document}$, $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\phi }_{k}={30}^{\circ }$$\end{document}$ and $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${t}_{f}=300$$\end{document}$ nm, with $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${S}_{N}=1.31\times {10}^{-5}$$\end{document}$ V/K when $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$\sigma =0$$\end{document}$ MPa, and $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${S}_{N}=1.09\times {10}^{-5}$$\end{document}$ V/K when $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$\sigma =1400$$\end{document}$ MPa.
Hence, we are able to describe through numerical calculations all the main features of the *V*~*ANE*~ response. Thus, we provide experimental evidence to confirm the validity of the theoretical approach to describe the magnetic properties and anomalous Nernst effect in ferromagnetic magnetostrictive films with uniaxial magnetic anisotropy and submitted to external stress.
Discussion {#Sec5}
==========
In summary, we have performed a theoretical and experimental investigation of the magnetic properties and anomalous Nernst effect in a flexible magnetostrictive film with induced uniaxial magnetic anisotropy and under external stress. Our findings have raised numerous interesting issues on the anomalous Nernst effect in nanostructured magnetic materials. In particular, they have shown how the magnetization behavior and the thermoelectric voltage response evolve with both, the magnetic field and external stress. Hence, by comparing our experiments with numerical calculations, we have elucidated the magnetic properties and thermoelectric voltage and demonstrated the possibility of tailoring the anomalous Nernst effect in a flexible magnetostrictive film. The quantitative agreement between experiment and numerical calculations has provided us evidence to confirm the validity of the theoretical approach to describe the magnetic properties and anomalous Nernst effect in ferromagnetic magnetostrictive films with uniaxial magnetic anisotropy and submitted to external stress. The results place flexible magnetostrictive systems as promising candidates for active elements in functionalized touch electronic devices.
Methods {#Sec6}
=======
Estimation of Δ*T*~*f*~ {#Sec7}
-----------------------
The relation between the temperature variation Δ*T* measured experimentally across the sample and the effective temperature variation Δ*T*~*f*~ in the film is given by^[@CR43]^$$\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$\Delta {T}_{f}=\frac{{t}_{f}{K}_{sub}}{{t}_{sub}{K}_{f}}\Delta T,$$\end{document}$$where *K*~*sub*~ and *t*~*sub*~ are the thermal conductivity and thickness of the substrate, while *K*~*f*~ and *t*~*f*~ are the respective quantities for the ferromagnetic film. Here, we consider $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${K}_{sub}=0.12$$\end{document}$ W/Km and $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${t}_{sub}=0.15$$\end{document}$ mm for the flexible substrate (Kapton®), while $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${K}_{f}=86.7$$\end{document}$ W/Km^[@CR44]--[@CR46]^ and $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${t}_{f}=300$$\end{document}$ nm for our flexible CoFeB film.
Estimation of *S*~*N*~ {#Sec8}
----------------------
We estimate the $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${S}_{N}={\lambda }_{N}{\mu }_{\circ }{m}_{s}$$\end{document}$ values for each applied stress *σ* level. This coefficient is obtained from the linear fitting of the experimental data of $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${V}_{ANE}^{Smax}$$\end{document}$ as a function of Δ*T*, as shown in Fig. [7](#Fig7){ref-type="fig"}. In particular, $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$\Delta {T}_{f}$$\end{document}$ is calculated by using Eq. ([6](#Equ6){ref-type=""}) and thus *S*~*N*~ is given by Eq. ([3](#Equ3){ref-type=""}). As a result, we assume $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${S}_{N}=1.31\times {10}^{-5}$$\end{document}$, $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$1.14\times {10}^{-5}$$\end{document}$ and $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$1.09\times {10}^{-5}$$\end{document}$ V/K for the calculations with $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$\sigma =0$$\end{document}$, 700, and 1400 MPa, respectively. Here, it is worth mentioning that the estimated *S*~*N*~ values may be different from the actual ones, given that the measured Δ*T* can include the parasitic contribution due to the interfacial thermal resistance between the sample and the heat sink.Figure 7Estimation of the coefficient *S*~*N*~. $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${V}_{ANE}^{Smax}$$\end{document}$ as a function of Δ*T* for distinct applied stress *σ*. The *S*~*N*~ value is obtained by using Eqs ([3](#Equ3){ref-type=""}) and ([6](#Equ6){ref-type=""}).
Sample preparation {#Sec9}
------------------
We investigate the anomalous Nernst effect in a Co~40~Fe~40~B~20~ (CoFeB) film with thickness of 300 nm grown onto a flexible (Kapton®) substrate. The film is deposited by magnetron sputtering onto a substrate with dimensions of 10 × 6 mm^2^. The deposition process is carried out with the following parameters: base pressure of 5 × 10^−6^ Torr, deposition pressure of 3.0 × 10^−3^ torr with a 99.99% pure Ar at 50 sccm constant flow, and with 50 W set in the DC source. By using these parameters, the deposition rate is 1.86 nm/s. During the deposition, a constant in-plane magnetic field of 1 kOe is applied perpendicularly to the main axis of the substrate in order to induce an uniaxial magnetic anisotropy.
Magnetization measurements {#Sec10}
--------------------------
Quasi-static magnetization curves are obtained within the range between $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\phi }_{H}={0}^{\circ }$$\end{document}$ (perpendicular) and $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\phi }_{H}={90}^{\circ }$$\end{document}$ (along the main axis of the film), in order to verify the magnetic properties. The curves are acquired at room temperature using a Lake Shore model 7404 vibrating sample magnetometer, with maximum in-plane magnetic field of ±300 Oe.
Thermoelectric voltage experiments {#Sec11}
----------------------------------
We employ a homemade experimental system to measure the thermoelectric voltage associated with the anomalous Nernst effect. In our experimental setup, the temperature gradient ∇*T* is normal to the film plane. A Peltier module is used to heat or cool the top of the sample while the substrate is kept in thermal contact with a heat sink, a copper block at room temperature. The temperature difference $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$\Delta T$$\end{document}$ across the sample (film and Kapton substrate) is measured with a differential micro-thermocouple, depicted in Fig. [8](#Fig8){ref-type="fig"}. The magnetic field is kept in the film plane, although its orientation can be modified by varying $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$${\phi }_{H}$$\end{document}$ from 0° up to 360°, having as reference the dashed line indicated in Fig. [1(b)](#Fig1){ref-type="fig"}. Finally, the *V*~*ANE*~ detection is performed using a nanovoltmeter with electrical contacts at the ends of the main axis of the film, also illustrated in Fig. [1(a,b)](#Fig1){ref-type="fig"}, whose separation distance is $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$L=5.0$$\end{document}$ mm.Figure 8Bending the sample. Kapton strips are place under the film to bend the sample. In particular, micro-thermocouples are used to infer the Δ*T* across the sample (film and Kapton substrate).
Stress application {#Sec12}
------------------
The magnetization measurements and thermoelectric voltage experiments are carried out in a CoFeB film grown onto a flexible substrate. In this context, we perform acquisitions for the sample with and without stress. The stress is applied by bending the sample, along the main axis, during the measurement. In particular, Kapton strips are inserted under the film substrate in order to induce the bending, as illustrated in Fig. [8](#Fig8){ref-type="fig"}. The magnitude of the stress *σ* is calculated following the procedure previously employed in refs^[@CR36],[@CR38]^. By bending the flexible substrate, we are able to induce tensile stress in the film, thus modifying the effective magnetic anisotropy^[@CR38],[@CR47],[@CR48]^.
Supplementary information
=========================
{#Sec13}
Supplementary Information
**Publisher's note** Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Supplementary information
=========================
is available for this paper at 10.1038/s41598-019-51971-7.
The authors acknowledge financial support from the Brazilian agencies CNPq and CAPES.
M.A.C. and F.B were responsible for the theoretical approach and performed the numerical calculations. A.B.O. and C.C. were responsible for the development of the experimental system. A.S.M. and R.D.D.P produced the samples and performed the experimental measurements. All authors contributed to interpret the results, and write and improve the text of the manuscript.
The authors declare no competing interests.
| {
"pile_set_name": "PubMed Central"
} |
INTRODUCTION {#sec1-1}
============
The family *Brassicaceae* (= *Cruciferae*) consists of 350 genera and about 3500 species, and includes several genera such as *Camelina*, *Crambe*, *Sinapis*, *Thlaspi*, and *Brassica*. The genus *Brassica* is the most important one within the tribe Brassiceae, which includes some crops and species of great worldwide economic importance such as *Brassica oleracea* L., *Brassica napus* L., and *Brassica rapa* L. The same species can be utilized for several uses according to different forms or types. The genus is categorized into oilseed, forage, condiment, and vegetable crops by using their buds, inflorescences, leaves, roots, seeds, and stems. Brassicaceae vegetables represent an important part of the human diet worldwide, and are consumed by people all over the world.\[[@ref1]\] The largest genera are *Draba* (365 species), *Cardamine* (200 species), *Erysimum* (225 species), *Lepidium* (230 species), and *Alyssum* (195 species). Numerous species have food economic importance such as cabbage, cauliflower, turnip, and rape.\[[@ref2]\] The species are utilized as salad plants due to their content of anti-scorbutic and low content of erucic acids. Many Cruciferous species are known for their use in folk medicine for the treatment of snake bites. Moreover, they are used as an antimicrobial agent for relief of biliary colic and wound sores. They have enhancing the detoxification effect of chemical carcinogen and some species exhibit hypoglycemic and hypotension effects.\[[@ref3][@ref4]\] *Carrichtera annua* is an endemic annual herb belonging to a family belonging to the Cruciferae (or Brassicaceae), growing up to 40 cm specially in north Sinai at Elarrish region, Egypt. *C. annua* in particular is known to be used by the native Bedouins as an antidiabetic and antispasmodic. Flavonoids are a major group of constituents and are assumed to be among the beneficial components. Recently, they have also received considerable interest as components of foodstuffs and nutraceuticals because of their antioxidant and anticancer properties.\[[@ref5]--[@ref7]\] The flavonoidal fraction of the seeds of *C. annua*, was investigated using LC/ESI-MS and nano-ESI-MS/CID/MS. The flavonoidal fraction was found to contain 12 flavonol *O*-glycosides, which were structurally related and of which 11 were acylated with one or more benzoyl, feruloyl, or sinapoyl groups.\[[@ref8]\] In a previous study different known and new flavonoids compounds were isolated from the seeds and the herb and investigated by LC/MS and or NMR.\[[@ref8]--[@ref10]\] In this study, we report the isolation and structure elucidation of a new acylated flavonol triglycoside.
MATERIALS AND METHODS {#sec1-2}
=====================
Plant material {#sec2-1}
--------------
The plant was collected from Al-Arish region, Sinai, Egypt in April 2008 during the seedling stage. The plant was collected and identified as described previously.\[[@ref8]\]
Apparatus and techniques {#sec2-2}
------------------------
All NMR spectra were run on a Bruker DRX-400 instrument operating at 400 MHz for ^1^H and at 100 MHz for ^13^C, using standard pulse sequences. Chemical shifts are reported on the δ scale in parts per million downfield from TMS. TLC was carried out on precoated Silica gel 60 F~245~ plates (Merck), developed with EtOAc-HOAc-HCOOH-H~2~O (30:0.8:1.2:8, v/v, upper phase (solvent 1), and 100:10:10:20) (solvent 2) and Neu\'s spray reagent (1% diphenylboric acid ethanolamine complex). Column chromatography was performed on Silica gel (Merck), polyamide 6S (Riedel, De HaΧn), and Sephadex LH-20 (Pharmacia).
Extraction and isolation of flavonoids {#sec2-3}
--------------------------------------
About 500 g of air-dried powdered seeds of *C. annua* were extracted with petroleum ether (br. 40--60°C) by maceration (1.2 L × 4). The defatted seeds were extracted with methanol (70%) for 3 days (1.5L × 3). The combined methanol extract was evaporated under reduced pressure at 45°C and the residue dissolved in hot distilled water (400 ml) and left in the refrigerator overnight and filtered the precipitated matters. The filtrate was partitioned with chloroform (500 ml × 2) and *n*-butanol (600 ml × 4). The butanol extract was concentrated and subjected to column chromatography (CC) on polyamide. Elution was started with distilled water and decreasing the polarity in 10% with methanol to 100% methanol. The fractions were collected by monitoring on thin layer chromatography (TLC) using solvent 1 and/or solvent 2 as developing solvents. The fraction eluted with aqueous methanol (40% and 50%) was found to contain three main flavonoids (1, 2, and 3), this fraction was subjected again to CC using 30% methanol for elution and increasing methanol to 70% as mentioned above. [Compound 1](#F1){ref-type="fig"} was isolated and purified by using Sephadex LH-20 column eluted with methanol to give 22 mg of [compound 1](#F1){ref-type="fig"}. Compounds 2 and 3 were further purified using preparative paper chromatography with 20% HOAc and passing over a small Sephadex LH-20 column using methanol as eluent (compound 2 was 18 mg and compound 3 was 16 mg, respectively). Compound 4 was isolated from the fraction eluted with 70% methanol (polyamide column) in the same manner to afford (10 mg).
RESULTS AND DISCUSSION {#sec1-3}
======================
The flavonoidal compounds were obtained from the *n*-butanol soluble fraction of the defatted aqueous methanol of *C. annua* powdered seeds. Acid hydrolysis of these compounds yielded quercetin and isorhamnetin as aglycones and arabinose and glucose as sugars. [Compound 1](#F1){ref-type="fig"} was isolated as an amorphous yellow substance, and its ^1^H and ^13^C NMR spectra showed the presence of a quercetine moiety, three sugar residues, sinapoyl moiety and two methoxyl groups. The chemical shifts of C-2 and C-3 (*δ* 156.9 and 135.2, respectively) indicated C-3 substitution of the quercetin moiety.\[[@ref11]\]
{#F1}
The ^1^H NMR spectrum showed, in the aromatic region, a pair of *meta* coupled doublets at δ 6.28 and δ 6.21 (*J* = 1.6 Hz), which were attributed to H-6 and H-8 of quercetin (A-ring), respectively, and the typical three-spin system of the 1,3,4-trisubstituted B-ring: a *meta* coupled doublet at δ 7.50 (*J* = 2 Hz) (H-2'), an *ortho*-coupled doublet at δ 6.79 (*J* = 8.8 Hz) (H-5'), and an *ortho* and *meta* coupled doublet of doublets at δ 7.55 (*J* = 8.8, 2 Hz) (H-6'). Substitution of quercetin in the position C-3 was evident from the chemical shift of C-2 (δ 156.9), whereas in flavonols with an unsubstituted hydroxyl functionality at this position C-2 is expected around δ 147.\[[@ref12]\]
The signal at *δ* 161.9 was assigned to C-5 and the assignment of C-6 and C-8 was confirmed through the long range coupling observed in the HMBC experiment with the ^1^H NMR signal at *δ* 6.28 (H-6) while the signal at *δ* 164.8 showed correlations to both H-6 and H-8 (*δ* 6.1) so the signal at 164.8 was assigned for C-7, [compound 1](#F1){ref-type="fig"} was a 3,7 di-substituted quercetin. The 2D NMR spectra allowed the assignment of all signals of the glucose moiety at C-7 which could be identified as a glucopyranoside\[[@ref13]\] and the β configuration of the anomeric carbon was cleared from the coupling constant of H-1'''''' (*J* = 7.5 Hz) observed in the ^1^H NMR spectrum.\[[@ref14]\] The presence of a sugar moiety at C-7 was further confirmed through the UV spectra of both the compound and it\'s aglycone after acid hydrolysis where band-II showed bathochromic shift (10 nm) in the sodium acetate spectrum of the aglycone and does not occur in the compound itself\[[@ref13]\]. In fact, the UV spectra with different shift reagents proved the presence of quercetin moiety with free hydroxyl groups at C-3', C-4', C-5, and substituted at C-3 where band-I (MeOH) appeared at 335 nm.\[[@ref9]\] . Two methoxy groups appeared in ^1^H NMR at *δ* 3.7 which correlated to C-3'''' and C-5''''. ^13^C NMR at *δ* 148.5. The C-7''''--C-8'''' double bond showed a large coupling constant (*J* = 14.9 Hz) which prove the *trans*-configuration of the double bond. The carbonyl group of the sinapoyl moiety occurring at *δ* 168.3 exhibited a long range ^13^C--^1^H correlation with a ^1^H NMR signal at *δ* 4.11 assigned to one of the H-6'''' proton of a hexose unit representing the second glucose moiety. The anomeric proton at *δ* 4.55 showed a long range correlation with a ^13^C NMR signal at *δ* 81.9 correspond to the proton at *δ* 4.19 in the HSQC spectrum. The anomeric proton H-1'' of arabinose was assigned at *δ* 5.52, so we can conclude that the glucose moiety is attached at C-2'' in arabinose. All data of ^1^H and ^13^C were presented in [Table 1](#T1){ref-type="table"}. The FAB mass spectrum of the compound displayed a molecular ion peak at *m*/*z* 987 \[M+ Na\]^+^ which is compatible with the molecular formula C~43~H~48~O~25~, this can confirm the presence of two hexose moieties, pentose moiety, and sinapoyl moiety attached to quercetin nucleus. The presence of the sinapoyl moiety was confirmed through the peak at *m*/*z* 207 and its fragments (at *m*/*z* = 192, 175) also the ion peak at *m*/*z* 523 attributed to \[Na^+^ sinapoyl-hexose--pentose\] part which attached to quercetin at *m*/*z* 303. The terminal hexose can be deduced from the ion at *m*/*z* 803 which matched with (M ^+^ + H-162)^+^. Since the loss of sinapoyl-hexose-pent part (*m*/*z* 465) is much more pronounced than the loss of terminal hexose (at *m*/*z* 803) so, we can conclude that these parts are at C-3 and C-7.\[[@ref8]\] These spectral data were in good agreement with those reported for quercetin 3-*O*-\[(6-feruloyl-β-glucopyranosyl)-(1→2)-β-arabinopyranoside\]-7-*O*-β-glucopyranoside which was isolated before from the herb of this plant.\[[@ref9]\]
######
H and 13C NMR data of [compound 1](#F1){ref-type="fig"} in DMSO
From all the previous data we can conclude the structure of [compound 1](#F1){ref-type="fig"} as quarecetin-3-*O* \[(6-sinapoyl-β-glucopyranosyl (1→2)-β-arabinopyroside\]-7-*O*-glucopyranoside,which is considered as a new compound. The other three compounds were identified as quarecetin-3-*O*-glucoside, isorhamnetin-3-*O*-runtinoside, and isorhamnetin by comparison with the literature data.\[[@ref10]--[@ref12]\] It is evident to note that the compound isorhamnetin-3-*O*-runtinoside was isolated before from the herb of this plant.\[[@ref10]\]
**Source of Support:** Nil,
**Conflict of Interest:** None declared.
| {
"pile_set_name": "PubMed Central"
} |
Related literature {#sec1}
==================
For related literature, see: Brahadeeswaran *et al.* (1998[@bb3], 1999[@bb2]); Jonie Varjula *et al*. (2007[@bb8]); Milton Boaz *et al*. (2005[@bb9]); Vesta *et al.* (2007[@bb11]).
Experimental {#sec2}
============
{#sec2.1}
### Crystal data {#sec2.1.1}
\[Ba(C~6~H~2~N~3~O~7~)~2~(H~2~O)~4~\]*M* *~r~* = 665.62Monoclinic,*a* = 11.6765 (4) Å*b* = 6.6878 (2) Å*c* = 27.0324 (9) Åβ = 95.608 (2)°*V* = 2100.86 (12) Å^3^*Z* = 4Mo *K*α radiationμ = 2.00 mm^−1^*T* = 293 (2) K0.20 × 0.20 × 0.15 mm
### Data collection {#sec2.1.2}
Bruker Kappa APEX2 diffractometerAbsorption correction: multi-scan (*SADABS*; Bruker, 1999[@bb4]) *T* ~min~ = 0.652, *T* ~max~ = 0.72320976 measured reflections3674 independent reflections3552 reflections with *I* \> 2σ(*I*)*R* ~int~ = 0.020
### Refinement {#sec2.1.3}
*R*\[*F* ^2^ \> 2σ(*F* ^2^)\] = 0.044*wR*(*F* ^2^) = 0.084*S* = 1.423674 reflections360 parameters234 restraintsH atoms treated by a mixture of independent and constrained refinementΔρ~max~ = 0.58 e Å^−3^Δρ~min~ = −0.53 e Å^−3^
{#d5e603}
Data collection: *APEX2* (Bruker, 2004[@bb5]); cell refinement: *APEX2* and *SAINT* (Bruker, 2004[@bb5]); data reduction: *SAINT* and *XPREP* (Bruker, 2004[@bb5]); program(s) used to solve structure: *SIR92* (Altomare *et al.*, 1993[@bb1]); program(s) used to refine structure: *SHELXL97* (Sheldrick, 2008[@bb10]); molecular graphics: *ORTEP-3* (Farrugia, 1997[@bb7]) and *Mercury* (Macrae *et al.*, 2006[@bb6]); software used to prepare material for publication: *SHELXL97*.
Supplementary Material
======================
Crystal structure: contains datablocks global, I. DOI: [10.1107/S1600536808002961/sg2220sup1.cif](http://dx.doi.org/10.1107/S1600536808002961/sg2220sup1.cif)
Structure factors: contains datablocks I. DOI: [10.1107/S1600536808002961/sg2220Isup2.hkl](http://dx.doi.org/10.1107/S1600536808002961/sg2220Isup2.hkl)
Additional supplementary materials: [crystallographic information](http://scripts.iucr.org/cgi-bin/sendsupfiles?sg2220&file=sg2220sup0.html&mime=text/html); [3D view](http://scripts.iucr.org/cgi-bin/sendcif?sg2220sup1&Qmime=cif); [checkCIF report](http://scripts.iucr.org/cgi-bin/paper?sg2220&checkcif=yes)
Supplementary data and figures for this paper are available from the IUCr electronic archives (Reference: [SG2220](http://scripts.iucr.org/cgi-bin/sendsup?sg2220)).
The authors thank the Sophisticated Analytical Instruments Facility, Indian Institute of Technology Madras, Chennai, for the X-ray data collection.
Comment
=======
Nitrophenol family of crystals are found to have high laser damage threshold, wide transparency windows, and high NLO co-efficients ((Brahadeeswaran *et al.*, 1998, 1999), (Milton Boaz *et al.*, 2005), (Vesta *et al.*, 2007), (Jonie Varjula *et al.*, 2007)). Nitrophenol groups are found to be good proton acceptors from the metallic hydroxide complexes. The title compound was synthesized as part of our ongoing research for synthesizing and characterizing new optically active materials. In the present work, the crystal structure of the compound (BaC~12~H~4~N~6~O~14~.4H~2~O) is reported for the first time.The reported compound is not optically active.
The title compound crystallizes in monoclinic system with space group *P*2~1~/*c*. *ORTEP* representation of the molecule with 50% anisotropic ellipsoids are shown in figure1.The asymmetric unit consists of two nitrophenolate moieties coordinated to barium through phenolate O atoms O7 and O14 and one nitro oxygen each from nitrophenolate moieties (O1 and O8), on one side. Four water molecules of the asymmetric unit coordinates to other side. The asymmetric unit and its inversion are linked to each other through nitro oxygen O5(symm: 2 - *x*, 2 - *y*, -*z*) coordinating to metal. The centrosymmetric pair and its a-translations are joined to each other through nitro O atoms O11 (symm: *x* - 1, *y*, *z*) to form an one dimensional infinite polymeric chain parallel to *a* axis (Fig.2).Thus, Barium is coordinated with 10 O atoms. The coordination distances around Ba vary from 2.728 Å to 3.138 Å. The one dimensional chains are further linked to each other (along b and c directions) through water mediated O---H···O hydrogen bonds (Fig.3). The crystal sample, on exposure to air at room temperature for many days, slowly looses the water and peels out as filaments.
Experimental {#experimental}
============
Picric acid (99%, 5.73 g ms) was dissolved in deionized water (100 ml) and then Ba(OH)2 (97%, 3.94 g ms) was added slowly with stirring to obtain saturated solution. The saturated solution kept at 305 K yielded fine yellow crystals in three days through spontaneous nucleation. The sample was purified further through recrystallization.
Refinement {#refinement}
==========
The aromatic H atoms were located in Fourier difference map and geometrically constrained at idealized positions (C---H = 0.93 Å) and were given riding model refinement with *U*~iso~ equal to 1.2 times *U*~eq~ of the parent carbon. All the water H atoms were located in difference Fourier map and refined isotropically with following restrints: O---H = 0.850 (1)Å and H···H = 1.380 (1) Å. These restraints were put to avoid bad geometry after refinement. The isotropic thermal parameters of H atoms H16A, H16B, H17A, H17B, H18A and H18B were constrained as 0.08 Å^-2^ during refinement.
Figures
=======
{#Fap1}
{#Fap2}
{#Fap3}
Crystal data {#tablewrapcrystaldatalong}
============
--------------------------------------- ---------------------------------------
\[Ba(C~6~H~2~N~3~O~7~)~2~(H~2~O)~4~\] *F*~000~ = 1304
*M~r~* = 665.62 *D*~x~ = 2.104 Mg m^−3^
Monoclinic, *P*2~1~/*c* Mo *K*α radiation λ = 0.71073 Å
Hall symbol: -P2ybc Cell parameters from 6588 reflections
*a* = 11.6765 (4) Å θ = 2.4--25.0º
*b* = 6.6878 (2) Å µ = 2.00 mm^−1^
*c* = 27.0324 (9) Å *T* = 293 (2) K
β = 95.608 (2)º Plate, yellow
*V* = 2100.86 (12) Å^3^ 0.20 × 0.20 × 0.15 mm
*Z* = 4
--------------------------------------- ---------------------------------------
Data collection {#tablewrapdatacollectionlong}
===============
--------------------------------------------------------- --------------------------------------
Bruker Kappa APEX2 diffractometer 3674 independent reflections
Radiation source: fine-focus sealed tube 3552 reflections with *I* \> 2σ(*I*)
Monochromator: graphite *R*~int~ = 0.020
*T* = 293(2) K θ~max~ = 25.0º
ω and φ scans θ~min~ = 2.4º
Absorption correction: multi-scan(SADABS; Bruker, 1999) *h* = −13→13
*T*~min~ = 0.652, *T*~max~ = 0.723 *k* = −7→7
20976 measured reflections *l* = −32→32
--------------------------------------------------------- --------------------------------------
Refinement {#tablewraprefinementdatalong}
==========
---------------------------------------------------------------- ----------------------------------------------------------------------------------
Refinement on *F*^2^ Secondary atom site location: difference Fourier map
Least-squares matrix: full Hydrogen site location: inferred from neighbouring sites
*R*\[*F*^2^ \> 2σ(*F*^2^)\] = 0.044 H atoms treated by a mixture of independent and constrained refinement
*wR*(*F*^2^) = 0.084 *w* = 1/\[σ^2^(*F*~o~^2^) + 9.4089*P*\] where *P* = (*F*~o~^2^ + 2*F*~c~^2^)/3
*S* = 1.42 (Δ/σ)~max~ = 0.002
3674 reflections Δρ~max~ = 0.58 e Å^−3^
360 parameters Δρ~min~ = −0.53 e Å^−3^
234 restraints Extinction correction: none
Primary atom site location: structure-invariant direct methods
---------------------------------------------------------------- ----------------------------------------------------------------------------------
Special details {#specialdetails}
===============
-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
Geometry. All e.s.d.\'s (except the e.s.d. in the dihedral angle between two l.s. planes) are estimated using the full covariance matrix. The cell e.s.d.\'s are taken into account individually in the estimation of e.s.d.\'s in distances, angles and torsion angles; correlations between e.s.d.\'s in cell parameters are only used when they are defined by crystal symmetry. An approximate (isotropic) treatment of cell e.s.d.\'s is used for estimating e.s.d.\'s involving l.s. planes.
Refinement. Refinement of *F*^2^ against ALL reflections. The weighted *R*-factor *wR* and goodness of fit *S* are based on *F*^2^, conventional *R*-factors *R* are based on *F*, with *F* set to zero for negative *F*^2^. The threshold expression of *F*^2^ \> σ(*F*^2^) is used only for calculating *R*-factors(gt) *etc*. and is not relevant to the choice of reflections for refinement. *R*-factors based on *F*^2^ are statistically about twice as large as those based on *F*, and *R*- factors based on ALL data will be even larger.
-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
Fractional atomic coordinates and isotropic or equivalent isotropic displacement parameters (Å^2^) {#tablewrapcoords}
==================================================================================================
------ ------------- ------------- --------------- -------------------- --
*x* *y* *z* *U*~iso~\*/*U*~eq~
C1 1.2555 (4) 1.1054 (8) 0.13511 (18) 0.0242 (11)
C2 1.1690 (4) 1.1014 (7) 0.09309 (19) 0.0238 (10)
C3 1.2195 (4) 1.1225 (7) 0.04666 (18) 0.0255 (11)
C4 1.3344 (4) 1.1314 (8) 0.04212 (19) 0.0280 (11)
H4 1.3617 1.1369 0.0110 0.034\*
C5 1.4101 (4) 1.1319 (8) 0.0851 (2) 0.0271 (11)
C6 1.3720 (4) 1.1214 (7) 0.13149 (19) 0.0258 (11)
H6 1.4238 1.1251 0.1599 0.031\*
C7 1.3226 (4) 0.6198 (7) 0.09934 (18) 0.0246 (11)
C8 1.4352 (4) 0.6309 (8) 0.0885 (2) 0.0280 (12)
H8 1.4527 0.6363 0.0557 0.034\*
C9 1.5217 (4) 0.6338 (8) 0.12707 (19) 0.0268 (11)
C10 1.4971 (4) 0.6212 (8) 0.1761 (2) 0.0301 (12)
H10 1.5561 0.6187 0.2018 0.036\*
C11 1.3855 (5) 0.6125 (8) 0.18597 (19) 0.0291 (12)
C12 1.2861 (4) 0.6108 (8) 0.1489 (2) 0.0270 (11)
N1 1.2203 (4) 1.0954 (7) 0.18515 (16) 0.0324 (10)
N2 1.5322 (4) 1.1492 (8) 0.0809 (2) 0.0409 (12)
N3 1.1438 (4) 1.1353 (7) 0.00065 (17) 0.0339 (11)
N4 1.2374 (4) 0.6143 (7) 0.05641 (16) 0.0284 (10)
N5 1.6402 (4) 0.6485 (8) 0.11597 (19) 0.0384 (12)
N6 1.3656 (4) 0.6031 (9) 0.23850 (18) 0.0426 (12)
O1 1.1282 (4) 1.0179 (7) 0.19220 (15) 0.0450 (11)
O2 1.2845 (4) 1.1681 (8) 0.21915 (15) 0.0524 (12)
O3 1.5656 (4) 1.1695 (9) 0.03987 (19) 0.0671 (15)
O4 1.5960 (4) 1.1459 (10) 0.1193 (2) 0.0709 (16)
O5 1.1760 (4) 1.0643 (9) −0.03692 (16) 0.0604 (14)
O6 1.0527 (4) 1.2218 (8) 0.00097 (17) 0.0616 (14)
O7 1.0641 (3) 1.0833 (6) 0.09589 (14) 0.0359 (9)
O8 1.1364 (3) 0.6458 (8) 0.06148 (16) 0.0566 (13)
O9 1.2699 (4) 0.5776 (7) 0.01567 (14) 0.0467 (11)
O10 1.6600 (4) 0.6531 (9) 0.07262 (19) 0.0672 (15)
O11 1.7160 (3) 0.6578 (8) 0.15020 (18) 0.0573 (13)
O12 1.2822 (5) 0.6819 (10) 0.25207 (18) 0.0770 (17)
O13 1.4367 (4) 0.5156 (8) 0.26667 (17) 0.0616 (14)
O14 1.1854 (3) 0.5971 (7) 0.15895 (15) 0.0419 (10)
O15 0.9006 (4) 0.4119 (7) 0.07251 (15) 0.0403 (10)
O16 0.8349 (4) 1.1003 (8) 0.13919 (18) 0.0522 (12)
O17 0.9062 (4) 0.8444 (9) 0.22439 (17) 0.0601 (13)
O18 0.9301 (5) 0.4269 (8) 0.19520 (18) 0.0639 (14)
Ba 0.96482 (2) 0.74829 (6) 0.130390 (11) 0.02924 (11)
H15A 0.955 (3) 0.334 (7) 0.067 (2) 0.07 (3)\*
H15B 0.852 (4) 0.415 (11) 0.0470 (14) 0.08 (3)\*
H16A 0.7657 (17) 1.072 (11) 0.144 (3) 0.080\*
H16B 0.835 (5) 1.182 (10) 0.115 (2) 0.080\*
H17A 0.852 (3) 0.930 (6) 0.222 (2) 0.080\*
H17B 0.948 (5) 0.866 (11) 0.2514 (15) 0.080\*
H18A 0.925 (6) 0.318 (5) 0.179 (2) 0.080\*
H18B 0.901 (6) 0.410 (10) 0.2224 (15) 0.080\*
------ ------------- ------------- --------------- -------------------- --
Atomic displacement parameters (Å^2^) {#tablewrapadps}
=====================================
----- -------------- -------------- -------------- --------------- -------------- --------------
*U*^11^ *U*^22^ *U*^33^ *U*^12^ *U*^13^ *U*^23^
C1 0.030 (3) 0.020 (2) 0.023 (2) 0.002 (2) 0.003 (2) 0.002 (2)
C2 0.019 (2) 0.017 (2) 0.035 (3) −0.002 (2) 0.003 (2) 0.001 (2)
C3 0.031 (3) 0.017 (3) 0.028 (3) −0.004 (2) 0.000 (2) 0.001 (2)
C4 0.033 (3) 0.023 (3) 0.030 (3) −0.004 (2) 0.010 (2) −0.003 (2)
C5 0.023 (3) 0.021 (3) 0.038 (3) 0.002 (2) 0.004 (2) −0.003 (2)
C6 0.024 (3) 0.019 (3) 0.033 (3) 0.003 (2) −0.003 (2) −0.002 (2)
C7 0.026 (3) 0.018 (2) 0.029 (3) 0.000 (2) −0.004 (2) 0.000 (2)
C8 0.027 (3) 0.025 (3) 0.032 (3) −0.001 (2) 0.006 (2) 0.002 (2)
C9 0.019 (2) 0.024 (3) 0.038 (3) 0.000 (2) 0.003 (2) 0.002 (2)
C10 0.026 (3) 0.027 (3) 0.036 (3) −0.002 (2) −0.007 (2) 0.003 (2)
C11 0.030 (3) 0.030 (3) 0.027 (3) 0.000 (2) 0.001 (2) −0.001 (2)
C12 0.022 (3) 0.021 (3) 0.037 (3) 0.000 (2) 0.001 (2) 0.001 (2)
N1 0.031 (2) 0.036 (3) 0.030 (2) −0.001 (2) 0.0047 (19) −0.002 (2)
N2 0.028 (2) 0.042 (3) 0.054 (3) 0.001 (2) 0.008 (2) −0.004 (2)
N3 0.033 (3) 0.036 (3) 0.032 (2) −0.010 (2) −0.0009 (19) 0.005 (2)
N4 0.027 (2) 0.023 (2) 0.034 (2) 0.0018 (19) −0.0030 (18) −0.0015 (19)
N5 0.023 (2) 0.041 (3) 0.051 (3) −0.002 (2) 0.005 (2) 0.003 (2)
N6 0.035 (3) 0.054 (3) 0.039 (3) −0.001 (3) 0.004 (2) 0.000 (3)
O1 0.040 (2) 0.058 (3) 0.039 (2) −0.016 (2) 0.0174 (18) −0.002 (2)
O2 0.050 (3) 0.075 (3) 0.032 (2) −0.011 (2) 0.0019 (19) −0.015 (2)
O3 0.042 (3) 0.095 (4) 0.069 (3) −0.010 (3) 0.029 (2) −0.006 (3)
O4 0.023 (2) 0.115 (4) 0.073 (3) 0.002 (3) −0.004 (2) 0.008 (3)
O5 0.046 (3) 0.102 (4) 0.033 (2) −0.012 (3) 0.0025 (19) −0.013 (3)
O6 0.058 (3) 0.064 (3) 0.058 (3) 0.024 (3) −0.019 (2) 0.002 (3)
O7 0.0214 (19) 0.039 (2) 0.047 (2) −0.0024 (17) 0.0017 (16) 0.0081 (19)
O8 0.025 (2) 0.091 (4) 0.051 (3) 0.015 (2) −0.0070 (18) −0.014 (3)
O9 0.042 (2) 0.070 (3) 0.028 (2) −0.006 (2) 0.0009 (17) 0.003 (2)
O10 0.040 (3) 0.106 (4) 0.059 (3) 0.001 (3) 0.019 (2) 0.013 (3)
O11 0.022 (2) 0.081 (3) 0.068 (3) −0.007 (2) −0.003 (2) −0.003 (3)
O12 0.070 (3) 0.117 (5) 0.047 (3) 0.026 (3) 0.019 (2) −0.006 (3)
O13 0.065 (3) 0.077 (4) 0.040 (2) −0.001 (3) −0.008 (2) 0.015 (2)
O14 0.024 (2) 0.055 (3) 0.047 (2) 0.0024 (19) 0.0071 (17) 0.003 (2)
O15 0.040 (2) 0.039 (2) 0.040 (2) 0.004 (2) −0.0036 (19) −0.0070 (19)
O16 0.033 (2) 0.054 (3) 0.069 (3) 0.006 (2) 0.004 (2) −0.005 (2)
O17 0.060 (3) 0.084 (4) 0.038 (2) −0.006 (3) 0.008 (2) −0.001 (2)
O18 0.082 (4) 0.059 (3) 0.048 (3) −0.010 (3) −0.007 (3) 0.011 (2)
Ba 0.02237 (16) 0.03288 (18) 0.03255 (17) −0.00217 (15) 0.00312 (11) 0.00155 (16)
----- -------------- -------------- -------------- --------------- -------------- --------------
Geometric parameters (Å, °) {#tablewrapgeomlong}
===========================
----------------------- ------------ --------------------------- -------------
C1---C6 1.378 (7) N3---O6 1.211 (6)
C1---C2 1.444 (7) N3---O5 1.214 (6)
C1---N1 1.453 (6) N4---O8 1.218 (6)
C2---O7 1.241 (6) N4---O9 1.224 (6)
C2---C3 1.445 (7) N5---O10 1.216 (6)
C3---C4 1.360 (7) N5---O11 1.218 (6)
C3---N3 1.457 (7) N6---O12 1.196 (7)
C4---C5 1.388 (7) N6---O13 1.220 (7)
C4---H4 0.9300 O1---Ba 3.010 (4)
C5---C6 1.374 (7) O5---Ba^i^ 3.138 (5)
C5---N2 1.446 (7) O7---Ba 2.728 (4)
C6---H6 0.9300 O8---Ba 2.947 (4)
C7---C8 1.377 (7) O11---Ba^ii^ 3.065 (4)
C7---C12 1.446 (7) O14---Ba 2.805 (4)
C7---N4 1.453 (6) O15---Ba 2.801 (4)
C8---C9 1.379 (7) O15---H15A 0.85 (4)
C8---H8 0.9300 O15---H15B 0.85 (4)
C9---C10 1.385 (7) O16---Ba 2.823 (5)
C9---N5 1.448 (6) O16---H16A 0.85 (3)
C10---C11 1.356 (7) O16---H16B 0.85 (6)
C10---H10 0.9300 O17---Ba 2.771 (4)
C11---C12 1.458 (7) O17---H17A 0.85 (4)
C11---N6 1.463 (7) O17---H17B 0.85 (5)
C12---O14 1.236 (6) O18---Ba 2.827 (5)
N1---O1 1.225 (6) O18---H18A 0.85 (4)
N1---O2 1.228 (6) O18---H18B 0.85 (5)
N2---O3 1.218 (6) Ba---O11^iii^ 3.065 (4)
N2---O4 1.218 (7) Ba---O5^i^ 3.138 (5)
C6---C1---C2 124.4 (4) C12---O14---Ba 141.3 (4)
C6---C1---N1 116.1 (4) Ba---O15---H15A 116 (3)
C2---C1---N1 119.5 (4) Ba---O15---H15B 124 (5)
O7---C2---C1 124.9 (5) H15A---O15---H15B 109 (5)
O7---C2---C3 123.5 (5) Ba---O16---H16A 111 (5)
C1---C2---C3 111.6 (4) Ba---O16---H16B 116 (5)
C4---C3---C2 124.9 (5) H16A---O16---H16B 109 (6)
C4---C3---N3 116.3 (5) Ba---O17---H17A 109 (5)
C2---C3---N3 118.8 (4) Ba---O17---H17B 131 (5)
C3---C4---C5 118.5 (5) H17A---O17---H17B 108 (6)
C3---C4---H4 120.7 Ba---O18---H18A 110 (4)
C5---C4---H4 120.7 Ba---O18---H18B 137 (5)
C6---C5---C4 121.8 (5) H18A---O18---H18B 109 (6)
C6---C5---N2 119.1 (5) O7---Ba---O17 105.91 (15)
C4---C5---N2 119.1 (5) O7---Ba---O15 124.63 (12)
C5---C6---C1 118.6 (5) O17---Ba---O15 128.47 (15)
C5---C6---H6 120.7 O7---Ba---O14 88.90 (12)
C1---C6---H6 120.7 O17---Ba---O14 97.68 (14)
C8---C7---C12 125.0 (5) O15---Ba---O14 93.02 (13)
C8---C7---N4 115.1 (4) O7---Ba---O16 66.02 (13)
C12---C7---N4 119.9 (4) O17---Ba---O16 63.03 (16)
C7---C8---C9 118.9 (5) O15---Ba---O16 126.89 (13)
C7---C8---H8 120.5 O14---Ba---O16 139.82 (13)
C9---C8---H8 120.5 O7---Ba---O18 158.07 (13)
C8---C9---C10 121.1 (5) O17---Ba---O18 62.93 (17)
C8---C9---N5 119.2 (5) O15---Ba---O18 71.85 (14)
C10---C9---N5 119.6 (5) O14---Ba---O18 74.86 (15)
C11---C10---C9 119.0 (5) O16---Ba---O18 118.10 (16)
C11---C10---H10 120.5 O7---Ba---O8 68.72 (14)
C9---C10---H10 120.5 O17---Ba---O8 151.36 (13)
C10---C11---C12 125.5 (5) O15---Ba---O8 67.76 (13)
C10---C11---N6 116.1 (5) O14---Ba---O8 55.05 (12)
C12---C11---N6 118.4 (5) O16---Ba---O8 130.63 (15)
O14---C12---C7 125.4 (5) O18---Ba---O8 111.25 (17)
O14---C12---C11 124.0 (5) O7---Ba---O1 55.28 (11)
C7---C12---C11 110.5 (4) O17---Ba---O1 63.37 (13)
O1---N1---O2 122.1 (5) O15---Ba---O1 155.78 (13)
O1---N1---C1 119.7 (4) O14---Ba---O1 63.12 (13)
O2---N1---C1 118.1 (4) O16---Ba---O1 76.72 (13)
O3---N2---O4 123.7 (5) O18---Ba---O1 103.48 (13)
O3---N2---C5 119.1 (5) O8---Ba---O1 93.27 (12)
O4---N2---C5 117.2 (5) O7---Ba---O11^iii^ 131.55 (13)
O6---N3---O5 122.5 (5) O17---Ba---O11^iii^ 64.06 (14)
O6---N3---C3 118.8 (5) O15---Ba---O11^iii^ 74.48 (13)
O5---N3---C3 118.6 (5) O14---Ba---O11^iii^ 137.85 (13)
O8---N4---O9 121.6 (4) O16---Ba---O11^iii^ 67.90 (14)
O8---N4---C7 120.0 (4) O18---Ba---O11^iii^ 62.99 (16)
O9---N4---C7 118.3 (4) O8---Ba---O11^iii^ 141.18 (13)
O10---N5---O11 122.6 (5) O1---Ba---O11^iii^ 125.55 (12)
O10---N5---C9 118.4 (5) O7---Ba---O5^i^ 66.63 (12)
O11---N5---C9 119.0 (5) O17---Ba---O5^i^ 119.27 (15)
O12---N6---O13 123.0 (6) O15---Ba---O5^i^ 77.31 (14)
O12---N6---C11 119.4 (5) O14---Ba---O5^i^ 139.51 (12)
O13---N6---C11 117.6 (5) O16---Ba---O5^i^ 59.42 (14)
N1---O1---Ba 133.3 (3) O18---Ba---O5^i^ 134.99 (15)
N3---O5---Ba^i^ 109.7 (4) O8---Ba---O5^i^ 85.35 (12)
C2---O7---Ba 123.8 (3) O1---Ba---O5^i^ 117.49 (13)
N4---O8---Ba 146.8 (3) O11^iii^---Ba---O5^i^ 77.80 (13)
N5---O11---Ba^ii^ 120.5 (4)
C6---C1---C2---O7 178.6 (5) C10---C11---N6---O12 146.4 (6)
N1---C1---C2---O7 −2.5 (8) C12---C11---N6---O12 −33.9 (9)
C6---C1---C2---C3 −1.8 (7) C10---C11---N6---O13 −32.8 (8)
N1---C1---C2---C3 177.2 (4) C12---C11---N6---O13 146.9 (6)
O7---C2---C3---C4 −176.1 (5) O2---N1---O1---Ba −164.9 (4)
C1---C2---C3---C4 4.2 (7) C1---N1---O1---Ba 15.9 (8)
O7---C2---C3---N3 3.8 (8) O6---N3---O5---Ba^i^ 17.7 (7)
C1---C2---C3---N3 −175.8 (4) C3---N3---O5---Ba^i^ −160.0 (3)
C2---C3---C4---C5 −3.9 (8) C1---C2---O7---Ba −65.4 (6)
N3---C3---C4---C5 176.2 (5) C3---C2---O7---Ba 115.0 (5)
C3---C4---C5---C6 0.8 (8) O9---N4---O8---Ba −178.2 (5)
C3---C4---C5---N2 −177.6 (5) C7---N4---O8---Ba 1.9 (10)
C4---C5---C6---C1 1.5 (8) O10---N5---O11---Ba^ii^ 7.9 (8)
N2---C5---C6---C1 179.9 (5) C9---N5---O11---Ba^ii^ −171.3 (4)
C2---C1---C6---C5 −0.9 (8) C7---C12---O14---Ba −42.3 (9)
N1---C1---C6---C5 −179.8 (5) C11---C12---O14---Ba 140.1 (5)
C12---C7---C8---C9 0.0 (8) C2---O7---Ba---O17 105.6 (4)
N4---C7---C8---C9 −179.1 (5) C2---O7---Ba---O15 −85.0 (4)
C7---C8---C9---C10 1.5 (8) C2---O7---Ba---O14 7.9 (4)
C7---C8---C9---N5 −179.1 (5) C2---O7---Ba---O16 155.6 (4)
C8---C9---C10---C11 −2.2 (8) C2---O7---Ba---O18 49.6 (6)
N5---C9---C10---C11 178.4 (5) C2---O7---Ba---O8 −44.7 (4)
C9---C10---C11---C12 1.5 (9) C2---O7---Ba---O1 65.4 (4)
C9---C10---C11---N6 −178.9 (5) C2---O7---Ba---O11^iii^ 174.8 (4)
C8---C7---C12---O14 −178.6 (5) C2---O7---Ba---O5^i^ −138.9 (4)
N4---C7---C12---O14 0.5 (8) C12---O14---Ba---O7 −25.0 (6)
C8---C7---C12---C11 −0.7 (7) C12---O14---Ba---O17 −130.9 (6)
N4---C7---C12---C11 178.4 (4) C12---O14---Ba---O15 99.6 (6)
C10---C11---C12---O14 177.9 (6) C12---O14---Ba---O16 −74.3 (6)
N6---C11---C12---O14 −1.8 (8) C12---O14---Ba---O18 169.9 (6)
C10---C11---C12---C7 0.0 (8) C12---O14---Ba---O8 39.5 (6)
N6---C11---C12---C7 −179.7 (5) C12---O14---Ba---O1 −76.1 (6)
C6---C1---N1---O1 −156.2 (5) C12---O14---Ba---O11^iii^ 169.6 (5)
C2---C1---N1---O1 24.8 (7) C12---O14---Ba---O5^i^ 25.7 (7)
C6---C1---N1---O2 24.6 (7) N4---O8---Ba---O7 85.4 (8)
C2---C1---N1---O2 −154.4 (5) N4---O8---Ba---O17 1.1 (10)
C6---C5---N2---O3 −176.0 (6) N4---O8---Ba---O15 −129.7 (8)
C4---C5---N2---O3 2.4 (8) N4---O8---Ba---O14 −18.9 (7)
C6---C5---N2---O4 2.7 (8) N4---O8---Ba---O16 110.0 (8)
C4---C5---N2---O4 −178.9 (6) N4---O8---Ba---O18 −71.1 (8)
C4---C3---N3---O6 −144.6 (5) N4---O8---Ba---O1 34.8 (8)
C2---C3---N3---O6 35.4 (7) N4---O8---Ba---O11^iii^ −144.0 (7)
C4---C3---N3---O5 33.1 (7) N4---O8---Ba---O5^i^ 152.1 (8)
C2---C3---N3---O5 −146.9 (5) N1---O1---Ba---O7 −45.6 (5)
C8---C7---N4---O8 −164.9 (5) N1---O1---Ba---O17 178.3 (6)
C12---C7---N4---O8 15.9 (7) N1---O1---Ba---O15 52.7 (7)
C8---C7---N4---O9 15.2 (7) N1---O1---Ba---O14 63.4 (5)
C12---C7---N4---O9 −164.0 (5) N1---O1---Ba---O16 −115.5 (5)
C8---C9---N5---O10 −1.7 (8) N1---O1---Ba---O18 128.4 (5)
C10---C9---N5---O10 177.7 (6) N1---O1---Ba---O8 15.6 (5)
C8---C9---N5---O11 177.6 (5) N1---O1---Ba---O11^iii^ −165.4 (5)
C10---C9---N5---O11 −3.0 (8) N1---O1---Ba---O5^i^ −70.9 (5)
----------------------- ------------ --------------------------- -------------
Symmetry codes: (i) −*x*+2, −*y*+2, −*z*; (ii) *x*+1, *y*, *z*; (iii) *x*−1, *y*, *z*.
Hydrogen-bond geometry (Å, °) {#tablewraphbondslong}
=============================
----------------------- ---------- ------------ ----------- ---------------
*D*---H···*A* *D*---H H···*A* *D*···*A* *D*---H···*A*
O15---H15A···O7^iv^ 0.85 (4) 2.20 (4) 2.939 (6) 145 (6)
O15---H15A···O6^iv^ 0.85 (4) 2.33 (4) 3.030 (7) 140 (5)
O15---H15B···O9^v^ 0.85 (4) 2.103 (6) 2.953 (6) 179 (7)
O16---H16A···O4^iii^ 0.85 (3) 2.08 (4) 2.806 (6) 142 (5)
O16---H16B···O15^vi^ 0.85 (6) 2.11 (3) 2.906 (7) 156 (6)
O17---H17A···O12^vii^ 0.85 (4) 2.45 (3) 3.258 (8) 158 (6)
O17---H17B···O18^vii^ 0.85 (5) 1.969 (17) 2.805 (7) 168 (7)
O18---H18A···O16^iv^ 0.85 (4) 2.04 (3) 2.822 (7) 153 (7)
O18---H18B···O1^viii^ 0.85 (5) 2.47 (3) 3.241 (7) 151 (6)
----------------------- ---------- ------------ ----------- ---------------
Symmetry codes: (iv) *x*, *y*−1, *z*; (v) −*x*+2, −*y*+1, −*z*; (iii) *x*−1, *y*, *z*; (vi) *x*, *y*+1, *z*; (vii) −*x*+2, *y*+1/2, −*z*+1/2; (viii) −*x*+2, *y*−1/2, −*z*+1/2.
###### Hydrogen-bond geometry (Å, °)
*D*---H⋯*A* *D*---H H⋯*A* *D*⋯*A* *D*---H⋯*A*
---------------------- ---------- ------------ ----------- -------------
O15---H15*A*⋯O7^i^ 0.85 (4) 2.20 (4) 2.939 (6) 145 (6)
O15---H15*A*⋯O6^i^ 0.85 (4) 2.33 (4) 3.030 (7) 140 (5)
O15---H15*B*⋯O9^ii^ 0.85 (4) 2.103 (6) 2.953 (6) 179 (7)
O16---H16*A*⋯O4^iii^ 0.85 (3) 2.08 (4) 2.806 (6) 142 (5)
O16---H16*B*⋯O15^iv^ 0.85 (6) 2.11 (3) 2.906 (7) 156 (6)
O17---H17*A*⋯O12^v^ 0.85 (4) 2.45 (3) 3.258 (8) 158 (6)
O17---H17*B*⋯O18^v^ 0.85 (5) 1.969 (17) 2.805 (7) 168 (7)
O18---H18*A*⋯O16^i^ 0.85 (4) 2.04 (3) 2.822 (7) 153 (7)
O18---H18*B*⋯O1^vi^ 0.85 (5) 2.47 (3) 3.241 (7) 151 (6)
Symmetry codes: (i) ; (ii) ; (iii) ; (iv) ; (v) ; (vi) .
| {
"pile_set_name": "PubMed Central"
} |
1. Genomic Stability, Inflammation, and Aging {#sec1}
=============================================
Aging has been defined as a progressive postmaturational decline in physiological capacity, accompanied by an increased susceptibility to disease and an increased mortality risk. Mechanisms to maintain genomic stability are thought to counteract the aging process, whereas inflammation is considered a driving force of human aging \[[@B1]\].
A large body of evidence supports the theory that genomic instability acts as a causative factor in the aging process, which is evident from the fact that most mouse models of premature aging as well as human progeria syndromes are related to defects in mechanisms of genomic maintenance \[[@B2]\]. This may be attributed to the fact that DNA serves as a blueprint of all cellular RNAs and proteins. Any acquired change in its sequence, which may arise from molecular damage, is permanent and thus may have irreversible consequences. For this reason nature invested in a sophisticated network of various mechanisms (i) to maintain genome integrity, such as DNA repair and cell cycle control, and (ii) to withdraw heavily damaged cells from the body, such as apoptosis and cellular senescence. However, even if these mechanisms may be very efficient they cannot cope with all the insults induced in the genome, leading to a gradual accumulation of DNA damage and mutations, thus contributing to organismic aging \[[@B2]\].
On the other hand, a direct relationship exists between physiological aging and increasing incidence of chronic inflammatory diseases. In its acute form, inflammation acts as a protective mechanism in response to pathogen invasion or tissue damage and helps to restore physiological integrity and function. However, in its chronic form, inflammation can exert detrimental effects on the cellular as well as the organismic level. Chronically inflamed tissue is characterized by infiltration of immune cells, neovascularization, fibrosis, and often tissue damage and necrosis \[[@B3]\]. The innate immune system, especially the mononuclear phagocyte system, is the most important mediator of chronic inflammation. Monocytes originate from the myeloid hematopoietic cell lineage in bone marrow. In the blood stream, monocytes are recruited by specific stimuli into different tissues, where they differentiate into phagocytic macrophages. Macrophages participate in the killing of invading microorganisms and emerging tumor cells through the production of reactive oxygen or nitrogen species (ROS and RNS). In addition, macrophages secrete cytokines, which play a key role in the regulation of multiple immune functions, especially inflammatory responses \[[@B3]\]. During aging, the continuous pressure on the immune system caused by repeated antigen stimulation, such as infections, food antigens, allergens, and self antigens, leads to an increase in activated cells and secretion of proinflammatory cytokines, such as TNF*α* \[[@B4]\]. These circulating proinflammatory factors may keep the immune system in a state of chronic low-level activation, a phenomenon described as "inflammaging" \[[@B5], [@B6]\]. Eventually, this causes "immunosenescence," that is, an age-related decline in the capacity of adaptive immunity, consisting of more specific responses carried out by B and T cells \[[@B7]\]. Thus, with advanced age, the immune system undergoes a gradual remodeling in the attempt to reestablish a new balance that assures survival, however, favoring the development of chronic inflammatory conditions \[[@B5], [@B6], [@B8], [@B9]\].
DNA damage and inflammation are inevitably linked by the production of reactive chemical species, such as ROS and RNS. Cellular ROS and RNS production occurs constantly under physiological as well as pathophysiological conditions as a consequence of electron leakage of the mitochondrial electron transport chain and via enzymes such as NADPH oxidase, nitric oxide synthases, and xanthine oxidase. The "free radical theory of aging" posits that aging and its related diseases are the net consequence of free radical-induced damage and the inability to counterbalance these changes by antioxidative defenses and sufficient DNA repair \[[@B10]\]. Chronic inflammation results in the generation of a broad spectrum of ROS and RNS by activated macrophages and neutrophils, which damage cellular macromolecules including DNA \[[@B11], [@B12]\]. Conversely, the generation of ROS and RNS activates redox sensitive transcription factors, such as NF-*κ*B, resulting in the generation of proinflammatory molecules. Moreover, DNA damage can induce cellular senescence, a tumor suppressive mechanism that is also associated with aging, leading to the secretion of inflammatory cytokines, a paracrine effect known as senescence-associated secretory phenotype (SASP) \[[@B5], [@B13]\]. Altogether, this can trigger a positive feedback loop that amplifies the processes of inflammation, damage, and destruction in target cells and organs, leading to an organismic decline and death over time. For example, chronic inflammation has been associated with an age-related decline in the function of hematopoietic and mesenchymal stem cells \[[@B14], [@B15]\] and has been implicated as a mediator of almost all of the aging-associated diseases, such as vascular diseases, diabetes, neurodegenerative diseases, and cancer \[[@B3], [@B5], [@B6], [@B9]\].
As discussed below, the nuclear enzyme PARP1 represents a factor that works at the interface between genomic maintenance and inflammation. Therefore, PARP1 may act in an antagonistic pleiotropic way, that is, functioning as a longevity assurance factor at younger age or in physiological conditions and as an aging-promoting factor at older age or in pathophysiological conditions. In this paper, we will discuss the numerous cellular functions of PARP1 in the context of mechanisms of longevity and aging and will put this into an organismic perspective by briefly summarizing *in vivo* studies in mice and humans.
2. PARP1 and Poly(ADP-ribosyl)ation {#sec2}
===================================
Poly(ADP-ribosyl)ation is a posttranslational modification of proteins that occurs in most eukaryotic organisms. The reaction is carried out by enzymes of the family of poly(ADP-ribose) polymerases (PARPs) by using NAD^+^ as a substrate to synthesize the linear or branched biopolymer poly(ADP-ribose) (PAR), which consist of up to 200 ADP-ribose subunits ([Figure 1](#fig1){ref-type="fig"}) \[[@B16]\]. PARP activation leads to covalent modification of various proteins with PAR including PARPs themselves, as some of them catalyze their automodification. Covalent linkage is mediated either through attachment on glutamate, aspartate, or lysine residues of the acceptor proteins \[[@B16]\]. Apart from covalent modification, some proteins can also bind preexisting ADP-ribose chains in a noncovalent fashion, and this binding is mediated via at least three different PAR binding motifs. Those include (i) a 20 amino acid motif, (ii) distinct macrodomains, and (iii) a PAR-binding zinc finger, all of which fulfill diverse cellular functions \[[@B17]--[@B22]\]. Whereas the PAR-binding macrodomains and zinc fingers are present in a limited number of human proteins (\<50), the 20-aa motif has been identified in several hundred human protein sequences \[[@B17], [@B18]\]. This weakly conserved motif consists of (i) a cluster rich in basic amino acids and (ii) a pattern of hydrophobic amino acids interspersed with basic residues \[[@B17], [@B18]\]. Most of the putative PAR-binding proteins identified are involved in a wide spectrum of cellular mechanisms such as genomic maintenance, chromatin remodeling, transcription, replication, RNA metabolism, inflammation, cell cycle control, and cell death \[[@B18]\]. In general, poly(ADP-ribosyl)ation modulates target protein function by modifying enzymatic activities or interactions with other macromolecules such as DNA, RNA, or proteins \[[@B23]\].
Importantly, the cellular existence of PAR is transient, since the polymer is rapidly hydrolyzed by PARPs catabolic counterpart, poly(ADP-ribose) glycohydrolase (PARG). PARG possesses both exo- and endoglycosidic activities and is encoded by a single gene giving rise to at least five different splice variants with distinct subcellular localizations \[[@B24]--[@B26]\]. In addition, a second enzyme was identified with weak PARG activity, that is, ADP-ribose-arginine protein hydrolase 3 (ARH3), with evidence that this enzyme is associated with PAR degradation in mitochondria \[[@B27], [@B28]\].
The PARP gene family consists of 17 homologues in the human genome \[[@B16]\]. PARP1 is the founding member of the gene family. It exhibits key roles in the regulation of nuclear and cellular functions and can be activated either by DNA damage, posttranslational protein modifications, or potentially by direct protein-protein interactions \[[@B29]\]. The strongest stimulation of PARP1 activity is mediated by its binding to DNA strand breaks which induces its catalytic activation as a monomer or dimer by several hundred-fold \[[@B30]--[@B33]\]. Under these conditions, PARP1 accounts for \>75% of the overall cellular poly(ADP-ribosyl)ation capacity \[[@B34], [@B35]\]. Apart from direct DNA damage-dependent PARP1 activation, its activity is also regulated by posttranslational modifications such as phosphorylation, acetylation, and sumoylation \[[@B36]--[@B40]\]. Moreover, PARP1 activity is subject to regulation by direct protein-protein interactions \[[@B41]--[@B43]\].
Three nonexclusive mechanisms of the cellular functions of PARP1 can be distinguished: (i) functions that rely on the enzymatic activity of PARP1 and the subsequent covalent modification or noncovalent interaction of nuclear proteins with PAR; (ii) direct interactions of proteins with PARP1 via protein-protein interaction, for example, via the BRCT domain; (iii) intervention in the cellular NAD^+^ metabolism by excessive PARP1 stimulation and potential signaling functions of free PAR or its derivatives. The consequences of these actions with regard to modulation of genomic maintenance, chromatin structure, inflammation, and cell death are discussed below.
3. PARP1 in Genomic Maintenance {#sec3}
===============================
It is estimated that thousands of DNA damage lesions occur in a mammalian cell per day, all of which need to be repaired to ensure genomic stability and longevity. In mammals, at least six major DNA repair pathways exist, that is, O^6^-methyl guanine methyltransferase (MGMT), base excision repair (BER), nucleotide excision repair (NER), mismatch repair (MMR), and DNA double-strand break (DSB) repair including the subpathways homologous recombination (HR) and nonhomologous end-joining (NHEJ) \[[@B2]\].
Except for the MGMT and MMR pathways, there is evidence that PARP1 is involved in all of these repair mechanisms, and therefore, PARP1 is considered a general caretaker of genomic stability \[[@B29]\]. Of note, the recruitment of PARP1 to sites of DNA damage and the subsequent production of PAR can occur within seconds and is one of the fastest DNA damage responses \[[@B44], [@B45]\]. Apart from this direct involvement in several DNA repair mechanisms, PARP1 participates in genomic maintenance through its role as a regulator of chromatin structure and cell cycle regulation ([Figure 2](#fig2){ref-type="fig"}).
Several cellular studies support a role of PARP1 as a general cell survival factor upon genotoxic stimuli: transdominant inhibition of PARP1 by overexpression of its DNA binding domain potentiates cytotoxicity upon treatment of cells with alkylating agents and ionizing radiation \[[@B46]\]. Moreover, PARP1-deficient cells exhibit an enhanced sensitivity to alkylating agents \[[@B47], [@B48]\] and show increased frequencies of sister chromatid exchanges, both under basal conditions and upon treatment with alkylating agents \[[@B49], [@B50]\]. Consistent with this, overexpression studies demonstrated that PARP1 acts as a negative regulator of alkylation-induced sister chromatid exchange \[[@B51]\], and *ex vivo* supplementation of human PBMC with the NAD^+^ precursor nicotinic acid enhances cellular poly(ADP-ribosyl)ation and improves cell viability upon induction of genotoxic stress \[[@B52]\].
3.1. PARP1 and Chromatin Regulation {#sec3.1}
-----------------------------------
Chromatin is a complex of DNA and proteins with a dynamic structure and is involved in replication, transcription, and other fundamental cellular processes. Structural and functional alterations of chromatin are widely associated with aging from yeast to mammals \[[@B53]\]. The molecular mechanisms leading to chromatin disturbances in aging are largely unknown, but may be related to accumulation of unrepaired DNA damage. On the other hand, alterations in chromatin structure increase the susceptibility to DNA damage, suggesting the presence of a positive feedback mechanism of DNA damage leading to chromatin rearrangements which, in turn, sensitizes DNA as a substrate for further damage. Moreover, there is evidence that chromatin defects lead to alterations in transcriptional programs thereby contributing to the aging process \[[@B53]\].
PARP1 acts as a structural and regulatory component of chromatin, both in undamaged cells and upon genotoxic stress. It may either regulate chromatin structure directly by poly(ADP-ribosyl)ation of chromatin components, or indirectly by controlling the recruitment of chromatin remodeling factors \[[@B54]\]. Many PAR acceptor and binding proteins contribute to chromatin and nuclear architecture such as histones, lamins, high-mobility group (HMG) proteins, heterochromatin protein 1 (HP1), and the DEK protein \[[@B54]--[@B59]\]. It was proposed that PARP1 induces a histone-shuttling mechanism, based on findings that poly(ADP-ribosyl)ation of polynucleosomes causes relaxation of chromatin structure and that activity of PARG degrades PAR from modified histones \[[@B60]--[@B63]\]. According to this model, DNA-bound histones dissociate from DNA upon poly(ADP-ribosyl)ation, causing an open chromatin structure and guiding repair factors to sites of DNA damage. Upon degradation of PAR by PARG, DNA reassociates with histones, thereby restoring the condensed chromatin structure. Moreover, upon DNA damage, PARP1 activation leads to the recruitment of the histone variant macroH2A1.1 to the site of the damage, which transiently causes chromatin rearrangements and dynamically modulates the DNA damage response \[[@B22]\]. Kim et al. reported that PARP1 itself can function as a component of chromatin \[[@B64]\], that is, histone H1 and PARP1 bind in a competitive and mutually exclusive manner to nucleosomes*in vitro*. Thereby, PARP1 promotes the local compaction of chromatin into higher order structures, which are associated with transcriptional repression. The authors suggested that PARP1 modulates the chromatin architecture and gene transcription through its intrinsic enzymatic activity in a DNA damage-independent manner; that is, PARP1 activation and automodification trigger its release from chromatin, thereby facilitating chromatin decondensation and gene transcription by RNA polymerase II. Subsequent cellular studies demonstrated that PARP1 could replace histone H1 at RNA polymerase II-transcribed promoters, which was associated with actively transcribed genes \[[@B65]\].
Apart from a functional interplay between PARP1 with histones, an interesting physical and functional interaction exists between PARP1 and DEK. The DEK protein is a major nonhistone chromatin component with functions in DNA metabolism and repair on a cellular, and carcinogenesis and autoimmunity on an organismic level. DEK is often found to be upregulated in tumor tissue, and high levels of DEK favor cell immortalization by inhibiting senescence and apoptosis. Consistently, DEK deficient cells are prone to induction of senescence in the response to genotoxic stress \[[@B66]\]. We and others have shown that PARP1 poly(ADP-ribosyl)ates DEK. Moreover, DEK interacts with PAR in a non-covalent manner which regulates its DNA binding affinity and multimerization with possible implications in response to genotoxic stress and gene transcription. In terms of gene transcription, DEK is released from chromatin upon poly(ADP-ribosyl)ation to permit transcriptional initiation \[[@B56], [@B59], [@B67]\]. Whether DEK itself or its interplay with PARP1 has a direct role in aging mechanisms remains to be clarified.
PARP1 activation is also necessary for the exchange of histone H1 with high-mobility group B (HMGB) proteins, which are non-histone chromatin-associated proteins that bend DNA and recruit transcription factors to their DNA targets \[[@B68]\]. Interestingly, during inflammation, HMGB1 can be secreted by activated cells, where it inhibits phagocytic uptake of dying cells by macrophages. HMGB1 secreted upon specific stimuli, that is, TLR4 stimulation is highly poly(ADP-ribosyl)ated, which enhances the inhibitory effect of HMGB1 on macrophage-dependent phagocytosis. This indicates a regulatory role of PAR in such inflammatory mechanisms with potential implications in mechanisms of aging as discussed below \[[@B69]\].
Importantly, not only structural components of the chromatin are regulated by poly(ADP-ribosyl)ation, PAR also serves as an important factor in the regulation of chromatin remodeling factors, such as ALC1 and NURD \[[@B21], [@B70]--[@B72]\]. For example, the recruitment of the NURD chromatin remodeling complex to sites of DNA lesions depends on the synthesis of PAR. Interestingly, this complex was identified as an important modulator of aging-associated chromatin defects, and loss of several NURD components and function was evident during human premature aging \[[@B73]\].
The role of PARP1 in gene transcription and chromatin remodeling was impressively demonstrated in a *Drosophila* study \[[@B74]\]. The authors revealed that PARP1 is crucial for puff formation in giant polytene chromosomes. Puff formation arises from local relaxation of the chromatin structure and is associated with actively transcribed regions \[[@B74]\]. Ju et al. provided interesting mechanistic evidence linking PARP1-dependent initiation of transcription and its function in DNA binding \[[@B68]\]. According to this work, PARP1 acts in concert with another binding partner, that is, topoisomerase II. Topoisomerase II introduces a transient double strand break at the promoter, which leads to PARP1 binding and activation. The subsequent rapid but transient poly(ADP-ribosyl)ation triggers chromatin relaxation and initiation of transcription.
Together, these findings suggest a functional interplay of PARP1 with chromatin components and associated remodeling factors, implying an active role of PARP1 in chromatin function and transcriptional regulation during the aging process. Gene profiling data support such a hypothesis, since PARP1 deficiency alters expression of genes involved in cell cycle progression, DNA replication, oxidative stress, cancer initiation, and aging \[[@B75], [@B76]\]. The detailed spatial and temporal characteristics of these mechanisms, however, remain to be determined.
3.2. PARP1 in DNA Repair {#sec3.2}
------------------------
As discussed above, a substantial body of evidence demonstrates a causative role of DNA repair and genome maintenance mechanisms in mammalian longevity.
Base excision repair (BER) is the major DNA repair pathway that acts on damage that occurs during cellular metabolism including damage from ROS, methylation, deamination, and hydroxylation. The levels of many of these lesions increase with age including the well-studied lesion 8-oxoguanine. Moreover, BER activity decreases with age in multiple tissues \[[@B77]\]. The core BER reaction is initiated by a DNA single-strand break (SSB) upon excision of the damaged bases by DNA glycosylases \[[@B78]\]. PARP1 detects such SSB via its second zinc finger (ZFII), thus triggering its enzymatic activation \[[@B79], [@B80]\]. Moreover, PARP1 physically cooperates with 8-oxoguanine-DNA glycosylase, which further stimulates PARP1 activity \[[@B81]\]. Importantly, the recruitment of the BER loading platform X-ray repair complementing factor 1 (XRCC1) is completely dependent on poly(ADP-ribsosyl)ation \[[@B82], [@B83]\]. Thus, PARP1 and PAR are required for the assembly and stability of XRCC1 nuclear foci after DNA damage \[[@B83]\]. Furthermore, XRCC1 and PARP1 interact with DNA polymerase-*β* and DNA ligase III, forming a multiprotein complex consisting of the major BER factors \[[@B84]--[@B86]\].
The finding that PARP1-deficient cells still synthesized PAR led to the identification of an additional nuclear PARP, that is, PARP2, which is also activated upon genotoxic stimuli \[[@B34], [@B35]\]. PARP1 and PARP2 homo- and heterodimerize and work at least partially in a redundant fashion, since only double-knockout mice show embryonic lethality \[[@B87], [@B88]\]. This notion is supported by the fact that PARP2 also participates in BER and interacts physically and functionally with XRCC1, DNA polymerase-*β*, and DNA ligase III. Recruitment studies indicate a role of PARP2 in later steps of BER repair, as proposed by the following model for spatiotemporal accumulation of BER factors: SSBs are detected by the DNA binding domain of PARP1, leading to its activation, production of PAR, and chromatin relaxation. Subsequently, additional PARP1 molecules are attracted, causing amplification of the signal. At the "point of repulsion," PARP1 then dissociates from the DNA, enabling the recruitment of the BER loading platform XRCC1, PARP2, and further DNA repair factors. This triggers resealing of the DNA lesion and reestablishment of genomic integrity \[[@B89]\].
Nucleotide excision repair is responsible for the removal of bulky helix-distorting DNA adducts, which are caused by UV irradiation and endogenous metabolites \[[@B78]\]. The functional role of the NER as a longevity assurance mechanism is impressively represented by the fact that patients with defects in a subset of NER proteins, that is, CSA and CSB (Cockayne syndrome) and XPB, XPD, TTDA (trichothiodystrophy), as well as corresponding mouse models, show in some tissues a strong premature aging phenotype \[[@B2]\]. Although the role of PARP1 in NER is not very well established, at least two NER factors, the DNA-dependent ATPase Cockayne syndrome group B (CSB) protein and the DNA lesion recognition protein xeroderma pigmentosum group A (XPA), were identified as PAR binding factors \[[@B17], [@B90], [@B91]\]. CSB also physically interacts with PARP1 and its ATPase activity is inhibited by poly(ADP-ribosyl)ation. Consistently, there is some evidence from cell culture studies that PARP1 is involved in NER of UV photo-damage products \[[@B92], [@B93]\].
DNA double strand breaks (DSBs) arise from ionizing radiation, free radicals, chemicals, or during replication of a SSB through collapsed replication forks. They represent the most cytotoxic form of DNA damage and, if unrepaired, they can trigger apoptosis, senescence, or genomic instability. Mammalian cells repair DSBs via two mechanisms: homologous recombination (HR) utilizes the sister chromatid or chromosome for error-free repair of the DSB, whereas nonhomologous end-joining (NHEJ) simply reattaches free DNA ends without using a template. For this reason, NHEJ is prone to microdeletions or insertions which can cause frameshift mutations \[[@B78]\]. Whether HR or NHJE is employed depends on the species, cell type, and cell cycle phase \[[@B94]\]. In both pathways, PARP1 already participates at very early stages. PARP1 and the DSB sensing complexes MRN (MRE11/Rad50/NBS1) (involved in HR) and Ku70/80 (involved in NHEJ) were shown to interact with and compete for binding at free DNA ends, with PARP1 potentially guiding these proteins to the damaged site \[[@B44], [@B95]\]. PARP1 also physically and functionally interacts with two phosphatidyl inositol 3-like protein kinases ATM (involved in HR) and DNA-PK~cs~ (involved in NHEJ), which are crucial for DSB signaling \[[@B96]--[@B99]\]. It was suggested that PARP1 serves as a general DNA-damage-detecting molecule, which potentially mediates a switch between the NHEJ and the HR pathways \[[@B94]\]. Consistent with this, PARP1 functions in a NHEJ back up pathway \[[@B95], [@B100]\], and several reports demonstrated an antirecombinogenic activity of PARP1 \[[@B101]--[@B103]\]. Some information on the role of PARP1 in DSB repair was obtained from studies investigating the possible use of PARP inhibitors as anticancer drugs following the concept of synthetic lethality. According to this concept, PARP1 inhibition alone shows no cytotoxic effect on HR proficient cells, but causes cytotoxicity in HR-defective cells, for example, *BRCA* mutant cancer cells. This cytotoxicity is thought to be induced by accumulation of unrepaired SSBs, which are converted to DSBs by collisions with the replication machinery which cannot be repaired due to the lack of HR, ultimately triggering cell death \[[@B104]\]. A recent study challenged this view by presenting an alternative model based on the finding that coinactivation of NHEJ rescued the synthetic lethal effect in BRCA1-deficient cells indicating that PARP1 catalytic activity regulates NHEJ activity, thus preventing NHEJ components from binding to sites of DNA damage \[[@B105]\]. Yet another study reported that loss of PARP1 activity itself can inhibit HR by suppressing the expression of BRCA1 and RAD51, two key factors in HR \[[@B106]\]. Moreover, PARP1 is activated at stalled forks to mediate Mre11-dependent replication restart by HR \[[@B107]\]. Another level of complexity is added by recent work demonstrating that SIRT6, a mammalian homolog of the yeast Sir2 deacetylase which functions as a longevity regulator in yeast, is recruited to sites of DSBs. There, SIRT6 appears to stimulate DSB repair via NHEJ and HR. Of note, SIRT6 interacts with PARP1 and stimulates its activity thereby enhancing DSB repair upon oxidative stress \[[@B108]\]. It is important to note that the results mentioned above are not necessarily mutually exclusive, but more work is necessary to define the exact molecular mechanisms by which PARP1 participates in DSB repair and which role this may have during aging.
3.3. PARP1, Telomere Maintenance, and the WRN Protein {#sec3.3}
-----------------------------------------------------
Telomeres are repetitive sequences at the end of the chromosomes and function as a buffer to prevent loss of coding sequences during DNA replication. They are capped by a protein complex known as shelterin, which tightly regulates the telomeric structure by interaction with several DNA repair proteins and the telomere-elongating reverse transcriptase, telomerase. Deterioration of telomeres represents an important factor during human aging \[[@B109]\].
The role of PARP1 in the regulation of telomere length is well established. *In vivo*, a substantial loss of telomeric DNA by 30% was observed in the first generation of *Parp*1^−/−^ mice \[[@B110]\]. Gomez et al. reported that PARP1 is dispensable for the capping of normal telomeres, but is specifically recruited to eroded telomeres, where it might help to protect chromosomes against end-to-end fusions and genomic instability \[[@B111]\]. Our group demonstrated in various cell culture systems that pharmacological inhibition of poly(ADP-ribosyl)ation or PARP1 knockdown via RNA interference leads to a rapid decrease in telomere length and stabilization at a lower level. Importantly, neither the length of the single-stranded telomeric overhang nor telomerase activity was affected by PARP1 inhibition. Interestingly, release from PARP inhibition led to a fast regain in telomere length in telomerase-positive cells indicating that PARP1 activity is an important determinant in telomere length regulation \[[@B112]\]. On a molecular level, the function of PARP1 in telomere length regulation presumably depends on its interaction with the telomeric repeat binding factor 2 (TRF2). TRF2 is a key component of the shelterin complex and is responsible for telomeric stability, length regulation, and suppression of unscheduled activity of the double-strand break repair machinery by maintaining the t-loop \[[@B113]\]. PARP1 interacts with and modifies TRF2, and the poly(ADP-ribosyl)ation of TRF2 affects its binding to telomeric DNA \[[@B111], [@B114]\].
Another PARP1 interaction partner that is involved in telomere regulation is the RecQ helicase WRN \[[@B115]\]. Patients with the rare autosomal recessive disorder Werner syndrome, in which the *WRN* gene is mutated, display genomic instability and telomere shortening on the cellular and premature aging on the organismic level with symptoms resembling normal human aging in many aspects including cataracts, graying of hair and alopecia, atherosclerosis, osteoporosis, and higher cancer incidence. The premature aging phenotype of these patients appears to be at least partially dependent on telomere length, since human symptoms were only recapitulated in mice with short telomeres, that is, WRN/telomerase double-knockout mice \[[@B115], [@B116]\]. (*N.B*. Mice usually exhibit considerably longer telomeres (\~40 kb) than humans (5--15 kb)). On a cellular level, fibroblasts derived from WS patients display genomic instability and a reduced replicative lifespan. This phenotype is in accordance with experimental data demonstrating that WRN is involved in multiple aspects of DNA metabolism, such as DNA replication, genomic maintenance, and telomere regulation \[[@B115]\]. WRN functions as a 3′-5′ helicase and additionally as a 3′-5′ exonuclease. Proper enzymatic activity of WRN seems to be crucial for maintaining genomic integrity, since pharmacological inhibition of WRN\'s helicase activity causes DSBs and apoptosis \[[@B117]\]. WRN and PARP1 directly interact with each other physically and PARP1 modulates WRN\'s exonuclease and helicase activities \[[@B118], [@B119]\]. Upon automodification of PARP1, the inhibition of WRN\'s exonuclease and helicase activities is released suggesting that PARP1 regulates the timing of WRN activity towards its substrates \[[@B113]\]. The regulation of PARP1 and WRN appears to be reciprocal, because poly(ADP-ribosyl)ation is impaired in WRN deficient cells indicating that WRN is required to regulate PARP1-dependent poly(ADP-ribosyl)ation \[[@B120]\]. Moreover, other factors than PARP1 and WRN are involved in these mechanisms, because WRN and PARP1 share many interaction partners, including DNA-PK, P53, and TRF2 ([Figure 3](#fig3){ref-type="fig"}). For example, PARP1, WRN, and DNA-PK (including Ku70/80 and DNA-PK~cs~) can form a complex, in which PAR-modified Ku70/80 inhibits WRN \[[@B121]\]. Furthermore, both PARP1 and WRN have positive impact on telomere length, presumably by regulating the binding of TRF2 to the t-loop. Genetic cooperation between PARP1 and WRN was demonstrated *in vivo*, because mice with deficiencies in both proteins display higher rates of chromatid breaks, chromosomal rearrangements, and cancer than each of the single-mutant mice \[[@B122]\]. Moreover, double mutants appear to have reduced median and maximum lifespan, despite the fact that these mice were on a telomerase-positive genetic background and telomere lengths of single-mutant MEFs did not differ significantly from the double-mutant MEFs. This finding suggests that telomere-independent functions of WRN and PARP1 exist in the mouse to maintain organismic longevity. (*N.B.* In contrast, wild-type MEFs showed 30--40% longer telomeres). In conclusion, since PARP1 and WRN share many interaction partners and both proteins participate in other DNA repair pathways such as BER and NHEJ, they probably synergistically collaborate to maintain overall genomic stability and ensure longevity.
3.4. PARP1 during DNA Replication, Mitosis, and Cell Cycle Control {#sec3.4}
------------------------------------------------------------------
The WRN helicase also participates in the response to replicative stress, a cellular stressor that was linked to mammalian aging due to its ability to drive cells, including stem cells, into senescence and apoptosis \[[@B123], [@B124]\]. Replication forks contain several proteins such as helicases and polymerases, forming the so-called replisome. Usually, progression of the replication fork continues until it encounters a replication fork barrier such as DNA-protein complexes or SSBs. In this case, the replicative helicase progresses much more slowly, so that the fork is "stalled." If this goes along with the disassembly of the replisome the fork "collapses" and a DSBs is formed \[[@B125]\]. WRN and PARP1 are involved in the reactivation of stalled replication forks. Specifically, PARP1 binds to and is activated at stalled replication forks and mediates the recruitment of Mre11, a key component of the MRN complex. Mre11 may collaborate with WRN helicase to resect DNA ends for RAD51 loading and subsequent HR repair to promote replication fork restart after release from replication blocks \[[@B44], [@B107], [@B125], [@B126]\]. In accordance with these data, a recent study demonstrated that PARP activity is required for effective replication fork restart upon treatment of cells with sublethal dosis of the replication stress-inducing topoisomerase 1 inhibitor camptothecin \[[@B127]\].
After DNA replication is completed, proper mitotic regulation is crucial to ensure genomic integrity \[[@B128]\]. During mitosis, the spindle pole formation requires the centrosome, whereas the centromere is the chromosomal region that organizes the kinetochore, thus enabling the attachment of the mitotic spindle microtubules. First, evidence for a role of poly(ADP-ribosyl)ation in spindle regulation was obtained from a study with *Xenopus laevis* egg extracts showing that PAR is a component of the mitotic spindle and is required for its assembly and function, although this was attributed to the enzymatic activity of another PARP family member, that is, tankyrase-1 \[[@B129], [@B130]\]. With regards to PARP1, it was shown that haploinsufficiency for PARP1 is related to centrosome duplication and chromosomal instability \[[@B131]\]. Consistent with this, PARP1 localizes to the centrosome \[[@B132], [@B133]\]. Moreover, PARP1 and PARP2 are present at centromeres and interact with the constitutive centromere proteins Cenpa, Cenpb and the spindle check point protein Bub3 \[[@B134], [@B135]\]. The physical and functional relationship of PARP1 to the centrosome and the centromere links DNA damage surveillance to the mitotic spindle checkpoint. The notion that mitotic spindle checkpoint proteins play an important role to ensure mammalian longevity is supported by studies demonstrating that mice with low levels of the mitotic checkpoint protein BubR1 and mice haploinsufficient for Bub3 and Rae1---another mitotic checkpoint gene---age prematurely (*N.B.* A complete knockout of these genes results in embryonic lethality in the mouse) \[[@B136], [@B137]\].
Because severe DNA damage or mitotic misregulation can cause genomic instability leading to tumor formation, a complex cellular security network has evolved to counteract carcinogenesis. This signaling network can stop the cell cycle at different stages, thereby either inducing DNA repair, or eradicating or neutralizing heavily damaged cells by apoptosis or senescence, respectively. To this end, apoptosis and senescence are powerful tumor-suppressive mechanisms, but on the other hand, both pathways can lead to depletion of the regenerative cell pool, thus promoting tissue degeneration and organ failure, which are hallmarks of aging \[[@B138]\]. One of the most important regulators of cell cycle progression and induction of senescence/apoptosis is the transcription factor P53. Consequently, mouse studies demonstrated that P53 deficiency leads to premature death due to tumor development, whereas constantly active P53 protects against cancer at the cost of a premature aging phenotype \[[@B138]\]. Consistent with the role of PARP1 and P53 as caretakers and guardians of the genome, respectively, PARP1 and P53 synergistically cooperate *in vivo* in telomere and chromosomal maintenance as well as in tumor suppression \[[@B139]--[@B143]\]. Many functional interactions between PARP1 and P53 during DNA damage response and apoptosis exist, such as delayed P53 transactivation potential in PARP1-deficient cells \[[@B144]--[@B147]\]. In addition to its function as a positive regulator of gene expression, P53 also acts as a gene-specific transcriptional transrepressor. Interestingly, P53-mediated transrepression of the *MTA1* gene (MTA1, metastasis associated protein 1), a component of a nucleosome remodeling complex which is associated with very aggressive tumor phenotypes, depends on functional poly(ADP-ribosyl)ation of P53 \[[@B148]\]. On the other hand, poly(ADP-ribosyl)ation of P53 is also able to inhibit its binding to its transcriptional consensus sequence, indicating that multifaceted regulatory mechanisms exist between PARP1 and P53 \[[@B149], [@B150]\]. Kanai et al. suggested a mechanism of PARP1-dependent regulation of P53 activity. According to this study, poly(ADP-ribosyl)ation induces structural changes in P53 that mask its nuclear export sequence, resulting in an accumulation of P53 in the nucleus, where it exerts its transactivational functions. Accordingly, a P53 mutant in which acceptor sites were mutated was localized to the cytoplasm to a greater extent than wildtype P53 \[[@B151]\].
In conclusion, there is ample evidence that PARP1 modulates P53 stability, intracellular localization, and transcriptional activity with likely implications in the induction of apoptosis and senescence on a cellular and therefore aging and longevity on an organismic level. However, studying the combined role of PARP1 and P53 in the aging process is complicated by the situation that mouse models with deficiencies in both tumor-suppressor genes show cancer-dependent premature death unrelated to other signs of premature aging. The development of sophisticated conditional mouse models with spatiotemporal-controlled expression of PARP1 and P53 may represent an approach to overcome these hurdles.
3.5. PARP1 as a Longevity Assurance Factor {#sec3.5}
------------------------------------------
As discussed in the preceding sections, PARP1 acts as a general caretaker of genomic stability and is associated with various factors, whose involvement in mechanisms of aging and longevity are well established. This indicates a potential role of PARP1 as a longevity assurance factor which is supported by *in vivo* studies as briefly discussed in the following.
There is a large body of evidence showing a positive correlation of poly(ADP-ribosyl)ation capacity and mammalian longevity. Previously, we demonstrated that poly(ADP-ribosyl)ation capacity in peripheral blood mononuclear cells (PBMCs) of 13 mammalian species strongly correlates with their maximum lifespan, for example, maximum poly(ADP-ribosyl)ation levels were five times higher in humans than in rodents \[[@B152]\]. Interestingly, these differences in poly(ADP-ribosyl)ation are not associated with different enzyme levels, but are rather influenced by an higher poly(ADP-ribosyl)ation capacity of the human PARP1 enzyme in comparison to its mouse ortholgue \[[@B153]\]. Moreover, poly(ADP-ribosyl)ation capacity in PBMCs declines with age in humans and rodents \[[@B152], [@B154]\]. Interestingly, humans exhibiting an exceptional long lifespan, that is, centenarians, display a significantly higher poly(ADP-ribosyl)ation capacity than the average population \[[@B155]\], which is comparable to those of young subjects \[[@B156]\].
Apart from these correlative studies, studies with*Parp*1^−/−^ mice indicate a role of PARP1 as a longevity assurance factor. Thus, *Parp*1^−/−^ mice and cells derived thereof are hypersensitive to DNA-damaging agents and *Parp*1^−/−^ cells display increased spontaneous genomic instability as measured by the frequency of sister chromatid exchanges, chromosome aberrations, and micronuclei formation, which confirmed that PARP1 functions as a general caretaker of the genome \[[@B157]\]. In support of the view that PARP1 counteracts the aging process is the finding that *Parp*1^−/−^ mice age is moderately faster compared to wild-type control animals \[[@B158]\]. Moreover, this and various other studies supported the notion that PARP1 acts as a tumor-suppressor gene, since PARP1 deficiency enhances carcinogenesis during aging and upon induction by DNA damaging agents \[[@B143], [@B158]--[@B161]\]. Consistently, data from human studies showed that a hypomorphic PARP1 polymorphism (V762A) serves as a risk factor in the development of some types of human cancers \[[@B162]--[@B166]\].
In conclusion, PARP1 participates in various genome maintenance mechanisms, such as chromatin remodeling, DNA repair, reactivation of stalled replication forks, telomere maintenance, and cell cycle control. Consistent with its role as a general caretaker of the genome, there is strong evidence from *in vivo* studies indicating that PARP1 indeed functions as a longevity assurance factor. On the other hand, the interaction of PARP1 with key regulators of immune function, such as NF-*κ*B, and its potential to induce cell death may contribute to aging-promoting mechanisms as discussed in the following sections.
4. PARP1 in Inflammation and Cell Death {#sec4}
=======================================
First, evidence that PARP1 contributes to inflammation and the development of related pathologies was revealed by genetic studies in *Parp*1^−/−^ mice, because these animals are protected from several inflammation and cell death-associated pathologies such as ischemic infarction, collagen-induced arthritis, and LPS-induced septic shock \[[@B157], [@B167]\]. Moreover, *Parp*1^−/−^ animals are resistant to MPTP-induced Parkinson\'s disease and streptozotocin-induced diabetes mellitus \[[@B168]--[@B171]\]. The molecular and cellular mechanisms underlying these phenotypic results and their possible implications in mechanisms of aging and longevity are discussed below.
4.1. PARP1, NF-**κ**B, and Inflammation {#sec4.1}
---------------------------------------
Various studies demonstrated that PARP1 participates in the regulation of transcriptional processes, either via general chromatin remodeling or through specific interaction and regulation of a wide range of transcription factors \[[@B171], [@B172]\].
Maybe the best studied interaction is that of PARP1 with NF-*κ*B. The transcription factor NF-*κ*B is considered a master regulator in controlling gene expression upon proinflammatory stimuli. NF-*κ*B is composed of dimeric combinations of Rel family members with the major subunits p65 and p50. In nonstimulated cells, NF-*κ*B is located in the cytoplasm via the binding to the inhibitory I*κ*B proteins. Upon proinflammatory stimuli, I*κ*B proteins are phosphorylated by I*κ*B kinases (IKKs), which cause their degradation by the ubiquitin/proteasome system. Subsequently, NF-*κ*B is translocated to the nucleus, where it can activate the transcription of a number of genes, especially inflammatory genes \[[@B173]\]. Apart from the regulation of NF-*κ*B by its subcellular localization, its action is tightly regulated within the nucleus by posttranslational modifications and interaction with transcriptional cofactors. Importantly, NF-*κ*B-dependent gene expression is associated with aging in the mouse as well as in humans \[[@B174]\]. Recently, it was shown that hyperactive NF-*κ*B signaling contributes to premature aging in the mouse \[[@B175]\], and blocking of NF-*κ*B in aged mice was sufficient to reverse some features of skin aging \[[@B174], [@B176]\]. In accordance with these studies, pharmacological inhibition of NF-*κ*B prolongs lifespan of *Drosophila melanogaster*by \~15% \[[@B177]\]. Cellular studies showed that NF-*κ*B-dependent gene transcription can be induced by genotoxic stress and gene transcription studies in conditionally immortalized human fibroblast suggested that NF-*κ*B signaling plays a causal role in the development of senescence \[[@B178]\]. In addition, NF-*κ*B signaling was implicated in maintaining cellular senescence, because NF-*κ*B-deficient fibroblasts escape senescence earlier and immortalize at a faster rate \[[@B179]\]. On the other hand, NF-*κ*B-dependent gene transcription can be induced by genotoxic stress and is required for the transcription of many SASP factors \[[@B5]\]. In summary, there is substantial evidence that NF-*κ*B plays a crucial role in aging and age-related diseases \[[@B180]\].
The expression and activation patterns of PARP1 and NF-*κ*B are remarkably similar in various tissues. A direct role of PARP1 in NF-*κ*B-mediated transcription was emphasized by the finding that expression of NF-*κ*B-dependent proinflammatory mediators, such as TNF*α*, IL6, or iNOS, is impaired in *Parp*1^−/−^ mice \[[@B167], [@B181]\]. PARP1 physically interacts with both major subunits of NF-*κ*B, that is, p65 and p50, and is required for NF-*κ*B-dependent gene transcription ([Figure 4](#fig4){ref-type="fig"}) \[[@B182]\]. Moreover, PARP1 is acetylated by the histone acetylase p300/CBP upon inflammatory stimuli, leading to a stronger association with NF-*κ*B \[[@B39]\]. Subsequent expression of proinflammatory mediators such as iNOS leads to the production of highly reactive chemical species that, in turn, cause extensive DNA damage in the target cell, potentially supporting a positive feedback mechanism. Importantly, in this study neither the DNA binding nor the enzymatic activity of PARP1 was necessary for direct transcriptional activation of NF-*κ*B \[[@B183]\]. On the other hand, inhibition of PARP\'s enzymatic activity is sufficient to decrease the expression of iNOS, IL6, and TNF*α* in cultured cells and to reduce the expression of inflammatory mediators in mice \[[@B184]\]. This is consistent with a recent study demonstrating that the PARP1-dependent activation of NF-*κ*B occurs at two levels ([Figure 4](#fig4){ref-type="fig"}). Thus, in addition to the nuclear coactivator function of PARP1 on NF-*κ*B activity, this study identified PARP1 as a trigger for the translocation of NF-*κ*B from the cytoplasm into the nucleus upon genotoxic stress ([Figure 4](#fig4){ref-type="fig"}) \[[@B185]\]. According to this model, PARP1 is recruited to DNA strand breaks and is automodified with PAR. Upon dissociation into the nucleoplasm, PARP1 then rapidly forms a signalosome composed of the SUMO1 ligase PIASy, IKK*γ* (NEMO), and ATM. The signalosome is stabilized by a network of direct protein-protein interactions as well as by PAR binding of PIASy and ATM through PAR binding motifs. PAR degradation by PARG causes subsequent destabilization of the signalosome, resulting in IKK*γ* SUMOylation, translocation to the cytoplasm, phosphorylation of I*κ*B proteins, and NF-kB activation. This mechanism directly links the DNA-damage-signaling functions of PARP1 to its role in inflammation-related mechanisms. Interestingly, PARP1-NF-*κ*B signaling seems also to contribute to the activation and maintenance of the secretory phenotype of senescent cells \[[@B186]\]. In consequence, the associated secretion of proinflammatory factors possibly changes the tissue microenvironment and forms a site of low-level chronic inflammation with tumor and aging-promoting properties.
4.2. PARP1 and Its Role in Cell Death {#sec4.2}
-------------------------------------
Historically, two major mechanisms of mammalian cell death are distinguished, that is, apoptosis and necrosis. Apoptosis is considered as the default pathway, where cell death occurs in a controlled manner resulting in the elimination of cells by macrophages without secondary damage of the surrounding cells. In contrast, necrosis is considered an uncontrolled process which leads to disruption of cells promoting tissue inflammation \[[@B187]\]. Several transition states between the two pathways exist such as apoptosis inducing factor-(AIF-) dependent cell death \[[@B188]\]. Cell death is an important factor contributing to organismic aging, because apoptosis can lead to depletion of the regenerative cell pool and necrosis can cause chronic inflammatory conditions. PARP1 is involved in necrosis as well as in apoptosis, depending on the cell type and the intensity of DNA damage. Excessive DNA damage, as it can be triggered by pathophysiological stimuli and during NF-*κ*B-dependent inflammatory responses, leads to an overactivation of PARP1, which induces the depletion of cellular NAD^+^ pools and subsequently of ATP pools \[[@B189]\]. This could affect energy-dependent cellular functions resulting in necrosis, which in turn reinforce tissue inflammation leading to a vicious cycle of PARP1 activation, necrosis, and inflammation. The role of PARP1 in apoptosis is manifold depending on the cell cycle state. Two major types of apoptosis exist: caspase-dependent and caspase-independent apoptosis. On the one hand, in proliferating cells, PARP1 contributes to classical caspase-dependent apoptosis through its regulatory activity on P53. Here, after an initial synthesis of PAR, PARP1 is cleaved by caspases 3 and 7 in a 24 kD and an 89 kD fragment \[[@B190]\]. This occurs potentially to inactivate PARP1 and to preserve cellular ATP pools for the apoptosis program \[[@B191]--[@B193]\]. On the other hand, it was shown that PARP1 contributes to caspase-independent apoptosis by releasing AIF from the mitochondria \[[@B194], [@B195]\]. PAR itself acts as a signaling molecule between nucleus and mitochondria, where it binds to AIF in a non-covalent manner and then triggers its release. AIF then translocates to the nucleus, where it causes chromatin condensation, large-scale DNA fragmentation, and finally cell death \[[@B196]--[@B198]\].
In conclusion, three interconnected cellular mechanisms have been proposed to be responsible for the involvement of PARP1 in cell death and inflammation-related, age-related pathologies. First, PARP1 overactivation by severe DNA damage upon an initial pathological insult can lead to NAD^+^ and subsequent ATP depletion causing necrotic cell death due to energy depletion \[[@B189]\]. Second, such an initial pathological insult or secondary necrotic disruption of cells can trigger an inflammatory response leading to further damage of the surrounding tissue. This process can be stimulated by the action of PARP1 as an essential transcriptional coactivator of the proinflammatory transcription factor NF-*κ*B. Products of NF-*κ*B regulated genes, for example, iNOS, participate in the production of ROS and RNS, which could support the aforementioned vicious cycle of DNA damage, subsequent PARP1 activation, and cell death potentiating inflammation and tissue damage. Third, the PAR-dependent release of apoptosis-inducing factor (AIF) from the mitochondria resulting in caspase-independent apoptosis may contribute to some extent to PARP1-dependent pathologies in particular neurodegenerative disorders \[[@B194], [@B197], [@B198]\]. Over time, these mechanisms can contribute to aging and the development of age-related pathological conditions.
4.3. PARP1 as an Aging-Promoting Factor {#sec4.3}
---------------------------------------
Evidence supporting a role of PARP1 as a driving force of inflammation on an organismic level is given by the fact that *Parp*1^−/−^ mice are protected from several inflammation- and cell-death-associated diseases and that *Parp*1^−/−^ mice and cells display lower expression levels of a whole spectrum of proinflammatory cytokines, adhesion molecules, and enzymes \[[@B167]\]. Consequently, given the role of inflammation during mammalian aging, PARP1 was postulated to act as an aging-promoting factor \[[@B199]\]. In line with this concept, PARP inhibition or ablation of gene transcription has beneficial effects on several age-related diseases, including aging-associated cardiac and vascular dysfunctions \[[@B184], [@B200], [@B201]\]. Apart from its functions as a regulator of NF-*κ*B-dependent gene transcription, PARP1 competes with the family of type III histone deacetylases of sirtuins (i.e., in humans and mice SIRT1-7) for the same substrate, that is, NAD^+^. In addition, PARP1 and SIRT1 interact physically and show an antagonistic interplay on a functional level \[[@B71], [@B202]\]. Sirtuins regulate the energy homeostasis by controlling the acetylation status and activity of various enzymes and transcriptional regulators and have been identified to act as longevity factors in various species \[[@B203]\]. Moreover, consumption of NAD^+^ links PARP1 and sirtuins to energy metabolism, which plays a fundamental role in aging mechanisms. *In vivo* studies showed that SIRT1-overexpressing mice are leaner, metabolically more active, show improved glucose tolerance, exhibit less inflammation, and are resistant to intestinal cancer development \[[@B204]--[@B208]\]. *Parp*1^−/−^ mice exhibit increased NAD^+^ content and enhanced SIRT1 activity in brown adipose and muscle tissue. Consequently, *Parp*1^−/−^ mice phenocopy many aspects of SIRT1 activation, such as a higher mitochondrial content, increased energy expenditure, reduced body weight, and protection against metabolic disease \[[@B209]\].
Conversely, this phenotype is mirrored by the phenotype of mice with ectopic expression of hPARP1 \[[@B210]\]. These mice develop sporadic obesity and show impaired glucose tolerance. Furthermore, hPARP1-expressing mice exhibit impaired survival rates, which are accompanied by premature development of several inflammation and age-associated pathologies, such as nephropathy, dermatitis, pneumonitis, myocardiopathy, and hepatitis. In support of this hypothesis, *hPARP1*mice develop normocytic, normochromic anemia and show an increase in the fraction of circulating monocytes, which is suggestive of anemia of chronic inflammatory disease \[[@B211], [@B212]\]. Moreover, *hPARP1* mice show typical signs of premature aging, such as early development of kyphosis and impaired hair regrowth. In addition to a potentially altered interplay between PARP1 and sirtuins in these mice, the pathological phenotype of *hPARP1* mice might be related to an altered PARP1-NF-*κ*B interaction leading to a continuous low-level increase in pro-inflammatory stimuli. Consistently, expression of NF-*κ*B-dependent target genes, such as TNF*α*, IL1, and IL6, is dysregulated in *hPARP1* animals. This may contribute to the premature development of typical age-related chronic diseases in these mice \[[@B210]\].
5. Summary {#sec5}
==========
Aging is a complex process which cannot be explained by a single pathway or even a set of closely related pathways. More likely, many diverse cellular functions will contribute to aging and they will do so in a highly interdependent manner \[[@B53]\]. As summarized here, this complexity is already represented at the level a single enzyme, that is, PARP1. PARP1 is a factor that connects DNA damage response and inflammatory mechanisms, both of which are closely associated with mammalian aging. Thus, under physiological conditions and mild genotoxic stress, PARP1 is thought to play an important role in genomic maintenance ([Figure 5](#fig5){ref-type="fig"}). On the other hand, under pathophysiological conditions, reactive chemical species are generated by activated immune cells potentially inducing DNA damage in an autocrine and paracrine fashion. If exceeding a threshold, DNA damage and the subsequent signaling can force cells into senescence. Senescent cells, in turn, possess the potential to secrete proinflammatory cytokines thereby reinforcing tissue inflammation. PARP1 is involved in these processes due to its close interplay with NF-*κ*B at different stages during NF-*κ*B activation. In addition, severe DNA damage can trigger overactivation of PARP1 resulting in cell death by apoptosis or necrosis ([Figure 5](#fig5){ref-type="fig"}). Debris of necrotic cells is phagocytosed by macrophages, thereby triggering a proinflammatory response, again inducing the generation of reactive chemical species. Interestingly, there is some evidence that chronic intestinal inflammation can induce systemic genotoxicity, for example, in leukocytes and hepatocytes, suggesting that local sites of inflammation can affect genomic stability and homeostasis even at an organismic level \[[@B213]\]. Over time, inflammation, senescence, and cell death contribute to the depletion of the regenerative cell pool and tissue dysfunction accumulating in the aging process.
There is ample evidence supporting a role of PARP1 as a longevity assurance factor on the one hand, but also as an aging-promoting factor on the other hand.The dual role of PARP1 in longevity and aging might be reflected in the moderate premature aging phenotype observed in cohorts of *Parp*1^−/−^ mice \[[@B158]\]. Thus, it is reasonable to assume that overall aging in these mice is kept nearly in balance, due to comprised genomic integrity on the one hand, but reduced inflammatory status on the other hand. The generation of *Parp*1^−/−^ mice with tissue specific reconstitution of PARP1 expression may be a suitable model to test such a hypothesis. Tissue-specific reexpression of PARP1 in cells of the innate immune system on an otherwise *Parp*1^−/−^ background may lead to a more drastic accelerated aging phenotype, since PARP1-overexpressing cells of the innate immune system are expected to exhibit an enhanced inflammatory status, while cells of the remaining *Parp*1^−/−^ tissues are genomically unstable. Another possibility explaining the moderate premature aging phenotype of *Parp*1^−/−^ mice may be that alternative mechanisms are able to compensate for the PARP1 deficiency. Such potential backup mechanisms rely most likely on PARP2 which shares some redundancy to PARP1, as it is evident by the finding that *Parp1/Parp2* double-deficient mice are not viable. The generation of conditional and inducible double-knockout mice may help to test this hypothesis.
Many theories of aging exist. Most of these are not mutually exclusive and although none of these is probably able to explain all characteristics of human aging, in all probability there is some truth in many of them. The "antagonistic pleitropy" theory of aging postulates the existence of pleiotropic genes and mechanisms having opposite effects on fitness at different stages of age. Thus, a gene or mechanism may be beneficial for survival in early life, when natural selection is strong, but harmful at later ages, when selection is weak or absent \[[@B214]\]. Mechanisms of DNA damage response as well as inflammation may support such a theory. DNA damage response, with its final end points, DNA repair, senescence, and apoptosis, is clearly beneficial at young age, as these mechanisms prevent cancer development. However, at older age, this may become detrimental, as depletion of the regenerative cell pool by senescence or apoptosis may contribute to tissue degeneration and aging. The same holds true for inflammation. At young age, inflammatory responses most likely fulfill beneficial functions, for example, acting as a first line defense against infections. (*N.B.* This is supported by the finding that some mouse models with deficiencies in NF-*κ*B signaling are hypersensitive to infectious diseases \[[@B215]\]). However, at older age, continuous pressure on the immune system caused by repeated antigen stimulation leads to remodeling of the immune system with pro-inflammatory properties reinforcing the aging process and the development of age-related disease. Because PARP1 fulfills key roles in mechanisms of DNA damage response and inflammation, it is conceivable that functions of this gene act in some aspects in an antagonistic pleiotropic way, with beneficial functions in the youth and detrimental functions at old age.
In conclusion, PARP1 and the synthesis of poly(ADP-ribose) are emerging as central factors in general cellular stress response with functions in a plethora of molecular mechanisms, such as chromatin remodeling, transcription, DNA damage signaling, DNA repair, cell cycle regulation, cell death, and inflammation. As reviewed here, there is ample evidence that PARP1 fulfills numerous direct as well as indirect roles in mechanisms of aging and longevity which renders it an interesting factor to study in order to better define mechanisms of the aging process.
Our experimental work was supported by the DFG-funded Collaborative Research Center (CRC) 969, the Konstanz Research School Chemical Biology (KoRS-CB) and the International Research Training Group (IRTG) 1331. Funding for open access publishing was provided by University of Konstanz.
{#fig1}
{#fig2}
![Interaction map between PARP1 and Werner syndrome protein (WRN). The two proteins share many overlapping interaction pathways. There is a reciprocal interaction with DNA-PK (double-headed arrow) and p53, stimulation of base excision repair (BER, one-headed arrow), and inhibition of TRF2-DNA binding (blocked arrow). PARP1 also inhibits WRN functions if in an unmodified state (reproduced from \[[@B113]\]).](OXIMED2012-321653.003){#fig3}
![Simplified model of PARP1-dependent mechanisms of NF-*κ*B activation. For details see text. Scheme based on \[[@B167], [@B185]\].](OXIMED2012-321653.004){#fig4}
{#fig5}
[^1]: Academic Editor: Paula Ludovico
| {
"pile_set_name": "PubMed Central"
} |
Introduction {#shil12374-sec-0001}
============
The UK has been recognised as a global leader in health inequalities research and policy, with recent government‐led policy efforts to reduce health inequalities being described as 'historically and internationally unique' (Mackenbach [2011](#shil12374-bib-0026){ref-type="ref"}: 1249). Yet, despite the raft of policies intended to reduce health inequalities, introduced between 1997 and 2010, the UK\'s health inequalities have continued to widen by most (though not all) measures -- see (Bambra [2012](#shil12374-bib-0002){ref-type="ref"}, Barr *et al*. [2012](#shil12374-bib-0006){ref-type="ref"}, Thomas *et al*. [2010](#shil12374-bib-0044){ref-type="ref"}, Thorlby and Maybin [2010](#shil12374-bib-0045){ref-type="ref"}). This failure has prompted multiple commentaries and calls for advocacy to facilitate greater public support for the kinds of upstream policies that available evidence suggests may be required (Bambra *et al*. [2011](#shil12374-bib-0004){ref-type="ref"}, Mackenbach [2011](#shil12374-bib-0026){ref-type="ref"}).
In Chapman\'s ([2007](#shil12374-bib-0009){ref-type="ref"}) terms, public health advocacy includes (amongst other things) working to place and maintain issues on public and political agendas (and exploiting opportunities to do so), discrediting opponents of public health objectives and working to frame evidence in persuasive ways (e.g., by using metaphors or analogies). In other words, advocacy involves strategically 'selling' public health objectives to a range of non‐academic audiences. This way of thinking about advocacy, which Carlisle ([2000](#shil12374-bib-0008){ref-type="ref"}) has termed representational, implies that health inequalities researchers would first need to achieve some kind of consensus on the policy (or societal) changes they are trying to sell. Yet there is currently little agreement as to what exactly it is that public health egalitarians ought to be advocating (Horton [2012](#shil12374-bib-0020){ref-type="ref"}). Other definitions of advocacy are more akin to Burawoy\'s ([2005](#shil12374-bib-0007){ref-type="ref"}) notion of public sociology, in which researchers engage in dialogue with members of the public, work collaboratively with organisations representing public interests and generally try 'to make visible the invisible' (Burawoy [2005](#shil12374-bib-0007){ref-type="ref"}: 264). Here, advocacy involves working with relevant communities to ensure that voices that might traditionally be ignored are given due regard; a 'facilitational' form of advocacy, in Carlisle\'s ([2000](#shil12374-bib-0008){ref-type="ref"}) terms. In either case, advocacy clearly involves something more than the widely accepted model of researchers working with senior civil servants to try to develop evidence‐informed policy responses to health inequalities (see, for example, Petticrew *et al*. [2004](#shil12374-bib-0032){ref-type="ref"}).
Calls for more advocacy to reduce health inequalities therefore raise important questions for researchers about the ways in which we work to effect change. As Scambler ([2012](#shil12374-bib-0039){ref-type="ref"}) points out, the phrase 'tackling health inequalities' means different things to different people:"Articulated as Weberian ideal types, for those coming from a 'policy sociology' perspective engagement for change involves working with people of influence; while for those coming from a 'critical sociology' perspective it can involve working against them (Burawoy [2005](#shil12374-bib-0007){ref-type="ref"}).(Scambler [2012](#shil12374-bib-0039){ref-type="ref"}: 139)"
This article applies a sociological lens to exploring these different ways of approaching health inequalities, reflecting Burawoy\'s suggestion that we should 'apply sociology to ourselves', in order that we become more conscious of the forces driving our research (Burawoy [2005](#shil12374-bib-0007){ref-type="ref"}: 285). It presents an analysis of the perspectives of 52 UK‐based researchers involved in health inequalities debates to explore how these researchers feel about the development of the field so far and what kinds of activities they think researchers should be undertaking. After briefly describing the methods and data sources, the article is divided into four sections. Firstly, we consider what the data suggest researchers feel they have learned (and not learned) from health inequalities research and policy experiences in the UK to date. Secondly, we identify areas of future research activity that the health inequalities researchers who participated in this project felt ought to be prioritized. Thirdly, we examine the kinds of activities our participants believed they ought to be undertaking in order to promote evidence‐informed change. The fourth, concluding section draws the findings together, arguing that the data identify some important divisions amongst health inequalities researchers over the kinds of work it is appropriate and desirable for researchers to undertake; schisms which appear to map onto epistemological, ontological and ideological differences. These differences require attention because they appear to be contributing to professional contestations, especially on research funding, and therefore seem likely to undermine efforts to achieve the kind of clear policy messages that public health advocacy requires. Indeed, the contrasting ways of thinking about health inequalities (and the wider world) lead not only to differing preferences for the future direction of health inequalities research but also to different understandings of the role that research, and researchers, play (and ought to play) in public policy debates.
The article builds directly on two earlier (linked) articles that explored the views of a small number of senior researchers and civil servants on evidence for tackling health inequalities (Petticrew *et al*. [2004](#shil12374-bib-0032){ref-type="ref"}, Whitehead *et al*. [2004](#shil12374-bib-0050){ref-type="ref"}). Like this earlier study, we have produced a companion article that compares the perspectives of policy actors to those of the researchers discussed in this article (Smith *et al.,* under review). As well as providing a contemporary update, our research considers the views of a broader range of researchers and policy actors representing a variety of academic disciplines, career stages and academic or policy‐related institutions (52 researchers and 58 policy actors compared with the nine senior researchers and seven senior civil servants involved in the earlier studies). Looking back on the findings of these earlier studies, the article focusing on the views of senior researchers concluded that there was, at that time, 'significant potential for rapid progress to be made in developing both evidence based policy and policy relevant evidence to tackle health inequalities' (Whitehead *et al*. [2004](#shil12374-bib-0050){ref-type="ref"}: 817). In light of subsequent claims that policy efforts to reduce health inequalities have been relatively unsuccessful (Mackenbach [2011](#shil12374-bib-0026){ref-type="ref"}), it seems appropriate to create space to once again reflect on the kinds of work that health inequalities researchers produce and engage in, and to consider progress and potential future directions. The planned changes to the UK\'s constitutional arrangements, including the devolution of key responsibilities from Westminster to some large urban local authorities in England and the governments in Scotland, Wales and Northern Ireland, as well as current fears over the likely impact of 'austerity'‐led reforms on health inequalities, underline the timeliness of this endeavour (Bambra *et al*. [2015](#shil12374-bib-0005){ref-type="ref"}, Pearce [2013](#shil12374-bib-0030){ref-type="ref"}, Reeves *et al*. [2013](#shil12374-bib-0035){ref-type="ref"}).
Methods {#shil12374-sec-0002}
=======
A 2‐day symposium was held in Edinburgh in December 2012 at which 87 participants from a range of sectors discussed the legacy of health inequalities research and policy and considered potential directions for future research. Participants were invited to the event on the basis of their involvement in health inequalities research, policy, practice or advocacy. Selection was informed by two previous qualitative research projects (Smith [2013a](#shil12374-bib-0041){ref-type="ref"}) and by the professional networks and expertise of a steering group (see acknowledgements). Efforts were made to ensure that researchers came from a range of disciplinary backgrounds, career stages, institutional locations and specific areas of expertise within health inequalities research.
A total of 14 one‐hour focus groups were undertaken during this event, in which 76 symposium attendants participated. There are many definitions of focus groups; we adopted a style based upon a loosely facilitated approach to generating organised discussion, as suggested by Kitzinger ([1994](#shil12374-bib-0021){ref-type="ref"}). For the first seven (morning) sessions, participants were divided according to their profession (Table [1](#shil12374-tbl-0001){ref-type="table-wrap"}). Academic researchers were allocated to three separate groups, according to their primary methodological expertise (which was ascertained via their website profiles and checked via e‐mails prior to the event): (i) quantitative, (ii) mixed methods and (iii) qualitative. This division was made to ensure that the focus groups provided a range of views from researchers working in different academic traditions and with a variety of epistemological perspectives. Researchers working in public sector and policy settings were allocated to other groups. In the context of this article, the term researcher includes individuals involved in research working across a range of settings (e.g., in knowledge broker organisations and the public sector, as well as academia) and applied to 52 symposium attendants (see Tables [1](#shil12374-tbl-0001){ref-type="table-wrap"} and [2](#shil12374-tbl-0002){ref-type="table-wrap"}). All these individuals had published articles in academic journals focusing on health inequalities or the social determinants of health, and health inequalities in the UK were a primary research focus for most. However, a small number of researchers were invited to participate because they had a relevant area of expertise that would otherwise not be represented (this included two health economists and two political scientists focusing on corporate policy influence, all of whom had an interest in health equity). The disciplinary backgrounds of participants varied widely and included anthropology, economics, epidemiology, geography, public health medicine, political science, social policy and sociology.
######
Focus group participants (*N *=* *76)
Participants in the first focus small group discussions
--------------------------------------------------------- ------------------------------------------------------------------------------------------------------------------
Focus group 1 11 researchers who primarily employ quantitative methods
Focus group 2 10 researchers who primarily employ mixed methods
Focus group 3 10 researchers who primarily employ qualitative methods
Focus group 4 11 (2 of whom were researchers), who primarily identified themselves as being involved in public health advocacy
Focus group 5 14 individuals (of whom 4 were researchers) working in public health policy and practice
Focus group 6 11 individuals (of whom 8 in one group were researchers) involved in public health knowledge exchange
Focus group 7 9 individuals (of whom 7 were researchers) in another group were involved in public health knowledge exchange
Total: 52 involved in research
John Wiley & Sons, Ltd
######
Participants in the second focus group discussions (*N *=* *70)
Total no. Involved in research Methodology of those involved in research Chosen topic
----------- ---------------------- ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- ----------------------------------------------------------------------------------
7 4 4 academics (all quantitative/mixed methods) Research agendas beyond health and what can we do to reduce health inequalities?
8 7 5 qualitative academics, 1 knowledge broker (with a quantitative background) and 1 individual working on research in a policy setting (with mixed methods experience) Lived experiences of health inequalities
10 8 4 quantitative academics, 1 qualitative academic, 1 individual working with research in a local policy setting, 1 researcher working in national policy contexts and 1 researcher working in a knowledge broker organisation. Participatory research and policy
10 7 4 mixed methods or quantitative academic researchers and 3 public sector researchers. Evaluation
7 5 4 quantitative academic researchers and 1 public sector researcher. Evaluation
14 6 3 primarily employ mixed methods, 1 quantitative and 2 qualitative Welfare reform/retrenchment
14 11 2 qualitative academics, 1 mixed methods academic, 2 quantitative academics and 6 researchers working in public sector/policy settings Encouraging researchers, policymakers and practitioners to work collaboratively
Total 48
John Wiley & Sons, Ltd
During the first set of focus group discussions, participants were asked to generate a list of suggestions for future health inequalities research agendas. Participants were then asked to indicate which topics they were most interested in discussing further and were allocated to one of seven afternoon focus group discussions on this basis (see Table [2](#shil12374-tbl-0002){ref-type="table-wrap"}). A small number of participants were unable to stay for the whole day (so 52 researchers participated in the first set of focus groups and only 48 in the second set). Each focus group was facilitated by a member of the steering group and participants were asked to discuss the questions outlined in Box [1](#shil12374-fea-0001){ref-type="boxed-text"}.
###### Main questions for participants in the focus group sessions
{#shil12374-sec-0020}
Questions for the first seven focus groups: What kinds of research should researchers working to reduce health inequalities focus on?What should health inequalities researchers be doing (if anything), beyond academic work, to support efforts to reduce health inequalities?
Questions for the second seven focus groups: What specifically would this new research agenda involve?What needs to happen to facilitate/enable this kind of research activity?What, if any, are the potential problems with developing this kind of research agenda?
Discussions took place under the Chatham House rule, which enabled all participants to share the content of discussions with others but only on a non‐attributable basis. All participants were asked to sign written consent forms enabling the focus groups to be digitally recorded and transcribed, before being anonymised (by KS). Two authors (KG and KS) read all 14 transcripts and jointly developed a thematic coding framework. The transcripts were then coded using NVivo 10 software (KG and KS each led on coding half the transcripts and then cross‐checked the other seven transcripts for consistency). The anonymised, coded transcripts were made available to the other steering group members for further analysis.
Results {#shil12374-sec-0003}
=======
What have we learned so far? Researchers\' reflections on the legacy of health inequalities research and policy in the UK {#shil12374-sec-0004}
-------------------------------------------------------------------------------------------------------------------------
### Post‐1997 policy advice and limited learning {#shil12374-sec-0005}
The findings reveal there is widespread concern among researchers about the failure to adequately learn from the multiple policies and interventions intended to tackle health inequalities in the post‐1997 UK context. Participants attributed this failure to the complexity of health inequalities and the policies intended to address these inequalities as well as to research funding and methodological constraints.
Several academic researchers specifically picked up on Mackenbach\'s ([2011](#shil12374-bib-0026){ref-type="ref"}) article, which had been circulated in advance and which argues that, despite some partial successes, the English strategy failed to reach its own targets of a 10 per cent reduction in inequalities in life expectancy and infant mortality. This claim was disputed by a small number of researchers, particularly with regard to health inequalities between Spearhead and other areas. The Spearhead group of 70 local authorities and 62 primary care trusts was introduced in 2004 in the 88 most health‐deprived areas in England as the focus of government interventions designed to reduce health inequalities (Department of Health [2004](#shil12374-bib-0012){ref-type="ref"}). More recent analysis seems to support Mackenbach\'s ([2011](#shil12374-bib-0026){ref-type="ref"}) assessment. Barr *et al*. ([2012](#shil12374-bib-0006){ref-type="ref"}), for example, present data which suggest that, despite some success, overall people in Spearhead areas did not experience the improvement in health that was promised. Indeed, one of the main areas of agreement among the researchers participating in the symposium seemed to be that efforts to tackle health inequalities in the UK had been less successful than people had hoped, and this was perceived to be at least partly because health improvements had been greater for those already doing well.
Beyond this broad assessment, there was a palpable frustration amongst many researchers that national‐level analyses have ignored the potential for learning lessons about what has (and has not) worked amongst the multitude of different policies and interventions that were implemented across the UK between 1997--2012:"I think we really need to pin down what\'s worked in the last 15 years. And when I re‐read the Mackenbach paper ... I felt a little bit annoyed ... . It must have been relatively easy ... to go back and pull out some key policy documents and then look at the targets. And you hear them say, 'okay, well folks this didn\'t work'. But ... I think we need to be really clear ... about what we\'ve learned over the past 15 years, and what we\'re confident about and what we don\'t know about. (Academic)"
As several participants noted, a failure to reduce health inequalities at a national level does not rule out the possibility that some interventions and policies were successful but that the impacts were countered by other policies, or that health inequalities policies prevented the gradient from getting even steeper. Responding to suggestions made by some policy actors at the event that there are examples of geographical areas in the UK in which health inequalities have been reduced, researchers widely agreed that not enough effort has gone into collating the multiple analyses and evaluations that have been undertaken of local and national policies and interventions in the UK. There was concern that, without this kind of detailed, comprehensive analysis, it would remain difficult for health inequalities researchers to adequately support future decision‐making.
### A lack of clear policy solutions {#shil12374-sec-0006}
In addition, some researchers expressed frustration at what they perceived to be an ongoing reticence within the research community to provide clear policy guidance as to the kinds of policies most likely to reduce health inequalities:AcademicI\'m remembering ... when the '97 Labour government came in and said they wanted to do something about health inequalities, they looked around at us researchers and said, 'well, what should we do?' AcademicIt was embarrassing. AcademicAnd we\'ve spent ... time since the Black Report \[Black, 1980\] knocking down all the criticisms of what we were doing. We haven\'t been developing our own agenda, so I think we want to forget about being defensive and we want to say, well, these are the things which the government, if it\'s serious, could do.
The comments quoted above refer to the independent commission of inquiry, commissioned by the UK Labour government that was elected in 1997 into health inequalities research (Acheson [1998](#shil12374-bib-0001){ref-type="ref"}), which made 39 recommendations. It had, therefore, proved possible to develop some policy guidance from the available research at this time. However, these recommendations were criticised at the time for being vague and uncosted (with no hierarchy for putting the 39 recommendations in place) and for underrepresenting structural and socioeconomic determinants of health (Davey Smith *et al*. [1998](#shil12374-bib-0011){ref-type="ref"}). The focus group discussions drawn on here suggest that this situation has not yet improved substantially enough to enable the kind of evidence‐based policy advice that both researchers and policymakers desire (Whitehead *et al*. [2004](#shil12374-bib-0050){ref-type="ref"}).
Proposals for future health inequalities research agendas {#shil12374-sec-0007}
---------------------------------------------------------
Reflecting the palpable concern with developing clearer policy guidance, when asked what researchers ought now to be focusing on, the most common responses centred on discussing means of improving knowledge about the actual and likely impacts of different kinds of policies and interventions. However, there were sharply contrasting perspectives on how best to enhance this kind of knowledge.
### Improving evaluations {#shil12374-sec-0008}
For some researchers, the need to improve knowledge about what works in reducing health inequalities necessitated a far more rigorous evaluation of interventions intended to address inequality, with a focus on their efficacy and cost‐effectiveness (see Egan *et al*. [2009](#shil12374-bib-0013){ref-type="ref"}, Wanless [2004](#shil12374-bib-0047){ref-type="ref"}). One senior academic went as far as suggesting that researchers should decline to be involved in assessing interventions or policies where a proper evaluation was not possible:"I think researchers should be tougher in saying \[to policymakers\], 'if you do it like that, then we\'re not going to be involved in the evaluation' ... . That\'s my feeling, that researchers should take a stronger line, and also the politicians should, but I think researchers have a role in actually saying, 'if you do that evaluation that way, even if you give us five million quid, we actually can\'t give you the answer'."
However, two linked concerns about a turn towards evaluating interventions were also evident. First, some participants claimed it was difficult to obtain resources and support to evaluate the impacts of macro‐level policy shifts or non‐health policies on health inequalities, particularly where policy changes might be expected to have a negative impact on health inequalities, such as the current welfare reforms. As a consequence, some researchers argued that a focus on evaluating interventions unintentionally leads to the widely discussed problem of 'lifestyle drift' (Popay *et al*. [2010](#shil12374-bib-0034){ref-type="ref"}):"Often the focus on ... randomised control trials \[and\] classical intervention studies means that most of the most important determinants \[are not assessed\]. I mean ... things to do with welfare reform and occupational structure are not going to be easily evaluated in that way.(Academic)"
Secondly, relating to this, several researchers argued that there has been a dearth of macro‐level policies implemented in the UK over the past 30 years that might be expected to reduce health inequalities and, therefore, there have been limited opportunities to evaluate policies likely to reduce health inequalities. For example:"We need to do more intervention research but part of the problem we\'ve got, I think, over the last decade or so is that there\'s so few promising interventions, in truth, that have happened at a policy level that we would want to really spend a lot of time evaluating. So we\'ve had health action zones, we\'ve had the tobacco ban \[i.e., the smoking ban in most indoor public places\], we\'re going to have, hopefully, minimum pricing, but beyond that there\'s very few big policy experiments.(Public sector researcher)"
The difficulties in evaluating policy shifts and interventions in the real world, where policies are rarely implemented in ways that enable randomised comparisons, has been widely discussed in public health (Perkins *et al*. [2010](#shil12374-bib-0031){ref-type="ref"}). Specific suggestions from participants as to how researchers might address this included: (i) paying more attention to the impacts of policy changes and interventions on known social determinants of health, without necessarily requiring health indicators to be captured (see Bambra *et al*. [2010](#shil12374-bib-0003){ref-type="ref"}), (ii) working more closely with policymakers to roll out major policy changes in ways which allow for a proper evaluation and (iii) expanding methodological approaches (see below).
Despite these suggestions, for some participants it seemed that the increasing funding opportunities for evaluation orientated research were actively 'damaging' health inequalities research by narrowing the focus onto behavioural and individualised interventions that are easier to evaluate using positivist quantitative methodological frameworks. The concerns raised by some participants are summed up in a recent essay by Ted Schrecker:"Use of the randomized controlled trial (RCT) as the gold standard for intervention research, sitting atop a hierarchy of evidence ... incorporates a set of methodological value judgments that merit reconsideration. Although examples exist of sound RCTs of large‐scale policy initiatives ... many kinds of interventions and policies cannot be assessed using RCTs, for reasons of ethics, costs, logistics, or all of these. Even when an RCT is conceptually possible, insisting on evidence from RCTs may build into intervention research a bias against larger‐scale, contextual interventions that are difficult to evaluate in this manner.(Schrecker [2013](#shil12374-bib-0040){ref-type="ref"}: 742)"
In line with Schrecker\'s account, several of the more qualitative researchers involved in discussions argued that a preoccupation with evaluating interventions was squeezing out other types of research, with one senior academic referring to her sense that health inequalities research was increasingly being guarded by a 'trial police'. For another senior academic, these concerns prompted a rejection of the very terminology associated with the evaluative turn in public health:"I\'ve got to the point now where, for a long time I\'ve not used the word lifestyle -- it\'s a refusal in me to actually even use the term and I\'m getting the same feeling now with the word intervention: I can\'t bear it when people start talking about interventions in relation to the health of deprived communities ... I can\'t bear the imposition of the straightjacket of the kind of work we do."
The data presented in this section underline the obvious methodological tensions within the multi‐disciplinary field of health inequalities research and suggest that researchers prioritise particular methodological approaches (usually their own) over others. A likely effect of this is the compartmentalisation of research funding in ways that potentially encourage partial investigations into complex issues. The final quotation suggests there may be a related division in the kind of language that different researchers choose to employ.
### Expanding our methodological toolkit {#shil12374-sec-0009}
Most participants agreed methodological innovation, including greater use of mixed methods, is required to develop better understandings of the impacts of policy change on health inequalities but more precise suggestions differed. Some researchers were concerned that the right things are not yet being measured (for a debate on this topic, see Frank and Haw [2011](#shil12374-bib-0016){ref-type="ref"}, McCartney *et al*. [2013](#shil12374-bib-0025){ref-type="ref"}) but several (more quantitative) researchers suggested emerging data linkage opportunities could afford new opportunities for understanding the impacts of different kinds of interventions and policy changes on health inequalities:"One of the other areas which is growing in capacity and interest is the data linkage. I simply see that as ... another way of being able to better evaluate social interventions, policy or whatever. Primary data collection, as we all know, is really expensive and if ... we could draw on comparative stuff, qualitative stuff, primary, quantitative and ... routine data across not just medical but across education, criminology, justice, then I think you could begin to get a ... very rich picture of what might be working and what might not be working.(Academic)"
Overall, there seemed to be a consensus that there is a need for a more interdisciplinary approach to studying health inequalities and that researchers need to become more imaginative at incorporating different kinds of data sources into analyses (see Table [3](#shil12374-tbl-0003){ref-type="table-wrap"}).
######
Researchers\' suggestions for future directions in health inequalities (HIs) research
------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
Topics requiring further research New/emerging topics to explore Suggested/conceptual approaches Suggested collaborations, links & syntheses
-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- ---------------------------------------------------------------------------------------------------------------------------------------------------------------- --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
Links between income, job status and HIs by studying the organisations in which we work (e.g., NHS bodies and universities);\ Likely (and actual) impacts of contemporary welfare cuts, austerity measures, etc. on health inequalities;\ More mixed methods research to develop comprehensive methodological approaches to studying health inequalities;\ Invest more resources in comprehensively collating/synthesising the various existing analyses of the impacts on HIs of interventions and policies.\
The impacts of social and structural violence^1^ on health inequalities;\ Likely (and actual) impacts of NHS reforms (in England) on health inequalities;\ More ethnographic and anthropological methods to better understand people\'s life‐worlds (including using innovative approaches such as 'human libraries');\ Work with policy scholars to better understand the policymaking process and the factors necessary to facilitate policy change;\
Community assets and resilience;\ Develop links between environmental/climate change & HIs research;\ Assess the impact of major policy shifts on HIs by employing more historical & international data;\ Work to better understand intersectionality within HIs (i.e., studying the interactions between different axes of health inequalities, such as class, ethnicity and gender);\
Intervention‐generated HIs (policies and interventions that unintentionally function to widen health inequalities);\ Public, political and policy perceptions and understandings of health inequalities;\ Work with policymakers to ensure commissioned evaluations are rigorous (consider refusing to undertake commissioned evaluations where this is not possible);\ Develop more interdisciplinary/multidisciplinary research teams for HIs projects (e.g., including anthropologists, economists, education and social policy scholars, historians, sociologists and political scientists).
How different kinds of upstream social factors lead to biological (including genetic) changes in individuals (i.e., better understanding the aetiological pathways that lead to health inequalities);\ Role of corporations in HIs debates and in shaping the policies impacting on health inequalities;\ Exploit emerging opportunities for data linkage (both in terms of quantitative data sets and linking quantitative data sets with available qualitative data);\
Links between 'class' (in sociological and political terms) and HIs (see Scambler, [2012](#shil12374-bib-0039){ref-type="ref"}). Identify more 'intermediate' structural determinants of HIs to enable better studies of the relationship between upstream changes and HIs (helping to guard against 'lifestyle drift');\ Exploit the increasing availability of secondary qualitative data sets, such as oral histories;\
Investigate the relationships between violence, drugs & HIs (esp. in Scotland);\ Undertake more participatory collaborative research with relevant communities.\
Research social and economic inequalities as well as HIs (given what is known about the links between these broader kinds of inequalities and health inequalities), both because this kind of research may be more informative and because local communities are more likely to engage with it. Work to integrate studies of HIs with relevant concepts in political science, sociology and social theory.
------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
1 Social and structural violence refers to political and economic inequality (Farmer [1997](#shil12374-bib-0014){ref-type="ref"}).
NHS, National Health Service.
John Wiley & Sons, Ltd
### New research directions {#shil12374-sec-0010}
When asked to consider new research directions, the most common theme throughout the discussions was the need for research to help understand the impacts of major policy reforms currently being implemented in the UK (e.g., public funding cuts and, in England, NHS reforms):"I think one of the things that\'s really important ... is the absolute reality that the impact of public sector cuts and the welfare cuts which are utterly savage, I mean savage ... I mean that\'s the reality and I think, as researchers within public health ... that\'s a really important issue to take on board, kind of morally.(Senior academic)"
In order to better understand the impacts of economic distress and what have been labelled austerity measures on people\'s health and wellbeing, researchers suggested: (i) exploring historical and international experiences of similar situations (as noted above) and (ii) undertaking more in‐depth qualitative research to better understand people\'s lived experiences of these changes:"How good ... is the research community at knowing and understanding communities and working with communities, quite genuinely listening to them? ... None of us do actually really listen ... . It\'s about actually listening to people about what it\'s ... like for \[them\] where they\'re living and what are the challenges they face on a daily basis -- which will often be different from what we think they are, because we don\'t know, because we\'re all middle class.(Researcher)"
However, another researcher challenged the above assertion, arguing instead that a particular focus on deprived communities can shift the focus away from the need for population‐level change, while also unintentionally stigmatising people living in communities labelled as deprived:"We do tend to conceptualise health inequalities as \[being\] about particular well‐defined communities, and of course it\'s all about the gradient. It\'s not only about the people in those communities, and sometimes I think we do harm by focusing so much on the deprived communities and calling them 'deprived communities' ... Sometimes you make the stigma worse ... because you\'ve defined them as deprived communities and all your documents say, 'we\'re focusing on people in these deprived communities' ... And I see the faces, sometimes, of the admin staff with some of the things that I write, where we say we\'re going to prioritise these deprived communities, and they live there! And they say, 'but wait a minute, actually we\'re not deprived, we live there -- it\'s a perfectly good community'. And I think we need to be a bit careful about saying there are some people who suffer from health inequalities and they\'re the deprived folk, then there\'s everybody else.(Researcher)"
These kinds of concerns reflect the findings of a recent seven‐country comparative study of the effects of poverty, undertaken by Walker *et al*. ([2013](#shil12374-bib-0046){ref-type="ref"}), which argues that the shame associated with poverty is one of the most important dimensions for understanding how and why poverty impacts negatively on people\'s lives. Likewise, Susan George, warns researchers against studying 'the poor and powerless', who 'already know what is wrong with their lives', noting that this kind of research can be used in ways other than those intended by the sympathetic researcher (George [1976](#shil12374-bib-0017){ref-type="ref"}: 289).
There is, therefore, a dilemma facing health inequalities researchers; while many participants suggested it would be extremely helpful to have more in‐depth insights into the everyday experiences of living in difficult circumstances to better understand health inequalities (with several noting how different researchers\' own lived experiences tend to be), there were also clear concerns around the potentially stigmatising consequences of this kind of research.
George\'s suggestion is that researchers ought to instead work to better understand 'the rich and powerful' (George [1976](#shil12374-bib-0017){ref-type="ref"}: 289) and several researchers supported this idea, calling for more research to understand how well‐resourced actors work to shape policies in ways that contribute to health inequalities. This is a strand of research that has so far been underexplored in health inequalities, although Scambler has drawn attention to the issue via his 'greedy bastards hypothesis' (Scambler [2007](#shil12374-bib-0037){ref-type="ref"}, [2009](#shil12374-bib-0038){ref-type="ref"}) which asserts that 'Britain\'s widening health inequalities can be seen as a largely unintended consequence of the voracious, strategic appetites' of capitalist power elites (Scambler [2012](#shil12374-bib-0039){ref-type="ref"}: 137). Specifically, several researchers argued that the role of corporations in generating health inequalities required exploration:"One thing that I think has been under‐investigated is the paradox that currently most of the immediate causes of premature mortality are due to overconsumption of, whether it\'s tobacco, alcohol, food and drugs, and yet the impact is the greatest on most disadvantaged groups. And it seems to me that we\'ve not really focused enough on the free market economy and on the big multinational companies and the way in which they can exploit, particularly the weakest, to create this sense that, 'I must have this, I need this to make me happy', and so on, with absolutely devastating results. (Knowledge broker)"
Beyond the above suggestions, which stimulated significant discussion, there was a wealth of further suggestions for topics deserving of more attention within health inequalities research. As it is not feasible to discuss each suggestion within the space of this article, Table [3](#shil12374-tbl-0003){ref-type="table-wrap"} provides a summary.
It is worth noting that some researchers with interests in qualitative methods claimed that funders were increasingly more inclined to fund quantitative studies and that, 'where qualitative aspects are included', there was 'a tendency for them to be seen as of marginal importance and stripped out'. This seems important given that most of the methodological suggestions put forward by researchers (as summarised in Table [3](#shil12374-tbl-0003){ref-type="table-wrap"}) involved qualitative or mixed methods research.
Suggestions for improving the influence of health inequalities research on policy and practice {#shil12374-sec-0011}
----------------------------------------------------------------------------------------------
There was a strong emphasis throughout the discussions on the need to improve links between health inequalities research, policy and practice but clear differences of opinion regarding the best way to achieve this.
### Getting the public on board {#shil12374-sec-0012}
Reflecting recent suggestions that health inequalities researchers ought to do more to ensure future governments have the necessary democratic mandate to implement some of the upstream measures that the evidence suggests may be required (Mackenbach [2011](#shil12374-bib-0026){ref-type="ref"}, Whitehead and Popay [2010](#shil12374-bib-0049){ref-type="ref"}), many participants emphasised the need to improve relationships between health inequalities researchers and public (as opposed to policy) audiences:"We\'re not hugely public and most of us don\'t write commentaries for newspapers or letters to the papers, or do those sorts of media things that economists do all the time ... There are enough of us really who have known for a long, long time that if you want to improve the public health and reduce health inequalities you\'ve got to deal with the structural determinants, and still most members of the public don\'t think that. They think people die early because they\'re feckless, reckless, smoke and drink too much. So, we haven\'t got it across, not just to policymakers, we haven\'t got it across to the public.(Academic)"
Other researchers challenged these claims, noting that there has been insufficient research in this area to really know what the public think about health inequalities. One researcher argued that the studies of lay perceptions of health inequalities point to a relatively sophisticated understanding of the social determinants of health, at least in communities bearing the brunt of these inequalities (Popay *et al*. [2003](#shil12374-bib-0033){ref-type="ref"}). Nonetheless, researchers involved in the symposium seemed to feel it was at least as important to communicate and engage with local communities and the public as they did to engage with policy audiences. Indeed, some of the discussions involved some reflection as to who, ultimately, health inequalities research is for:"I am now struggling to understand who my research is for. And do I want to waste my time doing research for a government who is either going to punish me for daring to study the effects of ill‐health in housing estates, or just disregard my research? And so now I\'m thinking it\'s kind of like an opportunity to reflect on what I\'m doing, and it was a bit uncomfortable really just sitting here and listening to \[symposium presentations and discussion\] because they all seemed to be saying, blatantly ... who is the research for?(Academic)"
The unspoken implication of these kinds of contributions seemed to be that the public, or local communities, are important agents of change, although how public and policy perceptions of health inequalities might link was not explicitly discussed.
Overall, there seemed to be a consensus that health inequalities researchers need to get better at engaging with various non‐academic communities. However, there was a distinction between researchers who emphasised the importance of listening to, and working collaboratively with, relevant communities (as the above participant went on to do) and those, such as the following researcher, who focused on the need to enhance general public interest in, and understanding of, health inequalities:"If you look back at the pioneers of social policy ... Richard Titmuss, Peter Townsend, Brian Abel‐Smith ... and all the rest of it, they would tell you that one op‐ed piece in a broadsheet newspaper is worth 25 peer review journal articles ... Or speaking to the local radio or whatever. So I think communicating to real people in a practical way needs to be part of what\'s seen as our legitimate activities.(Academic)"
These two differing ways of framing engagement map onto the contrasting definitions of advocacy outlined in the introduction; while the former seems to be more about helping people to get their voices heard (facilitational advocacy -- Carlisle [2000](#shil12374-bib-0008){ref-type="ref"}), the latter seems to be about selling particular issues and goals (representational advocacy -- Carlisle [2000](#shil12374-bib-0008){ref-type="ref"}, Chapman [2007](#shil12374-bib-0009){ref-type="ref"}).
### Getting political {#shil12374-sec-0013}
A few researchers went further, arguing that it is essential for health inequalities to work more strategically to achieve social and policy change, including by developing better advocacy skills:"There\'s a reluctance to recognise that public health is essentially a political discipline ... I do think it\'s indefensible that advocacy isn\'t a core public health competence and ... I don\'t think it\'s a coincidence that lots of the pressure for advocacy being integrated into the public health curriculum has come from people like Simon Chapman \[in tobacco control\] who\'ve long argued that, why is it that we\'re prepared to go on and make media appearance without any training, but will sit down and rehearse a conference paper that\'s going to be listened to by 20 people?(Academic)"
However, there was also palpable apprehension amongst some participants of the ability of academic researchers to take on advocacy roles:"I\'m uncomfortable with that because ... not because advocacy\'s wrong but because some people within the community are better suited and skilled to engage in advocacy than others. So I prefer the word grounded -- I think even if you\'re not good at advocacy, if you\'re engaged in this field you should ground yourself in real everyday experiences with people in communities.(Academic)"
Some participants suggested that, rather than becoming advocates or lobbyists themselves, researchers ought simply to develop better links with other kinds of campaigners and policy advocates, such as third sector organisations. However, others expressed concern that the topic‐based nature of many third sector organisations could mean such an approach might unintentionally exacerbate 'lifestyle drift', and place a greater focus on 'the ambulance at the bottom of the cliff ... than your fence at the top'. These kinds of debates suggest it may also be useful to work to better understand the role of third sector organisations in health inequalities debates.
Concluding discussion {#shil12374-sec-0014}
=====================
Reflecting on the views of the 52 researchers who participated in the symposium, it seems there is a strong desire to identify clear recommendations for action to reduce health inequalities and to work collectively to promote these recommendations. Beyond this, however, three distinct types of researcher seem identifiable, each of whom has definable preferences over the most promising areas of research, the most appropriate methodological approaches and the most important non‐academic audiences to engage with. The three types are described below in the manner of Weber\'s ([1962](#shil12374-bib-0048){ref-type="ref"}) ideal types (that is, the descriptions are logical constructions to help elucidate some of the coalitions and divisions that are evident in the data); in reality (as discussed further below), individual researchers often made contributions that would place them in more than one of these categories.
Policy‐focused positivists {#shil12374-sec-0015}
--------------------------
This type of health inequalities researcher is strongly committed to the idea that quantitative, experimental research designs provide the most useful insights into understanding what works to reduce health inequalities. In terms of moving health inequalities research forward, researchers in this category tended to promote the need for more (and better) evaluations of interventions and policies, including through the exploitation of data linkage opportunities. They came across as policy‐focused, in the sense that they want to ensure the research is useful for policy audiences (who were often conceived of in a relatively narrow sense, such as senior civil servants and ministers). However, in prioritising scientific independence and methodological rigour, it was also clear that they were wary of researchers becoming co‐opted by policymakers or interpreting data through a political or ideological lens.
This way of approaching health inequalities came under significant criticism during the symposium discussions for at least four reasons. Firstly, some researchers argued that, no matter how hard researchers might try to persuade policymakers of the benefits of rolling interventions out in ways that enable effective evaluation, major policy changes tend to be driven by reasons other than research, making it unlikely that this will change. A small number of hardline researchers in this category argued that researchers should simply refuse to engage in evaluations that they felt were likely to be hampered by the policy design or roll‐out. However, the most common, 'softer' response was to suggest that researchers needed to supplement traditional approaches to evaluation with broader, more sophisticated methodological approaches. Secondly, some researchers expressed concern that a greater focus on evaluation‐orientated research results, in practice, in lifestyle drift (Popay *et al*. [2010](#shil12374-bib-0034){ref-type="ref"}: 148), both because it is difficult to apply rigorous, quantitative evaluative methods to macro‐level (or even meso‐level) policies and because large‐scale policy shifts are relatively unusual, limiting opportunities to study the impacts of macro‐level policy changes. Thirdly, some researchers criticised the elite orientation of this way of working, which tends to focus on the views of experienced researchers and senior bureaucrats, paying little (if any) attention to the views of the communities most negatively impacted on by health inequalities (see Scambler [2012](#shil12374-bib-0039){ref-type="ref"}). Finally, some researchers simply noted that the benefits of this approach for reducing health inequalities remain unproven (i.e., that, despite investments in this way of working, advances in our knowledge of how to reduce health inequalities remain limited).
Empathetic ethnographers {#shil12374-sec-0016}
------------------------
Other participants argued that health inequalities researchers need to interrogate what health inequalities mean in people\'s social worlds (Scambler [2012](#shil12374-bib-0039){ref-type="ref"}: 144). These researchers called for more priority to be afforded to listening to, and working to understand, the experiences of communities experiencing the brunt of health inequalities. Reflecting Ruth Lister\'s assertion, these participants were often overtly critical of research in which *'*those with every day experience of living in poverty' are treated as research objects, rather than having 'their thoughts published' (Lister [2004](#shil12374-bib-0024){ref-type="ref"}: 2). This approach was much more ethnographic in nature, with preferred methods including in‐depth qualitative methods to capture the complexity of people\'s everyday experiences in order to help: (i) better understand the pathways linking upstream determinants and health inequalities and (ii) prioritise research issues overtly impacting on the people bearing the brunt of health inequalities (rather than assessing what is worthy of further research based on identifying gaps in the available evidence). This group of researchers seemed committed to advocacy in the sense of working to help voices that are often ignored to become more audible (Carlisle [2000](#shil12374-bib-0008){ref-type="ref"}), including by co‐producing research with communities (Letcher and Perlow [2009](#shil12374-bib-0023){ref-type="ref"}).
In the symposium discussions, this kind of approach to studying health inequalities seemed to enjoy broad support. However, many contributors said they felt that funders were unlikely to provide resources for this kind of research. Three criticisms of this approach were also put forward, including by researchers who seemed generally sympathetic to qualitative research. Firstly, some researchers cautioned that a focus on researching poorer communities exacerbates the tendency to conceptualise health inequalities as a matter of health deprivation (i.e., a problem caused by the poor health of poor people), rather than a social gradient of health (see Graham and Kelly [2004](#shil12374-bib-0019){ref-type="ref"}). Secondly and relatedly, some participants argued that this can be unintentionally stigmatising for the communities involved. Thirdly, some researchers questioned whether commitments to working with communities necessarily effect change (although some successful examples were cited -- see Martin *et al*. [1987](#shil12374-bib-0028){ref-type="ref"}, Roberts *et al*. [2011](#shil12374-bib-0036){ref-type="ref"}).
Critical materialists {#shil12374-sec-0017}
---------------------
Taking a rather different approach, some participants called for more attention to be paid to exploring how elite actors (including corporations) shape policies and debates in ways that create and exacerbate health inequalities. For these participants, the link between health inequalities and wider social and economic inequalities seemed undisputed so their priority for future research was less about trying to understand the aetiological pathways linking policies and interventions to everyday health experiences and more about working to reveal the extent to which society is becoming less equal, who is shaping these decisions and how power relations are enacted (see Coburn [2004](#shil12374-bib-0010){ref-type="ref"}, Navarro [2009](#shil12374-bib-0029){ref-type="ref"}, Scambler [2012](#shil12374-bib-0039){ref-type="ref"}). It was suggested this kind of work involved drawing on social and political theory, as well as empirical research (again, see Scambler [2012](#shil12374-bib-0039){ref-type="ref"}), another area in which there seemed to be some consensus that funding support was limited.
From this perspective, health inequalities research is, in itself, a political activity. The notion that researchers ought to be engaging in advocacy therefore tended to be supported by researchers in this category and involved working 'to provoke critical thinking' and public awareness of, and agitation against, the power imbalances underlying health inequalities (see Krieger *et al*. [2012](#shil12374-bib-0022){ref-type="ref"}). However, it was unclear that researchers in this category had a defined sense of what they were advocating (rather than what they were against). As Scambler ([2012](#shil12374-bib-0039){ref-type="ref"}) notes, the prospects for realising the conditions in which social and economic inequalities might be substantially reduced is a challenge more often posed than taken up. Without being able to offer obvious solutions, there are some important ethical questions to be addressed. For, if the communities in question do not have control over the levers shaping the structures that are negatively impacting on their lives, drawing attention to the health‐damaging effects of these structures may, in itself, have damaging implications (including the kind of stigmatising impacts discussed earlier).
Where do we go from here? {#shil12374-sec-0018}
=========================
The three types of health inequalities researcher outlined above are depicted as distinct from one another for heuristic purposes. In reality, it was evident from the discussions that the boundaries between these different types were not as distinct as the descriptions above might suggest; indeed, several researchers made contributions to discussions that could lead them to be positioned in more than one of these groups. Moreover, while the participants were selected with a view to ensuring a range of perspectives was represented, there are likely to be other perspectives that were not captured through this research (certainly some areas, such as health economics, were underrepresented in the sample, despite specific efforts by the steering group to attract health economists to the event). Nonetheless, the focus group discussions suggested that the three different ways of approaching health inequalities outlined above inform three different routes forward for future health inequalities research. This, in turn, appeared to be informing competing sources of policy advice (e.g., advice based on the best available evaluations of specific interventions versus advice informed by critical analyses of overarching policy paradigms) and contrasting views about with whom, beside other researchers, academics ought to be engaging (national level policymakers, local communities or the wider public).
The different approaches seem to relate to deeply held epistemological and ideological positions that are unlikely to shift. Researchers who are strongly committed to positivist approaches are unlikely, for example, to appreciate the benefits of ethnographic or co‐produced research and may dismiss critical materialist work as being overtly ideological (see Gieryn [1983](#shil12374-bib-0018){ref-type="ref"}, Smith [2013a](#shil12374-bib-0041){ref-type="ref"}). Focusing on trying to persuade each other of the benefits of our preferred way of working therefore seems ill‐fated. Yet, in the context of limited research funding, the data suggest that different types of researchers do view themselves as competing with one another. This is unsurprising, to the extent that it reflects well‐established phenomena within the sociology of science. Gieryn ([1983](#shil12374-bib-0018){ref-type="ref"}), for example, describes scientists undertaking what he calls 'boundary work' in order to distinguish between what is considered to be science and what is not, with a view to enhancing the credibility of their own research and, therefore, their access to resources. This kind of boundary work was evident in the focus group discussions, with some health inequalities researchers in each of the three categories criticising methodological approaches that differed from their own and challenging the legitimacy and funding of other types of research. Such divisions seem likely to promote the continued compartmentalisation of methodological approaches to studying health inequalities, despite the fact that many researchers acknowledged there was a need for more mixed methods research to better understand the complexity of interweaving and dynamic inequalities.
All this suggests that the field of health inequalities research is unlikely to appear cohesive to external (e.g., policy and advocacy) audiences. This is despite the fact that: (i) focus group participants generally agreed that the 2008 economic crisis and subsequent 2010 austerity agenda of public cuts were likely to have important (largely negative) impacts on health inequalities; (ii) a recent online survey of researchers suggests that there are areas of consensus around policy recommendations (Smith and Kandlik Eltanani 2014) and (iii) many of the researchers involved in the symposium made contributions that would place them in more than one of these categories.
Rather than pitting these three ways of approaching health inequalities against one another a more promising approach might be to better enable space for each of these different ways of working, particularly when it comes to studying the complex effects of clusters of policy changes, such as the current austerity‐led welfare reforms. Indeed, if we consider areas of public health in which significant policy and societal changes have been achieved, such as tobacco control, there is evidence that all three kinds of approach have made important contributions (Smith [2013b](#shil12374-bib-0042){ref-type="ref"}). Over time, it seems plausible that different approaches may lead to similar conclusions about the most desirable (or effective) kinds of policy responses, or that a combination of insights from the various approaches may collectively contribute to reductions in health inequalities. Moreover, as Petticrew *et al*. ([2004](#shil12374-bib-0032){ref-type="ref"}) note, different policy actors may be more or less persuaded by different kinds of evidence. This requires researchers to step back from boundary debates about what is, and what is not, considered legitimate (or even optimal) in health inequalities research. This is, however, likely to be particularly difficult in the context of an increasingly market‐based university system (Marginson and Considine [2005](#shil12374-bib-0027){ref-type="ref"}).
While the current situation does not make it impossible to advocate the reduction of health inequalities, it certainly makes it more challenging than recent calls imply (Mackenbach [2011](#shil12374-bib-0026){ref-type="ref"}), particularly as the distinct approaches to health inequalities identified in this article themselves evoke different ways of thinking about advocacy and policy change. For example (thinking back to the introduction), only empathetic ethnographers appeared to conceive of advocacy in the facilitational manner that Carlisle ([2000](#shil12374-bib-0008){ref-type="ref"}) describes; and, while policy‐focused positivists and critical materialists both appeared to favour a representational form of advocacy, their conceptualisations of the kind of audiences and work involved in this differed in ways reminiscent of the distinction Scambler ([2012](#shil12374-bib-0039){ref-type="ref"}) makes between policy sociology and critical sociology. Hence, without (at the very least) a greater appreciation of these different ways of working in the research community, it is likely to be difficult for researchers to contribute to developing effective advocacy for health equity (Farrer *et al*. [2015](#shil12374-bib-0015){ref-type="ref"}).
The symposium was jointly funded by the University of Edinburgh through a Challenge Investment award from the College of Humanities and Social Sciences, and by the Wolfson Research Institute for Health and Wellbeing, Durham University. KS was funded by an MRC‐ESRC Postdoctoral Fellowship during the symposium and its organisation (PTA‐037‐27‐0181) and by an ESRC Future Research Leaders award during the analysis and write up (ES/K001728/1). JP is supported by the European Research Council (ERC‐2010‐StG Grant 263501).The views in the article are those of the authors or participants and not necessarily of the funders. We particularly wish to thank the participants who gave up their time to take part in the symposium. We would also like to thank Heide Weishaar for assisting with the organisation of the symposium and other members of the steering group, Sarah Hill, David Hunter, Steve Platt and Niamh Shortt.
| {
"pile_set_name": "PubMed Central"
} |
Mit Beschluss vom 22. November 2019 hat der Gemeinsame Bundesausschuss (G-BA) eine Richtlinie zur Versorgung der hüftgelenknahen Femurfraktur erlassen. Nach Prüfung durch das Bundesministerium für Gesundheit und Veröffentlichung im Bundesanzeiger sollte die Richtlinie am 1. Juli 2020 in Kraft treten. Aufgrund der COVID-19-Pandemie wurde die Inkraftsetzung der Richtlinie auf 2021 verschoben.
Mit dem G-BA-Beschluss einer Richtlinie zur Versorgung der hüftgelenknahen Femurfraktur im Herbst 2019 werden allgemeine und spezifische Mindestanforderungen an die Struktur- und Prozessqualität in für die Versorgung zugelassenen Krankenhäusern als eine MaÃnahme zur Qualitätssicherung festgelegt. Dabei zielen die G-BA-Vorgaben auf die Operation coxaler Femurfrakturen innerhalb von 24 Stunden. Zur Erfüllung der Auflagen an die Prozessqualität werden den Krankenhäusern einige SOP (Standard Operating Procedures) abverlangt - insbesondere auch zur orthogeriatrischen Zusammenarbeit. Mit der Forderung nach einem orthogeria-trischen Ko-Management wird für viele unfallchirurgische Krankenhäuser Neuland betreten.
Anpassung Kriterienkatalog ATZ {#Sec1}
==============================
Zertifizierte AltersTraumaZentren (ATZ-DGU) erfüllen die Anforderungen des G-BA inhaltlich im Wesentlichen bereits jetzt. Dabei folgt der derzeit gültige Kriterienkatalog AltersTraumaZentrum DGU® (Version 1.2) in seiner Systematik nicht der des G-BA-Beschlusses. Andererseits gehen die inhaltlichen Anforderungen an ein AltersTraumaZentrum DGU® weit über die Vorgaben des G-BA hinaus. Einige allgemeine Mindestanforderungen des G-BA werden im Kriterienkatalog nicht explizit benannt - wie beispielsweise die Notwendigkeit eines verantwortlichen Arztes mit der Zusatzweiterbildung \"Klinische Notfall- und Akutmedizin\". Und einzelne vom G-BA benannte SOP sind im Katalog der ATZ-DGU als solche noch nicht aufgeführt.
Die DGU mit der Sektion Alterstraumatologie als herausgebende Stelle des Verfahrens sieht es als ihre Aufgabe an, dass mit der Zertifizierung als AltersTraumaZentrum DGU® per se und nachvollziehbar die Umsetzung der G-BA-Richtlinie gewährleistet ist. Dabei stehen die alle drei Jahre stattfindenden Audits für die Konformität der klinischen Praxis in ATZ-DGU mit den Anforderungen des Kriterienkatalogs.
Zweite Revision des AltersTraumaZentrum DGU® {#Sec2}
=============================================
Der Arbeitskreis AltersTraumaZentrum und AltersTraumaRegister der Sektion Alterstraumatologie hat nun den Kriterienkatalog AltersTraumaZentrum DGU® unter MaÃgabe der G-BA-Anforderungen überarbeitet. Die damit zweite Revision des Kriterienkatalogs in seiner Version 1.3 beinhaltet über die Anforderungen des G-BA hinaus insbesondere eine weitere Fokussierung und Konkretisierung des orthogeriatrischen Ko-Managements auf die unmittelbar perioperative Behandlungsphase des Alterstraumas. Besonderes Gewicht kommt dabei der perioperativen Risikostratifizierung, einem entsprechenden Risikomanagement und der vermehrten perioperativen Berücksichtigung des Delirs zu.
Der Kriterienkatalog steht ab 1. Juli 2020 auf der Website www.alterstraumazentrum-dgu.de der AUC - Akademie der Unfallchirurgie GmbH - zur Verfügung. Mit Gültigkeit des weiterentwickelten Kriterienkatalogs ab dem 1. Januar 2021 steht es interessierten Zentren frei, sich in der Ãbergangsphase des zweiten Halbjahres 2020 nach dem alten oder bereits neuem Kriterienkatalog zertifizieren zu lassen. Mit dem aktualisierten Kriterienkatalog bürgt somit ein zertifiziertes AltersTraumaZentrum DGU® auch für die vorbildliche und mit dem Auditprozess nachweisliche Umsetzung der G-BA-Vorgaben.
Muster-SOP unter Federführung der AUC {#Sec3}
======================================
Die Sektion Alterstraumatologie der DGU gibt hierzu Muster-SOP heraus, die den G-BA-Anforderungen umfänglich und sehr ausdifferenziert folgen. Diese SOP wurden von einer interdisziplinären Expertengruppe unter Federführung der AUC erstellt und sind so weit gefasst, dass sie nach entsprechender Anpassung an die örtlichen Gegebenheiten nicht nur in ATZ-DGU, sondern ebenso für Unfallchirurgien in Krankenhäusern unterschiedlicher Strukturen und Versorgungsstufen anwendbar sind.
Diskussion in der Webkonferenz {#Sec4}
==============================
Mit dem G-BA-Beschluss zur Behandlung der Hüftfrakturen befasste sich auch eine Online-Konferenz am 18. Juni 2020 - organisiert und veranstaltet von der Klinik für Allgemeine, Unfall- und Wiederherstellungschirurgie des LMU Klinikums München, der AUC und der Deutschen Gesellschaft für Geriatrie (DGG). Vor 145 Teilnehmern wurden die G-BA-Richtlinien diskutiert. Dies unter den Gesichtspunkten eines notwendigen orthogeriatrischen Ko-Managements, aus Sicht des Medizinischen Dienstes der Krankenversicherung (MDK), der Geriatrie und des G-BA selbst. Die von der Sektion Alterstraumatologie der DGU als Umsetzungshilfe herausgegebenen Muster-SOP wurden hier erstmals vorgestellt. Dabei wurde von den Autoren noch einmal ausdrücklich hervorgehoben, dass die SOP von den Anwendern natürlich an ihre spezifischen Rahmenbedingungen angepasst werden müssen. Die Muster-SOP können über das TraumaPortal heruntergeladen werden.
**Aufzeichnung des Webinars:**r https://www.youtube.com/watch?v=9TFE-ISIIak
**Die Muster-SOP können über das Trauma-Portal heruntergeladen werden:** http://www.traumanetzwerk-dgu.de/de/startseite_tnw.html
{#Sec5}
Für alle Krankenhäuser, die weiterhin an der Versorgung hüftglenknaher Femurfrakturen teilnehmen wollen, fordert der G-BA in seiner Richtlinie SOP auf zubesonderen Situationen der Einwilligungsfähigkeitpräoperativer Planung und Priorisierung von EingriffenOperationsverfahrenzum Umgang mit gerinnungshemmender Medikationpatientenorientiertem Blut- managementorthogeriatrischer Zusammenarbeitphysiotherapeutischen MaÃnahmen
{#Sec6}
München
Stellvertretende
Geschäftsführerin AUC
GmbH
{#Sec7}
Bochum, AUC GmbH
Programmkoordinator
AltersTraumatologie
| {
"pile_set_name": "PubMed Central"
} |
Introduction
============
During the routine care of children, painful invasive procedures such as venipuncture for the withdrawal of blood for hematological testing are usually inevitable in healthy and sick subjects. The importance of pain assessment and pain management is widely acknowledged ([@ref-7]), and alleviation of pain caused by minor invasive procedures in children is an important issue for humane reasons and in terms of their reactions to future painful events and acceptance of subsequent health care interventions; ([@ref-43]) moreover, unrecognized pain can become severe and difficult to control and lead to fear and stress ([@ref-33]).
Pain assessment is an ongoing and integral part of total pain management particularly in children, and includes such approaches as distraction, evaluation, reassessment and medical intervention ([@ref-6]; [@ref-8]; [@ref-27]; [@ref-39]). Children and adolescents often describe invasive procedures and their associated anticipatory anxiety as the most distressing aspect of illness or hospitalization ([@ref-43]). Venipuncture is one of the most feared and acute painful experience in children ([@ref-28]).
The main difficulty in assessing pain in children is the potential discrepancy between the perception and experience of pain and its expression ([@ref-1]; [@ref-6]; [@ref-8]; [@ref-27]; [@ref-33]; [@ref-39]; [@ref-43]). Self-report faces scales are widely used to evaluate pain intensity in children despite concerns regarding interpretability ([@ref-35]; [@ref-40]). Most scales have five to seven faces, which are intended to elicit an indication of pain intensity ([@ref-2]; [@ref-18]; [@ref-21]; [@ref-22]; [@ref-46]). This gives more information than a simple binary "pain"/"no pain" response.
Faces scales are frequently used as self report measures of pain intensity in research and clinical practice, and the Royal College of Nursing has identified the WBFS as suitable for peri-procedural pain ([@ref-33]).
Various pain assessment tools have been tested in the search for objective, specific physiologic measures of responses to pain in infants ([@ref-34]). One of the most investigated is skin conductance fluctuations (SCF) per sec measured in the palm of the hand or on the plantar aspect of the foot ([@ref-41]). Skin conductance fluctuations in these sites reflect emotional sweating due to skin sympathetic nerve activity. They occur within 1--2 s of the onset of emotional stressors such as pain ([@ref-14]; [@ref-45]). Because SCF are induced by acetylcholine acting on muscarinic receptors, they are not affected by environmental temperature, hemodynamic changes or respiratory rhythm medications such as beta blockers and neuromuscular blockers ([@ref-3]; [@ref-26]; [@ref-45]). Furthermore, changes in respiratory rhythm (including apnea) do not influence SCF ([@ref-13]; [@ref-15]; [@ref-42]). Thus, changes in skin conductance are considered useful to monitor nociceptive stimulation and pain ([@ref-36]). The test has also been used to identify increased emotional stress as reflected in changes in the sympathetic nervous system as a measure of discomfort in artificially ventilated children ([@ref-11]) and in other conditions ([@ref-9]; [@ref-16]; [@ref-17]; [@ref-24]; [@ref-29]; [@ref-32]; [@ref-37]; [@ref-38]). However, also such sympathetic nerve activity as skin temperature autoregulation cause skin conductance peaks that are generally observed in pain states ([@ref-42]).
Many pain management studies have focused on postoperative and chronic pain ([@ref-12]; [@ref-25]). However, although the simple insertion of a needle is one of the most frightening and distressing medical procedures for hospitalized children ([@ref-1]; [@ref-43]) the recognition and assessment of acute pain resulting from this procedure in children remains inadequate ([@ref-7]). We have evaluated simultaneously the efficacy of the Wong--Baker Faces Pain Rating Scale (WBFS), which is routinely used in our hospital and SCF per sec in measuring pain in children at different ages (with and without previous exposure to venipuncture), before, during and after venipuncture performed at the same anatomic location (antecubital fossa) ([@ref-47]).
Methods
=======
Study design
------------
This was a prospective, observational study designed to explore relationships and differences between the WBFS to and the SCF during venipuncture in children 5--16 years old.
Patients
--------
This analytical, observational study was undertaken in the Day Hospital of the Pediatric Specialist and Laboratory Analysis Service of the Children Hospital "Regina Margherita" (Turin, Italy). It was conducted between March 2010 and March (inclusive) 2011 and in accordance with good clinical practice and the Declaration of Helsinki. The protocol was approved by the Ethics Committee of the Azienda Ospedaliera, OIRM S. Anna -- Ospedale Mauriziano (Turin, Italy).
We know that WBFS and SCF values are hardly to compare, but we decided to use a conversion table provided by the developer of SCF tool which made possible to convert the SCF values in a graded scale from zero to ten similar to WBFS grades.
A convenience sample of pediatric patients (5--16 years old) was enrolled: eligible patients were children or adolescents candidates for venipuncture for the collection of diagnostic blood specimens for routine hematological testing. All the patients were recruited at the "Regina Margherita" Children Hospital: some were recruited in the outpatient clinic and others were enrolled in Day Hospital, where they were followed for previous gastrointestinal diseases, endocrine disorders or diabetes mellitus type I. Children with known cognitive impairments, developmental delays, sensory deficits, pathological conditions of the palm and children receiving analgesic drugs were excluded from the study. Written consent was obtained from the parents, and verbal assent was obtained from each child or adolescent, adequately informed about the purpose of the study and the tools used. Parents were informed that their acceptance or refusal of the study would not affect clinical service. Basic demographic data (age, gender, previous diseases) were recorded. All the eligible children had to be subjected to venipuncture according to a previous clinical prescriptions in order to obtain diagnostic information.
All the venipunctures for the collection of blood were performed in the same setting with the same environmental temperature and by the same person. In our hospital nonpharmacologic distraction techniques are routinely perfomed: for venipuncture all the subjects received parental holding and positioning.
Pain evaluation tools
=====================
Wong--Baker faces scale
-----------------------
The WBFS combines pictures and numbers to enable the user to rate pain ([Fig. 1](#fig-1){ref-type="fig"}). It can be used for children over the age of 3, and for adults. The faces range from a smiling face to a sad, crying face. A numerical rating is assigned to each face (from 0, "no hurt" to 10, "hurts worst") of the WBFS ([@ref-7]; [@ref-10]; [@ref-18]; [@ref-21]; [@ref-34]; [@ref-35]; [@ref-40]).
{#fig-1}
The WBFS also has adequate psychometric properties (reliability, validity), and it is easy and quick to. The greatest strength of this scale may be its acceptability, given the consistent finding that the WBFS was preferred by children (any age), parents, and practitioners when compared with other faces pain scales ([@ref-40]).
Concerning validity, WBFS has an high correlation (*r* \> 0.7) with other self-reported pain scale used at the same time and shows differences (*p* \< 0.05) in score between two comparable but different groups. Reliability has been proved by the use of "test and retest" (*r* \> 0.5) and by the concordance with simultaneous observational score (*r* \> 0.4). WBFS has a significant (*p* \< 0.05) responsiveness to pain-increasing (painful procedures) and pain-decreasing (analgesia) events ([@ref-40]).
Skin conductance test. Instruments
----------------------------------
Skin conductance fluctuations were measured by alternating current at 88 Hz. Low-frequency electrical conductance reflects the ionic conduction in the stratum corneum, which is largely determined by sweat duct filling. Resistance and conductance are two related quantities in measuring voltage changes in skin conductance. Siemens is the unit of electrical conductance in the same way as Ohm is the unit of electrical resistance ([@ref-36]). Conductance was preferred to resistance because of the parallel nature of the electrical polarization and conductance in the skin. A frequency of 88 Hz is sufficient to reduce considerably the requirements for low electrode polarizability but also low enough to ensure minimal influence from layers other than the stratum corneum. In this study we applied a voltage of 50 mV and used a 3-electrode system. The 3-electrode system consisted of a measuring electrode (M), a countercurrent electrode (C), and a reference voltage electrode (R), which ensured a constant applied voltage across the stratum corneum beneath the M electrode. The placement of the electrodes on enrolled children was on the palm of the hand according to the [Fig. 2](#fig-2){ref-type="fig"}.
{#fig-2}
The distance between each electrode has been at least 7 mm as recommended by the manufactor. The electrode named M was placed at the hypothenar emminence because this area on the palm gives highest stability and thus less movement artifacts; the electrode named R was placed below the third finger and the electrode named C was placed on the hypertenar side of the hand.
We used the Med-Storm software version 1.0.0.33 (Med-Storm Innovation AS, Oslo, Norway). The Med-Storm manual contains an index where the SC fluctuations per sec are transformed to a graded score from zero to ten ([Fig. 3](#fig-3){ref-type="fig"}) \[available at [www.med-storm.com/](www.med-storm.com/)\] . This transformation is empirical and based, among other things, on pain referred by adults ([@ref-23]); similar results have been reported for children: sensivity in 5--7 years old children is 97.0% while specificity is 72.9%; sensivity in 8--16 years old subjects is 85.2% while specificity is 67.1% ([@ref-19]).
{#fig-3}
Pain evaluation procedure with the WBFS and the skin conductance test
---------------------------------------------------------------------
The WBFS was applied in the first minute after venipuncture; at that time, the children were asked to indicate on the picture (see [Fig. 1](#fig-1){ref-type="fig"}) the level of pain they felt throughout the entire procedure. In the case of the SCT, electrodes were distributed over the palm of the hand. We divided the process of the measurement into three steps: preparation of the child; insertion of the needle; extraction of the needle. Skin conduction fluctuations per second were recorded at the same anatomic location, namely the antecubital fossa (see [Fig. 2](#fig-2){ref-type="fig"}), and by the same operator (VM) using the BD Vacutainer 367286 blood collection set (21 G × 3/4^″^ × 12^″^ \[0.8 × 19 mm × 305 mm\]; Becton Dickinson & Company, Plymouth, UK). SCF was measured in each subject during venipuncture at least 10 min. They were recorded for 1 min before preparation of the child, during insertion of the needle, and for 1 min after extraction of the needle ([@ref-36]). Children and parents provided consent to the procedure, and nurses carried out the distraction techniques routinely used in our children's hospital.
Statistical analysis
====================
The sample size was calculated to find a relationship in the evaluation of pain using the two methods, with an estimated correlation coefficient (based on a pilot study) of 0.175, α = 0.05 and β = 0.55, 143 patients were needed per group.
We tested the normality of the two variables (WBFS and SCF) using the Kolmogorov-Smirnov test. Because of the rejection of normality assumption, we report data as median and interquartile range and we performed non parametric tests. We used the Wilcoxon signed rank test and the Spearman correlation coefficient to compare the scores obtained with WBFS and SCF. We also compared the scores of children with or without previous exposure to venipuncture and children at different ages using the Wilcoxon sum rank test. To determine the relevance of the studied variables in relation to pain, we performed a multivariate linear regression model. Statistical analyses were performed by using SAS 9.2 (SAS Institute Inc., Cary, NC, USA).
Results
=======
Of the 195 patients assessed for eligibility, 150 children (78 girls and 72 boys) aged from 5 to 16 years were included in the study: 23 children did not meet the inclusion criteria, and 22 declined to participate in the study). One hundred children were undergoing venipuncture for routine hematologic tests for the first time ("never exposed"), while 50, who were affected by a chronic disease, had previously undergone venipuncture (see [Table 1](#table-1){ref-type="table"}).
10.7717/peerj.37/table-1
###### Characteristics of the 150 children enrolled in the study.
Age Mean years SD Median year Interquartile range
-------------------------- ------------ ---------- --------------------------------- -------------------------------------
(25th--75th)
Mean age (year) 10.49 0.36 11.00 7--13
Gender Boys Girls
72 (48%) 78 (52%)
Exposure to venipuncture
Never exposed 100 (66.7%)
Previously exposed 50 (33.3 %) of which diagnosis:
21 (42%) Endocrine diseases
18 (36%) Gastrointestinal disorders
11 (22%) Diabetes mellitus type 1
[Table 2](#table-2){ref-type="table"} shows the number of SCF 1 min before venipuncture, and during insertion and removal of the needle. Mean duration of procedures was 5.5 min (DS ± 1.2). The number of SCF per sec was significantly higher during venipuncture (0.33 fluctuations/s) than before venipuncture and during removal of the needle (*p* \< 0.0001).
10.7717/peerj.37/table-2
###### Skin conductance values in relation to health status before venipuncture, during insertion of the needle and during removal of the needle.
Before venipuncture Insertion of needle Removal of needle *P*
--------------------- --------------------- --------------------- ------------------- ----------
All children (150) 0.27 (0.20--0.33) 0.33 (0.27--0.40) 0.20 (0.13--0.27) \<0.0001
Never exposed (100) 0.20 (0.13--0.27) 0.40 (0.33--0.53) 0.17 (0.07--0.27) \<0.0001
Exposed (50) 0.27 (0.20--0.33) 0.33 (0.27--0.40) 0.22 (0.17--0.27) \<0.0001
**Notes.**
Unit of measurement: Peak per second of Siemens
[Table 3](#table-3){ref-type="table"} shows the WBFS and SCF results. The self-report pain score was significantly lower than the SCF score (median WBFS score: 2; median SCF score: 5, *p* \< 0.0001). This difference was consistent among all subgroups except for the younger children previously exposed to venipuncture, in whom the results of WBFS were consistent with the results of SCF (median score of 6 versus 6, *p* 0.27). Skin conductance fluctuations per sec were similar across all subgroups. Differently, in children below the age of 8 years, pain evaluated with the WBFS was less intense in those never exposed to venipuncture versus those with previous exposure (median of 2 versus 6, *p* 0.0001). Correlation analyses confirmed these results: WBFS scores are lowly correlated with SCF scores in older children and in children without previous experience of venipuncture ([Table 4](#table-4){ref-type="table"}).
10.7717/peerj.37/table-3
###### Wong--Baker Faces Scale score and skin conductance per sec scores according to previous exposure to venipuncture and age.
Wong--Baker scale Estimated pain scores (0--10) based on SC fluctuation *p*-value
----------------------------------------- ------------------- ------------------------------------------------------- -----------
median (range) median (range)
All children (150) 2 (0--4) 5 (4--6) \<0.0001
Status
Never exposed to venipuncture (100) 2 (0--4) 5 (4--6) \<0.0001
Previously exposed to venipuncture (50) 2 (2--6) 5 (4--6) 0.04
*p-values* 0.004 0.86
Age
\<8 years old (52) 4 (2--5) 6 (4--6) 0.002
8 + years old (98) 2 (0-4) 5 (4--6) \<0.0001
*p-values* 0.0001 0.30
Sex
Male (72) 2 (1--4) 6 (4--6) \<0.0001
Female (78) 2 (0--4) 5 (4--6) \<0.0001
*p-values* 0.19 0.13
\<8 years old
Never exposed to venipuncture (37) 2 (2--4) 6 (4--6) \<0.0001
Previously exposed to venipuncture (15) 6 (4--10) 6 (4--8) 0.27
*p-values* 0.0001 0.89
8 + years old
Never exposed to venipuncture (63) 2 (0--4) 5 (4--6) \<0.0001
Previously exposed to venipuncture (35) 2 (0--4) 5 (4--6) 0.0005
*p-values* 0.14 0.83
**Notes.**
Method of comparison used: Wilcoxon test (*p* \< 0.0001).
10.7717/peerj.37/table-4
###### Spearman correlation coefficient (*p*-value) between WBFS score and SC per sec score.
Spearman coefficient (*p*-value)
---------------------------------------- ----------------------------------
All subjects 0.30 (0.0002)
No previous experience of venipuncture 0.28 (0.05)
Previous experience of venipuncture 0.35 (0.0004)
Children \<8 years old 0.35 (0.0004)
Children \>8 years old 0.20 (0.15)
The results of the linear regression model ([Table 5](#table-5){ref-type="table"}) show that age and previous exposure are important determinants of WBFS. In fact, the score was 1.81 points lower in older than in young children, and 0.86 points lower in children never exposed than in exposed children (both statistically significant, *p* \< 0.0001). The *r*^2^ of the model is 0.23, which shows that age and health status are involved in WBFS pain assessment by children. On the contrary, neither age nor health status affected the SCF score (the *r*^2^ approaches zero). Sex does not seem to affect either scale.
10.7717/peerj.37/table-5
###### Multivariate linear regression for the Wong--Baker Faces Scale scores and skin conductance fluctuations per sec scores.
β *p*-value *r*^2^ of the model
---------------------------------------------- ------- ----------- ---------------------
Dependent variable = Wong--Baker faces scale
Age −1.81 \<0.0001
Previous experience of venipuncture −0.86 \<0.0001
Sex 0.65 0.07 0.23
Dependent variable = Skin conductance
Age −0.42 0.26
Previous experience of venipuncture −0.09 0.65
Sex −0.39 0.27 0.00001
Discussion
==========
Painful invasive procedures such as venipuncture are usually inevitable during the routine care of children whether healthy or sick. Pediatric pain experiences are a consequence of an intricate interplay of genetic, experiential, and developmental factors ([@ref-44]). Therefore, pain assessment should not be an isolated element, but an ongoing and integral part of total pain management particularly in children ([@ref-39]). Face scales are frequently used as self-report measures of pain intensity in research and clinical practice, and the Royal College of Nursing has identified the WBFS as suitable for peri-procedural pain ([@ref-33]; [@ref-40]). Of the various objective pain assessment tools, one of the most widely investigated procedures is SCF per sec. In this study, we assessed the pain perceived by children during venipuncture using the two tools simultaneously: the WBFS (as subjective scale) and SCF per sec (as objective scale). The same stimulus, insertion of a needle, was used for children of different ages with and without previous exposure to venipuncture. Since the stimulus was the same for all children enrolled in the study, it was expected that the responses would be similar for the two methods and across all patient groups. To our knowledge, this is the first study to evaluate the WBFS and the SCF test simultaneously, and the first to use SCF to measure pain during venipuncture.
We found that SCF per sec were significantly higher during venipuncture than before the procedure. Therefore, the increase in the number of SCF per sec can be interpreted as a response to a painful procedure related to the insertion of the needle, whereas the SCF per sec recorded before beginning insertion of the needle are related to emotional stress and fear perceived by children ([@ref-31]). In addition, we found that the increase in SCF per sec was not influenced by age or previous experience of venipuncture. Similarly, in previous studies conducted in infants, health status, gestational age and postnatal age did not affect SCF per sec during painful procedures or when the infant was calm ([@ref-29]; [@ref-31]; [@ref-32]; [@ref-37]). Furthermore, using skin conductance monitoring, Røeggen et al. reported a very low variation between and within infants on the same discomfort level (behavioral state 1) ([@ref-32]).
On the contrary, we found that the WBFS was affected by both age and previous experience of venipuncture although the correlation between the two methods is only low. In fact, the WBFS results differed between children below the age of 8 years with experience of venipuncture and older, venipuncture-naive children. This seems to be in line with the report by von Baeyer et al. that previous painful events play a role in the anticipation and evaluation of future pain experience ([@ref-43]). In the older, never-exposed children, the lower values of WBFS with respect to SCF probably reflect an underestimation of pain related to a lack of sensitization, together with good practice of the team nurse and a comfortable environment during the procedure ([@ref-43]). However, we cannot exclude that the distraction techniques used during application of the electrodes in the SCF test could have affected the children's pain scores.
Pagé et al. used various pain scales to assess acute postoperative pain in children and observed that children aged 8--18 years find the faces scales the easiest to use ([@ref-30]). Moreover, in a comparative prospective evaluation of responsiveness to self-report scales, Connelly et al. reported that patients tended to rate pain intensity higher with numerical rating scale than with a visual analog scale ([@ref-5]).
A recent systematic review of cross-cultural comparison studies of child, parent, and health professional outcomes associated with pediatric medical procedures revealed that cultural factors may be associated with children's pain experiences when elicited by medical procedural pain, specifically children's pain behavior, but the authors noted that research using more sophisticated research methods is needed to develop culturally sensitive behavioral pain measures ([@ref-20]). In this context, Drendel et al. reported that a variety of factors could affect the self-report of pain in children, and that optimizing the use of pain assessment remains a challenge for the health care provider in the emergency setting of pediatric practice ([@ref-7]).
Our findings using skin conductance monitoring are in line with those reported by Hullet et al. ([@ref-19]) who monitored postoperative pain using a postoperative pain score in children. In fact, they found that the best accordance between a self-reported pain score and SCF per sec occurred in children between the ages of 4 and 8 years. Interestingly, two studies of postoperative pain in children showed that anxiety did not affect the skin conductance analysis ([@ref-4]; [@ref-19]) and in one of the studies there was no correlation between skin conductance activity and WBFS scores ([@ref-4]), unlike our findings. The lack of correlation between the two procedures may be due to the low specificity of skin conductance to identify moderate pain or it could indicate that WBFS pain scores are influenced by age and health status as shown in our study.
A potential limitation of this study is the lack of previous scientific literature about the comparison between SCF and WBFS and this issue makes more difficult to establish the study design. Another limitation is the difficulty to convert SCF data in a numerical scale that could be compared with WBFS scores. Additionally there is no universally accepted lower age limit for the self-reporting of pain in the use of WBFS ([@ref-18]). Further, fear and anxiety may bias pain reporting and interfere with attempts at measuring pain intensity; although the WBFS are reported to have content validity ([@ref-10]).
Lastly our sample size calculation is probably not adequate to examine the sub group analysis (previously exposed to venipuncture versus not).
In conclusion, this simultaneous evaluation of the WBFS and SCF in the assessment of pain in children undergoing a routine medical procedure showed that, although both tools can be useful for research and in clinical practice, SCF produced uniform data, whereas the WBFS was affected by age and previous venipuncture experience thereby providing new data about school-age children. If verified in other studies, our results should be taken into account when using these tools to evaluate pain in children and for providing more adequate pain care routine. Additional researches on positioning and of nonpharmacologic and pharmacologic interventions are needed, including larger samples sizes and expanded ages. AbbreviationsSCFSkin conductance fluctuationsWBFSWong--Baker faces pain rating scale
The study was performed at the Day Hospital of the Pediatric Specialist and Laboratory Analysis Service of Children Hospital "Regina Margherita" of Turin, Italy. We thank the parents and the children who took part in this study and Valentina Marmora, MD who performed SC analysis.
We are grateful to the nurses in the Day Hospital and Outpatients Clinic who made the study possible. We also thank Ing P. Duva and Ing C. Marconi for their time and technical support.
Lastly, we are indebted to Jean Ann Gilder (Scientific Communication srl) for editing the text.
Additional Information and Declarations
=======================================
Francesco Savino is an Academic Editor for PeerJ. Otherwise, the authors declare no Conflicts of Interests.
[Francesco Savino](#author-1){ref-type="contrib"} conceived and designed the experiments, wrote the paper.
[Liliana Vagliano](#author-2){ref-type="contrib"} performed the experiments, wrote the paper.
[Simone Ceratto](#author-3){ref-type="contrib"} contributed reagents/materials/analysis tools, wrote the paper.
[Fabio Viviani](#author-4){ref-type="contrib"} performed the experiments.
[Roberto Miniero](#author-5){ref-type="contrib"} contributed reagents/materials/analysis tools.
[Fulvio Ricceri](#author-6){ref-type="contrib"} analyzed the data.
The following information was supplied relating to ethical approvals (i.e. approving body and any reference numbers):
Ethics Committee of the Azienda Ospedaliera, OIRM S. Anna -- Ospedale Mauriziano (Turin, Italy).
| {
"pile_set_name": "PubMed Central"
} |
Background {#Sec1}
==========
The stiff person syndrome (SPS) is a rare disorder characterized by progressive rigidity and stiffness. SPS characteristically affects the truncal or axial muscles, and may lead to chronic pain, spasms, postural deformities and impaired motility. Variants of SPS, including stiff limb syndrome, jerking man syndrome and paraneoplastic associations have been described.
Up to 80% \[[@CR1]\] of SPS patients have elevated glutamic acid decarboxylase (GAD) antibodies, but its exact role in the pathogenesis remains unclear. GAD antibody levels generally do not correlate with disease severity \[[@CR2]\], nor has it been determined as the sole cause of clinical manifestations.
GAD is a presynaptic molecule which exists in 2 main isoforms: GAD67 and GAD65. The latter is of interest as it is localized to the nerve terminals and synthesizes gamma-aminobutyric acid (GABA) for neurotransmission \[[@CR3]\]. Most SPS patients with GAD antibodies also have antibodies which inhibit GABA-receptor-associated protein, leading to GABA functional impairment and hence the clinical features \[[@CR4]\].
The occurrence of fatigue has been reported in SPS \[[@CR5]\] but may be overshadowed by other muscular complaints. Neuromuscular fatigue is well correlated with neuromuscular transmission defect. This may be postsynaptic, as in myasthenia gravis, or presynaptic, as in the Lambert Eaton and Miller Fisher syndromes \[[@CR6], [@CR7]\].
We describe an SPS patient presenting with longstanding fatigue and electrophysiological evidence of presynaptic neuromuscular transmission defect.
Case presentations {#Sec2}
==================
Patient 1 {#Sec3}
---------
A 62-year old previously healthy Caucasian female experienced a 20 year history of pain in the neck, truncal and limb muscles. The symptoms were exacerbated by exercise and had a fluctuating course. In addition, she has muscle cramps, stiffness and spasms which she needed periods of rest to allow symptoms to alleviate.
Notably, she reported severe fatigue symptoms for the same period of time, made worse with physical exertion and mental stress. She had consulted multiple doctors and given diagnoses, including chronic fatigue syndrome (CFS) or myalic encephalomyelitis, and fibromyalgia. These symptoms had severely affected her lifestyle and can last up to several days in duration.
There were no complaints of confusion, memory loss, seizures, vertigo, loss of appetite or weight.
Clinical examination showed a well and alert middle-aged woman. No signs of muscle wasting, tremor, dystonia or fasciculations were observed. There was no obvious muscle weakness or fatigability. Examination of cranial nerves, cerebellar system, sensation and position sense were unremarkable.
Nerve conduction study is not suggestive of sensorimotor polyneuropathy. Needle electromyography performed showed involuntary firing of motor units with normal morphology at rest. During voluntary activation, there was normal recruitment of motor units seen.
Stimulated single fiber electromyography (SFEMG) of the orbicularis oculi was achieved using a disposable monopolar needle electrode (TECA, Old Woking, United Kingdom) placed 2.5 cm away from the edge of the orbicularis oculi. Stimulation pulses of 0.01 ms at 10 Hz and 5 to 12 mA were administered. A 40-mm 9013K0872needle electrode (Dantec, Skovlunde, Denmark) was inserted at the edge of the muscle for single-fiber recordings. Filter settings were maintained at 500 kHz to 10 kHz. Single-fiber responses were selected on the basis of short rise times (\<300 us), clear separation from other discharges and stability of waveform. Mean jitter was calculated from 20 accepted single-fiber responses. All SFEMG studies were performed on a Dantec Keypoint EMG machine \[[@CR8]\]. SFEMG showed a mean jitter of 27.7 μs (normal \< 23), of which 13 of 42 fibers recorded jitter values above 30 μs.
Repetitive nerve stimulation (RNS) studiesof the ulnar nerve was performed with right abductor digiti minimi recording \[[@CR6]\] at the following frequencies in random order: 3 Hz at rest, 3 Hz post-exercise, 20 Hz and 50 Hz. Exercise consisted of 20 s of maximal muscle contraction. Each study was performed at 3-min intervals. The 3 Hz RNS consisted of 10 stimuli trains, while 20 Hz and 50 Hz RNS consisted of 30 stimuli trains. For 3 Hz RNS, negative peak amplitude percentage decrements were compared between the first and fourth compound muscle action potentials (CMAP). For 20 Hz and 50 Hz RNS, percentage increments between the first CMAP and the CMAP with the largest amplitude were calculated. For this patient, RNS showed amplitude decrement of −1% (normal \< − 8%) at 3 Hz, increment of +71% (normal \< 48%) at 20 Hz, and +67% at 50 Hz (normal \< 52%).
Acetylcholine receptor and anti-MUSK antibody titres were within normal limits. Autoantibody screening showed increased antinuclear factor titre of 1/100 but anti- DsDNA, ANCA, Anti-Ro (SSA), anti-La (SSB) and anti-Jo-1 antibody titres were normal.
She was subsequently found to have elevated anti-GAD titre of 19 U/mL (normal \< 0.8).
She underwent CT scan of her chest, abdomen and pelvis but no malignancy was detected. MRI of the brain and spine performed previously were unremarkable.
In view of her religious beliefs, she declined intravenous immunoglobulin (IVIg), but was administered oral prednisolone 30 mg daily over a 6 week period. This was effective in reducing muscle cramps and spasms. Pain score had declined from 6 to 2 on the visual analogue scale. To address her longstanding fatigue, a trial of pyridostigmine at 60 mg three times a day over a 4 week period resulted in significant reduction of fatigue symptoms. Using the Fatigue Severity Scale \[[@CR9]\], initial scoring of 49 declined to 28 after treatment. On a 10 point visual analogue fatigue scale, the initial score of 8 declined to 3.5.
Patient 2 {#Sec4}
---------
A 63-year old man presented with back pain and stiffness radiating to the anterior trunk, neck and chest. Clinically, no involuntary movements, muscle wasting, fasciculations or tremors were noted. He was neurologically normal on examination. He was extensively investigated with brain imaging, gastroscopy, colonoscopy, CT scans of chest and abdomen which were all unremarkable. MRI of the brain was unremarkable, but there were mild degenerative changes in the lumbar spine. An autoimmune screen detected elevated anti-GAD titer of 50 U/mL. EMG showed characteristic continuous normal motor unit activity at rest in the paraspinal and shoulder girdle muscles, without myokymia or myotonic potentials.
RNS was performed using a similar method as in Patient 1. RNS amplitude decrement of −1% at 3 Hz, increment of +31.1% at 20 Hz, and +13.1% at 50 Hz, all within normal limits, were noted. SFEMG findings were within normal limits.
He responded to 2 courses of IVIg administered over 5 days but did not record benefit with oral corticosteroids. Pain score declined from 8 to 4 after each IVIg administration.
Patient 3 {#Sec5}
---------
A 56-year old man complained of left lower limb stiffness and cramps developing over a 2 month period, which gradually involved the right side. Upon presentation, tendon reflexes were normal but power was difficult to assess due to stiffness. Investigations to rule out malignancy were negative. Autoimmune testing showed elevated anti-GAD levels of 130 U/mL (normal up to 0.8), but other antibody titres were not elevated.
EMG gain showed continuous normal motor unit activity at rest in the paraspinal, proximal limb and shoulder girdle muscles.
CT scans of the brain, chest, abdomen and pelvis did not reveal space occupying lesions. Mild degenerative changes were present in the cervical and thoracic MRI, but brain MRI was unremarkable.
RNS was performed using a similar method as in Patient 1. RNS amplitude decrement of -- 4.6% at 3 Hz, increment of +9.2% at 20 Hz, and +6.2% at 50 Hz, all within normal limits, were noted.
The stiffness responded to a 5 day course of IVIg but he experienced a relapse, requiring a second course of IVIg and oral corticosteroids. Upon review 8 weeks later, he continued to improve and oral medication dosages were reduced. Pain score declined from 8 to 3 after each treatment.
All 3 patients had unremarkable cerebrospinal fluid examination, without elevation of cells or protein.
Figures [1](#Fig1){ref-type="fig"} and [2](#Fig2){ref-type="fig"} are RNS and EMG tracings of Patient 1.Fig. 1RNS tracings of Patient 1 showing maximal amplitude increment of +71% at 20 Hz. Sweep speed and vertical gain are as shown Fig. 2EMG of the left deltoid of Patient 1 depicting spontaneous normal motor unit activity. Sweep speed and vertical gain are as shown
Conclusions {#Sec6}
===========
Our series of 3 patients highlights the findings of presynaptic neuromuscular transmission defect in Patient 1, evidenced by symptomatic response to pyridostigmine and electrophysiological testing. In contrast, Patient 2 with classical SPS and Patient 3 with the stiff limb syndrome variant \[[@CR10]\] both do not exhibit fatigue symptoms or positive electrophysiological evidence of neuromuscular transmission disturbance, thus serving as negative controls.
Incremental responses to RNS are characteristic of neuromuscular transmission defect at the presynaptic region \[[@CR11]\]. We have previously reported similar findings in the Miller Fisher syndrome as a subclinical phenomenon \[[@CR7]\]. However, the current Patient 1 is symptomatic and responded to pharmacological intervention. Fatigue is seldom reported in SPS, but Teggi et al. described an SPS patient with fatigue and recurrent vertigo who improved with immune therapy \[[@CR5]\]. However, it is unclear if this patient had neuromuscular transmission disturbances.
GAD65 is important for GABA synthesis in normal synaptic transmission. It has been shown in a *Drosophila* model that GAD mutants can result in defective synaptic transmission at the neuromuscular junction, as GAD is specifically required in the presynaptic neuron to induce a an appropriate postsynaptic response \[[@CR12]\].
In SPS, it is unclear the direct pathogenic role GAD antibodies play in its symptomatology. It has been suggested that reduction of GABA-mediated inhibitory effects leads to a pathological state of neuronal hyperexcitability. Injection of IgG-GAD antibodies in the lumbar region induced continuous motor activity of anterior horn cells \[[@CR13]\].
In the *Drosophila* model \[[@CR12]\], GAD is required in the presynaptic neuron to induce a postsynaptic glutamate receptor field, and the levels of postsynaptic receptors are closely dependent on presynaptic GAD function. Our electrophysiological findings point to a presynaptic neuromuscular transmission defect, but in human SPS, the effect of GAD antibodies is still unknown in the neuromuscular junction.
To date, GAD antibodies have been found in increased titer in Miller Fisher syndrome, a condition with a presynaptic neuromuscular transmission defect \[[@CR14]\], and myasthenia gravis \[[@CR15]\], where the defect occurs at the postsynaptic region. Hence, the effect of GAD antibodies may not be limited only to a single location in the nervous system.
Conversely, other autoantibodies against amphiphysin and gephyrin \[[@CR16]\] have been reported in SPS. Their immunological effects on the central and peripheral nervous system remain to be seen in this condition.
To our knowledge, this is the first report of presynaptic neuromuscular transmission defect occurring in a patient with chronic SPS. It follows that the diagnoses of CFS, myalgic encephalomyelitis and fibromyalgia are reasonable differentials here which justify further research into the role of neuromuscular transmission derangements in their respective pathophysiology.
CFS
: Chronic fatigue syndrome
CMAP
: Compound muscle action potention
GABA
: Gamma-aminobutyric acid
GAD
: Glutamic acid decarboxylase
IvIg
: Intravenous immunoglobulin
RNS
: Repetitive nerve stimulation
SFEMG
: Single fibre electromyography
SPS
: Stiff person syndrome
Not applicable.
Funding {#FPar1}
=======
Not applicable.
Availability of data and materials {#FPar2}
==================================
All relevant data available are in the published paper.
Authors' contributions {#FPar3}
======================
YLL conceptualized, wrote manuscript and managed patients. YET performed electrophysiology and helped manage patients. Both authors read and approved the final manuscript.
Competing interests {#FPar4}
===================
The authors declare that they have no competing interests.
Consent for publication {#FPar5}
=======================
Prior informed consent for publication was obtained from the patients.
Ethical approval and consent to participate {#FPar6}
===========================================
Ethics committee approval not needed as reports are on findings from clinical care.
| {
"pile_set_name": "PubMed Central"
} |
Introduction
============
Colorectal cancer (CRC) is the third leading cause of cancer worldwide. Metastatic CRC (mCRC) is one of the most common causes of cancer-linked death, and the anti-epidermal growth factor receptor (EGFR) antibodies cetuximab and panitumumab can be used to treat Kirsten rat sarcoma (*KRAS*) wild-type (WT) mCRC [@B1]. However, cetuximab is ineffective in CRC patients harboring *KRAS* mutations [@B2]. Moreover, almost all patients who initially respond to cetuximab become refractory, bearing evidence that acquired resistance to cetuximab is an important clinical problem. CRC cells usually overexpress EGFR, but most patients are resistant to cetuximab as a monotherapy or in combination with chemotherapeutic agents (5-FU, oxaliplatin, and irinotecan) [@B3]-[@B6]. Although mutation of*KRAS* and other genes, *BRAF* and*NRAF*, are found in primary and secondary resistance to cetuximab monotherapy, the pathways of tolerance are unclear. Cetuximab interacts strongly and competitively with EGFR, blocking the binding of its natural ligand EGF and transforming growth factor alpha (TGFα), causing receptor internalization [@B7], [@B8]. Combination therapy of cetuximab with anticancer drugs has been clinically approved for colorectal, head and neck, lung cancers [@B9]-[@B13]. Some of the proposed treatment strategies to overcome the resistance induced by downstream pathway reactivation are being examined in clinical trials combining anti-EGFR drugs with other targeted therapies. Preclinical research has reported that combined targeted treatments that lead to vertical interference of the EGFR pathway are a reasonable approach [@B13], [@B14]. However, further identification of tolerance pathways to anti-EGFR monotherapy is essential to develop productive therapies for CRC. Ultimately, understanding the molecular basis of the clinical feedback to cetuximab could lead to the identification of a subpopulation of patients who might benefit from cetuximab and avoid needless costs and drug toxicity.
Ephrin-related receptor tyrosine kinases (RTKs) have been involved in intercellular communication and signaling during embryonic development. Ephrin type-B receptor 3 (EPHB3), one of EPH transmembrane tyrosine kinase receptors (TKRs), has a critical function in tumor progression or regression in various cancers [@B15]-[@B20]. Chiu *et al*. [@B21] demonstrated that overexpression of EPHB3 suppressed tumor growth by enhancing cell-cell contact in CRC. In contrast, Zhang *et al*. [@B15] identified ephrinB3 as a negative regulator of cell proliferation and a positive regulator of cell survival in CRC cell lines. Consequently, the functions of EPH receptors in CRC remain obscure and controversial, and further study on this subject is warranted.
EGFR and hedgehog (HH) signaling have been described as key factors involved in the survival and proliferation of tumor cells [@B22]-[@B25]. HH signaling may also play an important role in the development of cancer-initiating (stem) cells (CSC) and drug resistance [@B26], [@B27]. EGFR and HH signaling synergize upstream of the GLI family zinc finger 1 (GLI-1) via extracellular signal-regulated kinase (ERK) / mitogen-activated protein kinase (MAPK) signaling [@B28]. EGF promotes the expression of *GLI-1* and target genes *PTCH1* and *BCL2* in advanced gastric cancer [@B29], and the HH ligand sonic hedgehog (SHH) signals via phosphoinositide 3-kinase (PI3K) and MAPK to enhance expression of HH-specific targets in renal cancer [@B30], [@B31].
In the present study, we found a novel cetuximab resistance mechanism in CRC. Elevated expression of the EPHB3 receptor leads to the activation of the phosphorylation EGFR pathway and the STAT3 signaling cascade via HH signaling and confers resistance to cetuximab in CRC. The results gathered in this study will increase our understanding of the role of EPHs/HH in drug resistance in cancer biology and contribute to the development of a feasible therapeutic option for CRC treatment.
Methods
=======
Cell culture and generation of resistant cells
----------------------------------------------
Human colon carcinoma cell lines SW48, DLD-1, HT29, HCT116, and Colo205 were obtained from the American Type Culture Collection (ATCC) and maintained according to the ATCC\'s instructions. SW48R cells were kindly provided by the MOGAM Institute. Cetuximab-resistant cells (HT29, DLD-1, and HCT116) were obtained by increasing the cetuximab dosage stepwise from 1 μg/mL to 10 μg/mL over 5 months. Oxaliplatin-resistant cells (DLD-1 and Colo205) were obtained by increasing the oxaliplatin dosage stepwise from 0.05 μg/mL to 5 μg/mL over 1 year.
Reagents and antibodies
-----------------------
Erbitux (cetuximab) was purchased from Merck Serono (Burlington, Massachusetts, USA). The EPHB3 inhibitor (LDN-211904) was purchased from Merck Millipore. GANT61 was purchased from Selleckchem (Houston, TX, USA). The FGFR2 inhibitor (AZD4547) was purchased from Astrazeneca. The PDGFR Tyrosine Kinase inhibitor (imatinib) and VEGFR inhibitor (bevacizumab) were purchased from Calbiochem. Ephrin-B3 Fc chimera biotinylated protein (EFNB3 protein) was purchased from R&D Systems. Drug treatments were accomplished by aspirating the medium and replacing it with new medium containing the drugs. Anti-GLI-3 (1:1000) antibody was purchased from Bethyl. Anti-SOX2 (1:1000), anti-N-Cadherin (1:1000), and anti-E-Cadherin (1:1000) antibodies were purchased from BD Biosciences. Protein G PLUS-Agarose and anti-SHH (1:500), anti-Smoothened (1:1000), anti-EpCAM (1:1000), anti-Snail (1:1000), anti-EFNB3 (1:1000), and anti-HHIP (1:1000) antibodies were purchased from Santa Cruz Biotechnology. Anti-CD133 (1:1000) antibody was purchased from MACS. Anti-Vimentin (1:1000) antibody was purchased from Dako. Anti-Nanog (1:1000), and anti-EPHB3 (1:500) antibodies were purchased from Abcam. Anti-GLI-1 (1:500), anti-GLI-2 (1:1000), anti-patched (1:1000), anti-p-STAT3 (1:500), anti-STAT3 (1:1000), anti-cleaved PARP-1 (1:1000), anti-OCT4 (1:1000), anti-EGFR (1:1000), anti-p-ERK (1:1000), anti-ERK (1:1000), anti-p-mTOR (1:1000), anti-mTOR (1:1000), anti-p-AKT (1:1000), anti-AKT (1:1000), anti-p-JNK (1:1000), anti-JNK (1:1000), anti-p-EGFR (Y1045, 1:1000), anti-p-EGFR (Y992, 1:1000), anti-p-EGFR (Y1068, 1:1000), anti-VEGFR2 (1:500), and anti-HER2 (1:1000) antibodies were purchased from Cell Signaling. Anti-actin (1:10000) antibody was purchased from Sigma. For the secondary antibodies, anti-mouse-IgG-horseradish peroxidase (HRP, 1:200) and anti-rabbit-IgG-HRP (1:200) were purchased from Cell Signaling.
Patients and tissue specimens
-----------------------------
Tissues from four cetuximab-resistant patients with colon cancer were collected from Korea University Guro Hospital tissue bank between 2009 and 2016. Four tissue samples before cetuximab treatment and another four which had developed resistance after cetuximab treatment were derived from colon cancer patients. This protocol was reviewed and permitted by the Institutional Review Board of Guro Hospital (KUGH16275-001).
Apoptosis assay (flow cytometry)
--------------------------------
The translocation of phosphatidylserine, one of the markers of apoptosis, was detected by the binding of allophycocyanin (APC)-conjugated annexin V. Briefly, SW48 and SW48R cells, untreated or treated with the EPHB3 inhibitor siSTAT3, cetuximab, or a combination of the two agents, were resuspended for 24 h in the binding buffer provided in the Annexin V-fluorescein isothiocyanate (FITC) Apoptosis Detection Kit (BioBud). Cells were mixed with 1.25 μL of Annexin V-FITC reagent and incubated for 15 min at room temperature in the dark. The staining was terminated and cells were immediately analyzed by flow cytometry.
Co-immunoprecipitation
----------------------
A total of 100Ø plates were washed with ice-cold PBS and incubated on ice for 5 min with 500 μL of lysis buffer (Cell Signaling, Cat. No. 9803) (1 mM phenylmethylsulfonyl fluoride (PMSF), protease inhibitor, phosphatase inhibitor). The cells were scrape-harvested, cellular debris was removed by centrifugation for 5 min at 15,000 × *g* at 4 °C, and the protein concentration was determined using a BCA kit (Thermo Scientific). Cell supernatants were incubated with a primary antibody overnight at 4 °C, followed by the addition of 50 μL protein G agarose beads (50% slurry) for 1 h at 4 °C. Beads were washed 5 times with ice-cold lysis buffer, separated by centrifugation for 30 s at 15,000 × *g* at 4 °C, and then heated for 10 min with 2× sample buffer (Biosesang, Korea) to release the immunoprecipitated proteins for subsequent electrophoresis and western blotting analysis.
EGF enzyme-linked immunosorbent assay (ELISA)
---------------------------------------------
Cells (3 × 10^5^/well) were seeded in 6-well plates and incubated overnight before treatment. After EGF treatment, cell culture medium was removed and stored at -80 °C. Levels of EGF protein in the medium were determined by ELISA using a commercial kit (RayBiotech) according to manufacturer\'s instructions [@B32]. All experiments were performed in triplicate.
Phospho-RTK assay
-----------------
Levels of tyrosine phosphorylation of human receptor tyrosine kinases (RTKs) were detected by X-ray film using a commercial kit (R&D Systems) according to manufacturer\'s instructions and our previous report [@B33]. Each membrane contained kinase-specific antibodies spotted in duplicate.
Microarray and pathway analysis
-------------------------------
Total RNA was isolated from second-passage cells using Trizol (Invitrogen) and purified using RNeasy columns (Qiagen, Valencia, CA, USA) according to the manufacturers\' instructions. A microarray was used to evaluate alterations of mRNA expression between the cetuximab-resistant group and the control group. Three-samples from each of the two groups were subjected to microarray analysis on the Illumina Human HT-12 v4 Expression BeadChip (Illumina, Inc., San Diego, CA) platform. After exporting the array data, quantile normalization was performed using the Limma package [@B6] in R software version 3.3.2 (R Foundation for Statistical Computing, Vienna, Austria) [@B7]. To identify pathway changes, hallmark gene sets were applied to the microarray data from the Molecular Signatures Database using gene set enrichment analysis (GSEA) [@B8]. Gene ontology (GO) analysis was also conducted.
Animal experiments
------------------
All animal procedures were carried out in accordance with animal care guidelines approved by the Korea University Institutional Animal Care and Use Committee (IACUC). Four-week-old female BALB/c nude mice were obtained from Orient Bio (Orient Bio, Korea) and housed in a pathogen-free environment. The animals were acclimated for 1 week before the study and had free access to food and water. All surgeries were performed under Zoletil^®^ 50/xylazine anesthesia, and animal suffering was minimized. SW48 and SW48R cells (1×10^6^) in 100 μL of culture medium were mixed with 100 μL of Matrigel and implanted subcutaneously in 5-week-old BALB/c nude female mice. When the tumor volume reached approximately 100 mm^3^, mice were randomized into four drug treatment groups: control (n = 7), cetuximab (n = 7), EPHB3 inhibitor (n = 7), and cetuximab + EPHB3 inhibitor (n = 7). The mice were intraperitoneally (i.p.) administered 10 mg/kg cetuximab and/or 0.1 mg/kg EPHB3 inhibitor three times a week for 21 days. Tumor volume was calculated every 3 days for 34 days according to the following equation: tumor volume (mm^3^) = π/6 × length × (width)^2^. Maximum tumor area and its corresponding section were calculated using MetaMorph software (Molecular Devices). Sections of formalin-fixed, paraffin-embedded tumor specimens were deparaffinized in xylene, rehydrated in graded alcohol and then subjected to immunohistochemical staining.
TUNEL assay
-----------
DNA fragmentation was visualized by the Terminal deoxynucleotidyl transferase dUTP nick end labeling (TUNEL) assay, as described by the manufacturer\'s protocol (Roche Applied Science). Fluorescent images were obtained using a Carl Zeiss confocal microscope (Weimar, Germany).
Immunofluorescence staining
---------------------------
Cells grown on glass coverslips were fixed with 3.7% formaldehyde for 15 min, followed by permeabilization with 0.5% Triton X-100 for 15 min at room temperature. Cells were then blocked 1 h with 3% bovine serum albumin. Primary antibodies were applied overnight at 4 °C, followed by incubation with secondary Alexa fluor-594-conjugated antibodies (Molecular Probes, Eugene, OR, USA) or FITC-conjugated secondary antibody (Sigma). The nuclei were counterstained with 4, 6-diamidino-2-phenylindole (DAPI). Images were acquired using a Carl Zeiss confocal microscope (Weimar, Germany). Quantification of the colocalization was performed from the captured fluorescence images and represented by a defined value provided by the ZEN 2011 software.
Colony formation assay
----------------------
Cells were seeded into 60Ø plates, treated with the cetuximab and incubated at 37 ° C for 24 h. The cultured cells were seeded again into 6-well plates (500 cells/well) and then cultured at 37 °C. Culture medium was changed every three days. After two weeks, the cells were washed with PBS, fixed with 4% paraformaldehyde for 30 min, and then stained with crystal violet for 30 min for visualization and counting. Cells dyed with crystal violet were photographed using a mobile phone, and the number of colonies was counted to draw a graph.
Immunoblotting
--------------
Cells were lysed in Radioimmunoprecipitation assay buffer (RIPA buffer) (50 mM Tris, 150 mM NaCl, 1% Triton X-100, 0.1% SDS, and 1% Na-deoxycholate \[pH 7.4\]) with protease inhibitor and phosphatase inhibitor cocktails, and then subjected to SDS-PAGE. The proteins were transferred onto nitrocellulose membranes (GE Healthcare life science), blocked with Tris-buffered saline (TBS) containing 0.2% Tween 20 and 5% skim milk, incubated with primary antibody, then incubated with horseradish peroxidase-labeled secondary antibody. Signals were detected using X-ray films.
Small interfering RNA (siRNA) and short hairpin RNA (shRNA)
-----------------------------------------------------------
GLI-1 siRNA (siGLI-1), STAT3 siRNA (siSTAT3), EPHB3 shRNA (h) lentiviral particles, and negative control siRNAs were obtained from SantaCruz Biotechnology. Cells were transfected with siRNA oligonucleotides using Lipofectamine RNAi Max reagents (Invitrogen) according to the manufacturer\'s introductions (200 nM).
Quantitative real-time PCR
--------------------------
Total RNA was extracted by using TRIzol reagent (Life Technologies). Amplification of transcripts was performed using a reverse transcriptase polymerase chain reaction kit (Life Technologies). Real-time PCR was performed using the QuantStudio 6 Flex system. Taqman probes were as follows: STAT3 (Hs00374280_m1), GLI-1 (Hs01110766_m1), EPHB3 (Hs00177903_m1) and GAPDH (Hs99999905_m1). GAPDH was used for the normalization of gene expression.
Survival assay (WST-1)
----------------------
10,000 cells were grown in tissue culture-treated 96-well plates and treated as described in the Results. Cells were then treated with EZ-Cytox (WST-1 assay, DOGEN, Korea) for 3 h at 37 °C in an atmosphere of 5% CO2. Absorbance at 450 nm was determined using an enzyme-linked immunosorbent assay plate reader.
Immunohistochemical staining
----------------------------
The collected patients\' tissues were cut into 5-μm sections, dewaxed in xylene and rehydrated in graded alcohol. Antigen retrieval was performed using microwave oven irradiation for 20 min in 0.01 M citrate buffer (pH 6.0). Endogenous peroxidase was blocked using 3% hydrogen peroxide in PBS for 12 min. The specimens were incubated with a protein-blocking solution consisting of PBS with 5% normal donkey serum for 30 min at room temperature. The sections were incubated with primary antibodies at 4 °C overnight. Primary antibodies for the following proteins were used: GLI-1 (1:30), p-STAT3 (1:30), EFNB3 (1:30), and EPHB3 (1:50). The levels of GLI-1, p-STAT3, EFNB3, and EPHB3 in the tumor cells was judged by an independent pathologist (Baek-Hui Kim) according to methods previously described (1). Briefly, the percentage of immunoreactive cells was used to divide the samples into five grades (percentage scores), as follows: 0 (\<10%), 1 (10-25%), 2 (26-50%), 3 (51-75%), and 4 (\> 75%). The staining intensity was divided into four grades (intensity scores) as follows: 0 (no staining), 1 (light brown), 2 (brown), and 3 (dark brown). Positive staining was determined using the following formula: overall score = percentage score × intensity score. Expression was classified into two groups (positive and negative) with a cut-off value based on the median value of the respective overall score.
Statistical analysis
--------------------
The statistical significance of differences between two groups was analyzed with the unpaired Student\'s *t*-test using GraphPad Prism (version 7.0; GraphPad Software, Inc., La Jolla, CA). For multiple group comparisons and repeated measures, analysis of variance (ANOVA) and repeated-measures ANOVA (RM ANOVA), followed by post hoc least significant difference (LSD) test, were used. All *P* values were two-sided. *P* values less than 0.05 were considered statistically significant. Detailed information regarding the other methods is available in the [Supplementary Materials](#SM0){ref-type="supplementary-material"} and Methods.
Results
=======
Cetuximab-resistant cells are associated with hedgehog, stemness, and EMT pathways
----------------------------------------------------------------------------------
To study the difference between SW48 parent (SW48P) cells and SW48 cetuximab-resistant (SW48R) cells in colon cancer, we first confirmed the viability of both cells with or without cetuximab (Figure [1](#F1){ref-type="fig"}A). SW48R cells had relatively high cell viability and colony formation ability compared with SW48P cells (Figures [1](#F1){ref-type="fig"}B and [1](#F1){ref-type="fig"}C). mRNA expression in both groups was measured using a microarray (Figure [1](#F1){ref-type="fig"}D). Gene set enrichment analysis (GSEA) was performed to confirm pathway changes between the two groups. The GSEA of hallmark gene sets showed the highest increase for hedgehog signaling (normalized enrichment score (NES): 1.55; false discovery rate (FDR): 0.11) and the third highest increase for the mechanistic target of rapamycin (mTOR) signaling (NES 1.44, FDR 0.17) (Figure [1](#F1){ref-type="fig"}E, Supplementary Figures [S2](#SM0){ref-type="supplementary-material"}C and S2D). To evaluate the activity of ephrin at the mRNA level, gene ontology (GO) data sets were applied again to the microarray data with GSEA. Interestingly, all ephrin-related ontologies were found to be increased in the cetuximab-resistant group (Figure [1](#F1){ref-type="fig"}F). Especially, ephrin receptor binding increased beyond statistical significance (NES 1.40, FDR 0.08). We confirmed that cetuximab specifically induced EPHB3 activation in CRC cells and only activated the EPHB3 receptor. There was an increase in EPHB3 protein levels when cetuximab resistance was obtained (Figure [1](#F1){ref-type="fig"}G and Supplementary Figure [S2](#SM0){ref-type="supplementary-material"}E). The expression of the EPHB3 protein was not increased in oxaliplatin-resistant cells derived from other cell lines.
Cetuximab activates stemness, hedgehog signaling, and EMT in CRC cells
----------------------------------------------------------------------
SW48 cetuximab-resistant cells showed increased levels of hedgehog (GLI-1, SHH, and SMO), stemness (p-STAT3 and SOX2) and epithelial-mesenchymal transition (Vimentin) pathway proteins (Figure [2](#F2){ref-type="fig"}A-C). On the other hand, we confirmed that protein expression of Nanog, OCT4, EpCAM, GLI-2, GLI-3, PTCH, Snail, and N-Cadherin did not change. Importantly, E-Cadherin levels increase in SW48R. These results suggest that binding of EPHB3 and its interacting factors is increased when resistance to cetuximab is acquired (addressed in the Discussion). To examine which proteins in SW48 cetuximab-resistant cells were affected, we confirmed the levels of EGFR downstream signaling proteins with and without cetuximab in the presence of EGF (Figure [2](#F2){ref-type="fig"}D). As a result, despite cetuximab resistance, p-STAT3 expression was affected by EGF. In addition, we confirmed that the increase was maintained with and without cetuximab. STAT3, which is already known to be associated with cancer, was thus selected as a target. In order to examine whether cetuximab resistance affected STAT3, we determined the phosphorylation of STAT3 associated with cetuximab resistance. Phosphorylation of STAT3 and GLI-1 are known to be correlated [@B34]. We knocked down the expression of STAT3 and GLI-1 using siRNAs (Figure [2](#F2){ref-type="fig"}E and [2](#F2){ref-type="fig"}F), which confirmed that GLI-1 was regulated according to the level of p-STAT3. The efficiency of the transfection with siGLI-1 and siSTAT3 is shown in Supplementary Figure [S3](#SM0){ref-type="supplementary-material"}G.
Effect of EPHB3 on EFNB3 in models of acquired resistance to cetuximab
----------------------------------------------------------------------
To assess the potential relevance of cetuximab resistance on colon cancer RNA levels, we conducted a microarray analysis for comparing RNA levels on samples from SW48P cells and cells with acquired resistance to cetuximab (Figure [3](#F3){ref-type="fig"}A). The results are shown in Table [S1](#SM0){ref-type="supplementary-material"} as SW48 Resistance (SW48R)/SW48P fold-change of gene expression. EFNB3, which is known to be associated with cancer, was selected as a target [@B35]. We confirmed that the expression of EPHB3, the receptor of EFNB3, increased as the EFNB3 protein level increased. Based on the results shown in Figure [1](#F1){ref-type="fig"}D-G, we confirmed that the expression of the EphB3 receptor increased in the SW48 resistant cells as the EFNB3 protein level increased (Figure [3](#F3){ref-type="fig"}B). We confirmed the cell viability of both cells types in response to treatment with the EFNB3 protein (Figure [3](#F3){ref-type="fig"}C). The EFNB3- -treated group and SW48R cells (Figure [1](#F1){ref-type="fig"}A) showed similar viabilities. Therefore, EPHB3 expression is increased by EFNB3 when cetuximab resistance is acquired.
To explore the effect of EPHB3, we examined the protein levels of both cell types with or without cetuximab. We confirmed that key proteins (EPHB3, GLI-1, SOX2, Vimentin) were significantly decreased in SW48P cells compared with SW48R cells (Figure [3](#F3){ref-type="fig"}D), and cetuximab significantly reduced the levels of these proteins in the SW48P cells. Similar results were obtained for HT29 cells, with increased expression of EPHB3 upon resistance to cetuximab. (Figure [1](#F1){ref-type="fig"}G and Supplementary Figure [S2](#SM0){ref-type="supplementary-material"}G). This suggested that the EPHB3 protein specifically reacts to cetuximab. In agreement with these results, immunofluorescence studies demonstrated that EPHB3 levels were significantly increased in SW48R cells, as compared with SW48P cells (Figure [3](#F3){ref-type="fig"}E). Cetuximab is effective in CRC patients that have wild‑type *KRAS*. We compared wild-type *KRAS* with mutant-type *KRAS* cell lines to confirm EPHB3 protein expression. EPHB3 levels were increased when cetuximab resistance was acquired only in wild-type *KRAS* cell lines (Figure [3](#F3){ref-type="fig"}F). In addition, we investigated the protein levels of EPHB3 in cells resistant to antibodies for another receptor (not cetuximab). The increase in EPHB3 was only dependent on cetuximab resistance (Figure [3](#F3){ref-type="fig"}G). Before acquiring the inhibitor resistance of each receptor, EPHB3 protein expression in cells treated with various inhibitors was reduced by cetuximab in the SW48P cells (Supplementary Figure [S2](#SM0){ref-type="supplementary-material"}F).
STAT3 reduction through EPHB3 inhibition overcomes cetuximab resistance
-----------------------------------------------------------------------
To study the different effects of EPHB3 in SW48P cells and SW48 cetuximab-resistant cells, we assessed the levels of STAT3 in both cell types, with or without EPHB3 (Figure [4](#F4){ref-type="fig"}A and Supplementary Figure [S3](#SM0){ref-type="supplementary-material"}D). Results showed that the levels of p-STAT3, GLI-1, SOX2 and Vimentin in SW48 resistant cells were decreased by EPHB3. We showed that cetuximab and the inhibition of EPHB3 in SW48R cells affects p-STAT3 and GLI-1 and induces apoptosis by increasing c-PARP. In addition, cetuximab and the inhibition of p-STAT3 modulated GLI-1 and increased c-PARP levels. These results suggest that EPHB3 affects EGFR and modulates EGFR down-signaling of p-STAT3 and GLI-1 (Figure [4](#F4){ref-type="fig"}B and [4](#F4){ref-type="fig"}C). Next, using GANT61, we confirmed that GLI-1 is the key gene in our hypothesis. The combination of cetuximab and GLI-1 inhibition in SW48R cells was shown to increase c-PARP (Supplementary Figure [S2](#SM0){ref-type="supplementary-material"}H). We also confirmed by flow cytometry that EPHB3 inhibition with siSTAT3 in combination with cetuximab increased apoptosis in SW48R cells (Figures [4](#F4){ref-type="fig"}D and [4](#F4){ref-type="fig"}E). This indicated that cetuximab resistance had been overcome. The same results were obtained using shEPHB3 (Supplementary figures [S3](#SM0){ref-type="supplementary-material"}E and S3F). The efficiency of the transfection with shEPHB3 is shown in Supplementary Figure [S3](#SM0){ref-type="supplementary-material"}G. We assessed cell viability in response to a combination of cetuximab with the EPHB3 inhibitor and a siRNA targeting STAT3 (siSTAT3) (Figure [4](#F4){ref-type="fig"}F and [4](#F4){ref-type="fig"}G). EPHB3 inhibition using siSTAT3 in SW48R cells decreased cell viability in combination with cetuximab.
Cetuximab resistance is caused by increased binding of EGFR and EPHB3
---------------------------------------------------------------------
We observed an increase in EPHB3 levels in SW48R, but not in p-EPHB3. These results suggest that the increase in p-STAT3 expression is due to the activation of EGFR rather than activation of EPHB3 (Supplementary figures [S3](#SM0){ref-type="supplementary-material"}A and S3B). To study how EPHB3 affects EGFR and STAT3 in CRC, we first confirmed that there is an interaction between EGFR and EPHB3. We confirmed the possibility of indirect binding of EGFR and EPHB3 through structural modeling and found that though they are both large in size, they can structurally bind (Figure [5](#F5){ref-type="fig"}A). Co-immunoprecipitation and immunofluorescence confirmed the increase in interaction and co-localization caused by the binding of EGFR and EPHB3 in SW48R cells (Figure [5](#F5){ref-type="fig"}B-D), whereas EFNB3 and HHIP did not bind to EGFR (Supplementary Figure [S3](#SM0){ref-type="supplementary-material"}H and S3I). Next, we examined the effect of EGF using ELISA. The concentration of EGF in the supernatant was higher in SW48R cells than in SW48P cells, indicating that the binding of EGF to its receptor was low in SW48R cells (Figure [5](#F5){ref-type="fig"}E). Thus, we assumed that EGFR was affected by its binding to EPHB3. Finally, western blotting confirmed activation only of the p-EGFR (Y1068) site, known as a STAT3-activating site, when resistance to cetuximab was obtained [@B36], [@B37]. Also, p-EGFR (Y1068) reacted with EGF and its activity was reduced by the EPHB3 inhibitor (Figure [5](#F5){ref-type="fig"}F-G and Supplementary Figure [S3](#SM0){ref-type="supplementary-material"}C). In conclusion, the binding of EGFR to EPHB3 in SW48R is increased, resulting in the activation of the p-EGFR (Y1068) site, which in turn, results in increased phosphorylation of STAT3.
EPHB3 inhibits and overcomes cetuximab resistance *in vivo*
-----------------------------------------------------------
Finally, we investigated whether EPHB3 could inhibit colorectal tumorigenicity and resistance *in vivo*. SW48 and SW48R cells were subcutaneously injected into nude mice. After 3 weeks of treatment, well-formed tumors in the SW48R cetuximab and EPHB3 inhibitor combination group were much smaller than those in the control group (Figure [6](#F6){ref-type="fig"}A-C). A terminal deoxynucleotidyl transferase (TdT) dUTP nick-end labeling (TUNEL) assay was performed to assess the apoptosis of SW48 and SW48R. The SW48R cetuximab and EPHB3 inhibitor combination group induced marked apoptosis, whereas only weakly positive staining could be observed in the control group (Figure [6](#F6){ref-type="fig"}D and [6](#F6){ref-type="fig"}E). Immunofluorescence staining was then performed to measure the expression of EGFR and EPHB3 in tumor tissues. Consistent with previous observations, we confirmed co-localization because of the increased binding of both factors in SW48R cells (Figure [6](#F6){ref-type="fig"}F). Taken together, these results suggest that EPHB3 could regulate cetuximab resistance in CRC *in vivo*.
We then examined the expression of proteins (EPHB3, EFNB3, GLI-1, and p-STAT3) in both pre-cetuximab-treatment and cetuximab-resistant patient-derived samples, based on previous western blotting results. First, immunohistochemical analysis was performed to confirm EPHB3, EFNB3, GLI-1, and p-STAT3 expression. As shown in supplementary Figure [S1](#SM0){ref-type="supplementary-material"}A-D, EPHB3 was significantly highly expressed in cetuximab-resistant patient samples compared with pre-cetuximab treatment patient samples. Expression of EFNB3, GLI-1, and p-STAT3 was also higher in cetuximab-resistant samples compared with pre-cetuximab treatment samples, but no significance was noted. Collectively, these data show that EPHB3 affects cell sensitivity to cetuximab.
Discussion
==========
Cetuximab plays a role by inhibiting EGF ligand binding to the extracellular domain of EGFR, thus blocking ligand-mediated EGFR signaling. Moreover, cetuximab elevates receptor internalization and degradation, and enhances antibody-dependent cell-mediated cytotoxicity. Cetuximab was the first FDA approved anti-EGFR antibody for CRC therapy, guided for combination with standard chemotherapy or radiotherapy in locally advanced, metastatic, and recurrent CRC [@B38], [@B39]. Despite promising advances in regional limit cure of CRC, further innovations of treatment are demanded. Despite the clinical success of cetuximab and panitumumab, the efficacy of these drugs is challenged by the development of resistance, which is achieved by activation of canonical and non-canonical signal transduction pathways [@B40]. There are many mechanisms of resistance to CRC treatment. Furthermore, CRC treatment via canonical and non-canonical Wnt signaling, crosstalk of pathways (e.g., NOTCH, mTOR, AKT/PI3K, NFκB) has also been reported to induce β-catenin activation. Therefore, in the present study, we focused on a microarray analysis and discovered previously unknown targets involved in overcoming intrinsic resistance to cetuximab.
The goals of this study were to identify whether resistance to cetuximab in CRC cells involved EPHB3 signaling, to disclose the molecular events underlying such changes, and to examine its role in tumor suppression. Finally, we used siRNAs and chemical inhibitors to validate the function of activated kinases in the proliferation of CRC cells and suggested several kinases that can overcome cetuximab resistance. Our results led us to suggest novel therapeutic targets against cetuximab resistance in regard to the kinome network derived from the results of phosphoproteomic analyses. We also found that alterations in cell behavior, morphology (Supplementary Figure [S2](#SM0){ref-type="supplementary-material"}A and S2B), and signaling molecules, resulting from EPHB3 signaling, contributed to both EMT and tumor suppression. Most importantly, we found that the EPHB3-ephrin-B interaction stimulated the junctional adhesion of E-Cadherin molecules, which are representative of tight junctions (Figure [2](#F2){ref-type="fig"}C).
EPH receptors constitute the largest subfamily of transmembrane TKRs, including 14 members of EPHAs and EPHBs [@B41]. Previous reports showed that EPHB3 is related to apoptotic signaling and acts as a tumor suppressor gene [@B17]. Zhang *et al*. found reduced expression of EPHB3 in human CRC tissues is related to poor survival and advanced recurrence [@B15], suggesting that EPHB3 might act as a tumor suppressor in CRC. However, we and another group found that EPHB3 has oncogenic functions in CRC patients treated with cetuximab [@B42]. In addition, recent studies have begun to report the interesting functions of EPH receptors in tumorigenesis. Overexpression of EPHs has been reported in numerous cancers, including breast, prostate, melanoma, and glioma [@B42]-[@B44]. For instance, EPHA2 is upregulated in breast cancer and activates tumor growth and invasion via increased Ras/MAPK signaling [@B45], [@B46]. However, due to the complicated functions of the EPH/ephrin system, recent reports show that the EPH/ephrin system may contribute as a tumor suppressor in certain cancer types and contexts [@B47]. A role in tumor suppression by the EPH/ephrin system is seen in CRC, in which EPHB signaling inhibits the transition of malignancy during CRC development by compartmentalizing the spread of cancer cells [@B48]. EPHB3 also interacts with integrin β1 to increase both tumor invasion and progress, raising the possibility that targeting EPHB3 may modulate integrin β1 signaling in CRC. Moreover, integrin β1 can trigger ligand-independent SRC-AKT signaling, inducing resistance to erlotinib in lung cancer [@B49]. However, the association of EPHB3 with integrin β1-SRC-AKT signaling, including its contribution to EGFR therapy resistance, remains uncertain in CRCs.
When activated by ephrin, the EPH receptor can transduce intracellular pathways that are associated with a variety of biological functions [@B50]. For instance, EPHB2 can promote cell proliferation via Abl-mediated increase in Cyclin-D1 expression and Src-mediated increase in STAT3 phosphorylation. In the present work, we found that EPHB3 can inhibit cancer cell growth in preclinical cancer models, including xenografts, in the absence of an EPHB3 inhibitor. Combination of an EPHB3 inhibitor and an EGFR inhibitor (cetuximab) led to increased anti-tumor effects in CRC cells. Furthermore, we confirmed the anti-tumor effects of EPHB3 inhibition using an EPHB3 inhibitor in mouse models as well as by resistance to cetuximab in tumor models that were not characterized by activating EGFR mutations (Figure [6](#F6){ref-type="fig"}A). Moreover, expression of EPHB3 in human CRC tissues that presented recurrence following cetuximab therapy demonstrated elevated EPHB3 levels compared with those from pretreatment CRC tissues. These observations suggest that inhibiting EPHB3 might improve the anti-tumor effects of EGFR inhibitors.
These findings reveal an important role for EPHB3 in cetuximab resistance of CRC cells. Our results showed that EPHB3 was a direct target of STAT3, and cell progression, including cell growth, invasion, cell cycle, and apoptosis, were regulated by increasing or decreasing EPHB3 expression. To explain whether the EPHB3 pathway was the main target in CRC *in vivo*, EPHB3 was suppressed by the EPHB3 inhibitor, which blocked tumor formation.
We also found increased expression of GLI-1 after EGFR suppression in EGFR-dependent SW48 cells. GLI-1 is a molecular factor for metastasis and cancer cell growth in various cancers [@B51]. In the present study, we discovered that "cross-talk" between EGFR and EPHB3 occurs through the HH/STAT3 cascade, but that HH inhibition might make cells more EGFR-dependent. However, the roles and underlying pathways of HH in drug-resistant cancer cells remain to be researched. EGFR and HH signaling synergize through the MAPK signaling in various tumor types [@B28], [@B52]. EGF stimulates the expression of *GLI-1* and its target genes*PTCH1*and *BCL2* in gastric cancer [@B53], and activates the HH ligand sonic hedgehog (SHH) pathway through PI3K and MAPK to increase the expression of HH specific targets [@B54]. Taken together, these results represent a powerful rationale for the development of combination therapy in xenograft models.
STAT3 is activated in various tumors [@B55]. STAT3 can be phosphorylated by the stimulation of upstream receptors and/or nonreceptor kinases including EGFR [@B56], IL-6/GP130 and Janus kinases (JAKs) [@B57], and SRC family kinases [@B58]. Continuous activation of STAT3 has been implicated in the development of resistance to conventional therapies in certain cancers. Additionally, overexpression of STAT3 has been observed in anticancer drug-resistant tumor cells [@B59]. For example, the activation of STAT3 has been reported in the tolerance to EGFR inhibitors in head and neck squamous cell carcinoma (HNSCC) and glioma [@B60], [@B61] and tolerance to EGFR inhibitors in the therapy of HNSCC patients was associated with increased expression of STAT3 [@B62]. These data suggest that activation of STAT3 induces tolerance to EGFR inhibitors in CRC, and a combination with STAT3 targeting may provide an effective therapeutic strategy.
Signaling of MEK/ERK and PI3K/AKT plays an important role in cetuximab resistance in CRC [@B63]. Our previous data and the present work found activation of ERK in HH-overexpressing gastric cancer cells [@B64]; however, this signaling was unchanged by manipulation of HH signaling in *KRAS*-WT CRC cells. Surprisingly, PI3K/AKT signaling can be activated in both*KRAS*-WT and *KRAS*-mutated CRC cells. The ERK signaling activated by SHH was responsible for PI3K/AKT activation in *KRAS*-mutated CRC cells. Our data demonstrated that targeting STAT3 using a STAT3 inhibitor in cetuximab-resistant cells sensitizes the cells to the EGFR inhibitor therapy.
We also performed an animal study using a combination of an EPHB3 inhibitor (LDN-211904) and cetuximab that could overcome cetuximab resistance and inhibit tumor growth (Supplementary Figure [S4](#SM0){ref-type="supplementary-material"}). Targeting the EPHB3 pathway has been shown to overcome cetuximab resistance in *KRAS*-WT CRC and our findings in cell and animal models showed that the EPHB3 pathway may be responsible for cetuximab resistance in CRC cells.
Conclusion
==========
In summary, our results indicate that HH may play a role as an oncogene in cetuximab-resistant CRC cells by targeting EPHB3. We have provided a scheme of the working model of cetuximab resistance-induced activation of EPHB3 and hedgehog signaling via a graphic abstract. Collectively, our results using cells, mice, and patients strongly suggest the hypothesis that STAT3 could serve as a reliable biomarker for drug resistance, tumor recurrence, and survival prediction in CRC. Furthermore, the combination of an EPHB3 inhibitor with cetuximab could be effective in inhibiting STAT3-activated CSC stemness and cetuximab resistance in CRC.
Supplementary Material {#SM0}
======================
######
Supplementary figures and tables.
######
Click here for additional data file.
SW48 cetuximab-resistant cells were kindly provided by the MOGAM Institute. We thank MOGAM Institute for its support of the cell line. This research is based on data used in Seong Hye Park\'s doctoral dissertation (Korea University).
Funding
=======
This work was supported by grant from the National Research Foundation (NRF) of Korea and funded by the Korean government (MSIP) (NRF-2017R1A2B2011684, NRF-2018M3A9G1075561).
Availability of data and materials
==================================
The data sets supporting the conclusions of this article are included within this article and the Supplementary Data.
Ethics approval and consent to participate
==========================================
All experiments were approved by the Ethics Committee of Korea University.
Contributions
=============
The study funders had no role in the design of the study; the collection, analysis, or interpretation of the data; the writing of the manuscript; or the decision to submit the manuscript for publication.
**SHP** conceived and designed the study, collected and assembled the data, analyzed and interpreted the data, and wrote the manuscript. **MJJ** and **YAJ** provided study materials. **YJN, HKY, DYK,** and **BGK** conceived and designed the study and analyzed and interpreted the data. **SHK** and **BRK** conceived and designed the study. **JLK** and **SYJ** collected and assembled the data, and analyzed and interpreted the data. **SCO** and **DHL** conceived and designed the study, provided financial support, collected and assembled the data, analyzed and interpreted the data, wrote the manuscript, and provided final approval of the manuscript. All authors discussed the results and commented on the manuscript.
CRC
: colorectal cancer
SMO
: smoothened
GLI-1
: Glioma-Associated Oncogene Homolog 1
PARP
: poly (ADP-ribose) polymerase
PBS
: phosphate-buffered saline solution
PUMA
: p53 upregulated modulator of apoptosis
MDM
: Mouse double minute 2 homolog
PI
: propidium iodide
SDS-PAGE
: sodium dodecyl sulfate polyacrylamide gel electrophoresis
PMSF
: phenylmethylsulfonyl fluoride
TBS
: Tris-buffered saline
TUNEL
: terminal deoxynucleotidyl transferase (TdT) dUTP nick-end labeling
DAPI
: 4,6-diamidino-2-phenylindole
ELISA
: enzyme-linked immunosorbent assay
FBS
: fetal bovine serum
EMT
: epithelial-mesenchymal transition
siRNA
: short interfering RNA
HRP
: horseradish peroxidase
RIPA
: radioimmunoprecipitation assay
TBS
: Tris-buffered saline
APC
: allophycocyanin
FITC
: fluorescein isothiocyanate
PMSF
: phenylmethylsulfonyl fluoride
GSEA
: gene set enrichment analysis
GO
: gene ontology
NES
: normalized enrichment score
FDR
: false discovery rate
HNSCC
: head and neck squamous cell carcinoma.
{#F1}
{#F2}
{#F3}
{#F4}
{#F5}
{#F6}
[^1]: Competing Interests: The authors have declared that no competing interest exists.
| {
"pile_set_name": "PubMed Central"
} |
Introduction {#Sec1}
============
Chromium occurs in the environment mostly in two species, i.e. as a trivalent (III) and hexavalent (VI) form. Chromium has been proposed as an essential element; however, recent studies suggest that it should be removed from the list of essential trace elements. It was shown that Cr(III) plays a key role in chromium allergy and causes DNA damage in cell-culture systems (Vincent [@CR43]). Cr(VI) (chromate) is known to be a genotoxic carcinogen due to the redox reactions that take place in cells which generate Cr(V)-1,2-diolato species (Chellan and Sadler [@CR7]).
Chromium can enter the environment both from natural and anthropogenic sources, e.g. via electroplating, leather tanning and the textile industries (Avudainayagam et al. [@CR2]; Johnson et al. [@CR22]; Unceta et al. [@CR42]; Dhal et al. [@CR10]). Anthropogenic emission of chromium to the atmosphere is substantial and was estimated at 336 t in the European Union in 2013 (EEA Technical report [@CR12]), 2700--2900 t in the USA and 21,000 t in China in 2009 (Cheng et al. [@CR8]). Chromium released into the atmosphere is carried in the air as particles or dust. It may be transported over long distances by the wind, but finally settles on the soil. Rain will also remove chromium particles from the atmosphere and deposit them in the ground, thus contaminating the soil. The disposal of chromium-containing commercial products (e.g. some inks, paints and paper, rubber and composite floor coverings or toner powders used in copying machines) and coal fly ash from electric utilities and other industries are major sources of chromium releases into the soil (Barceloux [@CR3]; Metze et al. [@CR31]; Dhal et al. [@CR10]). Solid waste and slag produced during chromate manufacturing processes as well as agricultural and food wastes, when disposed of improperly in landfills, can be another potential source of chromium exposure (Kimbrough et al. [@CR23]; Barceloux [@CR3]).
The Canadian Soil Quality Guidelines for the protection of human health recommend a maximum content for hexavalent chromium and total chromium in agricultural and residential land as 0.4 and 64 mg kg^−1^, respectively (Canadian Environmental Quality Guidelines [@CR6]). The Swedish Guidelines (Guidelines for Polluted Soils [@CR40]) suggest maximum concentrations for the most sensitive type of land use at 5 and 120 mg kg^−1^ for Cr(VI) and Cr(III), respectively. In Italy, the highest permissible Cr(VI) concentrations in soil are 2 and 15 mg kg^−1^, respectively, depending on the type of exploitation, i.e. parkland or industrial (Decreto Ministeriale n.471 [@CR9]; Pettine and Capri [@CR37]). In Poland, the limit for total chromium in agricultural and residential land was set at 150 mg kg^−1^ (Ordinance of the Minister of Environment of Poland [@CR33]). The content of Cr~Tot~ and Cr(VI) in contaminated soil can reach a level of several grammes per·kilogramme (Dhal et al. [@CR10]).
The behaviour of metals in soil and uptake by plants is controlled by element speciation and by soil properties, such as pH, particle size, cation-exchange capacity, content of organic matter, content and type of clay minerals and Al, Fe and Mn oxides, redox potential and microbiological activity (Kotaś and Stasicka [@CR25]; Dhal et al. [@CR10]; Krasnodębska-Ostręga et al. [@CR26]; Paldyna et al. [@CR34]). Chromium(III) in soil is mostly present as insoluble chromium(III) hydroxide and tends to be adsorbed on the soil surface in a pH range of 4--8. The solubility of Cr(III) in soil and its mobility may increase due to the formation of soluble complexes with organic matter in soil, e.g. citric acid, diethylenetriaminepentaacetic acid (DTPA) and fulvic acid. Hence, a lower soil pH potentially facilitates complexation (Avudainayagam et al. [@CR2]; Kotaś and Stasicka [@CR25]). The most mobile forms of Cr(VI) in soil are CrO~4~ ^2−^ and HCrO~4~ ^−^ ions, but insoluble species such as BaCrO~4~ and PbCrO~4~ may also be present (Kotaś and Stasicka [@CR25]). Oxidation and reduction of chromium species in soil can take place simultaneously. Cr(VI) may react with many inorganic reductants such as Fe(II) and sulphide as well as with a number of organic compounds, including carboxylic and hydroxycarboxylic acids, aldehydes, phenols or fulvic acid (Eckert et al. [@CR11]; Brose and James [@CR5]). Moreover, several microorganisms possess the ability to reduce Cr(VI) (Brose and James [@CR5]; Dhal et al. [@CR10]). In soil containing manganese oxides, Cr(III) can be oxidised to Cr(VI), especially at high pH values (Dhal et al. [@CR10]).
Various extraction procedures have been proposed in order to determine chromium(VI) species in solid environmental samples. The extraction conditions have to be carefully chosen as the leaching process may lead to interconversion of Cr species. The extraction solutions can be divided into the following groups: (a) acids at various concentrations, e.g. 0.43 mol L^−1^ acetic acid (CH~3~COOH) (Lillengen and Wibetoe [@CR28]), (b) buffered salts, e.g. ammonium acetate (NH~4~OAc) (Morales-Muñoz et al. [@CR32]) and K~2~HPO~4~ (Rüdel and Terytze [@CR39]; James et al. [@CR21]), (c) neutral salts such as CaCl~2~ (Béni et al. [@CR4]), Na~3~PO~4~ (Mandiwana [@CR30]) or Na~2~CO~3~ (Panichev et al. [@CR35]; Elci et al. [@CR13]), (d) chelating agents, e.g. ethylenediaminetetraacetic acid (EDTA) (Lillengen and Wibetoe [@CR28]; Korolczuk and Grabarczyk [@CR24]), DTPA (Grabarczyk et al. [@CR17]) and S,S-ethylenediamine-N,N′-trisodium salt (EDDS) (Grabarczyk [@CR16]) and (e) other extractants (Gitet et al. [@CR15]) proposed for routine soil testing. Some of these procedures have been designed in order to distinguish between soluble, exchangeable and slightly soluble forms of chromium(VI).
Alkaline media have been suggested for selective extraction of slightly soluble Cr(VI), as in such an environment, Cr(VI) is stable in the solution while Cr(III) species form insoluble hydroxides or carbonates. Insoluble forms of Cr(VI) are often extracted with suitable chelating agents by forming soluble complexes of cations. At the same time, Cr(VI) is transferred into the solution as a soluble salt. The most recommended method for determining Cr(VI) in solid matrices is the US EPA method 3060A, which was designed and validated by Vitale et al. ([@CR45], [@CR44]). In this method, a hot (90--95 °C) 0.28 mol L^−1^ Na~2~CO~3~ solution in 0.5 mol L^−1^ NaOH is used to extract the "total" amount of Cr(VI) from the soil and sediments. However, the presence of the reducing compounds results in an underestimated concentration of Cr(VI) in the extraction solution (Malherbe et al. [@CR29]). Conversely, some authors have observed partial oxidation of soluble Cr(III) and a positive error in Cr(VI) determination (Huo et al. [@CR19]; Huo and Kingston [@CR18]). A brief description of these procedures, their effectiveness and application for solid samples is presented in Table [1](#Tab1){ref-type="table"}.Table 1Extraction procedures of Cr(VI) from solid samplesExtracted formsExtraction solutionExtraction conditionsRecovery of Cr(VI), %Analysed matricesReference*m*:*V*TimeWater-solubleDeionised water1:10024 h shaking0.07^a^SoilElci et al. [@CR13]Naturally extractedDeionised water, CO~2~1:10024 h shaking0.06^a^SoilPanichev et al. [@CR35]Soluble0.43 mol L^−1^ CH~3~COOH1:5016 h shaking0.66CRM 483---sewage sludge-amended soilLillengen and Wibetoe [@CR28]Soluble0.04 mol L^−1^ (NH~4~)~2~SO~4~, 0.5 mol L^−1^ NH~4~OH (pH \~8)1:1014 min, 300 W (microwave extraction)93\
112.6River sediment + 50 μg g^−1^ Cr(VI) as K~2~CrO~4~\
River sediment + 30 μg g^−1^ Cr~2~O~3~Morales-Muñoz et al. [@CR32]Soluble0.12 mol L^−1^ K~2~HPO~4~ (pH \~8.2), 0.37 mol L^−1^ Al~2~(SO~4~)~3~, 0.94 mol L^−1^ Na~2~SO~3~1:530 min shaking12\
88Soil + 10 μg g^−1^ Cr(VI) as PbCrO~4~\
Soil + 12.5 μg g^−1^ Cr(VI) as K~2~Cr~2~O~7~Rüdel and Terytze [@CR39]Soluble and exchangeable0.1 mol L^−1^ KH~2~PO~4~-K~2~HPO~4~ (pH \~7)1:1016 h shaking29Soil + 10--20 mg Cr(VI) as BaCrO~4~James et al. [@CR21]Soluble and insoluble0.01 mol L^−1^ Na~3~PO~4~ (pH \~12)1:1005 min boiling98CRM 027---sandy loamMandiwana [@CR30]Soluble and insoluble0.1 mol L^−1^ Na~2~CO~3~ (pH \~10)1:10010 min boiling98CRM 545---welding dustElci et al. [@CR13]Soluble and insoluble0.28 mol L^−1^ Na~2~CO~3~, 0.5 mol L^−1^ NaOH (pH \~12), 4 mol L^−1^ MgCl~2~, 1 mol L^−1^ K~2~HPO~4~-KH~2~PO~4~ (EPA 3060A)1:2060 min heating at 90--95 °C95Soil spiked with Cr(VI)Gitet et al. [@CR15]Soluble and insoluble0.5 mol L^−1^ NaOH, 0.28 mol L^−1^ Na~2~CO~3~1:2060 min heating at 90--95 °C88\
100\
63CRM 013---paint chips\
CRM 013 + 2600 μg g^−1^ Cr(III) as CrCl~3~\
Soil + 10--20 mg Cr(VI) as BaCrO~4~Korolczuk and Grabarczyk [@CR24]Soluble and insoluble0.05 mol L^−1^ EDTA1:5060 min shaking1\
0.91CRM 483---sewage sludge-amended soil\
CRM 07411 Chinese soilLilleengen and Wibetoe 2002Soluble and insoluble0.01 mol L^−1^ EDTA in 0.05 mol L^−1^ NH~4~OH + (NH~4~)~2~SO~4~ (pH \~9.5)1:10030 min, 40 °C (ultrasound-assisted extraction)96.5--98.4\
86SiO~2~ spiked with 400 μg g^−1^ Cr(VI) as insoluble chromate\
CRM 013---paint chipsKorolczuk and Grabarczyk [@CR24]Soluble and insoluble0.02 mol L^−1^ DTPA in 0.05 mol L^−1^ NH~4~OH + (NH~4~)~2~SO~4~ (pH \~9.5)1:5010 min heating at 40 °C, stirring97SiO~2~ spiked with 1.125 μg g^−1^ Cr(VI) as BaCrO~4~ and SrCrO~4~Grabarczyk et al. [@CR17]*m* mass of sample, *V* volume of extraction solutionCRM 483---sewage sludge-amended soil (Cr content 3925 ± 195 μg g^−1^); CRM 027---sandy loam (Cr content 26.9 ± 0.9 μg g^−1^); CRM 545---welding dust loaded on filter (Cr(VI) content 40,200 ± 0.6 μg g^−1^); CRM 07411---Chinese soil (Cr content 61.8 ± 2.1 μg g^−1^); CRM 013---paint chips (Cr content 618 μg g^−1^)^a^Efficiency of Cr extraction
It should be noted that the extraction procedures are not always selective for Cr(VI), as differentiation between oxidised and reduced Cr species may be obtained by using specific analytical methods, e.g. the diphenylcarbazide (DPC) method for Cr(VI). However, many papers have indicated interference in spectrophotometric chromium(VI) detection caused by the presence of other metal ions, e.g. Cu(II), Mo(VI), Fe(III), V(V) and Hg(II), and humic acids released from the soil (Pettine and Capri [@CR38]). Determining the chromium via a specific spectroanalytic detection technique, e.g. atomic absorption spectrometry, provides more accurate results and a lower detection limit. However, in this case, selectivity of the extraction procedures towards Cr(VI) and Cr(III) species should be assessed.
The aim of this work was to appraise the procedures originally proposed for the leaching of chromium(VI) species from solid samples in terms of their selectivity towards trivalent and hexavalent chromium species. For this reason, mineral and organic soil was spiked with soluble and slightly soluble chromium compounds (CrCl~3~ 6H~2~O, K~2~Cr~2~O~7~, BaCrO~4~). The spikes were dosed in liquid and solid forms in order to evaluate the mobility of chromium from waste disposal sites. The recoveries of chromium obtained after treating the soil with seven extraction procedures were controlled by electrothermal atomic absorption spectrometry (ETAAS) and compared in order to select the most suitable procedure for extraction of Cr(VI) and Cr(III) species. The trueness of the procedure proposed for selective extraction of Cr(VI) was verified in an analysis of certified reference material of soil (CRM 041). The chosen extraction procedures were applied to determine Cr(VI) in soils sampled in the Podlasie Province (Poland).
Material and Methods {#Sec2}
====================
Instrumentation {#Sec3}
---------------
A Solaar M6 (Thermo Electron Corporation, UK) atomic absorption spectrometer equipped with a Zeeman-effect background correction system and an electrothermal atomizer (ELC graphite tubes) was used for chromium determination. A chromium hollow cathode lamp (Photron, Australia) was operated at 15 mA. The measurements were done at *λ* = 357.9 nm with a spectral bandpass of 0.5 nm. The following optimised heating programme was used to determine chromium in the soil extracts: drying at 110 °C for 15 s, ashing at 1200 °C or 1650 °C for 8 s, and atomisation at 2600 °C for 3 s. A temperature of 1650 °C was used to ash extracts containing K~2~HPO~4~ in the presence of Mg(NO~3~)~2~ (10 μL of 10 μg mL^−1^, 0.1 μg) as a chemical modifier. An inoLab pH Level 1 (WTW, Germany) pH meter equipped with a SenTix 21 electrode (WTW, Germany) was used for the pH measurements. A Shimadzu SSM-5000A TOC Analyser was used to determine the carbon content in the soil by catalytically aided combustion oxidation method. A ball mill (KM 1 type K142, MLW, Poland) was used for soil grinding.
Reagents {#Sec4}
--------
Stock solutions (20 g L^−1^) of Cr(III) as CrCl~3~ (Merck, Germany) and (1.001 g L^−1^) of Cr(VI) as K~2~Cr~2~O~7~ (Sigma-Aldrich, Germany) were used. Working standard solutions of chromium were prepared daily by appropriate dilution of the stock standards. Reagents used for chromium extraction: K~2~HPO~4~, Na~3~PO~4~, Na~2~CO~3~ and EDTA as well as reagents used for spiking of soil CrCl~3~ 6H~2~O, K~2~Cr~2~O~7~ and BaCrO~4~ were obtained from POCh (Poland). Acetic acid and KCl were obtained from Chempur (Poland), and DTPA was obtained from Sigma-Aldrich (Germany). Deionised water was obtained from the Milli-Q water purification system (Millipore, USA).
Samples and Procedures {#Sec5}
----------------------
### Model Soils {#Sec6}
Two different soils, i.e. in terms of the physico-chemical properties (agricultural type, pH, content of the organic matrix), were collected from the arable layer in the Podlasie Province (Poland). The soil was air-dried, homogenised and sieved using a 1-mm sieve. Both the content of organic carbon and the pH of the soils (in KCl) were determined by using standard methods. The mineral soil (M) contained 1.6% organic carbon (pH~KCl~ = 5.4), while the organic soil (O) contained 10.5% organic carbon (pH~KCl~ = 6.1). The samples were spiked with different chromium compounds and used for optimisation of the extraction procedure.
### Spiking Procedures {#Sec7}
Three portions of each soil (5 g) were spiked with 50 mg g^−1^ of Cr(III) or Cr(VI) in a solid form (CrCl~3~·6H~2~O, K~2~Cr~2~O~7~, BaCrO~4~). The samples were homogenised for 1 h in a ball mill. Next, 0.25 g of spiked soil was mixed with 4.75 g of natural soil and the sample was homogenised again for 1 h. One gramme of each soil was mixed with 49 g of natural soil for 16 h. In effect, each portion of soil (50 g) was spiked with 50 μg g^−1^ of chromium as CrCl~3~ 6H~2~O (O-Cr(III)s, M-Cr(III)s), K~2~Cr~2~O~7~ (O-Cr(VI)s, M-Cr(VI)s) and BaCrO~4~ (O-Cr(VI)s\*, M-Cr(VI)s\*). Two other portions of soil (50 g) were shaken with 50 mL of 50 μg mL^−1^ solution of Cr(III) or Cr(VI) compounds, such as CrCl~3~ 6H~2~O (O-Cr(III)~liq~, M-Cr(III)~liq~) and K~2~Cr~2~O~7~ (O-Cr(VI)~liq~, M-Cr(VI)~liq~)), for 2 h and dried in air. These samples were then used for further studies.
### Determination of the Total Content of Cr {#Sec8}
The content of total chromium in both the natural and spiked soils was determined after wet mineralisation of soil in a mixture of HNO~3~:HF (5 mL:1 mL). Samples (0.2 g) were heated in closed Teflon vessels in a microwave digestion system (Ethos Plus, Milestone, Italy) according to optimised microwave program: 250 W for 2.5 min, 500 W for 5 min and 700 W for 15 min. The process was repeated twice for total digestion of soil. The obtained solutions were transferred into polyethylene vessels, diluted with Milli-Q water to the final volume of 15 mL and analysed by ETAAS. The chromium content in the mineral soil (M) was 16.9 μg g^−1^, while in the organic soil (O) it was 18.3 μg g^−1^. The average content of chromium in mineral soil spiked with different chromium compounds was 34.2 ± 5.1 μg g^−1^ (the average spiking efficiency was equal to 68%), while in the spiked organic soil it was 38.4 ± 4.5 μg g^−1^ (the average spiking efficiency was equal to 76%).
### Soil Samples {#Sec9}
The soils examined in this work were collected from industrially contaminated area of a closed down leather tannery in Krynki (Podlasie Province, Poland). The content of Cr(VI) in soil was evaluated by using the developed procedure.
### CRM {#Sec10}
Certified reference material of sandy clay soil CRM 041 (Sigma-Aldrich, Germany) with certified Cr(VI) content (86.3 ± 2.96 μg g^−1^) was used for the accuracy studies.
### Extraction Procedures {#Sec11}
The extraction procedures used to evaluate the efficiency of chromium released from the soils are summarised in Table [2](#Tab2){ref-type="table"}. Uniform extraction conditions were always used. A total of 50 mL of one of the extraction solutions was added to each soil sample (mass of 1 g). The suspensions were rotated for 16 h at room temperature (22 ± 2 °C) and next centrifuged at 3000 rpm for 15 min. The tubes for the hot carbonate extraction, after the addition of extraction solution, were placed on a preheated hot plate and maintained at 90--95 °C for 10 min. The supernatants, after centrifugation and appropriate dilution (5--100 times), were used to determine the chromium concentration by ETAAS. All analyses were carried out in triplicates. Recovery of spiked chromium obtained in various extraction procedures was assessed as follows:$$\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$ R\left(\%\right)=\left({m}_{\mathrm{exsp}}{\textstyle \hbox{-} }{m}_{\mathrm{exnat}}\right)/{m}_{\mathrm{sp}}\times 100\% $$\end{document}$$where *m* ~exsp~---mass of Cr extracted from spiked soil
*m* ~exnat~---mass of Cr extracted from soil
*m* ~sp~---mass of spiked Cr
The mass of chromium extracted from soil was calculated on the basis of an external calibration graph prepared for the given extracting solution.Table 2The efficiency of Cr extraction from natural soil and the recovery of Cr(III) and Cr(VI) from spiked soil by using various extraction solutions (mass of soil 1 g, volume of extraction solution 50 mL, extraction for 16 h at room temperature)Extraction solutionType of soilEfficiency of Cr extraction, %Recovery of Cr, %Natural soilSoil spiked with Cr(III)Soil spiked with Cr(VI)0.43 mol L^−1^ CH~3~COOH (pH \~2.7)Organic0.281.9--13.114.6--23.2Mineral0.251.7--16.014.9--29.20.1 mol L^−1^ K~2~HPO~4~ with addition of 1 mL 0.37 mol L^−1^ Al~2~(SO~4~)~3~ + 1 mL 0.94 mol L^−1^ Na~2~SO~3~ (pH \~8.0; 1.56 mol L^−1^ H~3~PO~4~)Organic1.00.7--7.44.6--14.1Mineral0.40.9--1.45.5--12.10.01 mol L^−1^ Na~3~PO~4~ (pH \~11.0)Organic4.71.4--15.816.6--80.1Mineral4.32.1--17.034.7--77.90.1 mol L^−1^ Na~2~CO~3~ (pH \~10.0)Organic10.72.0--35.635.6--109.8Mineral1.71.5--8.134.7--111.5^a^0.1 mol L^−1^ Na~2~CO~3~ (pH \~10.0)Organic14.25.3--7368.7--98.2Mineral6.03.5--22.943.2--99.80.01 mol L^−1^ EDTA (pH \~9.5; 0.05 mol L^−1^ (NH~4~)~2~SO~4~ + 25% NH~4~OH)Organic7.73.2--37.542.9--106.4Mineral3.32.0--17.253.9--110.70.02 mol L^−1^ DTPA (pH \~9.5; 0.05 mol L^−1^ (NH~4~)~2~SO~4~ + 25% NH~4~OH)Organic12.83.8--48.334.0--98.9Mineral3.510.5--50.451.2--103.4^a^ extraction by heating of suspension at 90--95 °C for 10 min
Results and Discussion {#Sec12}
======================
An Appraisal of Extraction Procedures of Chromium Species from Soil {#Sec13}
-------------------------------------------------------------------
The selectivity of procedures originally proposed for leaching of chromium(VI) species from solid samples (listed in Table [1](#Tab1){ref-type="table"}) was tested. The mass of chromium extracted from mineral and organic soil spiked with different trivalent and hexavalent compounds of chromium (as CrCl~3~·6H~2~O, K~2~CrO~4~, BaCrO~4~) was determined, and recoveries were calculated (Table [2](#Tab2){ref-type="table"}). The obtained results are shown in Fig. [1](#Fig1){ref-type="fig"}. For a comparison, all procedures were also applied for extraction of native chromium from unpolluted soils. The efficiency of chromium extraction was higher from organic soil (0.3--13% of its total content) than from mineral soil (0.3 to 4.4%) (Fig. [1](#Fig1){ref-type="fig"}c).Fig. 1The recovery of chromium from soil spiked with Cr(III) and Cr(VI) in solid (s) or liquid (liq) forms extracted with using various reagents. **a** Organic soil. **b** Mineral soil. **c** The efficiency of chromium extraction from native soil with using various reagents (value ± standard deviation, *n* = 3)
Acetic acid at various concentrations was recommended in order to determine the exchangeable and carbonate-bound chromium fraction in the soil (Lillengen and Wibetoe [@CR28]; Krasnodębska-Ostręga et al. [@CR26]). The amount of native chromium extracted from natural soil with 0.43 mol L^−1^ solution of CH~3~COOH was very low (\<0.3%). The recovery of chromium from soils spiked with Cr(VI) forms was higher (14--29%) than from soils spiked with Cr(III) forms (1.7--16%). It was found that Cr(III) added to the soil in a liquid form was extracted in a higher amount (13--16%) than that added in a solid form (1.7--1.9%). The opposite correlation was observed for Cr(VI) (Fig. [1](#Fig1){ref-type="fig"}a, b).
Some papers mentioned that extraction of Cr(VI) in alkaline solution is preferable to acid extraction because it ensures better solubility of some of the chromate compounds (James et al. [@CR21]). The extraction procedure using phosphate buffer is devoted to the leaching of soluble Cr(VI) forms (Rüdel and Terytze [@CR39]). The addition of Al~2~(SO~4~)~3~ enhances precipitation of Cr(III) in the sulphate form, while Na~2~SO~3~ prevents oxidation reactions. The efficiency of chromium extraction from natural soil with 0.1 mol L^−1^ solution of K~2~HPO~4~ was slightly higher (0.4--1%) than that with CH~3~COOH. However, this procedure was less effective for the extraction of chromium(VI) from spiked soils as its recovery was in the range of 4.6--14.1%; surprisingly, it was lower for the soluble (as K~2~CrO~4~) than for the insoluble (as BaCrO~4~) form of Cr(VI) (Fig. [1](#Fig1){ref-type="fig"}a, b). Such low recovery of insoluble Cr(VI) from the soil (\<12%) was also reported by others (Szulczewski et al. [@CR41]). This procedure is then not recommended for the extraction of total Cr(VI), as was proposed in James et al. ([@CR21]). It is worth noting that up to 7.4% of spiked Cr(III) was also extracted with this reagent.
During extraction of soil with the Na~3~PO~4~ solution, insoluble Cr(VI) compounds (e.g. BaCrO~4~) are transformed into soluble forms (e.g. Na~2~CrO~4~) (Mandiwana [@CR30]). The procedure should remove all common metals, e.g. insoluble phosphates, oxides or hydroxides, so the solution should contain only CrO~4~ ^2−^ ions. The efficiency of chromium extraction from natural soil with 0.01 mol L^−1^ Na~3~PO~4~ was below 5%. Higher recovery of chromium was obtained from soil spiked with solid K~2~CrO~4~ (80%) than from soil spiked with BaCrO~4~ (20--40%). Recovery of Cr(III) forms was similar to that obtained with CH~3~COOH, namely 2% from soil spiked with solid CrCl~3~ and 17% from soil spiked with CrCl~3~ solution (Fig. [1](#Fig1){ref-type="fig"}a, b).
Leaching of soil with 0.1 mol L^−1^ Na~2~CO~3~ in alkaline solution was proposed for extraction of soluble and insoluble Cr(VI) forms (Vitale et al. [@CR45]). Such treatment of samples transformed insoluble Cr(VI) salts into soluble chromate ions, whereas all common metals, including Cr(III), were removed as insoluble carbonates, oxides or hydroxycarbonates. The efficiency of extraction of native Cr with this reagent was 2% from mineral and 11% from organic soil. The recovery of chromium added as a solid Cr(III) salt was about 2% from both types of soil, while added as a solution, it increased to 8 and 35% from mineral and organic soil, respectively. High recovery of chromium (80--110%) from both types of soil spiked with soluble and insoluble solid Cr(VI) salts was observed. Lower recovery of chromium added as a solution of CrO~4~ ^2−^ suggests that Cr(VI) was partly reduced by the sample matrix to insoluble Cr(III), which supports the observations of Vitale et al. ([@CR46]) that soluble CrO~4~ ^2−^ is reduced to insoluble Cr~2~O~3~. This effect was also observed in all other procedures (Fig. [1](#Fig1){ref-type="fig"}a, b). In order to shorten the procedure, the extraction of soil with Na~2~CO~3~ solution at a high temperature, i.e. 90--95 °C, for 10 min (Panichev et al. [@CR35]) was tested for Cr(VI) leaching. It was observed that the colour of the alkaline extract was much darker, indicating that at higher temperature, the organic components of soil, e.g. humic acids, were better solubilised. This effect was more visible for organic soil. A small increase was observed in the extraction efficiency of native chromium (\~4%) and the recovery of solid Cr(III) spike (\~2%) from organic and mineral soil. The recovery of Cr(VI) spikes added as a solid in soluble and insoluble forms from both types of soil was quantitative, which indicates that the presence of liberated organic compounds in the extract did not significantly influence the reduction process of Cr(VI) species (Vitale et al. [@CR46]). The highest increase in the recovery of analytes (\~35%) in comparison to unheated treatment was observed for organic soil spiked with solutions of Cr(III) and soluble Cr(VI) (Fig. [2](#Fig2){ref-type="fig"}). The increase in recoveries of analytes from mineral soil spiked with the same chromium forms was slightly lower, i.e. 15% for Cr(III) and 8% for soluble Cr(VI). The observed changes suggest that immobile chromium species were converted into mobile forms. Most probably under alkaline conditions, high temperature and the presence of carbonate and organic matter, Cr(III) was oxidised to Cr(VI), which was extracted more efficiently. Extraction in high temperature resulted in the release of chromium species adsorbed on soil particles due to the dissolution of the organic and inorganic substances present in the soil. The studies showed only a small increase in the extraction efficiency of chromium from native soil (by 4%) at an elevated temperature. This suggests that native chromium is present in a form that is more resistant to oxidation and solubilisation.Fig. 2The recovery of chromium from soil spiked with Cr(III) and Cr(VI) in solid (s) or liquid (liq) forms extracted with Na~2~CO~3~ solution at room temperature or heated to 90--95 °C (value ± standard deviation, *n* = 3)
Extraction of chromium with EDTA and DTPA solutions is based on the transformation of insoluble Cr(VI) compounds into soluble species by complexation of the metal ions that form insoluble chromates (Korolczuk and Grabarczyk [@CR24]; Grabarczyk et al. [@CR17]). The extraction of chromium with EDTA and DTPA was similar to that obtained with the Na~2~CO~3~ solution. The efficiency of extraction of chromium from natural soil was in the range of 3.3--12.8%. Chromium(VI) added in a solid form was quantitatively recovered from both soils. The recoveries of Cr(III) and Cr(VI) spiked as a liquid were in the range of 17--54%. A higher percentage of metal was extracted with EDTA than with DTPA from soil spiked with the Cr(VI) form. The results showed that EDTA and DTPA were the most efficient extractants of Cr(VI) among all the tested reagents; however, they also leached significant amounts of Cr(III) (Fig. [1](#Fig1){ref-type="fig"}a, b).
On the basis of the performed experiments, the tested reagents were sorted according to their extraction potential into Cr(III) and Cr(VI) species. For the Cr(III) species, the differences in the extraction power of the reagents used were observed depending on the phase of the introduced spike. For Cr(III) added as a solid to mineral and organic soil, the order of extraction reagents was as follows: K~2~HPO~4~ (\~1%) \< CH~3~COOH, Na~3~PO~4~, Na~2~CO~3~ (\~2%) ≤ EDTA (2--3.2%) \< DTPA (3.8--10%). When Cr(III) was introduced as a liquid, a different extraction order of reagents was observed for the two tested types of soil. The spike of Cr(III) from mineral soil was extracted with increasing efficiency by using the following reagents: K~2~HPO~4~ (1.4%) \< Na~2~CO~3~ (8%) \< CH~3~COOH, Na~3~PO~4~, EDTA (\~17%) \< DTPA (50%), whereas from organic soil the order was as follows: K~2~HPO~4~ (7%) \< CH~3~COOH, Na~3~PO~4~ (\~15%) \< Na~2~CO~3~, EDTA (\~36%) \< DTPA (48%). For organic soil, the extraction of Cr(III) with hot Na~2~CO~3~ was the most efficient procedure (recovery: 5.3% for solid spike and 73% for liquid spike). For mineral soil, this procedure had a higher extraction power than CH~3~COOH, Na~3~PO~4~ and EDTA, but still lower than DTPA (recovery: 3.5% for solid spike and 22% for liquid spike). It should be noted that none of the tested procedures allowed for total recovery of Cr(III), which indicates a strong interaction of the spike with the matrix of the soil. The highest recovery of Cr(III) from both types of soil was obtained with DTPA solution (\~50%).
Spikes of Cr(VI) introduced into the organic soil in either a liquid or solid phase were extracted with increasing efficiency by using the following reagents: K~2~HPO~4~ (5--14%) \< CH~3~COOH (15--23%) \< Na~3~PO~4~ (17--80%) \< DTPA, Na~2~CO~3~, EDTA (35--110%). The order of reagents for mineral soil spiked with insoluble Cr(VI) was the same as for organic soil. In the case of mineral soil spiked with soluble Cr(VI) (as a liquid and solid form), similar power of Na~3~PO~4~ and Na~2~CO~3~ towards Cr(VI) was observed; therefore, the order of the reagents was slightly different: K~2~HPO~4~ (5--7%) \< CH~3~COOH (15--25%) \< Na~3~PO~4~ = Na~2~CO~3~ (35--78%) \< DTPA, EDTA (51--105%). The total recovery of solid Cr(VI) spikes from organic soil was obtained using Na~2~CO~3~, DTPA and EDTA solutions, whereas from mineral soil it was obtained with DTPA and EDTA solutions. The most efficient extraction of Cr(VI) from both samples was obtained with hot Na~2~CO~3~, thus allowing for quantitative recovery of the solid Cr(VI) spike.
The order of the tested reagents towards extraction of native chromium from organic soil was the same as for the spike of Cr(VI) and was as follows: CH~3~COOH (0.3%) \< K~2~HPO~4~ (1%) \< Na~3~PO~4~ (4.7%) \< EDTA (7.7%) \< Na~2~CO~3~ (11%) \< DTPA (12.8%) \< Na~2~CO~3~ hot (14.2%). For mineral soil, higher extraction power towards native chromium was demonstrated by the Na~3~PO~4~ solution and hot Na~2~CO~3~; therefore, the order of extraction reagents was as follows: K~2~HPO~4~, CH~3~COOH (0.3%) \< Na~2~CO~3~ (1.6%) \< EDTA, DTPA (3.5%) \< Na~3~PO~4~ (4.4%) \< Na~2~CO~3~ hot (6%).
It was found that the CH~3~COOH and K~2~HPO~4~ solutions exhibited the lowest extraction power towards all spikes of chromium, which is consistent with the literature data suggesting that these reagents be used for extraction of soluble Cr(VI) forms (Lillengen and Wibetoe [@CR28]; Rüdel and Terytze [@CR39]). Lower recovery of Cr(VI) spiked as a soluble chromate was obtained with K~2~HPO~4~ than was reported in Rüdel and Terytze ([@CR39]) (\~12 versus 88%), whereas a similar recovery was obtained for insoluble chromates (\~12%). Among the other reagents proposed for extraction of soluble and insoluble Cr(VI) forms, only Na~3~PO~4~ was ineffective, even when applied by others (Mandiwana [@CR30]). The results obtained in our experiments in procedures using Na~2~CO~3~, EDTA and DTPA solutions were comparable to the results presented in the literature for samples spiked with Cr(VI) in the form of soluble and insoluble chromate (see Table [1](#Tab1){ref-type="table"}) (Korolczuk and Grabarczyk [@CR24]; Grabarczyk et al. [@CR17]; Lillengen and Wibetoe [@CR28]). The efficiency of extraction of native chromium with CH~3~COOH and EDTA as obtained in our work was the same as that reported by Lillengen and Wibetoe ([@CR28]) for CRM 483 sewage sludge-amended soil and CRM 07411 Chinese soil.
The choice of procedure for selective extraction of Cr(VI) using 0.1 mol L^−1^ Na~2~CO~3~ at room temperature is highly recommended for mineral soil samples, as it provided quantitative recovery of Cr(VI) and slight extraction of Cr(III) (8%). This procedure may also be appropriate for natural organic soil, despite the higher recovery of Cr(III), which was probably the effect of better solubility of the spiked form of chromium (CrCl~3~). Taking into account the efficiency of extraction of native chromium from both mineral and organic soil, in which Cr(III) forms predominate, and the advantage of shortening the extraction time, the extraction procedure using Na~2~CO~3~ at high temperature may also be recommended for mineral soil.
As was discussed above, releasing chromium from native (model soils) and anthropogenic contaminated soil (spiked soils) was different. Therefore, the characteristics of the environmental samples and the sources of their contamination should be known to properly select the extraction procedure.
An Impact of Physical Form of Waste on Chromium Mobility in Soil {#Sec14}
----------------------------------------------------------------
The studies performed here show that the differences in the recovery of spiked chromium forms depend not only on the chromium oxidation state and type of soil but also on the physical phase of the spike. Cr(III) forms, considered to be immobile and insoluble, were partly extracted from the soil, though less efficiently when Cr(III) was spiked in a solid form than in a liquid form. The recovery of Cr(III) added as a solid compound was only 4% from the organic soil and 2% from the mineral soil, irrespective of the extraction procedure that was used (Fig. [1](#Fig1){ref-type="fig"}a, b). Higher recovery of chromium with the DTPA solution (up to 10%) was observed, which might be an effect of the weak dissolution of the solid spike due to a shift in the chemical equilibrium towards complexation of Cr(III) ions with the chelating reagent. Such an effect was not observed for extraction with EDTA due to the slow rate of the formation of Cr(III)-EDTA complexes. Cr(III) added to the soil in the form of a solution was extracted to a higher extent. Its recovery from mineral soil reached 17%, while from organic soil it was 37%. Also in this case, the recovery of chromium with DTPA solution from both types of soil was highest (up to \~50%). Low recoveries of added Cr(III) ions suggest the occurrence of a strong interaction between the spike and the soil. Probably, Cr(III) cations, added as a spike, are immobilised/sorbed on the negatively charged surface of particles of the soil components. The precipitation of Cr(OH)~3~, which gradually undergoes dehydration and crystallisation as Cr~2~O~3~, is also possible. Higher recoveries of Cr(III) (35--48%) obtained from the organic soil by Na~2~CO~3~, EDTA and DTPA solutions suggest that some soluble organic complexes of Cr(III) with fulvic/humic or citric acid could also be formed in the presence of humic substances.
Since the phenomenon of inter-conversion of chromium oxidation forms under environmental conditions has already been reported (Kotaś and Stasicka [@CR25]; Metze et al. [@CR31]; Dhal et al. [@CR10]), oxidation of the added Cr(III) to Cr(VI) in the tested soil could not be excluded. Such oxidation may arise only under certain circumstances in the presence of MnO~2~ or molecular oxygen (Apte et al. [@CR1]; Pantsar-Kallio and Oksanen [@CR36]). Dhal et al. ([@CR10]), in their review, stated that oxygen does not react appreciably with Cr(III), while oxidation of Cr(III) by MnO~2~ depends on the concentration of water-soluble chromium, pH of soil, amount of organic matter and drainage of the soil. For the studied samples, the presence of Cr(III) in the aqueous phase of soil was limited due to its adsorption on the soil particles, the precipitation and formation of stable complexes with humic substances. Therefore, oxidation of the added spike of Cr(III) during extraction at room temperature was less probable. However, during heating of soil with Na~2~CO~3~ solution, the probable oxidation of Cr(III) to mobile Cr(VI) forms arose, as higher recovery of Cr spikes was observed. The alkaline conditions, high temperature and presence of organic matter led to oxidation of Cr(III) to the Cr(VI) form.
The recoveries of Cr(VI) forms, which are considered to be more mobile than Cr(III) forms, were also dependent on the physical state of Cr(VI) added to the soil. Generally, Cr(VI) added as a solution was less extractable than when added as a solid compound (soluble and insoluble salts). Probably, such a phenomenon was the effect of the reduction of Cr(VI) to Cr(III) by the matrix components of the soil. Although the reduction of Cr(VI) to Cr(III) by the soil organic compounds and the oxidation of Cr(III) to Cr(VI) by the manganese oxides result from thermodynamically spontaneous reactions (James [@CR20]), the reduction process arises more easily than oxidation under environmental conditions (Dhal et al. [@CR10]; Brose and James [@CR5]). The concentration of the reducing agents such as organic matter, sulphides or Fe(II)-bearing minerals in the soil and the soil pH affect the redox behaviour of chromium (Apte et al. [@CR1]; Metze et al. [@CR31]). The rate of reduction of Cr(VI) by the humic acids increases along with the decrease in pH (Dhal et al. [@CR10]). The content of organic matter (18.1%) and acidic conditions in organic soil (pH 6.1) facilitated the reduction of the added Cr(VI) to Cr(III) to a large extent. In aerobic soil, which contains appropriate organic substances for the redox process, the reduction of Cr(VI) in a slightly alkaline environment is also possible. The formed Cr(III) species were probably immobilised on the soil particles. Therefore, in all of the tested extraction solutions, the recovery of Cr(VI) added as a solution to the organic soil was almost the same as the recovery of the liquid spike of Cr(III). For the mineral soil, the acidic conditions enhanced the rate of release of the Fe(II) species from the minerals, which could then react with Cr(VI) in the aqueous phase. However, higher recovery of Cr(VI) than Cr(III) (both added in the liquid form) indicates that a smaller amount of Cr(VI) was converted into Cr(III) in the mineral soil.
The obtained results demonstrate the transformation of added chromium forms in the soil, which leads to a significant change in their initial mobility and finally in the recovery of added forms of chromium. The most important conclusion from this experiment is that the mobility of chromium species depends not only on its oxidation state but also on its physical form as introduced into the environment, which is of great significance when chromium waste is disposed of into the environment.
Validation of Extraction Procedures of Cr(VI) from Soil {#Sec15}
-------------------------------------------------------
Validation of the extraction procedures based on using 0.1 mol L^−1^ Na~2~CO~3~ solution, shaking the suspension of the sample for 16 h at room temperature or heating the suspension of the sample at 90--95 °C for 10 min was performed in order to obtain reliable and accurate results of chromium species extracted from the soil samples. During method validation, the following parameters were estimated: linearity, limit of detection (LOD) and limit of quantification (LOQ), precision, trueness and uncertainty of measurements of Cr(VI) content in soil extracted with Na~2~CO~3~.
In order to verify the linearity of the calibration graph, Cr(VI) standards in a concentration range of 1--50 ng mL^−1^ were prepared in a 100 times diluted Na~2~CO~3~ extraction solution and their absorbance was measured by ETAAS. Then the calibration graph was constructed as *A* = *f*(*C* ~Cr(VI)~) and the correlation factor was used to verify their linearity. It was found that the correlation factor (*R*) was higher than 0.995 for the calibration graph prepared from the standard solutions at a concentration range of 1--30 ng mL^−1^. The obtained equation of the calibration graph was *y* = 0.0173*x* + 0.0195 (*R* = 0.9992). The extraction solution was used as a blank sample to calculate the limit of detection of Cr(VI). The value of the limit of detection (LOD) was calculated according to the following equation: LOD = 3SD~blank~/*b*, where *b* is the slope of the calibration graph. The limit of quantification (LOQ) was calculated as LOQ = 6SD~blank~/*b*. The LOD obtained for the extraction solution was 0.35 ng mL^−1^, while the LOQ was 0.71 ng mL^−1^. The volume of the extraction solutions and the mass of the soil samples were used for the calculations in order to evaluate these parameters for the soil samples. The LOD for Cr(VI) in soil was 17.5 μg kg^−1^, while the LOQ for soil was 35.5 μg kg^−1^.
The precision of measurements (expressed as the relative standard deviation (RSD)) was evaluated by analysing the standard solutions of Cr at a concentration of 4 ng mL^−1^ in extraction solution on the same day. The obtained value of RSD for measurements of six independent standards was 1.2%, which means that the precision of the measurements was satisfactory. The repeatability of extraction of chromium from the soil samples was evaluated by analysing extracts obtained by using the extraction solution at room temperature and heated to 90--95 °C. Repeatability was expressed as RSD for six independent extractions of the same sample. It was found that for both extraction procedures, the values of RSD were in the range of 1--10% for both analysed types of soil; however, slightly lower values were obtained for the mineral soil.
The trueness of extraction procedures of chromium was evaluated by analysing the certified reference material of soil CRM 041 (contaminated sandy clay soil) with a certified value of the Cr(VI) content determined by using the normalised EPA 3060A procedure (60 min of boiling the sample at 90 °C in a solution of 0.28 mol L^−1^ Na~2~CO~3~, 0.5 mol L^−1^ NaOH, 4 mol L^−1^ MgCl~2~ in 1 mol L^−1^ of phosphoric buffer). Good agreement of the content of Cr(VI) determined in 0.1 mol L^−1^ Na~2~CO~3~ extract obtained at room temperature (82.2 ± 5.5 μg g^−1^, *n* = 3) and after heating (86.6 ± 1.6 μg g^−1^, *n* = 3) with the certified value (86.3 ± 2.96 μg g^−1^) indicates good accuracy of the selected extraction procedures. The recovery of Cr(VI) in CRM 041 was 95.2 ± 6.3 and 100.4 ± 1.8%, respectively, which proved that the developed procedures may be applied for analysis of natural soil. We did not observed any influence of matrices of CRM 041 (as a mineral soil with 15 times lower content of manganese than the total content of chromium) on the recovery of Cr(VI) obtained after using extraction procedure at elevated temperature. It is worth noting that the developed procedures are simpler than normalised EPA 3060A in terms of the composition of the extraction medium (only Na~2~CO~3~ solution), conditions of extraction (room temperature) and time of extraction (10 min in the case of heating the suspension of the soil sample).
The uncertainty of measurements of Cr(VI) content in soil extracted with Na~2~CO~3~ solution by developed procedures were evaluated in accordance with the Guide to the Expression of Uncertainty in Measurement ([@CR14]), similarly to the scheme presented by Leśniewska et al. [@CR27]. The modelling approach, based on a model equation, in which individual components of uncertainty that contribute to uncertainty of measurement are quantified, was used for estimation of combined standard uncertainty of measurements. The obtained expanded uncertainty (*U*) of measurements of Cr(VI) content in soil extracted with Na~2~CO~3~ solution at room temperature (82.2 ± 8.0 μg g^−1^; *U* = 9.7%, *k* = 2) was slightly higher than for extraction procedure at elevated temperature (86.6 ± 5.5 μg g^−1^; *U* = 6.3%, *k* = 2) due to a higher value of standard uncertainty of Cr(VI) recovery from CRM 041.
Analysis of Soil Samples {#Sec16}
------------------------
Samples of soil (1--7) were collected from an industrially contaminated area of an old leather tannery where chromium(III) sulphate was used during the tanning process. The tannery had functioned there since the 1960s to the late 1990s. Samples 1, 2 and 3 were collected near the entrance to the wet, tanning and finishing departments (at a distance of 2--4 m from the building), respectively. Sample 4 was collected from the opposite side of the tannery building, probably along the route of transport of leather between the above-mentioned departments. Samples 5, 6 and 7 were collected at a distance of 1--2 m outside the tannery area. From all location, three mixed sub-samples were collected from a surface layer at a depth of 0--10 cm. All samples were categorised as mineral soil with various contents of organic matter and pH values (Table [3](#Tab3){ref-type="table"}). The samples were classified mostly as non-contaminated soil, as the total content of chromium did not exceed its permissible limit for agricultural soil in Poland (Ordinance of the Minister of Environment of Poland [@CR33]). Despite the use of large amounts of Cr(III) compounds in tannery, only two samples from its area (3 and 4) were classified as contaminated soil. The content of Cr(VI) in the samples was assessed by two developed and recommended procedures based on extraction with 0.1 mol L^−1^ Na~2~CO~3~ solution at room and elevated (90--95 °C) temperature. The results did not differ significantly (Table [3](#Tab3){ref-type="table"}). The share of Cr(VI) in the total content of chromium in the analysed soil was very low, i.e. in the range of 0.8 to 4.5%. This low content of extracted Cr(VI) in the analysed soil, despite the high total content of chromium, indicates that chromium is present in very immobile forms, probably as stable Cr(III) compounds.Table 3The characteristic of soil samples collected from contaminated area of old leather tannery and the results of Cr(VI) extraction with Na~2~CO~3~ solutionSamplepH~KCl~Organic matter, %Total content of Cr, μg g^−1^, *n* = 3Extraction with Na~2~CO~3~ at room temperature, *n* = 3Extraction with Na~2~CO~3~ at 90--95 °C, *n* = 3Content of Cr(VI) ± SD, μg g^−1^Percentage of Cr(VI) in total content ± SD, %Content of Cr(VI) ± SD, μg g^−1^Percentage of Cr(VI) in total content ± SD, %Soil 17.81.8668.6 ± 5.00.88 ± 0.031.29 ± 0.051.55 ± 0.062.27 ± 0.08Soli 27.37.34141.9 ± 6.13.54 ± 0.022.50 ± 0.024.36 ± 0.053.07 ± 0.03Soil 37.210.14283.9 ± 7.912.90 ± 1.434.54 ± 0.5012.18 ± 0.834.29 ± 0.29Soil 47.75.382336 ± 3018.21 ± 0.500.78 ± 0.0224.49 ± 0.731.05 ± 0.03Soil 57.32.9041.8 ± 3.20.54 ± 0.061.30 ± 0.140.45 ± 0.011.09 ± 0.03Soil 67.43.1026.0 ± 0.90.57 ± 0.052.20 ± 0.180.55 ± 0.022.13 ± 0.08Soil 77.44.3496.9 ± 2.82.38 ± 0.152.46 ± 0.161.94 ± 0.082.00 ± 0.08
Conclusions {#Sec17}
===========
The procedures proposed in the literature for the extraction of various forms of Cr(VI) from solid matrices were tested under uniform conditions in order to evaluate their selectivity and efficiency towards Cr(VI) and Cr(III) species. None of the tested reagents showed suitable extraction power towards quantitative extraction of Cr(III) species. The procedures based on Na~2~CO~3~ (at room and elevated temperatures) were found to be suitable for extraction of Cr(VI) species from mineral soil. The procedure based on extraction of Cr(VI) with Na~2~CO~3~ solution at room temperature is recommended for organic soil analysis, as partial co-extraction and oxidation of Cr(III) arises at a higher temperature. The proposed procedures were validated and accuracy was proved by analysis of CRM of soil (CRM 041).
The studies conducted here show that the physical state of waste, initial form and oxidation state of chromium as well as soil properties had an influence on the final form of chromium species in soil, which affected its mobility. It was found that Cr(III) introduced into soil as a solid waste was immobile, while introduced as a liquid, it may became mobile depending on the environmental conditions. Cr(VI) compounds disposed in the form of solid waste enter the soil after solubilisation. Compounds of Cr(VI) disposed as liquid waste undergo various reactions with the soil matrix and their reduction may occur; therefore, data on the sources of pollution and original forms of chromium in wastes are necessary for better prediction and estimation of the presence of chromium species in soil.
The analysis of soil collected from the contaminated area of the old tannery revealed that the share of Cr(VI) was very low (only 0.8--4.5%) despite the high total content of chromium. This confirms that chromium was present in the soil in immobile forms, probably as Cr(III) compounds.
The authors kindly acknowledge financial support from the Polish National Science Centre (DEC-2012/07/B/ST4/01581).
| {
"pile_set_name": "PubMed Central"
} |
1.. Introduction
================
The incubation period is a common feature of infection by pathogenic viruses. It is defined as the time between infection by a pathogen and the onset of symptoms. Determining the incubation periods of different pathogens assists health authorities control and track the progress of an infectious disease, thus limiting the spread of the pathogen and a possible epidemic. The length of the incubation period varies according to the infectious agent, the host immunological fitness, and previous immunological experience. In humans, it is difficult to determine the length of the incubation period since the exact time of infection is usually unknown. A thorough review of the literature by Lessler *et al.* \[[@b1-viruses-02-02541]\] showed that the reported incubation periods for human respiratory viruses ranges from around two days for influenza and human rhinovirus (HRV) to 10 days or more for measles virus (MeV).
2.. Termination of the Incubation Period---Onset of Symptoms is Mediated by the Immune Response
===============================================================================================
The abrupt onset of symptoms following infection with respiratory viruses marks the termination of the incubation period. Flu-like symptoms are varied and described by patients as fever and chills, malaise, myalgia, sneezing, cough, runny nose, sinus pain, congestion, headache and others \[[@b2-viruses-02-02541],[@b3-viruses-02-02541]\]. These symptoms are associated with the secretion of type I interferons (IFNs), interleukin 6 (IL-6), interleukin 8 (IL-8), interleukin 1 (IL-1), tumor necrosis factor α (TNF-α), macrophage inflammatory protein-1β (MIP-1β), interferon-γ (IFN-γ) and other cytokines \[[@b4-viruses-02-02541]--[@b6-viruses-02-02541]\].
While some of the symptoms may be directly related to the virus' cytopathic effect (shedding of damaged epithelium can lead to airway obstruction), most of the symptoms during influenza, MeV and HRV infections are the result of the immune response to the infection \[[@b7-viruses-02-02541]\]. The cause of the symptoms following respiratory syncytial virus (RSV) infection is controversial and it appears that both direct virus infection and the immune response play a role \[[@b8-viruses-02-02541],[@b9-viruses-02-02541]\].
Cytokines are usually observed prior to tissue damage generated by cytotoxic T cells or direct tissue damage caused by the virus infection. Patients treated with type I IFN, TNF-α, IL-1β, IL-1α or IL-6 for various illnesses report many flu-like symptoms without actually presenting with a respiratory virus infection \[[@b10-viruses-02-02541]--[@b14-viruses-02-02541]\]. An example of the immune system's contribution to the flu-like symptoms is the fact that administration of TNF-α or type I IFNs can cause headaches \[[@b10-viruses-02-02541],[@b11-viruses-02-02541],[@b13-viruses-02-02541]\].
Fever is mediated by the cytokines mentioned above, mainly IL-1, and is one of the best-understood interactions between the immune system and the nervous system. Although some aspects of the relay signals are unknown, it is largely thought that these cytokines signal the hypothalamus via the peripheral nervous system to increase the thermal set point \[[@b15-viruses-02-02541]--[@b18-viruses-02-02541]\]. Other symptoms also result from the cross talk of the immune system with the nervous system. Sneezing is mediated by the trigeminal nerve. This signal is relayed to the brain stem in response to histamines secreted by leukocytes \[[@b19-viruses-02-02541],[@b20-viruses-02-02541]\]. Coughing is mediated by the vagus nerves below the larynx and results from an inflammatory response in the lower respiratory tract \[[@b15-viruses-02-02541],[@b21-viruses-02-02541],[@b22-viruses-02-02541]\].
Nasal discharge (rhinorrhoea) is a combination of goblet cell secretion, gland secretion, plasma exudate, and contains dead leukocytes such as monocytes and neutrophils. The observed color change (from yellow to green) is due to the granule content of these cells \[[@b23-viruses-02-02541],[@b24-viruses-02-02541]\].
Many other cytokines, chemokines and growth factors are present at elevated levels in the virus-infected lung and in the serum, demonstrating similar kinetics to the above-mentioned cytokines. The cellular sources of these cytokines are still not completely known but both epithelial and hematopoietic cells are involved.
3.. Cellular Sensors for Viral Recognition
==========================================
Before an anti-viral response can take place in infected cells or cells that have been exposed to viral components, viral presence must be sensed. Toll-like receptors (TLRs), Retinoic acid inducible gene I (RIG-I) like receptors (RLRs) and the inflammasome complex take part in this process.
3.1.. The TLR System
--------------------
Specialized TLRs for viral sensing are TLR-3 that recognizes dsRNA and localizes to the plasma membrane or endosome \[[@b25-viruses-02-02541],[@b26-viruses-02-02541]\]. The endosomal TLR-7 and TLR-8 recognize viral single-stranded RNA (ssRNA) \[[@b27-viruses-02-02541],[@b28-viruses-02-02541]\]. TLR-9 recognizes unmethylated CpG DNA of bacteria and viruses \[[@b29-viruses-02-02541],[@b30-viruses-02-02541]\]. Some evidence suggests that TLR-4, TLR-6 and TLR-2 play a role in recognition of RSV \[[@b31-viruses-02-02541],[@b32-viruses-02-02541]\] while MeV hemagglutinin is recognized by TLR-2 \[[@b33-viruses-02-02541]\] ([Figure 1](#f1-viruses-02-02541){ref-type="fig"}).
3.2.. The RLR System
--------------------
The cytosolic mediators of viral sensing, the RLRs, include the RIG-I and melanoma differentiation-associated gene 5 (MDA5). RIG-I is activated by ssRNA or 5′-triphosphate double stranded RNA (dsRNA) and MDA5 by dsRNA \[[@b34-viruses-02-02541]--[@b36-viruses-02-02541]\]. Both MDA5 and RIGI signal through the mitochondrial-associated protein known as interferon beta promoter stimulator-1 (IPS-1) \[[@b36-viruses-02-02541]--[@b40-viruses-02-02541]\] ([Figure 1](#f1-viruses-02-02541){ref-type="fig"}). The role of a third member of the RLR family, the RNA helicase Lgp2, is less understood. Lgp2 has been implicated both as a negative and positive regulator of MDA5 and RIG-I function \[[@b41-viruses-02-02541]--[@b43-viruses-02-02541]\].
3.3.. Nod-like Receptor (NLRP3) Inflammasome
--------------------------------------------
The inflammasome is a protein complex composed of a number of proteins, among them caspase-1 and different Nod-like receptors (NLR)s. The main inflammasome complex involved in the response to the RNA viruses discussed in this review is NLRP3.
The inflammasome complex is required to generate the active form of the cytokines IL-1β, IL-18 and IL-33. The production of these cytokines requires two signals. Signal one is given by recognition of viral RNA as described above. This leads to increased levels of cytokines mRNA. Signal two activates the inflammasome and is sensed by NLRP3. NLRP3 is activated after exposure to ATP, dsRNA, poly I:C and various crystals such as monosodium urate \[[@b44-viruses-02-02541]--[@b47-viruses-02-02541]\].In order to produce activated cytokines, pro-IL-1β, pro-IL-18 and pro-IL-33 must be cleaved by caspase-1. Caspase-1 is part of the inflammasome complex that contains NLRP3 and the adapter apoptosis-associated speck-like protein containing a CARD (ASC) \[[@b48-viruses-02-02541]\]. It has been shown that the NLRP3 inflammasome is required for the production of IL-1β and IL-18 during influenza infection *in vivo* \[[@b49-viruses-02-02541]\]. It remains unclear whether the inflammasome physically senses these compounds. Recently it was suggested that influenza virus M2, an ion channel, causes changes in ionic concentration in cellular compartments which lead to NLRP3 activation \[[@b50-viruses-02-02541]\].
4.. Production and Signaling of Type I and III IFNs in Response to Virus Infection
==================================================================================
The first indication of an immune response to virus infection is the secretion of type I IFNs. Type I IFNs belong to a family of cytokines consisting of one subtype of IFN-β, 13 subtypes of IFN-α and also IFN-ω, IFN-κ, IFN-ε and IFN-ν. Type III IFNs (IFN-λ) are also produced quickly after infection, and although their function and regulation is less studied than that of type I IFNs, they share similar functions. The existence of multiple IFN-α genes and the fact that virtually all viruses encode proteins that antagonize the production or response to type I IFNs emphasizes their importance during the anti-viral immune response. As we will discuss in more detail later in this review, type I IFNs secretion is delayed *in vivo* until a few days after infection and is coincident with the end of the incubation period.
4.1.. Transcriptional Regulation of Type I IFNs
-----------------------------------------------
The transcriptional regulation of type I IFNs has been comprehensively reviewed \[[@b51-viruses-02-02541]\]. In short, IFN-β is the first type I IFN to be induced following viral recognition. Transcription of IFN-β mRNA requires binding of three groups of transcription factors to the regulatory domain of the IFN promoter; NFκB, activating transcription factor 2 (ATF2)/c-Jun and interferon regulatory factors 3 and 7 (IRF-3 and IRF-7). The activation of all these factors in response to virus infection is induced by triggering either the RLR or TLR systems ([Figure 1](#f1-viruses-02-02541){ref-type="fig"}).
4.2.. Type I IFNs Signaling
---------------------------
Type I IFNs signaling through its receptor leads to transcription of many interferon responsive genes (ISGs) that limit the virus replication and enhance the immune response. Secreted type I IFNs signal through the IFN-α/β receptor complex (IFNAR), composed of two transmembrane protein subunits, IFNAR1 and IFNAR2, which are present on the surface of every nucleated cell. Sensing of type I IFNs can enhance the production of type I IFNs and other inflammatory cytokines \[[@b52-viruses-02-02541],[@b53-viruses-02-02541]\]. The dimerization of the two subunits of the IFNAR with IFN-α or IFN-β leads to activation of the intracellular kinases Jak1 and Tyk2, which phosphorylate the STAT transcription factors leading to the generation of STAT homodimers (STAT1) and heterodimers (STAT1 with STAT2). Phosphorylated STAT1 and STAT2, together with IRF-9, form a complex called interferon-stimulated gene factor 3 (ISGF3) that translocates to the nucleus and activates the transcription of ISGs \[[@b54-viruses-02-02541]\] ([Figure 1](#f1-viruses-02-02541){ref-type="fig"}).
4.3.. Type III IFNs
-------------------
Similarly to type I IFNs, type III IFNs (IFN-λ), which in humans include IL-29, IL-28α and IL-28β, are expressed by many cell types after virus infection or TLR ligand stimulation and have similar effects to those observed with type I IFNs \[[@b55-viruses-02-02541],[@b56-viruses-02-02541]\]. The receptor for IFN-λ (IFN-λR) is composed by one IFN-λR chain and one IL-10Rβ chain. IFN-λR also signals through the JAK-STAT pathway \[[@b57-viruses-02-02541]--[@b59-viruses-02-02541]\]. Expression of IFN-λR appears to be restricted to non-hematopoietic cells such as epithelial cells.
5.. Inhibition of Innate Immunity by Viral Antagonists
======================================================
Given that mammals have evolved a sophisticated detection and response system to viral infections, viruses have adapted to inhibit the initial recognition by the host's immune system. Once the anti-viral response is initiated by type I IFNs signaling, it is rapidly amplified, and thus it is of great importance for the virus to delay this response as long as possible.
5.1.. Inhibition of Interferon Induction
----------------------------------------
Viruses have evolved to inhibit IFN induction in a number of ways; the many functions of the influenza A non-structural protein 1 (NS1) have been recently reviewed \[[@b60-viruses-02-02541]\]. Influenza NS1 inhibits RIG-I and IPS-1 signaling by forming a complex with RIG-I and ssRNA \[[@b34-viruses-02-02541],[@b61-viruses-02-02541]--[@b63-viruses-02-02541]\]. This explains the inhibition of IRF-3, NFκB, and c-Jun/ATF-2 activation observed upon infection with influenza viruses \[[@b64-viruses-02-02541]--[@b66-viruses-02-02541]\]. In addition, influenza A NS1 blocks virus detection by binding to dsRNA, thereby masking it from detection by RIG-I \[[@b67-viruses-02-02541],[@b68-viruses-02-02541]\]. Influenza NS1 also inhibits the cellular response by interfering with the processing and export of cellular mRNA \[[@b69-viruses-02-02541],[@b70-viruses-02-02541]\].
The paramyxoviruses' ability to inhibit IFN has been reviewed elsewhere \[[@b71-viruses-02-02541]\]. In brief, Sendai virus (SeV), MeV and Mumps virus (MuV) viruses V protein can block the activation of MDA5 \[[@b72-viruses-02-02541]--[@b75-viruses-02-02541]\]. Several V proteins of paramyxoviruses can inhibit IRF-3 activation \[[@b76-viruses-02-02541]\], for example RSV NS1 and NS2 also block IRF3 activation \[[@b77-viruses-02-02541]\]. RSV NS2 can block type I IFN induction by binding RIG-I and inhibiting downstream signaling \[[@b78-viruses-02-02541]\].
TLR agonists are potent inducers of cytokine production. It is, therefore, surprising that very little evidence exists for inhibition of the TLR signaling pathway by the viruses discussed above. It has been suggested that certain RSV strains and MeV can inhibit type I IFN induction by TLR-7 and TLR-9 signaling. In the case of MeV, the V protein acts as a decoy substrate for the kinase IκB kinase α, competing with IRF7 \[[@b79-viruses-02-02541]--[@b81-viruses-02-02541]\]. No evidence exists for such inhibition by Influenza, HRV or human parainfluenza virus (hPIV) ([Figure 1](#f1-viruses-02-02541){ref-type="fig"}).
5.2.. Inhibition of Type I IFN Signaling
----------------------------------------
Respiratory paramyxoviruses can inhibit the IFN signaling pathway. The C protein of hPIV1 inhibits the translocation of STAT-1 and STAT-2 to the nucleus and the activation of IRF-3 \[[@b82-viruses-02-02541],[@b83-viruses-02-02541]\], while the C protein of hPIV3 inhibits the phosphorylation of STAT-1 \[[@b84-viruses-02-02541]\]. Some evidence suggests the C protein of MeV acts to inhibit IFN signaling response \[[@b85-viruses-02-02541]\]. The V protein of MeV appears to form complexes with different signaling proteins in the IFN response pathway preventing either nuclear translocation or their phosphorylation \[[@b86-viruses-02-02541]--[@b91-viruses-02-02541]\]. The NS1 and NS2 proteins of RSV can both block type I IFN and IFN-λ responses \[[@b92-viruses-02-02541],[@b93-viruses-02-02541]\]. It is thought that STAT2 is actually degraded by NS1 and NS2 \[[@b94-viruses-02-02541]\] ([Figure 1](#f1-viruses-02-02541){ref-type="fig"}).
Many of the proteins involved in viral recognition, type I and III IFN induction and type I IFN signaling, such as RIG-I, MDA-5, IRF7, STAT1, *etc.*, are themselves type I IFN inducible genes. By blocking IFN induction and signaling the virus also limits the enhancement of the response to infection.
6.. Control of the Length of the Incubation Period *in vivo*
============================================================
As discussed above, influenza NS1 inhibits the detection of the virus by the host thereby preventing the production of type I IFNs and other cytokines \[[@b95-viruses-02-02541]\]. The inhibition of type I IFN production is of particular importance, since the sensing of type I IFN by neighboring cells generates an anti-viral state in these cells that limits virus propagation. Studies describing the viral proteins required for respiratory virus antagonism are limited to *in vitro* experiments, in most cases due to poor replication of antagonist deficient viruses *in vivo*. While it is difficult to extrapolate these observations to the events taking place during a natural infection, studies of influenza NS1 antagonism *in vivo* provide a model for respiratory virus inhibition of innate immunity.
6.1.. Influenza NS1 Antagonism in vivo
--------------------------------------
A close examination of an *in vivo* influenza virus infection in mice showed that the virus replicates in the lung for almost two days without inducing an innate immune response. We defined this period between early, undetected virus infection and the first signs of an immune response as the "stealth phase". Our group showed that the NS1 protein of influenza is responsible for the "stealth phase" by hampering cytokine production *in vivo*. Infection with a virus lacking NS1 triggers an immediate vigorous lung inflammation \[[@b96-viruses-02-02541]\]. Two days after infection with an NS1 competent virus, a robust and abrupt immune response is initiated in the infected lungs. This event demarcates the initiation of innate immunity. The lung innate response includes the production of cytokines (e.g., IL-6, TNF-a, type I IFNs, IFN-γ and IL1-α chemokines (e.g., CCL-2, CCL-20 and KC), the recruitment of diverse cells of the immune system, and the migration of dendritic cells (DCs) to the draining lymph nodes leading to the triggering of T cell responses. This abrupt rise of chemokine is responsible for the recruitment of mononuclear phagocytes, granulocytes and other leukocytes to the site of infection. These recruited cells will play a major role in the eventual clearance of the virus.
6.2.. Overcoming Viral Antagonism in vivo
-----------------------------------------
Based on several studies, there are a number of possible mechanisms by which the immune system can be stimulated to initiate inflammation.
### 6.2.1.. Cell Death
In an inflamed tissue, the sensing of "danger signals" \[[@b97-viruses-02-02541]\] in the form of factors released from infected necrotic or apoptotic cells may stimulate neighboring cells to produce cytokines and chemoattract other immune cells from the blood \[[@b98-viruses-02-02541],[@b99-viruses-02-02541]\]. Viral RNA released from dying cells may stimulate TLR-7 or TLR-3 upon phagocytosis by plasmacytoid DCs (pDCs), macrophages and other cells culminating in type I IFNs production \[[@b100-viruses-02-02541]\]. The TLR system avoids viral antagonism by rapidly sensing the virus inside endosomal compartments in uninfected phagocytes that culminates in type I IFNs and cytokine production \[[@b101-viruses-02-02541],[@b102-viruses-02-02541]\]
### 6.2.2.. Errors in Virus Replication
Intracellular purine metabolites are released from damaged cells and include uric acid and ATP, which can stimulate the inflammasome complex to cleave pro-IL-1β and intensify the innate response \[[@b47-viruses-02-02541],[@b103-viruses-02-02541]--[@b109-viruses-02-02541]\]. The inflammasome can also activate type I IFNs production \[[@b45-viruses-02-02541]\] and type I IFNs itself can upregulate AIM2, a protein that contains a pyrin motif that is necessary for promoting IL-1β production. IFN-γ is also involved in this signaling cascade since it can upregulate components of the inflammasome complex \[[@b110-viruses-02-02541],[@b111-viruses-02-02541]\].
The natural process by which viruses replicate may contribute to the culmination of the stealth phase. The viral polymerase of many of the viruses discussed here is error-prone. From an evolution or natural selection standpoint, this property has the advantage of promoting rapid mutations in the viral genome, avoiding recognition by the adaptive immune response. However, it is also possible that such a process promotes mistakes in viral replication, such as the generation of mutated, less efficient viral antagonists and defective interfering virus particles (DIs) that may lead to immune recognition. It has been shown in mice that stocks of SeV with high DI content enhance the immune response \[[@b112-viruses-02-02541]\] and MeV vaccine strains induction of type I IFNs correlates with high DI content \[[@b113-viruses-02-02541]\].
### 6.2.3.. Priming by Type I IFNs
*In vitro* studies show that cells primed with type I IFNs are able to mount an innate response to an infecting virus, despite viral antagonism. Type I IFN signal transduction turns on transcriptional programs within cells that can decrease the inhibitory effects of the viral antagonists upon infection. Not only does the virus replicate poorly in cells primed with type I IFNs, but also primed cells can secrete pro-inflammatory cytokines more efficiently. It is known that pre-exposure of DCs to type I IFNs upregulates costimulatory molecules and major histocompatibility class I and II (MHCI and MHCII) molecules, improving their function as antigen presenting cells \[[@b114-viruses-02-02541]--[@b116-viruses-02-02541]\].
*In vivo*, lung secreted cytokines and chemokines also promote systemic awareness to the virus infection. Type I IFNs can signal to developing leukocytes and memory T cells in primary and secondary lymphoid organs such as the bone marrow and spleen to acquire an anti-viral state and enhance their function \[[@b117-viruses-02-02541],[@b118-viruses-02-02541]\]. Such an anti-viral state is thought to functionally improve cells of the immune system before they infiltrate the lungs. Type III IFNs are also induced after respiratory virus infection and likely limit virus spread in epithelial cells \[[@b119-viruses-02-02541]\]. Therefore, type III IFN might be induced at the end the of the stealth phase complementing the function of type I IFN. The speed at which this process occurs is controlled by the ability of the virus to suppress inflammation. This observation points out that immune modulation by the pathogen not only targets local lung immunity but also the external intervention of pre-programmed leukocytes with advantageous antiviral machinery.
Finally, the multifaceted inflammatory response can also affect non-hematopoietic cells such as uninfected epithelial cells, protecting them from infection and allowing a more vigorous response upon stimulation.
7.. Viral Antagonism Delays the Initiation of Adaptive Immune Response
======================================================================
In close contact to the epithelial border is a tight network of lung DCs \[[@b120-viruses-02-02541]\] that sense viruses and migrate along a CCR7-mediated chemokine gradient \[[@b121-viruses-02-02541]\] to the lung draining mediastinal lymph nodes (MLNs). In the MLNs, the DCs trigger the proliferation and differentiation of virus-specific T cells \[[@b96-viruses-02-02541],[@b122-viruses-02-02541]--[@b124-viruses-02-02541]\]. Activated virus-specific effector T cells will eventually circulate back to the bloodstream and are then recruited to the respiratory tract to terminate the infection and clear the virus \[[@b125-viruses-02-02541],[@b126-viruses-02-02541]\].
Studies tracking DC migration from the lung to the MLNs during influenza infection using fluorescent reagents that induce unspecific inflammation have shown that DCs migrate from the lung to the MLNs rapidly \[[@b127-viruses-02-02541]--[@b129-viruses-02-02541]\]. It is likely that the viral antagonist is unable to inhibit the inflammation triggered by these inflammation inducing fluorescent reagents. However, when no inflammatory agent is present in the tracking reagent, the kinetics of DC migration from the lung to the MLNs during influenza virus infection is quite slow and correlates with the termination of the "stealth phase". DC migration begins around two days after infection when small numbers of DCs carrying viral antigens are first seen in the MLNs and reach a plateau around 3--4 days after infection \[[@b96-viruses-02-02541],[@b122-viruses-02-02541],[@b124-viruses-02-02541],[@b130-viruses-02-02541]\]. Therefore, inhibiting inflammation for two days not only affects innate immunity but also delays the initiation of adaptive immunity.
8.. Conclusions
===============
The incubation period is a helpful definition that describes the time between virus infection and the onset of symptoms. Based on new findings, we propose a model that describes a mechanism of the delayed symptoms (innate immune response) that is likely common to almost all known respiratory virus infections ([Figure 2](#f2-viruses-02-02541){ref-type="fig"}). The delayed rise of the innate immune response to a respiratory virus is explained by the suppression of immunity by the viral antagonist *in vivo*. The "stealth phase" is terminated by an initiating event or breakthrough that triggers type I IFN and other cytokines that serve to stimulate cells before they are infected. Type I IFN primed cells are protected from viral antagonism allowing the innate immune response to proceed. Much work must still be done to determine the factors, the sequence of events, and cell types that are relevant to accomplish the end of the incubation period.
The authors wish to thank Karla Tapia for assistance with graphic design and to Sharon Czelusniak for reading the manuscript. This work was supported by NIH/NIAID grants AI041111 and AI082970 to T.M.M. and grants A1083481 and A1083284 to C.B.L.
{#f1-viruses-02-02541}
{#f2-viruses-02-02541}
[^1]: These authors contributed equally to this work.
| {
"pile_set_name": "PubMed Central"
} |
Introduction {#s0001}
============
Inflammation is a complex process mediated by the activation of various immune cells. Many studies have shown that inflammation is associated with various human diseases, including cancer (Pan and Ho [@CIT0030]). Macrophages play an important role in various inflammatory responses by upregulating the expression of pro-inflammatory cytokines and enzymes such as tumour necrosis factor-α (TNF-α), interleukin, inducible nitric oxide synthase (iNOS) and cyclooxygenase-2 (COX-2) (Larsen and Henson [@CIT0014]; Lawrence et al. [@CIT0015]; Sica et al. [@CIT0040]). Mechanistically, lipopolysaccharide (LPS), a component of the cell wall of Gram-negative bacteria, interacts with toll-like receptor 4 (TLR4), and triggers the activation of monocytes and macrophages involved in infection response (Aderem and Ulevitch [@CIT0001]; Saluk-Juszczak and Wachowicz [@CIT0036]; Takeda and Akira [@CIT0046]; Pan et al. [@CIT0029]). Low levels of NO have many biological functions, including neurotransmitters, vascular homoeostasis and wound repair and have antimicrobial activity against bacterial pathogens. NO can be synthesized from [L]{.smallcaps}-arginine by a family of NO synthases (NOS). An inducible isoform of NOS (iNOS) is only expressed after exposure to pro-inflammatory conditions. Once expressed, iNOS generates large amounts of NO, which plays an important role in acute and chronic inflammation (Denlinger et al. [@CIT0005]; Weisz et al. [@CIT0048]). iNOS is widely expressed in various cells, including vascular smooth muscle cells, hepatocytes and Kupffer cells and is highly expresses in LPS-activated macrophages (Rockey et al. [@CIT0035]). Cyclooxygenase-2 (COX-2) is another inducible enzyme that catalyses the biosynthesis of prostaglandins (PGEs), particularly PGE~2~, which contributes to pathogenesis of various inflammatory diseases, invasion, angiogenesis, and tumour growth (Claria [@CIT0004]). COX-2 is also pre-eminently expressed in inflammatory cells stimulated by LPS, pro-inflammatory cytokines and tumour promoters (Meric et al. [@CIT0021]).
Ceramides are a type of sphingolipid and consist of a sphingoid base and a saturated fatty acid moiety. Ceramides are present as a dominant lipid in the stratum corneum (SC), the most upper layer of the epidermis of the skin and play a crucial role in its water-holding and barrier function (Takeda et al. [@CIT0047]). Until recently, more than 12 types of ceramide have been designated in human SC (Masukawa et al. [@CIT0019]). Ceramides in the epidermis are synthesized by several enzymes such as serine palmitoyltransferase (SPT) (Hanada [@CIT0010]), ceramide synthase (CerS) (Levy and Futerman [@CIT0016]), glucosylceramide synthase (GCS) (Hamanaka et al. [@CIT0009]), β-glucocerebrosidese (GBA) (Takagi et al. [@CIT0045]), sphingomyelin synthase (SMS) (Tafesse et al. [@CIT0044]), and acid sphingomyelinase (ASM) (Jenkins et al. [@CIT0011]). SPT and CerS are involved in the *de novo* synthesis of ceramides. SPT catalyses the condensation of serine and palmitoyl-CoA as the first step of *de novo* synthesis. The term \`Filaggrin\' (derived from \`filament-aggregating protein\') was first coined in 1981 to describe a class of structural protein that had been isolated from the stratum corneum (Steinert et al. [@CIT0042]). By aggregating keratin filaments into keratin fibrils within the cytoskeleton of corneocytes, filaggrin is responsible for the mechanical strength and integrity of the stratum corneum (O\'Regan et al. [@CIT0025]). Filaggrin has also been designated as a natural moisturizing factor protein that contributes to the permeable barrier as an aggregated particle comprised of profilaggrin (Kezic et al. [@CIT0012]).
Plants that have been used worldwide for a long time in traditional medicine have been constantly reviewed as resources for the development of new drugs to control various diseases. Among them, *Citrus unshiu* Markovich (Rutaceae) peel is the dried skin of the *Citrus unshiu* fruit, which is produced primarily in Jeju Island, Korea, and in the southern regions of China and Japan. *Citrus unshiu* peel and dried peels have been used as traditional medicines to treat common colds, bronchial discomfort, and indigestion and have been reported to possess pharmacological effects on inflammation, allergies, diabetes, and viral infections (Suzuki et al. [@CIT0043]; Oh et al. [@CIT0026]; Park et al. [@CIT0032], [@CIT0033]; Min et al. [@CIT0022]). Here, we have investigated an herbal agent that can help skin moisturization and anti-inflammation by using fermented dried *Citrus unshiu* peel extracts. In order to use natural products as raw materials for cosmetics, researchers have primarily investigated potential to improve skin moisturization and block inflammation (Kim et al. [@CIT0013]; Eom et al. [@CIT0007]; Spilioti et al. [@CIT0041]; Shen et al. [@CIT0039]). However, the effects of fermented *Citrus unshiu* peel extracts on RAW 264.7 mouse macrophage-mediated inflammation and moisturizing effect in HaCaT keratinocytes remain unknown.
In our study, we attempted to elucidate the anti-inflammatory potential of fermented *Citrus unshiu* peel extracts in LPS-stimulated RAW 264.7 macrophages. Here, we evaluated the inhibitory effect of fermented *Citrus unshiu* peel extracts on inflammatory biomarkers such as NO and PGE~2~ production, expression of iNOS, COX-2, and pro-inflammatory cytokines in LPS-stimulated RAW 264.7 cells. Furthermore, fermented *Citrus unshiu* peel extracts can stimulate the production of hyaluronic acid and expression of filaggrin and SPT in HaCaT keratinocyte cells, which is an indicative of moisturizing effect.
Materials and methods {#s0002}
=====================
Reagents {#s0003}
--------
LPS (*Escherichia coli* 055:B5), 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT), Tris base, glycine, NaCl, sodium dodecylsulphate (SDS), Griess reagent and bovine serum albumin (BSA) were purchased from Sigma-Aldrich (St. Louis, MO, USA). RPMI 1640, foetal bovine serum (FBS) and antibiotic-antimycotic mixture were obtained from Thermo Fisher Scientific Inc. (Waltham, MA, USA). Trypan blue vital stain (0.4%) was obtained from Life Technologies (Grand Island, NY, USA). PGE~2~, TNF-α, IL-6 ELISA kits were obtained from R&D Systems (Minneapolis, MN, USA). Anti-COX-2 and anti-iNOS antibodies were obtained from BD Biosciences (San Diego, CA, USA). Anti-filaggrin, anti-serine palmitoyltransferase (SPT), anti-β-actin, and horseradish peroxidase (HRP)-conjugated secondary antibodies were obtained from Santa Cruz Biotechnology (Santa Cruz, CA, USA).
Isolation of *Bacillus subtilis* from Korean fermented soybean (cheonggukjang) {#s0004}
------------------------------------------------------------------------------
Samples of Cheonggukjang (1 mL), serially diluted, were plated on tryptic soy broth (TSB) agar (Becton-Dickinson, Franklin Lakes, NJ, USA) and placed at 30 °C for 24 h under anaerobiosis to isolate presumptive mesophilic *Bacillus subtilis*. At least 10 colonies, possibly with different morphology, were isolated on the TSB plates. Two *Bacillus subtilis* were identified by TLC on activated silica gel plates using *n*-butanol: acetic acid: water (5:2:2, v/v/v). Isolated *Bacillus subtilis* strains were named '1-4' and '2-1', respectively.
Preparation of FCU using *Bacillus subtilis* {#s0005}
--------------------------------------------
The isolated *Bacillus subtilis* strains ('1-4' and '2-1') were cultured in tryptic soy broth (TSB) at 37 °C. For the production of fermented dried *Citrus unshiu* peel, we first added water to the dried *Citrus unshiu* peel at a ratio of 1:20 (w/v) and heated for 2 h at 95 °C. After autoclaving at 121 °C for 15 min, a dried *Citrus unshiu* peel extracts were obtained and the *Bacillus subtilis* ('1-4' and '2-1') culture broth was added to the extracts at a concentration of 10% (v/v), which was well mixed and then incubated for 96 h at 37 °C. The fermented extracts were filtered using an aspirator, water and ethanol were added, and filtration was performed again. The supernatant was then dried and stored at 4 °C. The intact extracts (control) had no *Bacillus subtilis*. The sample table for each fermentation condition is shown in [Table 1](#t0001){ref-type="table"}.
######
Sample table for each fermentation condition.
Name of sample *Bacillus subtilis* Filtrate solution
---------------- --------------------- -------------------
WE None H~2~O
WE(1-4) 1-4 strain H~2~O
WE(2-1) 2-1 strain H~2~O
AL None EtOH
AL(1-4) 1-4 strain EtOH
AL(2-1) 2-1 strain EtOH
Cell lines {#s0006}
----------
The RAW 264.7 macrophage and HaCaT cell lines were obtained from Korean Cell Line Bank (KCLB, Seoul National University College of Medicine, 28 Yongon-dong, Chongno-gu, Seoul 110-744, Korea). These cells were maintained at subconfluence in a 95% air, 5% CO~2~ humidified atmosphere at 37 °C. RAW 264.7 cells were cultured in RPMI 1640 medium containing 10% FBS. HaCaT cells were cultured in Dulbecco's Modified Eagle's medium (DMEM) containing 10% FBS. All media were also supplemented with 100 U/mL of penicillin and 100 μg/mL of streptomycin.
MTT assay {#s0007}
---------
Cell viability was measured by an MTT assay to detect NADH-dependent dehydrogenase activity. Thirty microliters of MTT solution (2 mg/mL) in 1× phosphate-buffered saline (PBS) was directly added to the cells, which were then incubated for 3 h to allow MTT to metabolize to formazan. Absorbance was measured with an automated spectrophotometric plate reader at a wavelength of 570 nm. Cell viability was normalized as relative percentages in comparison with untreated controls.
Nitrite assay {#s0008}
-------------
The RAW 264.7 macrophage cells were plated at a density of 2 × 10^5^ cells per well in a 24-well plate. The cells were pre-treated with the indicated concentrations of various extracts for 2 h, and then induced with a 1 μg/mL concentration of LPS for an additional 22 h. Nitrite accumulation in the culture was measured colorimetrically by the Griess reaction using a Griess reagent (Sigma-Aldrich, St. Louis, MO, USA). For the assay, equal volumes of cultured medium and Griess reagent were mixed, and the absorption coefficient was calibrated using a sodium nitrite solution standard (Sigma-Aldrich, St. Louis, MO, USA). The absorbance of each sample after the Griess reaction was determined by an ELISA plate reader at 540 nm.
Measurement of PGE~2~ release by the RAW 264.7 macrophage cells {#s0009}
---------------------------------------------------------------
The RAW 264.7 macrophage cells were plated at a density of 2 × 10^5^ cells per well in a 24-well plate. The cells were pre-treated with the indicated concentrations of WE(2-1) and AL(2-1) for 2 h, and then induced with 1 μg/mL of LPS for an additional 22 h. The level of PGE~2~ production from endogenous arachidonic acid metabolism was measured in cell culture supernatants of the RAW 264.7 cells by enzyme-linked immunosorbent assay (ELISA) kit (R&D Systems, Minneapolis, MN, USA).
Measurement of pro-inflammatory cytokines (TNF-α and IL-6) production {#s0010}
---------------------------------------------------------------------
The inhibitory effect of WE(2-1) and AL(2-1) on the production of proinflammatory cytokines (TNF-α and IL-6) from LPS-treated RAW 264.7 cells was determined. The supernatants were subsequently employed for the proinflammatory cytokine assays using a mouse enzyme-linked immunosorbent assay (ELISA) kit (R&D Systems, Minneapolis, MN, USA).
Real-time PCR {#s0011}
=============
Real-time quantitative PCR was performed using the Universal SYBR Green Master Mix (Applied Bio systems, USA). Amplification of the cDNA was performed as follow: 95 °C for 15 min followed by 40 cycles at 95 °C for 30 s, at 59 °C for 30 s, and at 72 °C for 30 s. The real-time PCR analysis was performed on an Applied Bio-systems StepOne system (Applied Bio-systems, USA). In this study, quantification based on the relative expression of a target gene versus GAPDH gene (2^−ΔΔCt^) was utilized to determine the level of mRNA expression.
Reverse transcription polymerase chain reaction (RT-PCR) {#s0012}
--------------------------------------------------------
Cells were washed and suspended in Trizol reagent. Total RNA was extracted according to the manufacturer's instructions (Invitrogen, Life Technologies, Carlsbad, CA, USA). One microgram of total RNA was converted to cDNA by superscript reverse transcriptase and then amplified by a Taq polymerase using reverse transcription polymerase chain reaction (RT-PCR) (TAKARA, Tokyo, Japan). The relative expression of iNOS, COX-2, filaggrin, and SPT were analyzed using PCR with glyceraldehyde-3-phosphate dehydrogenase (GAPDH) as an internal control. The following pairs of forward and reverse primer sets were used: iNOS, 5′-TCTTTGACGCTCG-GAACTGTAGCA-3′ and 5′-CGTGAAGCCATGACCTTTCGCATT-3′. COX-2, 5′-TTGCTGTACAAGCAGT-GGCAAAGG-3′ and 5′-AGGACAAACACCGGAGGGAATCTT-3′. Filaggrin, 5′-CAAATCCTGAAGAATCCAGATGAC-3′ and 5′-TGCTTGAGCCAACTTGAATACC-3′, serine-palmitoyltransferase, 5′-TTTCCGGTTTAAAAGTGGTG-3′ and 5′-CTGATGCTTGGAGGAGGAAG-3′. The cDNA reaction was performed at 45 °C for 60 min and 95 °C for 5 min. PCR products were run on 1% agarose gel and then stained with Loading Star (Dynebio, Seongnam, Korea). Stained bands were visualized under UV light and photographed.
Western blotting {#s0013}
----------------
After the cells were treated with the indicated concentrations of WE(2-1) and AL(2-1), the cells were lysed and the total protein concentrations were determined by Bradford reagent (Bio-Rad, Hercules, CA, USA). Equal amounts of lysates resolved on sodium dodecyl-polyacrylamide gel electrophoresis (SDS-PAGE) were transferred to a nitrocellulose membrane, and the membrane was blocked with 1× TBS containing 0.1% Tween 20 and 5% skim milk or 2% BSA for 1 h at room temperature. After the blocking, the membranes were incubated overnight at 4 °C with the respective primary antibodies. The membranes were washed twice and incubated with diluted horseradish peroxidase (HRP)-conjugated secondary antibodies (1:10000) for 1 h at room temperature. After three washes, the membranes were detected using an enhanced chemiluminescence (ECL) kit (Millipore, Bedford, MA, USA).
Monitoring of cell growth with the RTCA DP instrument {#s0014}
-----------------------------------------------------
Cell growth behavior was continuously monitored for 120 h using the xCELLigence RTCA DP Instrument (Roche Diagnostics GmbH, Germany). Background impedance was measured in 100 μL cell culture medium per well. The final volume was adjusted to 200 μL cell culture medium, including 5 × 10^3^ cells/well. After plating, impedance was recorded in 15 min intervals. All experiments were performed in triplicates. Cell Index (CI) values were normalized to the time point of indicated concentration of WE(2-1) and AL(2-1) administration (referred to as normalized CI).
Hyaluronic acid measurement {#s0015}
---------------------------
The cellular hyaluronic acid levels were measured by Hyaluronan Quantikine ELISA Kit (R&D Systems) according to the manufacturer's protocol.
Analysis of flavonoids in WE(2-1) and AL(2-1) {#s0016}
---------------------------------------------
The chemical profiles of the WE(2-1), AL(2-1) and the chemical changes of constituents were monitored by the liquid chromatography-mass spectrometry system consisted of a Thermo Scientific Vanquish UHPLC system (Thermo Fisher Scientific, Sunnyvale, CA, USA) with a Shim-pack GIS-ODS (3 µm, 3.0 × 100 mm, Shimadzu) and a Triple TOF 5600+ mass spectrometer system (Triple TOF MS; QTOF, SCIEX, Foster City, CA, USA). The QTOF MS, equipped with a Duospray™ ion source, and was used to complete the high-resolution experiment. The LC gradient used a mobile phase a containing 0.1% formic acid in water and a mobile phase B of acetonitrile with 0.1% formic acid. The flow rate was kept constant at 0.8 mL/min and the injection volume was 2 μL. The gradient elution system was as follows: 5--100% B for 12 min. Authentic standards; narirutin, narigenin, rutin, hesperidin, hesperetin, neoponcirin, tangeretin, and nobiletin were purchased from Chemface (Wuhan, China). HPLC-grade acetonitrile (ACN) and methanol (MeOH) were purchased from Honeywell Burdick & Jackson (Morristown, NJ, USA). Analytical-grade formic acid (99% purity) was obtained from Sigma-Aldrich (St. Louis, MO, USA). The chemical profiles of six extracts were analyzed and compared using LC-QTOF. Eight major bioactive components were identified with chromatographic information (retention time, mass; Da, and fragmentation ions) of authentic standards. Data acquisition and processing were carried out using Analyst TF 1.7, PeakVeiw 2.2 and MasterView software (SCIEX, Foster City, CA, USA).
LC: Liquid chromatography, QTOF: Quadurpole Time-of-Flight mass spectrometry.
Statistical analysis {#s0017}
--------------------
All numeric values are represented as the mean ± SE. Statistical significance of the data compared with the untreated control was determined using the Mann--Whitney U-test. Significance was set at *p* \< 0.05.
Results {#s0018}
=======
Cytotoxic effect of the six FCU in RAW 264.7 macrophages {#s0019}
--------------------------------------------------------
We investigated for the first time the anti-inflammatory effects of FCU in RAW 246.7 macrophages. Here, we examined the cytotoxic effects of six FCU and cell viability was assessed using the MTT assay. We found that WE, WE(1-4), AL, and AL(1-4) had no effect on the cytotoxicity except for WE(2-1) and AL(2-1) at a concentration of 10 or 100 μg/mL ([Figure 1(A)](#F0001){ref-type="fig"}).
{#F0001}
Inhibition of LPS-induced NO production by six fermented extracts in RAW 264.7 macrophages {#s0020}
------------------------------------------------------------------------------------------
In order to investigate the anti-inflammatory effects of six fermented extracts, we first investigated its effects on nitrite production in LPS-stimulated RAW 264.7 macrophages. The effects of six fermented extracts on LPS-induced NO production in RAW 264.7 macrophages were investigated by measuring the accumulated nitrite in the culture medium as estimated by the Griess reaction. Cells were pre-treated with the indicated concentrations of six fermented extracts for 2 h and then induced with LPS (1 μg/mL) for 22 h. LPS-treated cells significantly increased the concentration of NO. As shown in [Figure 1(B)](#F0001){ref-type="fig"}, in cells which were pre-treated with various concentrations of six fermented extracts and also together with 1 μg/mL of LPS for 22 h, significant concentration-dependent inhibition of nitrite production was found in WE(2-1) and AL(2-1) treated cells.
Inhibition of LPS-induced PGE~2~ secretion by WE(2-1) and AL(2-1) in RAW 264.7 cells {#s0021}
------------------------------------------------------------------------------------
To examine the potential anti-inflammatory properties of WE(2-1) and AL(2-1) on LPS-induced prostaglandin E~2~ (PGE~2~) production in RAW 264.7 cells, cells were pre-treated with or without WE(2-1) (1, 3, 5, 10 μg/mL) and AL(2-1) (10, 30, 50, 100 μg/mL) for 2 h and then stimulated with LPS (1 μg/mL) for 24 h. PGE~2~ concentrations were measured in the culture supernatants by the ELISA assays. We found that WE(2-1) and AL(2-1) substantially suppressed LPS-induced PGE~2~ production in a concentration-dependent manner ([Figure 2(A)](#F0002){ref-type="fig"}).
{#F0002}
Inhibition of LPS-induced pro-inflammatory cytokines by WE(2-1) and AL(2-1) in RAW 264.7 cells {#s0022}
----------------------------------------------------------------------------------------------
Pro-inflammatory cytokines such as TNF-α, and IL-6 play important roles in immune responses to a variety of inflammatory stimuli. Therefore, the inhibitory effect of WE(2-1) and AL(2-1) on TNF-α, and IL-6 in LPS-stimulated RAW 264.7 cells were evaluated using ELISA kits. The treatment of RAW 264.7 cells with LPS alone resulted in significant increases in these cytokine production, however, WE(2-1) and AL(2-1) pre-treated cell repressed LPS-induced TNF-α, and IL-6 production in a concentration-dependent manner ([Figure 2(B and C)](#F0002){ref-type="fig"}). When RNA was isolated and quantitative real-time PCR was performed to examine the effects of WE(2-1) and AL(2-1) on gene expression, treatment of RAW264.7 cells with LPS alone showed a significantly increases in TNF-α and IL-6 mRNA expression. In the co-incubation of WE(2-1) with LPS, TNF-α expression level was found to decrease from 7.9- to 7.2-, 5.1-, 4.2-, 2.1-fold in a concentration-dependent manner as compared with the LPS alone treatment group. In the co-incubation of AL(2-1) with LPS, TNF-α expression level was found to decrease from 7.5- to 4.7-, 2.5-, 1.3-, 1.2-fold in a concentration-dependent manner ([Figure 2(D)](#F0002){ref-type="fig"}). In the co-incubation of WE(2-1) with LPS, IL-6 expression level was found to decrease from 7.2- to 6.1-, 5.8-, 4.4-, 2.8-fold in a concentration-dependent manner as compared with the LPS alone treatment group. In the co-incubation of AL(2-1) with LPS, IL-6 expression level was found to decrease from 7.5- to 3.7-, 2.9-, 1.7-, 0.8-fold in a concentration-dependent manner ([Figure 2(E)](#F0002){ref-type="fig"}).
Inhibition of LPS-induced iNOS and COX-2 protein and its mRNA expression by WE(2-1) and AL(2-1) {#s0023}
-----------------------------------------------------------------------------------------------
The effects of WE(2-1) and AL(2-1) on iNOS and COX-2 protein expression in RAW 264.7 macrophages were examined by Western blot analysis. As shown in [Figure 3(A)](#F0003){ref-type="fig"}, the cells expressed extremely low detectable levels of iNOS and COX-2 proteins under an unstimulated condition; however, iNOS and COX-2 expressions were highly increased in response to LPS (1 μg/mL) after 22 h. Pretreatment of cells with WE(2-1) (1, 3, 5, 10 μg/mL) and AL(2-1) (10, 30, 50, 100 μg/mL) for 2 h dramatically suppressed LPS-induced iNOS and COX-2 expression in a concentration-dependent manner ([Figure 3(A)](#F0003){ref-type="fig"}). The individual treatment of WE(2-1) (10 μg/mL) and AL(2-1) (100 μg/mL) alone did not affect the basal iNOS and COX-2 expressions.
{#F0003}
Also, we have attempted to determine whether the suppression in the expression of iNOS and COX-2 proteins paralleled their inhibition at mRNA levels. After a 2 h pre-treatment of WE(2-1) and AL(2-1), RAW 264.7 macrophages were stimulated with LPS for 22 h. Then, they were harvested and assayed for iNOS and COX-2 mRNA expressions by RT-PCR. The results demonstrate the upregulation of its mRNA levels upon stimulation with LPS for 22 h. In unstimulated macrophages, there was no detectable mRNA. Pre-incubation of cells with WE(2-1) and AL(2-1) plus LPS caused a suppression of iNOS and COX-2 mRNA induction after 22 h of incubation. RT-PCR analysis showed that WE(2-1) and AL(2-1) suppressed the LPS-induced iNOS and COX-2 mRNA levels in a concentration-dependent manner ([Figure 3(B)](#F0003){ref-type="fig"}).
Cytotoxic effect of the WE(2-1) and AL(2-1) in HaCaT keratinocyte cells {#s0024}
-----------------------------------------------------------------------
We next examined the cytotoxic effects of WE(2-1) and AL(2-1) and cell viability was assessed using the MTT assay. We found that WE(2-1) and AL(2-1) had little effect on the cytotoxicity in HaCaT cells ([Figure 3(C)](#F0003){ref-type="fig"}).
WE(2-1) and AL(2-1) had little effect on cell proliferation in HaCaT cells {#s0025}
--------------------------------------------------------------------------
To specifically examine the cell proliferation activity of WE(2-1) and AL(2-1) on HaCaT cells, the cells were treated with indicated concentrations of WE(2-1) and AL(2-1), and then cell viability was analyzed every 15 min time intervals using the xCELLigence RTCA DP Instrument (Roche Diagnostics GmbH, Germany). As shown in [Figure 3(D)](#F0003){ref-type="fig"}, WE(2-1) and AL(2-1) had little effect on cell proliferation in HaCaT cells in a time-dependent manner.
WE(2-1) and AL(2-1) stimulate the production of hyaluronic acid in HaCaT cells {#s0026}
------------------------------------------------------------------------------
We next set out to determine the effect of WE(2-1) and AL(2-1) on hyaluronic acid production in HaCaT cells. Hyaluronic acid is known to act as a sponge in the skin to attract and hold water (Draelos [@CIT0006]). After treatment for 24 h, WE(2-1) and AL(2-1)-induced an increased production of hyaluronic acid, which is an indicative of moisturizing effect ([Figure 4(A)](#F0004){ref-type="fig"}).
{#F0004}
WE(2-1) and AL(2-1) induces the expression of filaggrin and SPT in HaCaT cells {#s0027}
------------------------------------------------------------------------------
Filaggrin and SPT acting as a key factor for skin hydration, we examined whether WE(2-1) and AL(2-1) regulates the expression of these proteins. As shown in [Figure 4(B)](#F0004){ref-type="fig"}, WE(2-1) and AL(2-1) led to increased expression of filaggrin and SPT in HaCaT cells at the protein level. WE(2-1) and AL(2-1) also enhanced mRNA level of filaggrin and SPT in a concentration-dependent manner in HaCaT cells ([Figure 4(C)](#F0004){ref-type="fig"}).
Identification of flavonoids in WE(2-1) and AL(2-1) {#s0028}
---------------------------------------------------
Among flavonoids, dominant flavanone glycosides in all six samples are three flavanone glycosides (hesperidin, narirutin, and neoponcirin) identified by comparing retention times of analytes in authentic standards and samples ([Figure 5(A)](#F0005){ref-type="fig"}), as well as according to the characteristic exact mass and tandem mass (ms/ms) spectra. Hesperidin and narirutin are well known as analytical markers of *Citrus unshiu* peel. Rutin flavonol glycoside was identified in all samples. Polymethoxyflavones (nobiletin and tangeretin) and flavanones (hesperetin and narigenin) were authentically identified. As shown in [Figure 5(B) and (C)](#F0005){ref-type="fig"}, the peak between nobiletin and tangeretin at 11.70 min is putatively identified as heptamethoxyflavone. The list of major compounds detected by LC-QTOF is shown in [Table 2](#t0002){ref-type="table"}.
{#F0005}
######
List of major compounds detected by LC-QTOF.
Compound Formula Retention Time (min) \[M + H\]^+^ Mass (Da) \[M + H\]^+^ Found At Mass (Da)
------------- ----------------- ---------------------- ------------------------ ---------------------------------
Narirutin C~27~H~32~O~14~ 5.06 581.1865 581.1851
Narigenin C~15~H~12~O~5~ 9.99 273.0758 273.0751
Rutin C~27~H~30~O~16~ 4.41 611.1607 611.1590
Hesperidin C~28~H~34~O~15~ 5.57 611.1971 611.1955
Hesperetin C~16~H~14~O~6~ 10.29 303.0863 303.0856
Neoponcirin C~28~H~34~O~14~ 7.97 595.2021 595.2020
Tangeretin C~20~H~20~O~7~ 11.86 373.1282 373.1280
Nobiletin C~21~H~22~O~8~ 11.42 403.1387 403.1383
Discussion {#s0029}
==========
The aim of this study was to investigate the effects of FCU on the suppression of LPS-induced iNOS, COX-2 expression, TNF-α, IL-6, and PGE~2~ production that regulate inflammation in RAW 264.7 murine macrophages. Moreover, fermented *Citrus unshiu* peel extracts can cause stimulate the production of hyaluronic acid and expression of filaggrin and SPT in HaCaT keratinocyte cells, which is an indicative of moisturizing effect.
Inflammation is the body\'s first response of the immune system to infection or irritation. During the inflammatory process, measurable quantities of the inflammatory mediators NO and PGE~2~ are regulated by the inducible isoforms of iNOS and COX-2 (Posadas et al. [@CIT0034]). Overproduction of NO by iNOS occurred in various cell types after stimulation with cytokine and endotoxin, and are also involved in different inflammatory diseases and tumorigenesis (Nathan and Xie [@CIT0023]; Ohshima and Bartsch [@CIT0027]). High levels of PGE~2~ synthesized by COX-2 also occurred in various cancer tissues and implicated in angiogenesis, proliferation, and tumour growth (Claria [@CIT0004]; Meric et al. [@CIT0021]). Thus, there is a causal relationship between inflammation and cancer, iNOS and COX-2 are considered potential candidates for anti-inflammatory drugs (Pan et al. [@CIT0031]). We demonstrated that FCU inhibited LPS-induced NO production via the suppression of iNOS expression. In addition, in an LPS-stimulated macrophage cells, FCU also suppressed PGE~2~ production through the suppression of COX-2 expression. Several studies have demonstrated that the expression of iNOS is induced by pro-inflammatory cytokines, such as IL-1β and TNF-α and TNF-α-induced IL-6 secretion is a prerequisite for increased NO production (Marcus et al. [@CIT0018]; Schrader et al. [@CIT0038]). They are also considered to be important initiators of the inflammatory response and mediators of the development of a variety of inflammatory diseases (Glauser [@CIT0008]; Mannel and Echtenacher [@CIT0017]). We found that FCU results in a concentration-dependent decrease in the LPS-induced secretion of TNF-α, and IL-6. This indicates that the inhibitory effect of FCU on NO production stimulated by LPS is associated with the inhibition of iNOS and COX-2 through the reduction of pro-inflammatory cytokine production.
The skin barrier function is essential and important to maintain the homoeostasis of the skin because of its role in suppressing the penetration of chemicals and the invasion of microorganisms from the outside environment and also to its role in reducing water evaporation from inside the body. Filaggrin is a key structural protein that promotes the formation of the stratum corneum (SC), the outermost layer of the skin, and the terminal differentiation of the epidermis (Armengot-Carbo et al. [@CIT0002]). The stratum corneum is essential to minimize water loss through the epidermis and prevent entry of pathogens, allergens and toxic chemicals (O\'Regan et al. [@CIT0024]; Osawa et al. [@CIT0028]). Filaggrin is initially synthesized as a giant inactive precursor protein, profilaggrin, a \> 400-kDa in human, which is a complex, highly phosphorylated, and insoluble polypeptide (Sandilands et al. [@CIT0037]; Brown and McLean [@CIT0003]). Filaggrin is degraded into amino acids, including histidine and glutamine, which are subsequently modified into trans-urocanic acid and pyrrolidone carboxylic acid, respectively. These amino acids and derivatives contribute to the formation of natural moisturizing factors. They bind water in the stratum corneum and limit moisture loss in the skin, as well as acidify the skin mantle and protect the epidermis (McAleer and Irvine [@CIT0020]). The first step in *de novo* biosynthesis of ceramides is catalyzed by the serine palmitoyl transferase (SPT), which is a condensation reaction of L-serine with palmitoyl-CoA. SPT has been suggested as a key enzyme controlling the level of sphingolipid involved in ceramide generation (Hanada [@CIT0010]). We found that FCU results in a concentration-dependent increases mRNA level of filaggrin and SPT. This can cause stimulate the production of hyaluronic acid and expression of filaggrin and SPT proteins in HaCaT keratinocyte cells. Hyaluronan, an extracellular matrix component, is important for the water content of the skin. So, FCU increases the components which manage the barrier function and water content of the skin.
Conclusions {#s0030}
===========
Overall, our data clearly indicates that FCU exhibits an anti-inflammatory activity that is dependent on its ability to regulate the production of NO, PGE~2~, and pro-inflammatory cytokines in LPS-induced RAW 264.7 cells. Also, we found the FCU increases filaggrin and SPT on HaCaT cells that play a key role in skin hydration and integrity and are involved in skin appearance, metabolism, mechanical properties, and barrier function.
Disclosure statement {#s0031}
====================
The authors have no conflicts of interest to disclose.
[^1]: These authors contributed equally to this study.
| {
"pile_set_name": "PubMed Central"
} |
Background {#Sec1}
==========
The use of pedigreed colonies remains a powerful resource to study many phenotypes of interest in great detail, yet among companion animals, specifically dogs and cats, large well-maintained pedigrees for such use are rare and not readily available. Given this rarity, their optimal use in the understanding of health and behavioral well-being is of crucial importance. Disease surveillance is a critical component of comprehensive veterinary care programs to detect and prevent the spread of disease within animal colonies, thereby enhancing the quality of life of these animals. Veterinary health checks routinely include the collection of samples that can provide a means to detect existing or future health problems and thus provide appropriate care directly to benefit the animal. With the advances underway in the collection of electronic medical records for companion animal patients, mimicking efforts in human clinical practice, the ability to return to banked samples for basic disease research or clinical testing to provide optimal care has veterinarians excited about these health management opportunities. Data collection such as a whole genome or targeted sequencing, immunoassays, metabolite profiling, fixed genotyping and others, collectively or in isolation can drive discovery of the sources of trait diversity linked to genetic variation.
Information provided by genetic data can aid breeding programs by reducing introduction and propagation of health problems in a pedigree. There are many diseases in animals that have been associated with gene variants, for some of those variants, there are commercial genetic tests available \[[@CR1]\]. The ability to select individuals based on genetic information circumvents the issue of producing progeny with health issues that can be unfavorable for these animals. Especially in occasions where disease symptoms appear later than breeding age, without genetic information, those animals will be included in the breeding program resulting in dissemination of undesirable traits. Polycystic kidney disease (PKD) in Persian cats is an example of a disease in which symptoms appear after breeding age \[[@CR2]\]. Lyons et al. \[[@CR3]\] have identified the causative mutation for PKD in the gene *PKD1*, and a commercial genetic test is available enabling Persian cat breeders to make mating decisions based on genetic information.
WGS although still costly for companion animal veterinary practice is the most comprehensive method for detection of an individual's genetic variation. WGS has enabled enormous progress in understanding disease in human and animals. Moreover, it allows extensive evaluation of genetic diversity which is essential for the maintenance of a healthy pedigree. However, the interpretation of the enormous amount of genetic information generated remains a difficult task and reference assembly quality for the cat presents additional variant detection challenges \[[@CR4]\]. The cat has a reference genome that was first assembled with a 1.9X coverage genome sequence of an inbred Abyssinian cat \[[@CR5]\]. Additional sequencing of the same cat to 14X and other cat breeds have allowed enhancement of reference and the identification of common variation in the cat genome \[[@CR6]\]. However, the cat reference still has flaws, such as gaps and unplaced sequences (not in chromosomal regions) and these problems often hinder the discovery of variants associated with phenotypes.
Genetic variant interpretation has been a massive challenge in genetic studies of any species. However, large-scale human disease cohort sequencing projects and even more important the development of databases containing common variation and variants associated with disease have eased this burden of following false positive candidates. In dogs and cats, there have been variants deposited into databases, such as dbSNP, but there is no information on frequency, breed or health status of the individuals from which variants were discovered. Without such data, future efforts to associate putatively damaging variants with disease outcome are much less efficient.
In this preliminary study of a feline colony pedigree, we generated WGS data from three related cats within the pedigree, which contains historical data from \~800 cats, in order to survey segregation of potential disease variants and genetic diversity. We evaluated single nucleotide variants (SNVs) and then compared to databases containing information on genes associated with a disease. Additionally, we calculated runs of homozygosity and inbreeding coefficients on each as a measure of genetic health.
Methods {#Sec2}
=======
Animal descriptions {#Sec3}
-------------------
Three cats were selected for WGS that are part of a pedigreed population maintained by Nestlé Purina as a resource to study behavior preferences and nutritional developments. The pedigree consists of Domestic shorthair cats. We selected cats that were placed at an intermediate position in the total pedigree structure. Also, these cats were directly related allowing observation of accumulation of damaging variants and if there is a decrease in genetic diversity. The health of the cat colony is provided by a veterinary team with a proactive attitude towards disease management. All cats have regular health screening tests depending of their age risk and individual cases. Veterinary care is provided in the same manner and principles than to any individually owned house cat visiting a veterinary clinic, and all individual clinical histories are recorded.
An example of an extended family within this pedigree is depicted in Fig. [1](#Fig1){ref-type="fig"}, including the three WGS cats (Cat I, Cat II, and Cat III). Cat I is an 8-year-old female that is the dam of Cat II and has eight other offspring in the pedigree. Cat II is a 6-year-old female that has about 40 half siblings on the pedigree. Also, Cat II is the dam of Cat III and has two other offspring. Cat III is a 5-year-old female with no offspring, but it has 15 half siblings in the pedigree. All three cats were healthy based on annual veterinarian physicals and no observed disease symptoms during routine care. Samples of whole blood were collected from each cat, by trained veterinary staff, into an Acid Citrate Dextrose vacutainer tube. DNA was isolated using the MagNA Pure 96 (Roche Diagnostics) automated instrument according to the manufacturer's instructions.Fig. 1Pedigree and inbreeding coefficients. The pedigree is showing close related individuals to the three WGS cats (red circles). Squares represent males and circles represent females. Diagonal lines across symbols represent deceased cats. The asterisk indicates cats that have SNPchip data. On the left are the inbreeding coefficients calculated based on pedigree, SNPchip and WGS data
Whole-genome sequencing and variant detection {#Sec4}
---------------------------------------------
Cats were sequenced on an Illumina HiSeq X10 instrument with 350 to 550 bp PCR-free libraries to 150 bp read length. The sequence data for each cat were aligned to the chromosomes of the domestic cat reference assembly (*Felis catus* 8.0) using Speedseq \[[@CR7]\]. Variants, specifically SNVs and small indels (\<10 bp), were called using The Genome Analysis Toolkit (GATK) HaplotypeCaller and GenotypeGVCFs \[[@CR8]\]. Samtools flagstat \[[@CR9]\] and GATK DepthofCoverage were used to extract sequence alignment statistics. SNVs were then extracted from all variants for further evaluation. All SNVs were annotated using the Variant Annotation, Analysis and Search Tool (VAAST 2) \[[@CR10]\]. Next, SNVs predicted to severely disrupt protein-coding genes, loss of function (LoF) variants, shared between individuals or unique to each were reported. In this study, we only considered LoF variants with the most likely deleterious impact: splice sites disruption, stop gain or loss, and frameshifts.
Variant validation and genotyping {#Sec5}
---------------------------------
To estimate our number of false SNVs detected, we selected five variants to be validated by Sanger sequencing in all three cats. We found two were homozygous, and three were heterozygous genotypes. Additionally, two closely related cats and five unrelated cats from the pedigree were genotyped for the same variants to evaluate expected genotypes. Sequences were amplified by PCR using specific primers (Additional file [1](#MOESM1){ref-type="media"}: Table S1) and Amplitaq Gold polymerase kit (Thermo Scientific) according to the manufacturer's protocol with the following modifications: AmpliTaq Gold concentration was 2.5 U/reaction, the forward and reverse primer concentrations were 0.4 μM (final concentration), and the denature time was 30 s. PCR products were purified using PureLink PCR purification kit (Thermo Scientific) following manufacturer's protocol. After purification PCR products were sent to GeneWiz (South Plainfield, NJ) for Sanger sequencing. In addition, a set of 20 heterozygous SNVs (Additional file [1](#MOESM1){ref-type="media"}: Table S2) genotyped on the Illumina feline 63 K SNP BeadChip \[[@CR4]\] from the same three WGS cats were validated for equivalency of SNVs calls.
Detection of variant impact {#Sec6}
---------------------------
The databases Online Mendelian Inheritance in Man (OMIM) \[[@CR11]\], Online Mendelian Inheritance in Animals (OMIA) \[[@CR12]\] and the database of essential genes (DEG) \[[@CR13]\] were consulted for information on disease-causing genes. We only considered genes in OMIM that were associated with a phenotype and in OMIA, only genes that were associated with phenotypes in cats. Also, genes considered essential for mice were incorporated from DEG in the analysis. All LoF variants in disease-causing genes were manually evaluated with the Integrative Genomics Viewer (IGV) \[[@CR14]\], and SNV effect on each gene was assessed by comparison of protein translation to other mammals with NCBI blastp using default parameter settings.
Known disease variants screening {#Sec7}
--------------------------------
Variants associated with disease in cats that have a commercial DNA test available were screened in our three cats by inspection of genotypes or presence of deletions for the specific positions on sequence data. The list of variants screened is in Additional file [1](#MOESM1){ref-type="media"}: Table S3 with respective positions on cat reference version 8.0.
Inbreeding coefficient estimation {#Sec8}
---------------------------------
Pedigree-based inbreeding coefficient (IC) was calculated using the Wright's equation:$$\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$ {F}_X=\sum \left[{\left(\frac{1}{2}\right)}^{n_1+{n}_2+1}\left(1+{F}_A\right)\right] $$\end{document}$$
Where *F* ~*x*~ is the inbreeding coefficient of the cat in question, *F* ~*A*~ is the inbreeding coefficient of the common ancestor, *n* ~1~ is the number of generations from the sire to the common ancestor, and *n* ~2~ is the number of generations from the dam to the common ancestor. We used the known information about the pedigree (Fig. [1](#Fig1){ref-type="fig"}), however, there were unknown ancestors. IC was calculated for Cat I based on three generations, Cat II based on five generations and Cat III based on 6 generations.
Genotype data from the Illumina feline 63 K SNP BeadChip (SNPchip) and WGS data based inbreeding coefficients were calculated using PLINK v1.07 \[[@CR15]\] \--het function. The SNPchip data was analyzed with 55,053 SNPs, while the WGS data was analyzed with 13,455,757 SNPs, both from autosomes only. Since we have SNPchip data from 297 cats that are part of the pedigree, we calculated IC using data for all the cats to compare scores when calculating with only the three cats.
Runs of Homozygosity analysis {#Sec9}
-----------------------------
To calculate runs of homozygosity (RoH) for SNPchip and WGS data we used PLINK v1.07 \--homozyg function. For SNPchip data, we defined our RoH segments as five or more consecutive homozygous SNVs per individual. For WGS data we used a window size of 250 kb since this approximate window size is roughly equivalent to five homozygous SNVs on SNPchip data.
Results {#Sec10}
=======
Whole genome sequencing and variant validation {#Sec11}
----------------------------------------------
We generated WGS for each cat with an average range of 27-32X coverage. The number of sequences that properly mapped to Felis_catus-8.0 reference was between 92 and 97%, and duplicates were between 13 and 15% (Table [1](#Tab1){ref-type="table"}). A total of 18,137,177 variants were identified in all three cats, 14,088,779 were SNVs and 4,048,398 were indels (Table [2](#Tab2){ref-type="table"}). For this study we only report putatively deleterious SNVs since indels are known to have high rates of false positives \[[@CR16]\].Table 1Whole-genome sequencing results summaryCat ICat IICat IIINumber of reads639,766,569575,104,522552,808,448Average coverage32.1327.8627.79Duplicates15.52%13.76%14.78%Mapped reads97.19%92.83%96.70%Properly paired94.84%89.99%93.96% Table 2The number of variants identified for each cat, including SNPs and Indels. Variants are divided by annotation categoriesTotalNon-synonymousFrameshiftSplice siteStop gain/lossCat I all13,791,28228,9842394656650 homozygous4,722,22585021106237179 unique1,947,0674256290104110Cat II all13,663,92128,9612414636627 homozygous4,808,54786051115255170 unique1,142,03523992447058Cat III all13,635,21528,4672380632622 homozygous4,882,66788691115230189 unique1,708,60233993179279Shared all9,196,14919,3771730434407 homozygous2,597,9484947898131108
We selected five SNVs (Additional file [1](#MOESM1){ref-type="media"}: Table S1) for validation and genotyping in seven additional cats. The two homozygous SNVs (chrE1:40,235,385 and chrE1:40,235,189) were validated and also present on the additional seven cats. The three heterozygous SNVs were not validated, all cats were homozygous for the reference allele. It was surprising that all three heterozygous SNVs were false positives. Examination of the regions containing the SNVs revealed that two of those regions were located within sequences homologous to other sequences in the cat genome, which may suggest that misalignment created the false positives. The convergence between SNPchip heterozygous genotypes and WGS SNP showed 100% agreement between calls, indicating our SNV calls are of high confidence for further study.
Variant functional evaluation {#Sec12}
-----------------------------
Genes that harbored LoF variants were cross-referenced with three databases containing information on phenotypes associated with genes: OMIM, OMIA, and DEG. Next, we manually checked variants in IGV and the protein translation similarity to other mammals. From the genes associated with a phenotype, there was one homozygous and two heterozygous shared LoFs between the three cats, while there was one heterozygous unique LoF. We first investigated the LoF variants that were shared between the three cats in a homozygous and heterozygous state. Only one homozygous LoF was identified in a gene that matched OMIM and DEG databases; it is an SNV that changes the splice site sequence on the huntingtin-associated protein 1 (*HAP1*) gene (Fig. [2](#Fig2){ref-type="fig"}). *AHR* and *CTNNA2*, which are genes in the DEG database, both have a heterozygous LoF predicted to disrupt splice sites. Furthermore, we investigated unique LoFs homozygous and heterozygous in each cat and no unique homozygous LoFs in genes that matched the databases were found. There was only one heterozygous LoF in Cat III that creates a stop five amino acids before the end of the immunoglobulin mu binding protein 2 (*IGHMBP2*) protein. The positions and types of LoFs are described in Additional file [1](#MOESM1){ref-type="media"}: Table S4. Overall, we observed a high number of false positive LoFs at \~42%, based on the variants manually inspected, that we attribute to inaccurate gene models.Fig. 2Homozygous LoF on *HAP1* gene. The variant in *HAP1* gene is homozygous on all three cats and it changes the splice donor sequence from GT to GG. This variant has been previously identified in other cats and has been deposited in dbSNP
Screening for known cat disease variants {#Sec13}
----------------------------------------
To ascertain if our screening would find known disease alleles, even though our cats were deemed healthy by veterinarian exams, we screened our three cats for variants that match these causative alleles. Each has accompanying commercial DNA tests that could confirm their putative disease carrier status in our pedigree. Some of these diseases are breed specific while all cats that are part of this pedigree are mixed breed so their frequency would be expected to be rare and in some cases may not present disease phenotype in a mixed genetic background. The diseases screened for relevant variants were: Gangliosidosis 1 \[[@CR17]\], Gangliosidosis 2 \[[@CR18]\], Cardiomyopathy \[[@CR19], [@CR20]\], Hypokalemia \[[@CR21]\], Progressive retinal atrophy \[[@CR22]\], Polycystic kidney \[[@CR3]\], and Spinal muscular atrophy \[[@CR23]\]. The occurrences of some of these diseases in our pedigree lead us to screen these variants even though the three cats selected are healthy. We found no instances matching disease variants.
Inbreeding coefficients {#Sec14}
-----------------------
We calculated IC to gauge genetic diversity using available pedigree relationships, SNPchip and WGS data for all three cats (Fig. [1](#Fig1){ref-type="fig"} **)**. New male cats were frequently introduced to the pedigree to keep genetic diversity high but in most cases, these sires lack ancestry information as shown in Fig. [1](#Fig1){ref-type="fig"}. Pedigree based IC was calculated on available information on ancestors using the Wright's equation. Cat III had the highest pedigree based IC, most likely due to this cat having more pedigree information than the others. IC from SNPchip and WGS data were calculated on the observed versus expected number of homozygous genotypes. SNPchip and WGS based ICs were equivalent, as is expected since both were calculated by the same method. The IC was remarkably low in all three analyses. Since there is missing ancestry information for some of the cats, IC calculated based on SNPchip and WGS data are more reliable than the pedigree-based. However, IC calculated by PLINK is more accurate when calculated with larger sample size. When IC was calculated with 297 individuals the scores were higher (Aditional file 1: Table S5) while still negative indicating high rate of heterozygosity (CatI: −0.508, −0.057; CatII: −0.473, −0.034; CatIII: −0.456, −0.022; IC calculated with three cats and IC calculated with 297 cats respectively).
Detection of RoH in SNPchip and WGS data {#Sec15}
----------------------------------------
The estimation of the level of homozygosity in our pedigree associated cats was carried out with SNPchip and WGS data using different window sizes, five consecutive homozygous SNPs and 250 kb, respectively. WGS analysis was done in window size instead of a number of consecutive homozygous SNPs, because the distances between SNPs are on average less than 500 bp apart compared to SNPchip markers that are on average 50 kb apart. SNPchip data analysis detected 20--27 RoH segments (Additional file [1](#MOESM1){ref-type="media"}: Table S6) while WGS data analysis detected 24--57 RoH segments (Additional file [1](#MOESM1){ref-type="media"}: Table S7). The higher number of RoH identified by WGS data is expected given the higher resolution of detected variants. Comparison of the analysis of the two datasets shows that there was considerable overlap for both, but the length of the RoH was much higher for SNPchip data (average 6344 kb SNPchip, average 324 kb WGS). RoH segments identified by SNPchip data were frequently broken into smaller RoH segments identified by WGS when there was overlap. Some of the segments are fully or partially shared between cats for each data set. Cat III has the higher number of RoH for both data sets (27 for SNPchip and 57 for WGS) compared to the other two cats. The low numbers of RoH segments identified corresponds to the low IC observed for theses cats calculated with SNPchip and WGS data.
Discussion {#Sec16}
----------
Genetic diversity plays an important role in maintaining a healthy pedigree. While the success of a genetically diverse pedigree relies on an effective breeding program, mating decisions with inadequate genetic information could have unintended consequences in future generations. The traditional process of selecting individuals for breeding involves calculations of inbreeding prior to mating in order to optimize hybrid vigor as well as consideration of traits and symptoms when they became apparent in the sire or dam and offspring. Unlike the mating decisions made in food-producing animals, determined largely by the need to improve phenotypes of economic interest, health is the major priority in managing companion animal pedigrees. The identification of genetic variants within genes implicated in clinically relevant phenotypes provides a new means to avoid the spread of unintended alleles with harmful outcomes before breeding. Already veterinarians are attempting to utilize genetic information as a diagnostic and clinical management aid. However, limited validation of numerous putative alleles of clinical significance has hampered their abilities. In a well-maintained pedigree, undesirable recessive phenotypes can be avoided by selective breeding to circumvent the production of homozygous individuals, thus preventing propagation of individuals with potentially adverse health conditions. We contend that the use of WGS data to assess sire and dam mutational profiles and to determine genetic diversity can help to improve animal health and in maintaining offspring genome diversity in the pedigree.
In our limited study of pedigree associated cats, we first analyzed potential clinical relevant variants. Several variants have been associated with diseases in cats \[[@CR1]\] and some have commercially available genetic tests for breeders to screen their animals. Nonetheless, not all tests are applicable breed wide, for example, the Gangliosidosis 2 *HEXB* variant \[[@CR18]\] is specific for the Burmese breed. The mixed breed cats in our pedigree do show occurrences of a few of these diseases in the pedigree, such as Polycystic kidney disease (PKD). PKD was estimated to have a prevalence of 30--38% worldwide in Persian and closely related cat breeds \[[@CR24]--[@CR26]\]. Therefore, there was a possibility that the cats in our pedigree carry one or more of the PKD disease alleles. However, screening for all known disease variants revealed that our cats do not carry any of these causative alleles, but continued surveillance is needed for the appearance of new disease causative alleles.
The tremendous expansion of variant knowledge among human studies can reveal shared genic events, at least within the same gene, that may be of clinical relevance in veterinarian care. In this report, we find several cases of shared variants that could lead to future health consequences but most often is undetected phenotypically. Shared among the three cats we have identified a homozygous splice-site SNV in the gene *HAP1*. This SNV has been identified previously in other cats according to dbSNP, rs784247714, and it was also identified in the additional seven cats we genotyped from the pedigree. The HAP1 protein interacts with the huntingtin protein \[[@CR27]\], which is associated with Huntington disease \[[@CR28]\]. However, *HAP1* itself has not been directly linked to Huntington disease. Chan et al. \[[@CR29]\] have shown that *Hap1* knockout mice exhibit strikingly depressed feeding behavior and are unable to gain body weight after birth. The mice often die after day 2--3, but the ones that survive displayed growth retardation with apparent normal brain and behavioral development suggesting an effect only in early postnatal feeding behavior \[[@CR30]\]. In our three cats, no similar abnormal feeding behavior has been observed, which may suggest that this mutation does not affect the protein function. Alternatively, the creation of a protein isoform that skips one exon is fully functional.
We identified two heterozygous LoF variants shared by the three cats that matched *AHR* and *CTNNA2*, which are considered essential genes for survival in mice according to DEG. The aryl hydrocarbon receptor plays important roles in the developmental remodeling of vascular architecture in the liver \[[@CR31]\], regulates the toxicity of halogenated dioxins \[[@CR32]\] and controls the adaptive up-regulation of xenobiotic metabolizing enzymes in response to polycyclic aromatic hydrocarbons \[[@CR33]\]. In *Ahr*-null mice disruption of AHR signaling pathway causes fetal necrosis and consequent liver deformation which persists through adulthood \[[@CR34]\]. The CTNNA2 protein links the classical cadherins to the neuronal cytoskeleton and is expressed only in the central nervous system in mice \[[@CR35]\]. Mice lacking part of the CTNNA2 protein are ataxic and show abnormal lobulation of the cerebellum and cerebellar hypoplasia \[[@CR36]\]. These phenotypes haven't been observed in the pedigree so far, however, our discoveries highlight how pedigree breeding management would provide a means to avoid the propagation of these alleles in subsequent generations.
Additionally, we explored unique LoFs for each cat. We identified a heterozygous stop gained SNV on *IGHMBP2* gene unique to Cat III. Mutations in this gene are reported to cause distal spinal muscle atrophy type 1 \[[@CR37]\] and Charcot-Marie-Tooth disease type 2 \[[@CR38]\]. The stop gained, found in our cat, was 5 amino acids before the end of the protein, it is most likely that this SNV do not affect protein function. Given its haploid state in Cat III and no observed health abnormalities this variant is not considered to be a risk for disease development, therefore not affecting the decision to include this cat in the breeding pool. The effect of variants in gene function is not easily predicted. Despite the tools that classify variants as benign or damaging, it is of substantial advantage to having access to a common variants database, where information on health status and breed are recorded for each variant identified. In this example, we would greatly benefit from that information to determine if this SNV is benign or damaging. The need for a robust repository for variants in cats is critical for research in disease or trait variant discovery. In humans, great efforts have been made to create databases recoding variant information with different levels of evidence implicating variants in disease risk or causation, such as ClinVar. Also, guidelines for associating variants to disease have been described to avoid a proliferation of false positive findings \[[@CR39]\].
To access inbreeding status of the three cats, IC was calculated based in three data sets: pedigree information, SNPchip, and WGS data. Pedigree information was limited for part of the ancestors because cats are frequently introduced to the pedigree. ICs based on pedigree were higher than the ones calculated with SNPchip and WGS data. SNPchip and WGS ICs were interchangeable for each of the cats; the ICs were negative indicating a high rate of heterozygosity relative to their reference population. However, we observed that larger sample sizes are recommended for higher accuracy when calculating IC with genetic data. Our results for a small sampling of the pedigree reveal efforts to maintain diversity is successful thus far. The analysis of runs of homozygosity (RoH) with SNPchip and WGS data has shown that SNPchip analysis overestimates length and underestimates the number of RoH. WGS has a better resolution for this type of analysis because it genotypes every position while SNPchip analysis doesn't consider heterozygosity between markers and markers are in average 50 kb apart. The number of RoH segments identified for both datasets are in agreement with the low ICs observed.
Conclusions {#Sec17}
===========
In summary, we describe the assessment and possible use of genomic information of three cats from a large pedigree. We are cognizant of the limitations that three cat genomes could provide to the management of large pedigrees, nonetheless, our genome variant data has enabled us to identify possible disease causing variants, plan more cost-effective screening assays using this data and obtain an estimate of pedigree genetic diversity. The decision on inclusion and exclusion of cats in the breeding pool based on genetic variants must be carefully considered. First and foremost, variants that are known to cause disease should be regarded as most important during breeding management. The variants predicted to be damaging should be complemented by clinical observations of animals to conclude that they impact health. Since genetic diversity is as crucial as avoiding the spreading of disease-causing variants, it is necessary to balance the breeding pool via mating selections that safeguard genetic diversity while minimizing the accumulation of damaging variants. Once disease variants are discovered, they can be cost effectively screened as part of marker panels, much akin to human clinical disease screening protocols, to better manage pedigree health. The expectation is genome data strategically collected can be a powerful tool to improve animal health.
Additional file {#Sec18}
===============
Additional file 1:Additional tables. **Table S1.** Validation primer sequences and PCR annealing temperatures; **Table S2.** Heterozygous SNPs selected from SNPchip data for cross validation; **Table S3.** Variants associated with diseases in cats with a commercial DNA test available; **Table S4.** LoF variants identified in genes associated with disease; **Table S5.** Inbreeding coefficient calculated with SNPchip data of only 3 cats and with SNPchip data of 297 cats ; **Table S6.** RoH identified with SNPchip data on the three cats; **Table S7.** RoH identified with WGS data on the three cats. The last column has the overlap with SNPchip RoH number. (DOCX 42 kb)
DEG
: Database of Essential Genes
IC
: Inbreeding Coefficient
IGV
: Integrative Genomics Viewer
LoF
: loss of function
OMIA
: Online Mendelian Inheritance in Animals
OMIM
: Online Mendelian Inheritance in Man
PKD
: Polycystic kidney disease
RoH
: Runs of Homozygosity.
SNV
: Single nucleotide variant
WGS
: Whole genome sequence
**Electronic supplementary material**
The online version of this article (doi:10.1186/s12917-017-1144-y) contains supplementary material, which is available to authorized users.
We would like to thank Cynthia Steeby, Patricia Turpin and Holly Ambrose for their help with DNA extractions.
Funding {#FPar1}
=======
This work was supported by Nestlé Purina Research under the Comparative Genomics Postdoctoral Fellowship.
Availability of data and materials {#FPar2}
==================================
All the data supporting the results are included in the article. Whole genome sequence data for the three cats are available in the NCBI Sequence Read Archive (SRA) under bioproject PRJNA393717: (<http://www.ncbi.nlm.nih.gov/bioproject/393717>).
WCW, FHGF and RM conceived and designed the study. CT performed quality control of sequence data. Sequenced analysis was performed by FHGF and CT. FHGF executed variant analysis, inspection, and comparison between whole-genome sequencing and SNPchip data. Variant genotype was done by JL. FHGF and WCW were the major contributors in writing the manuscript, with input from all authors. All authors read and approved the final manuscript.
Ethics approval {#FPar3}
===============
All the study procedures involving cats were reviewed and approved by the Nestlé Purina Animal Care and Use Committee according to US regulations.
Consent for publication {#FPar4}
=======================
Not applicable.
Competing interests {#FPar5}
===================
Jeffrey Labuda, Gerardo Perez-Camargo, and Rondo Middleton are employees of Nestlé Purina.
Publisher's Note {#FPar6}
================
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
| {
"pile_set_name": "PubMed Central"
} |
The images for Figs 3 and 4 are incorrectly switched. The image that appears as Fig 3 should be Fig 4, and the image that appears as Fig 4 should be Fig 3. The figure captions appear in the correct order.
{#pone.0176577.g001}
{#pone.0176577.g002}
| {
"pile_set_name": "PubMed Central"
} |
Documentation, analysis, and prevention of the harmful effects of armed conflict on populations are established public health priorities \[[@pmed-0050243-b001]\]. Although public health research on war is increasingly framed in human rights terms \[[@pmed-0050243-b006]\], general public health methods are typically applied without direct links to laws of war. Laws of war are international humanitarian laws and customary standards regarding the treatment of civilians and combatants, mainly described in the four Geneva Conventions of 1949 and their Additional Protocols I and II regarding international and civil conflicts \[[@pmed-0050243-b014]\]. With notable exceptions \[[@pmed-0050243-b011],[@pmed-0050243-b015]\], absolute numbers are usually reported (e.g., number of persons killed), without systematic description of the proportional effects of armed conflict, thereby limiting the utility of findings and scope of interpretation.
In this paper, we introduce the "Dirty War Index" (DWI): a data-driven public health tool based on laws of war that systematically identifies rates of particularly undesirable or prohibited, i.e., "dirty," war outcomes inflicted on populations during armed conflict (e.g., civilian death, child injury, or torture). DWIs are explicitly linked to international humanitarian law to make public health outcomes directly relevant to prevention, monitoring, and humanitarian intervention for the moderation of war\'s effects. After choosing the particular outcome to be measured, a DWI is calculated as:
###### Summary Points
- War, a major public health problem, is a situation where the interests of public health, human rights, and humanitarian law intersect.
- The DWI is a data-driven public health tool that identifies rates of particularly undesirable or prohibited, i.e., "dirty," outcomes inflicted on populations during war (e.g., civilian death, child injury, or torture).
- A DWI is calculated as: (Number of "dirty," i.e., undesirable or prohibited cases/Total number of cases) × 100.
- DWIs are designed for direct, easy translation of war\'s public health outcomes into the human rights, policy, and interdisciplinary work needed to address war\'s practice.
- DWIs support monitoring, deterrence, and humanitarian intervention by explicit links to international humanitarian laws and by exposing rates of unacceptable combat outcomes (DWI values) from different weapons or combatant groups.
For example: In [Table 1](#pmed-0050243-t001){ref-type="table"}, we measure the DWI ratio of "Number of civilians killed/Total number of civilians and opponent combatants killed" using a casualty dataset for Colombia\'s civil conflict \[[@pmed-0050243-b018]\]. [Table 2](#pmed-0050243-t002){ref-type="table"} links this DWI to relevant laws of war. DWI values of 99 for illegal paramilitaries, 46 for guerrillas, and 45 for government forces show that paramilitaries are "dirtiest" in terms of proportion of civilians constituting their victims of unopposed attacks (chi-square = 5,010, degree of freedom \[df\] = 2, *p* \< 0.001). 99% of paramilitary victims were civilians and only 1% were military opponents. This finding, combined with the paramilitaries\' methods (execution by close-range gunfire in massacres), suggests intentional targeting of civilians that requires recognition in Colombia\'s paramilitary demobilization, disarmament, and reintegration process \[[@pmed-0050243-b019]\].
###### Dirty War Index for Attacks by Actors in the Colombian Civil Conflict, 1988--2005: Civilian Versus Opponent Combatant Mortality
###### DWIs Suggested for Measuring Rates of Undesirable or Prohibited Outcomes from Aggression in Armed Conflict
As ratios, DWIs complement absolute numbers and lend themselves to comparisons over time, between wars, between weapons, and between warring combatant groups to identify better versus worse performers. Noncombatant wounded-to-killed ratios can provide evidence of war crimes \[[@pmed-0050243-b016]\]. Proportional "atrocity statistics" \[[@pmed-0050243-b020]\] from a Darfur survey substantiated US Secretary of State Colin Powell\'s declaration of genocide and the referral of Darfur\'s situation to the International Criminal Court \[[@pmed-0050243-b020],[@pmed-0050243-b021]\]. By facilitating clear, systematic comparisons, DWIs can help analyze and expose how combatants engage in war and affect populations, thereby increasing the accountability of military and political leaders. This paper describes the theoretical basis and practical applications of the DWI, with brief examples from armed conflicts. More detailed DWI analyses of specific conflicts are planned for future papers.
Calculating and Using DWIs {#s2}
==========================
A DWI can be easily used and understood, facilitating interdisciplinary communication and research on war\'s effects. DWIs can measure rates of undesirable outcomes from accepted methods (e.g., civilian casualties from aerial bombing of military targets). They can also measure rates of using prohibited, illegitimate methods (e.g., torture), and rates of applying illegitimate methods to especially vulnerable populations (e.g., torturing children) to describe rates of exceptional atrocity. However, the mere application of DWI analysis to a combatant group does not indicate that it is "dirty": a DWI ratio simply identifies how often, if at all, the group is linked with the particular undesirable outcome being measured, facilitating comparisons. To illustrate, we draw on data from B\'Tselem (<http://www.btselem.org/english/statistics/Index.asp>), a nongovernmental organization that monitors casualties from the Israeli-Palestinian conflict. We apply a "female mortality DWI" (Number of females killed/Total number killed) to conflict-related killings from September 29, 2000 to April 30, 2007: Israeli security forces killed 213 females among 4,057 Palestinians (DWI = 5). Palestinians killed 283 females among 705 Israeli civilians (DWI = 40). Palestinians killed 10 females among 317 Palestinians (DWI = 3). Comparison of actors\' DWIs shows significantly higher discrimination of female from male targets by Israeli security forces and by Palestinian actors when targeting Palestinians, and lower discrimination of female from male targets when Palestinian actors target Israeli civilians (chi-square = 833, df = 2, *p* \< 0.001).
The best possible DWI value is 0, indicating that the objectionable outcome is identified in no measured cases. The worst possible DWI value is 100, indicating that the objectionable outcome is identified in 100% of measured cases. Any rate above 0 for prohibited actions or war crimes is unacceptable, and eliminating violations is imperative. DWIs for undesirable outcomes are less straightforward. The highly undesirable outcome of civilian harm is not prohibited by laws of war if combatants do everything feasible to distinguish between civilians and military targets (the principle of distinction), if they attempt to minimize incidental harm to civilians, and if they intend to avoid harming civilians in excess of anticipated military goals (the principle of proportionality) \[[@pmed-0050243-b001],[@pmed-0050243-b022],[@pmed-0050243-b023]\]. Civilian harm is also balanced against the "military necessity" of objectives \[[@pmed-0050243-b024]\]. Though what is feasible, proportional, or necessary is highly subjective \[[@pmed-0050243-b022]\], clearly the lowest possible rates should be sought for undesirable outcomes such as "incidental" civilian death. High DWI values for undesirable outcomes indicate extreme destruction, signal the need for close scrutiny, and may suggest war crimes.
[Tables 2](#pmed-0050243-t002){ref-type="table"} and [3](#pmed-0050243-t003){ref-type="table"} list specific DWIs, their pertinent laws of war, and example calculations. [Table 2](#pmed-0050243-t002){ref-type="table"} lists DWIs for undesirable or prohibited aggression in armed conflict. DWIs can be analyzed by demographic subgroup for indiscriminate warfare, disproportionate effects of targeting, or particular vulnerability to weapons. For example, with "casualties" defined as injuries or deaths, a "child casualty DWI" (Number of child casualties/ Total number of casualties) applied to weapons-casualty data from Chechnya \[[@pmed-0050243-b025]\] gives the following child casualty ratios for different explosive devices: antitank landmines (34/223, DWI = 15), antipersonnel landmines (223/1,004, DWI = 22), booby traps (65/214, DWI = 30), and other unexploded ordnance (UXO) (255/892, DWI = 29). DWIs indicate that in Chechnya, UXO and booby traps are more dangerous to children than landmines and significantly "dirtier" in this respect (chi-square = 25.0, df = 3, *p* \< 0.001).
###### DWIs Suggested for Measuring Rates of Unacceptable Endangerment in Armed Conflict
[Table 3](#pmed-0050243-t003){ref-type="table"} lists DWIs for unacceptable endangerment in armed conflict \[[@pmed-0050243-b014],[@pmed-0050243-b023],[@pmed-0050243-b024]\]. To illustrate, we apply the last DWI listed, "Destroying infrastructure essential for civilian survival (food, water, hospitals)," to survey data from eastern Burma where the Burmese military junta is in conflict with ethnic minority groups. The Burmese military regime destroyed or stole food from 472 of 1,813 surveyed households \[[@pmed-0050243-b011]\]. The Burmese military\'s DWI of 26 indicates a 26% rate of committing the humanitarian violation of destroying civilian food sources, associated in the study with significantly greater odds of household landmine injury (perhaps due to foraging for food), child malnutrition, and death \[[@pmed-0050243-b011]\].
In [Table 4](#pmed-0050243-t004){ref-type="table"} we analyze the Northern Ireland conflict for two complementary DWIs: aggressive acts (killing civilians) and endangerment to civilians (by not wearing uniforms). Combatants who blur distinctions between themselves and civilians transfer their risk onto civilians \[[@pmed-0050243-b023],[@pmed-0050243-b024]\]. Endangerment of noncombatants can be a byproduct of a method, as when guerrilla forces hide "among the people," taking the battlefield to civilians \[[@pmed-0050243-b023],[@pmed-0050243-b024],[@pmed-0050243-b026]\]. Endangerment can also be a direct goal. As described by Viet Cong leaders \[[@pmed-0050243-b027]\] and American soldiers \[[@pmed-0050243-b023]\] in the Vietnam War, Viet Cong forces trained children to throw grenades at South Vietnamese and American soldiers, partly to provoke opponents to shoot children and bring shame to themselves and their force. Child soldiers are more often killed or injured than adult soldiers, being deployed at the front line, to lay or clear mines, or as suicide bombers because they provoke less suspicion \[[@pmed-0050243-b003],[@pmed-0050243-b028],[@pmed-0050243-b029]\]. To illustrate the issue of variable access to valid data for DWI applications, precise data for calculating child solider DWIs ([Table 3](#pmed-0050243-t003){ref-type="table"}) may be difficult to obtain for some conflicts. However, DWIs for using child soldier suicide bombers ([Tables 2](#pmed-0050243-t002){ref-type="table"} and [3](#pmed-0050243-t003){ref-type="table"}) could be highly accurate due to extensive media coverage of suicide attacks.
###### The Northern Ireland Conflict, 1969--2001: Complementary DWI Analyses for Unacceptable Aggression and Endangerment by Actors
DWI analysis can use any data source (media reports, epidemiological surveys, coroners\' reports) as long as the data are adequately valid, accurate, and comprehensive. DWI analysis can be applied to event-based data or to aggregated data covering, for example, a year, a phase, or a whole conflict. Analysis of all DWIs supported by good data provides fuller description of a conflict and combatant behavior. A qualitative understanding of a conflict\'s nature and context is necessary for DWI application and interpretation. When possible, analysis should recognize when combatants avoid inflicting dirty outcomes, i.e., "clean" combat. DWIs suggest valuable data for prospective inclusion in conflict monitoring.
Considerations {#s3}
==============
When DWIs are used to compare combatant groups or methods, it should not be assumed that those with the highest values are simply the dirtiest. Nor should it be assumed that lower DWI values "don\'t count." A group may have a low DWI for recorded civilian mortality, but high DWIs for assassinating civilian leaders and disappearances. Another group may have low DWIs generally, and a very low DWI for torturing prisoners, but torture breaches the precepts of humanity utterly so that to have a measurable rate at all is deplorable.
DWIs reflect, in part, local conditions. For example, the lethality of civilian injuries reflects local treatment technology and access. It may therefore seem incorrect to compare DWIs for civilian lethality when health services differ. Similarly, it may seem unfair to compare child mortality DWIs between a conflict where children comprise a large proportion of the population and so are more likely to be killed and a conflict where children are few. However, researchers should not adjust for such factors when comparing DWIs across settings. This is because actors in armed conflict know, or are morally obliged to know, local resources and demographics and their implications for civilian harm. Combatants are obliged to take proportionately more care not to kill children when waging war in a child-dense population. Responsibility for dirty outcomes is not ameliorated by local conditions.
As for any conflict analysis \[[@pmed-0050243-b002],[@pmed-0050243-b003],[@pmed-0050243-b030]\], DWI selection, application, and interpretation must recognize the potential, varied biases of data sources and of particular DWIs. Conflicts are highly politicized, and combatants, supporters, and detractors have always tried to manipulate reports of war outcomes. Combatants may attempt to construct more favorable DWIs not only by decreasing dirty combat, but by concealing dirty outcomes, or by misrepresenting or provoking opponents\' dirty outcomes. For example, a group might attempt to raise an opponent\'s child mortality DWI by using child soldiers or children as human shields.
Some DWI outcomes, such as injuries, may tend to be under-reported \[[@pmed-0050243-b016]\]. War-associated rape may be difficult to measure due to stigma and under-reporting, though substantial reports exist \[[@pmed-0050243-b021],[@pmed-0050243-b024],[@pmed-0050243-b031],[@pmed-0050243-b032]\]. Although bias can affect DWI values, as ratios DWIs are relatively less affected by under- or over-counting than absolute numbers. For example, if a population generally under-reports war-related rape by 40%, this does not bias comparing rates between different combatant groups.
DWIs, complemented by absolute numbers, can suggest strategic aspects of actors\' methods. For example, systematic civilian targeting is suggested by combined findings of: many events killing or injuring civilians; high ratios of civilian versus combatant mortality; frequent use of methods causing high civilian casualties; frequent use of methods causing high civilian lethality; and high rates of civilian harm from methods that are inherently "targeted" (handguns, machetes). Such proportional and numerical findings on civilian casualties have been used as evidence in International Criminal Court trials to establish systematic patterns indicating war crimes \[[@pmed-0050243-b033]\].
DWIs Measure Outcomes, Not Justifications or Intentions {#s4}
=======================================================
DWIs focus on whether the practice of war is just (*jus in bello*) and ignore whether the reason for war is just (*jus ad bellum*), separating two logically distinct moral issues in war \[[@pmed-0050243-b023]\]. We focus on practical outcomes because justifications for war are contested, are used to legitimize dirty combat, and can bias examination of war\'s impact \[[@pmed-0050243-b024],[@pmed-0050243-b027],[@pmed-0050243-b034]\]. Combatants and their supporters may believe or describe methods as just, whether the method is suicide bombing \[[@pmed-0050243-b024],[@pmed-0050243-b037],[@pmed-0050243-b038]\] or the World War II targeting of civilians by Germany and by the Allies with carpet bombing, fire-bombing, and atomic bombs directed at cities \[[@pmed-0050243-b023],[@pmed-0050243-b024],[@pmed-0050243-b026],[@pmed-0050243-b027]\].
Although intentions affect combat outcomes, such as civilian mortality rates \[[@pmed-0050243-b016],[@pmed-0050243-b027],[@pmed-0050243-b034],[@pmed-0050243-b035]\], we separate DWIs from intentionality for the following reasons. Intentions are contested, obscured, and distorted \[[@pmed-0050243-b003],[@pmed-0050243-b035]\]. Dirty outcomes can result from malicious intent, beneficent intent, or recklessness (lack of intent to take due care). Frequently, combatant violence that appears wanton, sadistic, or vengeful (e.g., rape, mutilation) is mobilized by political actors for hidden strategic aims \[[@pmed-0050243-b024],[@pmed-0050243-b027],[@pmed-0050243-b035],[@pmed-0050243-b039],[@pmed-0050243-b040]\]. Combatants\' intended effects may be disrupted by targets or adversaries \[[@pmed-0050243-b041]\]. Individual combatant behavior reflects overriding goals and sociocultural aspects of larger groups \[[@pmed-0050243-b034],[@pmed-0050243-b035],[@pmed-0050243-b038]\].
Accommodating intentions or justifications in DWIs would imply that good intentions or a "just war" attenuate responsibility for bad outcomes; an implication that is morally and legally refuted \[[@pmed-0050243-b036]\]. DWIs therefore only recognize the crucial matter of outcomes: the killing, injury, or abuse of individuals and populations who should be protected from war.
Potential Deterrent Effect of the Dirty War Index {#s5}
=================================================
We choose the term "Dirty War Index" for three reasons. First, it unites moral, humanitarian, and scientific values inherent to most armed conflict research. Second, it avoids euphemisms that sanitize descriptions of war-induced public harm \[[@pmed-0050243-b012],[@pmed-0050243-b013],[@pmed-0050243-b027],[@pmed-0050243-b042]\]. Third, emotional and cultural implications of "dirty" versus "clean" may heighten the sensitivity of combatant groups to the index, increasing its potential deterrent effect. No nation or combatant group wants to be considered "dirty" or described as dirtier than others.
Increased accountability can have a deterrent effect in armed conflict and encourages adherence to international humanitarian law; an important element in preventing violence towards noncombatants \[[@pmed-0050243-b001],[@pmed-0050243-b024]\]. DWIs increase scrutiny and accountability specifically for dirty war methods. DWIs are analogous to corruption and bribery indices used by nongovernmental organizations and the World Bank to improve international governance through public monitoring and ranking governments by corruption \[[@pmed-0050243-b043],[@pmed-0050243-b044]\]. In *Better: A Surgeon\'s Notes on Performance* \[[@pmed-0050243-b045]\], Atul Gawande describes how systematic analysis of war casualties reveals problems and suggests solutions, and how identifying exemplary performers can improve general performance. The DWI is developed for systematic, data-driven identification of relatively good versus bad performance, heightening its potential to stimulate positive change.
Military and political leaders not only want to win wars. They also seek superior moral authority \[[@pmed-0050243-b023]\]. Moral authority has social currency, creating better access to material resources, support, and security within local and international communities. To improve behavior in combatants and politicians insufficiently motivated by altruism, harnessing such self-interest is crucial. Exposure of atrocities through DWIs can put reputation, legitimacy, future resources, threat of retaliation, or power itself at stake \[[@pmed-0050243-b024]\].
As comparative rates, DWIs evoke the potential for change. The possibility of becoming "cleaner" may appeal to some offenders \[[@pmed-0050243-b024]\]. Actors may compete for better outcomes relative to military opponents, relative to in-group political competitors, or relative to themselves over time. A DWI\'s potency can be increased by engagement with social, cultural, and religious values of actors and their communities: honor versus dishonor \[[@pmed-0050243-b024]\], gaining versus losing "face," shame versus pride, dignity versus humiliation \[[@pmed-0050243-b037],[@pmed-0050243-b046]\], sacred versus profane \[[@pmed-0050243-b037]\], and valuing mercy and the lives of innocents \[[@pmed-0050243-b047]\]. Terms other than "Dirty War Index," e.g., the "Dishonorable War Index," could be used to greater effect in different contexts.
War and its destruction trigger emotions and self-interests that can obscure analysis by threatening us so that we revert to familiar prejudices, reactions, and cognitive frameworks. Through a public health approach using valid, precise proportional rates as outcomes, DWIs can help us and our audiences to detach from political biases and break through psychological denial when considering actors or methods in war. DWIs can present conflict data from a new perspective, thereby encouraging actors in war to reassess their combat methods, accountability, and interests.
###### Linked Perspectives
This Policy Forum is further discussed in two *PLoS Medicine* Perspectives:
Taback N (2008) The Dirty War Index: Statistical issues, feasibility, and interpretation. PLoS Med 5(12): e248. doi:[10.1371/journal.pmed.0050248](10.1371/journal.pmed.0050248)
Sondorp E (2008) A new tool for measuring the brutality of war. PLoS Med 5(12): e249. doi:[10.1371/journal.pmed.0050249](10.1371/journal.pmed.0050249)
We thank Hamit Dardagan of Iraq Body Count for suggesting the mistreatment of captured combatants as Dirty War Indices and the term "Dishonorable War Index," and for his helpful comments.
Madelyn Hsiao-Rei Hicks is an Honorary Lecturer with the Sections of Community Mental Health and Cultural Psychiatry, Health Service and Population Research Department, Institute of Psychiatry, King\'s College London, London, United Kingdom. Michael Spagat is a Professor in the Department of Economics, Royal Holloway College, University of London, Egham, United Kingdom.
**Funding:**The authors received no specific funding for this article.
**Competing Interests:** The authors have declared that no competing interests exist.
**Provenance:** Not commissioned; externally peer reviewed
df
: degree of freedom
DWI
: Dirty War Index
UXO
: unexploded ordnance
| {
"pile_set_name": "PubMed Central"
} |
Introduction {#S5}
============
Asthma, a chronic respiratory disease characterized by inflammation and bronchospasms, affects both children and adults. Typical symptoms include wheezing, dyspnea, and tachypnea. Characteristics of the disease are frequent exacerbations, which may require systemic therapy and hospitalization. Based on data from the National Health Interview Survey conducted by the Centers for Disease Control and Prevention (CDC), the prevalence of asthma in the United States is 6.3% among males and 9.0% among females ("CDC Asthma Surveillance: Data, statistics and surveillance," 2013). In addition, about half of children with asthma in the United States miss at least one school day each year, accounting for 13.8 million missed school days in 2013 ("Asthma Stats: Asthma-related missing school days among children 5 to 17 years," 2013).
Various social determinants of health affect the burden of asthma, which falls disproportionately on women and African-Americans. These include tobacco use in the home, neighborhood disorders, environmental pollution, pests in the home, stress, socioeconomic status, parental mental health problems, country of origin, and family structure ([@R3]; [@R6]; [@R7]; [@R8]; [@R9]; [@R10]). These factors affect the prevalence of asthma, its clinical outcomes, and the likelihood of complications.
In the current study, the most recent surveillance data were used to report the burden of asthma in the state of Georgia. Also reported are the effects of social determinants of health, including socioeconomic status, race, rurality, usual source of healthcare, insurance status, and education, on the prevalence of asthma.
Methods {#S6}
=======
Since this was a secondary analysis of de-identified, publicly available data under a contract with the Georgia Department of Public Health, institutional review board approval was not required. Data sources for the state of Georgia were from the Behavioral Risk Factors Surveillance Survey (BRFSS), the Georgia Asthma Call-back Survey (ACBS), and the Georgia hospital and emergency department survey for patients with a diagnosis of asthma. All data were from the years 2011 to 2014.
The BRFSS is an annual, stratified, random-digit dial telephone interview conducted by the CDC (<http://www.cdc.gov/brfss/>). Participants are non-institutionalized Georgia residents aged 18 years or older who are asked about their health-related behaviors, chronic conditions, and medical coverage. With the BRFSS, weighted prevalence estimates of asthma among Georgia residents were calculated by use of SAS 9.4 and SUDAAN 11.0.1 software. In addition to common determinants of interest (e.g., sex, age, education), the association of asthma with certain healthcare related variables were also calculated. Univariate and multivariate logistic regressions were performed with SUDAAN to determine the adjusted odds ratios (aOR) of selected variables for the risk of having asthma.
The ACBS is a survey conducted two weeks after a participant has responded to the BRFSS (<http://www.cdc.gov/brfss/acbs/>). In the BRFSS, if the participant answered 'yes' to ever having been diagnosed with asthma, they become eligible for the ACBS but must have given consent for the follow-up call. The ACBS is used to gather additional information from asthma participants regarding their control measures, effect on daily living, occupation, and environment. Weighted prevalence estimates of asthma were calculated using SAS and SUDAAN software.
De-identified data regarding hospital inpatient discharge and emergency department (ED) visits, collected by the Georgia Department of Public Health from non-federal acute care hospitals in Georgia were used to determine the burden of patients hospitalized or seen in EDs with a primary diagnosis of asthma (ICD-9 493.0 - 493.9).
Results {#S7}
=======
[Table 1](#T1){ref-type="table"} summarizes the prevalence of asthma by sex, age, race, and other demographic factors by year. The prevalence of asthma was highest among non-Hispanic blacks, and lowest among Hispanics, although there was a secular trend of increasing prevalence among Hispanics between 2011 and 2014. In 2014, asthma was more prevalent for women (10.0%; 95% CI: 8.7%-11.4%) than for men (6.7%; 95% CI: 5.5%-8.2%) and among persons with less than a high school education (12.0%; 95% CI: 9.2%-15.6%) than college graduates (6.7%; 95% CI: 5.5%-8.1%). An annual household income below \$25,000 was also associated with a higher prevalence of asthma. [Figure 1](#F1){ref-type="fig"} shows the prevalence of asthma by health districts from 2012 to 2014 (a list of districts is also shown). The prevalence was higher in south and northwest Georgia than in other parts of the state.
Regarding clinical and lifestyle factors ([Table 2](#T2){ref-type="table"}), the prevalence of asthma in 2014 was higher among respondents who were current smokers, who had a body mass index (BMI) \> 30 kg/m^2^, and who did not exercise outside of work. It was also higher among those who had poor self-reported physical health (13.4%; 95% CI: 11.6%-15.4% vs 5.7%; 95% CI: 4.8%-6.9%), those with poor self-reported mental health (11.7%; 95% CI: 10.0%-13.8% vs 6.5%; 95% CI: 5.6%-7.7%), and those whose activities were limited due to physical, mental, or emotional problems (17.1%; 95% CI: 14.6%-19.9% vs 6.4%; 95% CI: 5.5%-7.5%).
[Table 3](#T3){ref-type="table"} summarizes data regarding the prevalence of asthma by healthcare access, coverage, and satisfaction with care. Those who had been without coverage for more than three years in 2014 had a higher prevalence of asthma (12.5%; 95% CI: 8.3%-18.3%) than patients who were insured or had been uninsured for less than 6 months. Among persons with asthma ([Table 4](#T4){ref-type="table"}), the percentage of those aged 18 to 64 years without insurance declined from 2012 to 2014, but more than 1 in 5 adults in Georgia with asthma still lacked health insurance. In addition, more than half (54.6%) of those with asthma who did not have insurance had been without it for more 3 years. Each year, about one-third of Georgians with asthma were unable to see a doctor at least once because of cost; a similar percentage each year delayed getting care for reasons other than cost, and more than a third were currently paying off medical bills. Approximately one quarter did not report having a personal physician, and asimilar percentage reported having more than one year since their last check-up. Nevertheless, most asthma patients were either very satisfied (54.7%) or somewhat satisfied (38.9%) with the healthcare that they received. Utilization was high, with 40.8% reporting more than five doctor visits in the previous year.
[Table 5](#T5){ref-type="table"} summarizes the univariate logistical regression analyses. Women had a 1.53 (95% CI: 1.18-1.99) times higher odds of asthma than men, but race and age did not show any significant differences between groups. Higher income and education levels decreased odds of asthma (OR 0.60 if \>\$75k income; OR 0.52 if college graduate). Regarding lifestyle, individuals who reported not exercising in the last 30 days, who were smokers, or who had physical or mental problems had increased odds of asthma.
In the multivariate logistical regression analyses that controlled for socioeconomic variables, smoking, and BMI, education and race were not significantly associated with prevalence of asthma ([Table 6](#T6){ref-type="table"}). Women (adjusted odds ratio \[aOR\] 1.61), smokers (aOR 1.54), and persons with a higher BMI (aOR 1.56) were independently associated with having asthma. Although higher income was generally associated with lower odds of asthma, those with incomes greater than \$75,000 per year were not significantly different from those with incomes less than \$15,000 per year (OR 0.77; 95% CI: 0.46-1.29). However, those with incomes ranging from \$25,000 to \$74,000 per year had decreased adjusted odds of asthma relative to those with incomes less than \$15,000 per year (0.49 to 0.58 versus 1.0, the referent).
In Georgia during 2014, there were 6,601 asthma-related hospitalizations of adults, a rate of 87 per 100,000 and a decrease from 2012 of about 12 per 100,000. ED visits also decreased slightly from 443 per 100,000 (31,929 total ED visits) to 415 per 100,000 (31,595 total ED visits). Although both the total number and rate of ED visits decreased from 2012, the charges for these visits increased by \$14.2 million between 2012 and 2014, from \$68.1 million to \$82.3 million. The rate for hospitalization of women (121/100,000) due to asthma was more than two times the rate for men (49/100,000). The rate of hospitalizations increased as age increased (25/100,000 for those 18-24 years vs 149/100,000 for those 65+ years), whereas ED visits decreased with increasing age (547/100,000 for those 18-24 years vs 195/100,000 for those 65+ years). The rates of hospitalizations and ED visits for African Americans were two and four times higher than rates for Whites, respectively.
Discussion {#S8}
==========
Asthma creates a serious health burden for the state of Georgia, and the impact varies by age, sex, race, educational attainment, and insurance status. The prevalence of asthma decreased with age, although rates of hospitalization and ED use increased, and asthma was more common for patients uninsured for more than 3 years and those with less educational attainment. Asthma was also more common for persons who smoke, those who are obese, and those who did not exercise.
Some of the univariate associations may represent reverse causality (such as the association with exercise), confounding, or effect modification. Multivariate analyses found that female sex, being a current smoker, and having a BMI \>= 30 kg/m2 were independent predictors of having asthma. After adjustment for other clinical variables, race and educational attainment were no longer predictors of asthma, and incomes between \$25,000 and \$74,000 per year were protective. This may be because smoking is more common in persons with lower levels of educational attainment, and lower educational attainment is associated with lower income ([@R4]).
Regarding secular trends from 2011 to 2014, the overall prevalence of asthma remained relatively constant. The prevalence of asthma among Hispanics increased from 3.5% to 9.0% (the latter value similar to that of other groups). This may be due to greater health care access following implementation of the Affordable Care Act and leading to greater ascertainment of disease that had previously been present in the population but had been undiagnosed. Asthma was more prevalent in south and northwest Georgia, rural parts of the state that suffer from challenges to healthcare access and the availability of primary care physicians.
There were some favorable trends. For example, there was a trend toward fewer patients being uninsured, likely due to passage of the Affordable Care Act in 2009 and its rollout across the country in subsequent years. However, many patients with asthma were still uninsured, a problem that could be addressed by expanding access to Medicaid in Georgia. There were also fewer hospitalizations and ED visits, but overall costs increased. The price of metered dose inhalers, even for drugs long on the market, has increased, with out-of-pocket costs rising from \$13.60 to \$25.00 after the ban on chlorofluorocarbon led to reformulation and the loss of generic inhalers ([@R5]).
The results have implications for public health and medical systems in the state of Georgia. Approximately a third of respondents with asthma had delayed seeking care, had not seen a doctor in the past 12 months, or were paying off medical bills. These factors create barriers to appropriate primary care and specialty management of asthma that could prevent hospitalizations and ED visits. Potential solutions include broader access to health insurance, including Medicaid; provision of self-management education; and support for more federally qualified health centers and primary care residency training programs that accept patients with Medicaid as well as those without insurance. These efforts are particularly relevant in the rural south and northwest parts of the state, where there is a shortage of primary care physicians.
Conclusions {#S9}
===========
In Georgia, asthma is independently associated with female sex, obesity, and being a current smoker. Higher rates of tobacco use may mediate the associations with lower educational attainment and low income. The prevalence is also higher in south and northwest Georgia, which are generally rural parts of the state.
The authors thank the staff of the Georgia Department of Public Health for their assistance accessing de-identified data.
{#F1}
###### Prevalence of asthma by key demographic variables based on BRFSS data
Variable 2011 2012 2013 2014 2011-2014
------------------------------------------------------------------------ -------------------------------------- -------------------------------------- -------------------------------------- -------------------------------------- --------------------------------------
**Number of participants** 9895 6068 8082 6309 30,354
**Overall asthma prevalence (95% CI)**[\*](#TFN1){ref-type="table-fn"} 9.6% 8.2% 8.4% 8.4% 8.7%
**Sex**
Male[a](#TFN2){ref-type="table-fn"} 7.5% 5.1% 5.2% 6.7% 6.1%
Female 11.6%[¥](#TFN3){ref-type="table-fn"} 11.0%[¥](#TFN3){ref-type="table-fn"} 11.3%[¥](#TFN3){ref-type="table-fn"} 10.0%[¥](#TFN3){ref-type="table-fn"} 11.0%[¥](#TFN3){ref-type="table-fn"}
**Race**
White, non-Hispanic[a](#TFN2){ref-type="table-fn"} 9.9% 7.7% 7.9% 8.5% 8.5%
Black, non-Hispanic 10.9% 10.7% 10.5% 8.0% 10.0%[¥](#TFN3){ref-type="table-fn"}
Hispanic 3.5%[¥](#TFN3){ref-type="table-fn"} 3.4%[¥](#TFN3){ref-type="table-fn"} 5.9% 9.0% 5.5%[¥](#TFN3){ref-type="table-fn"}
Other, non-Hispanic 7.8% 6.2% 6.6% 9.8% 7.6%
**Age**
18-24[a](#TFN2){ref-type="table-fn"} 10.0% 9.3% 13.0% 9.3% 10.4%
25-34 10.6% 7.8% 6.7%[¥](#TFN3){ref-type="table-fn"} 7.2% 8.1%
35-44 9.3% 6.6% 7.9%[¥](#TFN3){ref-type="table-fn"} 8.1% 8.0%
45-54 9.3% 9.0% 7.3%[¥](#TFN3){ref-type="table-fn"} 8.0% 8.4%
55-64 9.1% 9.0% 9.0% 9.9% 9.3%
65+ 9.3% 7.8% 7.7%[¥](#TFN3){ref-type="table-fn"} 8.3% 8.3%
**Income (yearly)**
\<\$15k[a](#TFN2){ref-type="table-fn"} 16.0% 14.4% 18.0% 12.5% 15.2%
\$15k-\$24k 12.0% 10.9% 8.9%[¥](#TFN3){ref-type="table-fn"} 11.6% 10.9%[¥](#TFN3){ref-type="table-fn"}
\$25k-\$34k 8.6%[¥](#TFN3){ref-type="table-fn"} 7.5%[¥](#TFN3){ref-type="table-fn"} 8.8%[¥](#TFN3){ref-type="table-fn"} 5.9%[¥](#TFN3){ref-type="table-fn"} 7.8%[¥](#TFN3){ref-type="table-fn"}
\$35k-\$49k 8.8%[¥](#TFN3){ref-type="table-fn"} 4.6%[¥](#TFN3){ref-type="table-fn"} 6.0%[¥](#TFN3){ref-type="table-fn"} 6.6%[¥](#TFN3){ref-type="table-fn"} 6.5%[¥](#TFN3){ref-type="table-fn"}
\$50k-\$74k 7.7%[¥](#TFN3){ref-type="table-fn"} 5.9%[¥](#TFN3){ref-type="table-fn"} 5.7%[¥](#TFN3){ref-type="table-fn"} 5.8%[¥](#TFN3){ref-type="table-fn"} 6.3%[¥](#TFN3){ref-type="table-fn"}
\>\$75k 6.5%[¥](#TFN3){ref-type="table-fn"} 5.6%[¥](#TFN3){ref-type="table-fn"} 5.4%[¥](#TFN3){ref-type="table-fn"} 7.8% 6.4%[¥](#TFN3){ref-type="table-fn"}
**Education**
Did not graduate HS[a](#TFN2){ref-type="table-fn"} 13.4% 12.2% 11.0% 12.0% 12.2%
Graduated HS 9.8% 7.4%[¥](#TFN3){ref-type="table-fn"} 8.2% 8.1% 8.4%[¥](#TFN3){ref-type="table-fn"}
Some college 9.0% 8.9% 8.3% 8.2% 8.6%[¥](#TFN3){ref-type="table-fn"}
Graduated college 7.4%[¥](#TFN3){ref-type="table-fn"} 5.8%[¥](#TFN3){ref-type="table-fn"} 7.0%[¥](#TFN3){ref-type="table-fn"} 6.7%[¥](#TFN3){ref-type="table-fn"} 6.7%[¥](#TFN3){ref-type="table-fn"}
All prevalence estimates are weighted.
Reference for group.
Prevalence estimate is significantly different from reference, based on non-overlapping 95% confidence intervals.
###### Asthma prevalence by clinical and lifestyle factors
Variable 2011 2012 2013 2014 2011-2014
--------------------------------------------------------------------------- -------------------------------------- -------------------------------------- -------------------------------------- -------------------------------------- --------------------------------------
**Current smoker**
Yes[a](#TFN4){ref-type="table-fn"} 13.5% 9.8% 10.4% 13.1% 11.7%
No 8.5%[¥](#TFN5){ref-type="table-fn"} 7.6% 8.1% 7.8%[¥](#TFN5){ref-type="table-fn"} 8.0%[¥](#TFN5){ref-type="table-fn"}
**Body mass index (kg/m^2^)**
\<25[a](#TFN4){ref-type="table-fn"} 8.4% 6.6% 7.6% 7.0% 7.4%
\>=25 and \<30 8.3% 7.3% 7.5% 7.4% 7.6%
\>=30 12.9%[¥](#TFN5){ref-type="table-fn"} 10.5%[¥](#TFN5){ref-type="table-fn"} 10.8%[¥](#TFN5){ref-type="table-fn"} 10.7%[¥](#TFN5){ref-type="table-fn"} 11.2%[¥](#TFN5){ref-type="table-fn"}
**Exercise or other physical activity in last 30 days other than job**
Yes[a](#TFN4){ref-type="table-fn"} 8.6% 7.5% 7.6% 7.6% 7.8%
No 12.2%[¥](#TFN5){ref-type="table-fn"} 10.5%[¥](#TFN5){ref-type="table-fn"} 10.3%[¥](#TFN5){ref-type="table-fn"} 11.0%[¥](#TFN5){ref-type="table-fn"} 11.0%[¥](#TFN5){ref-type="table-fn"}
**Activities limited because of physical, mental, or emotional problems**
Yes[a](#TFN4){ref-type="table-fn"} 20.3% 15.9% 16.1% 17.1% 17.5%
No 6.2%[¥](#TFN5){ref-type="table-fn"} 6.1%[¥](#TFN5){ref-type="table-fn"} 6.6%[¥](#TFN5){ref-type="table-fn"} 6.4%[¥](#TFN5){ref-type="table-fn"} 6.3%[¥](#TFN5){ref-type="table-fn"}
**Reported physical health was not good in the last 30 days**
Yes[a](#TFN4){ref-type="table-fn"} 17.7% 14.9% 13.5% 13.4% 14.8%
No 5.7%[¥](#TFN5){ref-type="table-fn"} 4.9%[¥](#TFN5){ref-type="table-fn"} 5.6%[¥](#TFN5){ref-type="table-fn"} 5.7%[¥](#TFN5){ref-type="table-fn"} 5.5%[¥](#TFN5){ref-type="table-fn"}
**Reported mental health was not good in the last 30 days**
Yes[a](#TFN4){ref-type="table-fn"} 15.5% 12.6% 12.6% 11.7% 13.1%
No 6.5%[¥](#TFN5){ref-type="table-fn"} 6.1%[¥](#TFN5){ref-type="table-fn"} 6.4%[¥](#TFN5){ref-type="table-fn"} 6.5%[¥](#TFN5){ref-type="table-fn"} 6.4%[¥](#TFN5){ref-type="table-fn"}
Reference for group
Prevalence estimate is significantly different from reference, based on non-overlapping 95% confidence intervals.
###### Asthma prevalence by healthcare access, coverage, and satisfaction. Data not available for 2011 and 2012 for some variables
Variable 2011 2012 2013 2014 2011-2014 or 2013-2014 [\*](#TFN6){ref-type="table-fn"} (95% CI)
-------------------------------------------------------------------- ------- ------- ------- ------- ------------------------------------------------------------------
**Primary source of health care coverage**
Employer or union 6.9% 6.9% 6.9% (6.1%-7.9%)
Buys own 6.1% 6.8% 6.5% (4.7%-8.8%)
Medicare n/a 12.4% 12.4% (10.3%-14.9%)
Medicaid 17.8% 12.9% 16.2% (12.5%-20.6%)
Tricare 8.0% 7.5% 7.8% (5.5%-11.1%)
Other source 7.9% 13.8% 9.9% (6.9%-13.9%)
None 9.6% 19.6% 10.1% (7.5%-13.4%)
**Consider one person as personal doctor or health care provider**
Yes, only 1 9.8% 8.2% 8.5% 8.6% 8.8% (8.3%-9.3%)
More than 1 13.1% 11.7% 9.2% 11.3% 11.3% (9.7%-13.0%)
No 8.2% 7.0% 7.7% 7.3% 7.6% (6.6%-8.6%)
**Length of time since last check up**
Within the last year 10.2% 8.4% 8.4% 8.5% 8.9% (8.4%-9.4%)
More than 1 year, less than 2 years 8.1% 7.4% 8.7% 9.0% 8.3% (7.0%-9.8%)
More than 2 years, less than 5 years 7.2% 9.0% 8.7% 8.7% 8.4% (6.8%-10.4%)
5 or more years ago 10.0% 7.5% 7.3% 6.6% 7.8% (6.1%-9.9%)
**Did not have health coverage in last 12 months**
Yes 10.9% 6.5% 8.4% (6.2%-11.2%)
No 7.9% 8.8% 8.3% (7.6%-9.1%)
**How long since last had health coverage**
6 months or less 5.5% 6.0% 5.7% (3.3%-9.7%)
More than 6 months, less than 1 year 11.8% 0.4% 6.7% (3.5%-12.5%)
More than 1 year, less than 3 years 7.2% 7.9% 7.6% (5.0%-11.4%)
More than 3 years 13.4% 12.5% 13.0% (10.1%-16.5%)
Never 4.2% 7.6% 5.9% (3.4%-10.1%)
**Number of doctor visits in last 12 months**
1 5.7% 4.5% 5.1% (3.9%-6.7%)
2 7.2% 4.8% 6.0% (4.8%-7.4%)
3 5.9% 9.6% 7.8% (6.2%-9.7%)
4 8.4% 8.7% 8.6% (6.9%-10.6%)
5 11.5% 8.4% 10.0% (7.6%-13.1%)
\>5 14.5% 15.9% 15.2% (13.5%-17.1%)
**Satisfaction with health care received**
Very satisfied 7.6% 7.6% 7.6% (6.9%-8.4%)
Somewhat satisfied 9.7% 9.9% 9.8% (8.6%-11.1%)
Not at all satisfied 12.3% 11.9% 12.1% (9.1%-10.6%)
Results shown are for 2013-2014 where data were not available for 2011 and 2012.
n/a -- Medicare was not an option for respondents in 2013.
###### Healthcare access, coverage, and satisfaction for patients with asthma
Variable 2011 2012 2013 2014 2013-2014
-------------------------------------------------------------------------------- ------- ------- ------- ------- -----------
**Any form of insurance respondents aged 18-64**
Yes 71.7% 65.8% 70.4% 77.7% 73.9%
No 28.3% 34.2% 29.6% 22.3% 26.1%
**Primary source of health care coverage**
Employer or union n/a n/a 48.5% 43.5% 46.0%
Buys on own n/a n/a 9.3% 9.4% 9.4%
Medicare n/a n/a 20.2% 31.4% 25.8%
Medicaid n/a n/a n/a 6.7% 6.7%
Tricare n/a n/a 5.9% 3.6% 4.8%
Other source n/a n/a 5.8% 4.6% 5.2%
None n/a n/a 10.3% 0.8% 5.6%
**Could not see doctor because of cost**
Yes 33.9% 33.1% 32.5% 27.8% 30.1%
No 66.1% 66.9% 67.5% 72.2% 69.9%
**Delayed getting medical care reasons other than cost in the last 12 months**
Yes n/a n/a 38.0% 26.9% 32.4%
No n/a n/a 62.0% 73.1% 67.6%
**Consider one person as personal doctor or health care provider**
Yes, only 1 66.6% 66.1% 64.3% 66.9% 65.6%
More than 1 10.5% 12.6% 9.7% 8.3% 9.0%
No 22.9% 21.4% 26.0% 24.8% 25.4%
**Length of time since last check up**
Within the last year 76.4% 72.6% 73.4% 73.6% 73.5%
More than 1 year, less than 2 years 10.0% 12.1% 11.5% 12.4% 12.0%
More than 2 years, less than 5 years 6.2% 9.1% 7.8% 8.3% 8.0%
5 or more years ago 7.4% 6.3% 7.4% 5.6% 6.5%
**Did not have health coverage in last 12 months**
Yes n/a n/a 9.1% 7.0% 8.0%
No n/a n/a 90.9% 93.0% 92.0%
**How long since last had health coverage**
6 months or less n/a n/a 7.9% 9.0% 8.4%
More than 6 months, less than 1 year n/a n/a 12.4% 0.4% 7.0%
More than 1 year, less than 3 years n/a n/a 14.3% 18.0% 16.0%
More than 3 years n/a n/a 56.3% 52.4% 54.6%
Never n/a n/a 9.1% 20.2% 14.1%
**Number of doctor visits in last 12 months**
1 n/a n/a 14.0% 10.4% 12.2%
2 n/a n/a 17.4% 11.7% 14.5%
3 n/a n/a 10.8% 17.9% 14.3%
4 n/a n/a 11.3% 11.6% 11.5%
5 n/a n/a 8.1% 5.3% 6.7%
\>5 n/a n/a 38.4% 43.1% 40.8%
**Satisfaction with health care received**
Very satisfied n/a n/a 54.7% 54.8% 54.7%
Somewhat satisfied n/a n/a 38.9% 38.0% 38.4%
Not at all satisfied n/a n/a 6.5% 7.2% 6.9%
**Currently paying off medical bills**
Yes n/a n/a 31.3% 34.1% 32.7%
No n/a n/a 68.7% 65.9% 67.3%
###### Univariate Analyses of Selected Variables on Odds for Having Asthma
Variable OR Lower 95% CI Upper 95% CI p-value
--------------------------------------------------------------------------- ------ -------------- -------------- -----------
**Sex**
Male 1.00 reference
Female 1.53 1.18 1.99 \<0.01
**Race**
White, non-Hispanic 1.00 reference
Black, non-Hispanic 0.94 0.71 1.24 0.67
Hispanic 1.06 0.62 1.84 0.82
Other, non-Hispanic 1.17 0.60 2.28 0.65
**Age**
18-24 1.00 reference
25-34 0.76 0.44 1.32 0.34
35-44 0.86 0.51 1.46 0.58
45-54 0.85 0.52 1.40 0.53
55-64 1.08 0.68 1.71 0.76
65+ 0.89 0.57 1.40 0.61
**Income**
\<\$15k 1.00 reference
\$15k-\$24k 0.92 0.62 1.36 0.68
\$25k-\$34k 0.44 0.26 0.74 \<0.01
\$35k-\$49k 0.50 0.03 0.81 0.01
\$50k-\$74k 0.43 0.27 0.70 \<0.01
\>\$75k 0.60 0.40 0.89 0.01
**Education**
Did not graduate HS 1.00 reference
Graduated HS 0.65 0.44 0.94 0.02
Some college 0.65 0.44 0.96 0.03
Graduated college 0.52 0.36 0.75 \<0.01
**Current smoker**
No 1.00 reference
Yes 1.77 1.30 2.41 \<0.01
**Body mass index (kg/m2)**
\<25 1.00 reference
\>=25 and \<30 1.06 0.76 1.48 0.73
\>=30 1.58 1.15 2.16 \<0.01
**Exercise or other physical activity in last 30 days other than job**
Yes 1.00 reference
No 1.51 1.16 1.96 \<0.01
**Limited activities because of physical, mental, or emotional problems**
No 1.00 reference
Yes 3.01 2.33 3.88 \<0.01
**Reported physical health was not good in the last 30 days**
No 1.00 reference
Yes 2.54 1.97 3.29 \<0.01
**Reported mental health was not good in the last 30 days**
No 1.00 reference
Yes 1.90 1.48 2.44 \<0.01
OR = odds ratio; CI = confidence interval
###### Multivariate Analysis of Socioeconomic Variables, Smoking, and BMI on Odds of Asthma
Variable aOR Lower 95% CI Upper 95% CI
----------------------------- ------ -------------- --------------
**Sex**
Male 1.00 reference
Female 1.61 1.19 2.18
**Race**
White, non-Hispanic 1.00 reference
Black, non-Hispanic 0.89 0.64 1.23
Hispanic 1.11 0.57 2.15
Other, non-Hispanic 1.31 0.60 2.87
**Annual Household Income**
\<\$15k 1.00 reference
\$15k-\$24k 0.95 0.63 1.43
\$25k-\$34k 0.49 0.28 0.85
\$35k-\$49k 0.58 0.34 0.98
\$50k-\$74k 0.49 0.28 0.85
\>\$75k 0.77 0.46 1.29
**Educational Attainment**
Did not graduate HS 1.00 reference
Graduated HS 1.08 0.68 1.71
Some college 1.03 0.64 1.67
Graduated college 0.88 0.52 1.51
**Current smoker**
No 1.00 reference
Yes 1.54 1.09 2.19
**Body mass index (kg/m2)**
\<25 1.00 reference
\>=25 and \<30 1.15 0.80 1.65
\>=30 1.56 1.10 2.21
aOR = adjusted odds ratio; CI = confidence interval
| {
"pile_set_name": "PubMed Central"
} |
The dorsal raphe nucleus (DRN) of the mesencephalon/rostral pons contains the majority of neurons in the brain that use serotonin (5-HT) as a neurotransmitter[@b1]. Through activation of a diverse assortment of receptors distributed throughout the forebrain (14 5-HT receptors subtypes have been described to date in mammals), 5-HT released from DRN neurons modulates a broad range of physiological functions including learning and memory, reward and punishment, conditioned fear, stress, sleep and circadian rhythms, and affective behavior[@b2][@b3][@b4][@b5][@b6][@b7][@b8][@b9][@b10]. DRN 5-HT neuronal activity is regulated by the complex interaction between glutamatergic excitatory and GABAergic inhibitory neurotransmission arising from both extra-raphe and local sources[@b11][@b12][@b13][@b14][@b15]. The glutamatergic afferents to the DRN from the medial prefrontal cortex and lateral habenula begin to illustrate this complexity. These afferents target DRN GABAergic interneurons, which in turn inhibit DRN 5-HT neurons via activation of GABA~A~ receptors[@b16][@b17][@b18]. These same cortical and brainstem structures also send direct excitatory inputs to DRN 5-HT neurons[@b14]; determining how these and other afferent signals are integrated by DRN 5-HT neurons remains an ongoing challenge.
In addition to the many glutamatergic afferents from sites in the forebrain and brainstem, the DRN also receives glutamatergic excitatory input from the retina, although details regarding DRN-projecting retinal ganglion cells (RGCs) have only recently begun to emerge[@b19]. RGC projections to the DRN have been described in several species including primates but have been examined most extensively in the Mongolian gerbil. Of the many types of ganglion cell that have been described in the mammalian retina, in the gerbil more than 85% of the DRN-projecting RGCs have alpha morphologic and Y-like physiologic characteristics[@b20]. The DRN-projecting alpha/Y-like RGCs are very similar to the classic alpha/Y RGC type described originally in the cat retina (i.e., large soma, transient responses and non-linear spatial summation) and now considered to be a highly conserved RGC type found in species ranging from frogs to primates[@b21].
Subpopulations of alpha/Y RGCs have dendrites that stratify in either the proximal or distal inner plexiform layer of the retina corresponding to their ON-center and OFF-center receptive fields (ON RGCs increase firing in response to light increments whereas OFF RGCs increase firing in response to light decrements). OFF alpha/Y RGCs have narrower receptive-field centers and thus to cover the retina OFF cells typically outnumber their ON partners nearly 2-fold[@b22]. In the gerbil retina, both ON-center and OFF-center alpha/Y-like RGCs project to the DRN but the proportion of ON and OFF cells is strongly biased toward ON cells (≈4 ON: 1 OFF)[@b20]. The predominance of ON alpha/Y-like RGCs innervating the DRN is therefore unexpected considering the prevalence of OFF alpha/Y RGCs in the retina. It has been suggested that the excess of OFF cells in the retina is related to information available in natural scenes which contain more regions of negative than positive contrast[@b22]. It would seem that principles that apply to retinal organization regarding image-forming visual circuits may not apply equally to non-image forming retinal circuits that supply information to non-visual targets in the brain such as the DRN. Perhaps the DRN neurons that ON and OFF alpha/Y-like RGCs communicate with (synapse on) is more important in determining their relative proportion in this circuitry than is the contrast information in the natural scene.
Currently the phenotype of DRN neurons innervated by ON and OFF alpha/Y-like RGCs is unknown. Indirect evidence suggests that DRN-projecting OFF alpha/Y-like cells may contribute to the regulation of DRN 5-HT production and/or neurotransmission. Apoptosis of the outer retina can be induced by *N*-methyl-*N*-nitrosourea (MNU), an alkylating agent that results in the rapid elimination of all rod and cone photoreceptors after a single parenteral injection[@b23][@b24]. In the absence of rods and cones the spontaneous firing rate of all OFF RGCs increases dramatically[@b25][@b26]; when photoreceptor to ON bipolar synaptic transmission is blocked pharmacologically spontaneous activity in ON RGCs is abolished[@b27][@b28]. In MNU-treated animals with OFF RGCs firing at a high rate, it was found that DRN serotonin levels were increased while depressive-like behavior was reduced; silencing OFF RGCs or eliminating DRN-projecting RGCs reversed these effects[@b26]. These findings suggest a potential direct OFF alpha/Y-like RGC projection to DRN 5-HT neurons.
Presently we sought to determine if DRN-projecting RGCs innervate 5-HT and/or GABA neurons. We used intraocular injection of Cholera toxin B subunit (CTB) to label retinal processes in the DRN in conjunction with immunostaining for the glutamatergic receptor postsynaptic scaffold protein PSD-95, combined with immunocytochemical identification of 5-HT and GABA neurons and confocal microscopy. DRN 5-HT and GABA activity was also evaluated using c-Fos expression in animals maintained under different lighting conditions and in the model of MNU-induced photoreceptor apoptosis in which OFF RGC spontaneous firing rate is increased. We report a previously unknown level of complexity in the retinal input to the DRN: the less numerous DRN-projecting OFF Y RGCs primarily activate 5-HT neurons, whereas the more numerous DRN-projecting ON Y RGCs mainly activate GABAergic neurons.
Materials and Methods
=====================
Animals
-------
Male Mongolian gerbils (*Meriones unguiculatus*) (2--3 months old, 65--77 g) were used. Animals were individually housed and maintained in 12 light: 12 dark (LD) conditions (lights on at 0900 h) with food and water provided *ad libitum*. All procedures were performed in accordance with Jinan University guidelines for animal research and the *Association for Research in Vision and Ophthalmology* Statement for the Use of Animals in Ophthalmic and Vision Research. The experimental animal protocol was approved by Jinan University Institutional Animal Care and Use Committee.
Light stimulation
-----------------
Animals were randomly assigned to two groups: 1) a light stimulation group (L), in which the animals were kept in complete darkness overnight from light offset at 2100 h until 0900 h the next day when they were exposed to 90 min of white light (3000 lux) in their home cage; and 2) a dark control group (D) was treated similarly except lights remained off. At 1030 h, all animals were anesthetized (0.25 g/kg, Tribromeoethanol, IP) and perfused transcardially with 0.9% saline followed by 4% paraformaldehyde (PFA) in 0.1 M phosphate-buffered saline (PBS). Brains and eyes were removed and c-Fos immunocytochemistry was performed (see below). c-Fos expression was determined at 90 min after light onset based on the literature and our previous work examining light-induced c-Fos expression in the brain and retina[@b29][@b30].
Specific inhibition of the retinal ON pathway
---------------------------------------------
Specific inhibition of the retinal ON pathway was performed as previously described[@b29]. Briefly, at approximately 0840 h animals were anesthetized by isofluorane (2.5--5%) inhalation anesthesia. Under dim red light (5 lux), 2 μl of 1 mM L-AP4 (L-(+)-2-amino-4-phosphonobutyric acid; Tocris Bioscience) in sterile 0.9% saline was injected into the vitreous body of both eyes using a glass micropipette attached to a Nanoject II (Drummond Scientific). After recovering from anesthesia (≈10 min), animals were exposed to 90 min of light stimulation and killed at 1030 h as described above.
Selective elimination of retinal photoreceptors
-----------------------------------------------
Selective elimination of rod and cone photoreceptors was performed as previously described[@b26]. Briefly, animals were anesthetized, and then received an IP injection of MNU (N1517, Sigma-Aldrich) (80 mg/kg) and returned to their home cage maintained under LD conditions for at least 7 days. Some animals were killed 7 days after MNU-treatment and 40 μm cryostat sections were stained with DAPI to document photoreceptor loss.
Physiological Recording of RGCs
-------------------------------
CTB injections into the DRN and recording of DRN-projecting RGCs were previously described[@b26]. Recorded CTB-labeled RGCs that showed no spontaneous activity (n = 37) in MNU-treated animals were filled and subsequently determined to have dendrites stratifying in the proximal inner plexiform layer of the retina (i.e., ON RGCs). Recorded CTB-labeled RGCs that were spontaneously active (n = 10) in MNU-treated animals were filled and subsequently determined to have dendrites stratifying in the distal inner plexiform layer of the retina (i.e., OFF RGCs).
Anterograde labeling of axonal terminals of dorsal raphe projecting retinal ganglion cells
------------------------------------------------------------------------------------------
Briefly, animals were anesthetized, 0.5% Proparacaine hydrochloride was applied to the cornea, and a 35 g needle attached to a 5.0 μl Hamilton microsyringe was inserted intravitreally at the temporal cornea-conjunctival margin. 2.0 μl of 2% (w/v) CTB-conjugated Alexa Fluor 488 (C-22841, Molecular Probes, Invitrogen) dissolved in 2% dimethyl sulfoxide were slowly injected over 3 min. The needle was held in place for 5 min, withdrawn and the injected site was washed with saline and Bacitracin was applied.
Immunocytochemistry
-------------------
All animals were anesthetized (0.25 g/kg, Tribromeoethanol, IP) and perfused intracardially with 0.9% saline followed by 4% paraformaldehyde in phosphate-buffered saline (PBS). Brains and eyes were removed.
Double and triple-labeling was performed on free-floating 40 μm thick cryostat sections incubated in blocking solution for 1 h before primary antibodies were applied. For c-Fos and TPH or GABA double-labeling, were placed in blocking solution for 1 h before incubation in a mixture of primary antibodies against c-Fos (rabbit, 1:30,000; PC38T, Calbiochem) and tryptophan hydroxylase (mouse, TPH; 1:1000; T8575, Sigma-Aldrich), or c-Fos and GABA (mouse, 1:1000; A0310, Sigma-Aldrich) (36 h at 4 °C). Sections were then incubated with corresponding secondary antibodies at a dilution of 1:400 for 6 h at room temperature: goat anti-rabbit Alexa 488(107909, Jackson ImmunoResearch) and goat anti-mouse Alexa 594 (115-587-003, Jackson ImmunoResearch), and cover-slipped in anti-fading aqueous mounting medium (EMS, Hatfield, PA).
For triple-labeling of CTB/TPH/PSD-95 or CTB/GABA/PSD-95, 7 days after CTB intraocular injection animals were perfused with 0.9% saline followed by 4% paraformaldehyde in PBS. After blocking solution a mixture of primary antibodies either against TPH and PSD-95 (rabbit, 1:200, 51-6900, Invitrogen), or GABA and PSD-95 was applied for 36 h (4 °C) and then secondary antibodies at a dilution of 1:300 for 6 h at room temperature: goat anti-mouse Alexa 594 (as above) and donkey anti-Rabbit Alexa 647 (A31573, Life technologies), rinsed in 0.1 M PBS 6X 10 min and cover-slipped (as above).
Retinal whole mounts from a subset of animals used for triple-labeling experiments were immunostained for c-Fos expression. After the animals were anesthetized, the eyes were removed and the retina was dissected from the eyecup, mounted on filter paper, post-fixed 1 h in 4% PFA, removed from the filter paper, washed in 0.1 M PBS 3X 10 min, incubated in CAS-Block^TM^ (1673905A, Life technologies) containing 0.3% Triton-X-100 for 1 h, incubated in c-Fos antibody (as above) for 48 h at 4 °C followed by 3X PBS before incubation with Dylight 488 goat-anti-rabbit IgG (Vector Laboratories) at 1:400 for 6 h at room temperature.
The method for immunocytochemical staining of cholinergic amacrine cells in the retina was performed as previously described[@b31]. Briefly, retinas were fixed for 1 hour in 4% paraformaldehyde in 0.1 M PBS followed by rinsing in 0.1 M PBS (3X 10 min) and placed in 10% normal goat serum containing 2% Trition-X-100 for one hour at room temperature. Retinas were then incubated in goat-anti-ChAT antibody (1:200, AB144P, Millipore) for 48 hours at 4 °C. This was followed by 6 times rinsing in 0.1 M PBS (6X 10 min) and then incubation with a secondary antibody Alexa Fluor 594 donkey anti-goat IgG (1:400, A-11058, Molecular Probes) for 6 hours at room temperature. Finally, all retinas were rinsed in 0.1 M PBS and cover-slipped (as above).
Confocal microscopy and three-dimensional reconstruction
--------------------------------------------------------
Images were collected with a confocal laser-scanning microscope (Zeiss, LSM700). For triple-labeling the Z-axis interval was set at 0.3 μm and areas of interest were scanned with a 63X oil immersion objective, and zoomed 4 times by digital magnification. A montage of optical section stacks was created, projected to a 0° X--Y plane and a 90° X--Z plane to obtain a 3-D reconstruction. Using Image J and Photoshop CS5 (Adobe Corp., San Jose, California, USA), contrast and brightness were adjusted and images were pseudo-colored.
Statistical analysis
--------------------
For quantification of c-Fos+ cells and c-Fos+ GABA+ or TPH+ cells, 4 serial sections through the caudal DRND/L were examined/animal and analyzed using a one-way ANOVA. Data are expressed as the mean ± SEM. Statistical significance was set at p \< 0.05.
Results
=======
Retinal ganglion cells form synaptic connections with GABA and serotonin neurons in the DRN
-------------------------------------------------------------------------------------------
We examined first whether axonal varicosities of DRN-projecting RGCs formed synaptic connections with DRN 5-HT and/or GABA neurons. Following intraocular injection of CTB-Alexa Fluor 488, labeled axonal processes were found distributed mainly in the dorsal and lateral subdivisions of the caudal DRN (DRND/L)[@b32] ([Fig. 1A](#f1){ref-type="fig"}). Only a relatively small number of CTB-labeled retinal fibers were observed in individual tissue sections through the DRND/L. However, it is very likely that the density of retinal fibers in the DRN is under represented by CTB-Alexa Fluor 488 anterograde tracing[@b33]. Tissue sections of the DRND/L with CTB-labeled processes were subjected to immunocytochemical labeling for tryptophan hydroxylase (TPH), the rate limiting enzyme in 5-HT synthesis, or GABA, together with the postsynaptic protein PSD-95, a scaffolding protein associated with glutamatergic receptors. CTB-labeled varicosities of DRN-projecting RGCs were noted to form synaptic connections with 5-HT and GABA neurons ([Fig. 1B--K](#f1){ref-type="fig"}). Varicosities of the glutamatergic CTB-labeled retinal processes abutted the postsynaptic PSD-95 protein and contacts were confirmed in rotated and magnified images ([Fig. 1D--F,I--K](#f1){ref-type="fig"}) suggesting strongly that DRN-projecting RGCs form synaptic contacts with 5-HT and GABA neurons in the DRND/L.
Light stimulation induces c-Fos expression in the DRN
-----------------------------------------------------
After establishing that DRN-projecting RGCs were in direct contact with 5-HT and GABA neurons in the DRND/L using morphological procedures, we sought to determine whether RGC input to the DRN regulates neuron activity using light-induced c-Fos expression as a functional readout. Two groups of animals (n = 10/group) were used: animals in the light stimulated group (L) were sacrificed 90 minutes after light onset whereas the control group (D) remained in the dark and was sacrificed at the same time of the day; the DRND/L in both groups was examined for c-Fos expression. The number of c-Fos positive (+) neurons/animal in the DRND/L in the L group was significantly greater than the number of c-Fos positive neurons in the D group (37.4 ± 3.6 vs 201.6 ± 8.1; n = 10/group; p \< 0.0001); light stimulation (3000 lux) produced a \>5-fold increase in c-Fos+ expressing neurons in the DRND/L ([Fig. 2G](#f2){ref-type="fig"}).
To investigate whether 5-HT and/or GABA DRND/L neurons expressed c-Fos in response to light exposure, five animals from each D and L group were randomly selected and further examined for c-Fos and GABA co-localization and the remaining 5 animals in each group were analyzed for c-Fos and TPH co-localization. A vast majority (87%) of the c-Fos+ cells in the D group were either GABA+ or TPH+: 79.2 ± 2.8% of c-Fos+ cells were GABA+ neurons and 7.8 ± 2.2% of c-Fos+ cells were TPH+ neurons ([Fig. 2A,D,H](#f2){ref-type="fig"} and [Table 1](#t1){ref-type="table"}). Light exposure increased the number of c-Fos+/GABA+ cells ≈5-fold compared to the D group (35.6 ± 4.1 vs 171.2 ± 16.9, n = 5/group; p \< 0.001), whereas the number of c-Fos + /TPH+ cells increased ≈13-fold following light stimulation (2.6 ± 0.9 vs 33.0 ± 3.2, n = 5; p \< 0.001) ([Fig. 2B,E,H](#f2){ref-type="fig"} and [Table 1](#t1){ref-type="table"}). Thus, light exposure significantly increased the number of GABA+ and TPH+ DRND/L cells expressing c-Fos compared to animals in the dark. Although light had a disproportionately greater effect on 5-HT neurons (13-fold increase vs 5-fold increase in GABA cells, [Fig. 2H](#f2){ref-type="fig"}), following light stimulation the number of c-Fos + /GABA+ cells (171.2 ± 16.9) was significantly greater than the number of c-Fos + /TPH+ cells (33.0 ± 3.2) ([Table 1](#t1){ref-type="table"}). This could mean that GABA cells receive a greater number of RGC inputs than TPH cells and/or that activated GABA cells inhibit TPH cells. Since DRN-projecting ON Y-like RGCs outnumber DRN-projecting OFF Y-like RGCs approximately 4-fold, ON DRN-projecting RGCs may preferentially innervate DRN GABA neurons. To begin to break down the contributions of ON and OFF DRN-projecting RGCs in the regulation of neurons in the DRND/L, additional experiments were conducted.
ON and OFF retinal pathways differentially regulate DRN GABA and serotonin neurons
----------------------------------------------------------------------------------
Two different types of experiments were performed: the first examined light-induced Fos expression in the DRN following acute pharmacological suppression of the retinal ON pathway whereas the second examined Fos expression in the DRN after elimination of rods and cones resulting in chronic silencing of ON RGCs while concurrently increasing the spontaneous activity of OFF RGCs.
### L-AP4 blockade of the retinal ON pathway
It is well documented that L-AP4, an agonist of the mGluR6 receptor blocks transmission of visual signals between photoreceptors and ON-bipolar cells thus blocking input to ON RGCs[@b34]. To examine the role of OFF RGC input to the DRN in the absence of ON RGC input, 10 animals were intraocularly treated with 1 mM L-AP4 and light-induced c-Fos expression was examined in the DRN. L-AP4 application significantly decreased the total number of light induced c-Fos+ cells in the DRND/L compared to untreated light-stimulated control animals (97.0 ± 8.6 vs 201.6 ± 8.1; p \< 0.001) and this was due to the number c-Fos + /GABA+ being reduced to levels comparable to control animals in the dark (35.6 ± 4.1vs 28.8 ± 3.6). Conversely, the number of c-Fos + /TPH+ cells in the DRND/L was significantly increased in L-AP4 treated animals compared to untreated light-stimulated control animals (33.0 ± 3.2 vs 68.0 ± 5.6, p \< 0.001) ([Fig. 2H](#f2){ref-type="fig"} and [Table 1](#t1){ref-type="table"}). These results are consistent with the interpretation that DRN-projecting alpha/Y-like RGCs innervate both GABA and 5-HT cells in the DRND/L but that ON DRN-projecting RGCs predominately innervate GABAergic neurons.
### MNU-induced rod and cone apoptosis
Abolishing the maintained firing activity in the ON pathway following acute L-AP4 application may have also acutely increased the spontaneous firing rate of OFF RGCs; L-AP4 has been reported to increase OFF alpha cell firing rate ≈2-fold in an *in vitro* preparation[@b27][@b28]. To investigate further the role of ON and OFF DRN-projecting RGCs in the regulation of neural activity in the DRND/L in a model in which RGC activity is chronically altered, 20 animals received a single parenteral injection of MNU, which could result in the rapid and specific ablation of rod and cone photoreceptors in the outer retina[@b26]. Ablation of rods and cones results in silencing of ON RGCs and chronic activation of DRN-projecting OFF RGCs ([Fig. 3](#f3){ref-type="fig"})[@b26]; 16 animals injected with a similar volume of saline served as controls. Saline and MNU-treated animals were examined for c-Fos and GABA or TPH co-localization in the dark and after light stimulation.
The number of c-Fos+ cells in the DRND/L of MNU-D group was significantly increased (≈13-fold) over the number of c-Fos+ cells in saline-D control group (44.8 ± 2.6 vs 566.5 ± 8.3, p \< 0.001), and the number of c-Fos+ cells in MNU-L group was significantly increased (≈3 fold) over the number of c-Fos+ cells in saline-L group (201.2 ± 8.2 vs 556.6 ± 6.8, p \< 0.001) ([Fig. 4E](#f4){ref-type="fig"}; [Table 1](#t1){ref-type="table"}). The expression of c-Fos in the DRND/L was indistinguishable between the MNU-L (556.6 ± 6.8) and MNU-D (566.5 ± 8.3) groups ([Fig. 4E](#f4){ref-type="fig"}; [Table 1](#t1){ref-type="table"}). Thus, MNU-treatment evoked greater DRND/L c-Fos expression than did light stimulation. Since MNU-treatment renders DRN-projecting ON RGCs silent ([Fig. 3](#f3){ref-type="fig"}) while increasing the spontaneous firing rate of DRN-projecting OFF RGCs approximately 6-fold[@b26], the increase in DRND/L c-Fos expression in MNU-treated animals is attributed to the increased spontaneous activity of DRN-projecting OFF RGCs.
Immunocytochemical analyses of c-Fos and TPH or GABA co-labeling revealed that MNU treatment increased the number of c-Fos + /GABA+ cells only slightly over saline-L values (saline-L: 165.7 ± 7.8 vs MNU-L: 199.6 ± 5.3), whereas the number of DRND/L c-Fos + /TPH+ neurons was increased 10-fold (saline-L: 34.7 ± 3.8 vs MNU-L: 343.6 ± 12.5) ([Fig. 4](#f4){ref-type="fig"}; [Table 1](#t1){ref-type="table"}). Since MNU-treatment specifically increases the firing rate of OFF DRN-projecting RGCs, these RGCs appear to preferentially innervate DRN 5-HT neurons. Thus, in MNU-D or MNU-L animals, DRND/L c-Fos + /TPH+ cells outnumbered light activated DRN c-Fos + /GABA+ cells almost two-to-one ([Table 1](#t1){ref-type="table"}). The opposite is the case in the light exposed animals with intact retinas (saline-L), where DRND/L c-Fos + /GABA+ cells outnumber c-Fos + /TPH+ cells five-to-one ([Table 1](#t1){ref-type="table"}).
The retinas of the L-AP4-treated animals, saline control animals and MNU-treated animals lacking photoreceptors in the outer retina ([Fig. 5A,E](#f5){ref-type="fig"}) were also examined for c-Fos expression in the ganglion cell layer. In retinal whole mounts there were more c-Fos+ cells in the saline-L, L-AP4-L, MNU-D and MNU-L compared to the saline-D controls as expected ([Fig. 5](#f5){ref-type="fig"}). There was also no difference in the number of c-Fos+ cells in the retinas of MNU-D and MNU-L animals consistent with the absence of a response to light due to the ablation of outer retina photoreceptors ([Fig. 4H](#f4){ref-type="fig"} and [Table 1](#t1){ref-type="table"}). In addition, compared with saline-L group, the number of c-Fos+ cells in the retinas of MNU-treated animals was significantly decreased ([Fig. 5H](#f5){ref-type="fig"} and [Table 1](#t1){ref-type="table"}). The reduced expression of c-Fos in the retinas of MNU-treated animals is due to the fact that MNU treatment renders all ON RGCs silent while activating OFF RGCs ([Fig. 3](#f3){ref-type="fig"}) whereas light stimulation activates both ON and OFF RGCs.
The anatomical and functional data described above taken together with our previous work[@b20][@b26] and the known role of DRN GABAergic interneurons suggest that most ON DRN-projecting RGCs innervate DRND/L GABA neurons which in turn innervate serotonin neurons. Most OFF DRN-projecting RGCs innervate DRND/L serotonin neurons directly ([Fig. 6](#f6){ref-type="fig"}).
Discussion
==========
In a previous study we demonstrated for the first time that retinal input to the DRN can modulate 5-HT levels and depressive-like behavior[@b26]. These observations prompted us to investigate the postsynaptic targets of retinal afferents in the DRN. In the present set of experiments we demonstrated that CTB-labeled RGC presynaptic terminals are in close (synaptic) apposition with the postsynaptic protein PSD-95 (which is associated with glutamatergic receptors) on both 5-HT and GABA neurons in the DRND/L. In addition, we observed that light stimulation evoked increases in c-Fos expression in the DRND/L almost exclusively in 5-HT and GABA neurons; more than 97% of Fos+ cells were GABA or 5-HT neurons. We also report robust c-Fos expression in DRND/L 5-HT neurons in MNU-treated animals suggesting that the population of OFF DRN-projecting RGCs preferentially innervates 5-HT neurons whereas the majority of ON DRN-projecting RGCs (ON DRN-projecting RGCs outnumber OFF DRN-projecting RGCs cells 4:1) preferentially innervates GABA neurons. It is also likely that DRND/L GABA neurons receiving direct input from ON (and OFF) RGCs inhibit DRND/L 5-HT neurons. The anatomical and functional data presented herein taken together with our previous work suggest that RGCs differentially innervate DRND/L 5-HT and GABA neurons thereby adding to the complex glutamatergic regulation of DRN 5-HT activity.
The primary excitatory drive of DRN 5-HT neurons comes from a wide range of glutamatergic inputs from cortical and subcortical regions[@b12] and as demonstrated in this study, the retina. The origins of glutamatergic afferents to the DRN can be distinguished by the presence of the specific type of vesicular glutamate transporter used to load glutamate into synaptic vesicles. Axons from subcortical structures and the retina express the VGLUT2 isoform[@b35][@b36][@b37][@b38][@b39][@b40] and among axons in the DRN expressing VGLUT1, VGLUT2 or VGLUT3, those expressing VGLUT2 are found most often in close apposition to PSD-95 on 5-HT and non-5-HT neurons[@b41]. Thus, our data are consistent with previous work indicating that glutamatergic axons expressing VGLUT2 provide excitatory synaptic input to 5-HT and non-5-HT (e.g., GABA) neurons in the DRN.
In the current study, evidence is provided that the retina projects directly to 5-HT and GABA neurons in the DRN. Although the gerbil retinoraphe projection is robust compared to that described for other species the number of CTB-labeled retinal processes observed in the DRN is relatively low[@b19]. It is very likely however, that retinal processes in the DRN are under-represented when CTB is used as an anterograde tracer. For example, labeled retinal fibers were not reported in the DRN following intraocular CTB injection in the cat whereas a retinoraphe projection was unambiguously described in the cat using intraocular injection of ^3^H-amino acids and by retrograde labeling of RGCs after horseradish peroxidase was injected into the DRN[@b42][@b43]. We have also recently reported that retrograde labeling of DRN-projecting RGCs reveals more cells than predicted from CTB anterograde labeling of retinal fibers in the DRN[@b44]. When comparing retinal afferent fiber labeling in the rat DRN using different tracers, it was reported that CTB was a less sensitive tracer than CTB conjugated to HRP[@b33]. Since retinal fibers are likely under-reported following intraocular CTB application, no attempt was made to quantify retinal synapses on neurons in the DRN. High-resolution neuroanatomical techniques could be employed to quantify retinal synapses in the DRN[@b12] if the apparent under representation of the retinoraphe projection can be overcome using more sensitive approaches.
5-HT and GABA neurons in the DRND/L were activated by light. However, it is well documented that GABA neurons predominate in the DRND/L and play a pivotal role in the modulation of serotonergic tone through direct synaptic connections with 5-HT neurons[@b45][@b46][@b47]. Thus, in our simple model presented in [Fig. 6](#f6){ref-type="fig"}, we postulate that DRND/L 5-HT neurons may receive indirect inhibitory retinal input via GABAergic interneurons and direct excitatory retinal input, similar to the indirect inhibitory input and direct excitatory inputs to the DRN from the medial prefrontal cortex and lateral habenula[@b14]. In other studies examining c-Fos expression in the DRN following diverse non-light related stimulation (e.g., REM sleep, swim stress, depression), c-Fos expression was observed almost exclusively in DRN GABA neurons, which could then modulate serotonergic tone through inhibitory synaptic connections with 5-HT neurons[@b48][@b49][@b50][@b51]. The number of GABA and 5-HT cells expressing c-Fos suggests that ON and OFF DRN-projecting RGCs project to both GABA and 5-HT neurons but that ON RGCs primarily innervate GABA neurons and OFF RGCs primarily innervate 5-HT neurons in the DRND/L ([Table 1](#t1){ref-type="table"}). It is possible that ON and OFF RGCs might even target the same DRN cells. The specific synaptic architecture between glutamatergic retinal afferents and their postsynaptic targets, such as the presynaptic gating of glutamatergic inputs by GABA in the DRN[@b13], remains to be determined.
OFF RGCs increase their spontaneous firing rate when synaptic input via rods and cones is severely reduced or eliminated following degeneration of rods and cones[@b25], pharmacological blockade[@b27][@b28] or after MNU-induced apoptosis of the outer retina[@b26]. MNU has been shown to specifically target rod and cone photoreceptors in several species causing rapid apoptosis while sparing RGCs[@b23][@b24][@b26][@b52] ([Fig. 5](#f5){ref-type="fig"}). In MNU-treated animals, acute silencing of OFF RGC firing reverses the effects on depressive-like behavior[@b26]. Thus, c-Fos expression observed in the DRN in MNU-treated animals is attributed to increased activity of OFF DRN-projecting RGCs rather than an indirect effect of MNU treatment.
ON DRN-projecting Y RGCs dominate the retinoraphe projection even though OFF Y cells out number ON Y cells 2:1. The reason for this disparity in ON and OFF DRN-projecting RGCs is unknown. OFF RGCs cells are more numerous in the retina because natural images contain more dark information and thus the retina apparently devotes more resources to processing dark contrasts[@b22]. Since the retinoraphe projection is not a 'visual' pathway concerned with visual information, the retinoraphe projection may not need to follow the same pattern as classic retinal circuits. On the other hand, the ON and OFF RGC projection to the DRN shares some features with the visual projections to the thalamus. ON and OFF pathways that begin in the retina maintain their segregation when they innervate neurons in the dorsal lateral geniculate nucleus (dLGN)[@b53]. The postsynaptic targets of ON and OFF RGCs in the dLGN are glutamatergic relay neurons and GABAergic interneurons[@b54]. ON RGCs innervate dLGN GABAergic interneurons that synapse on dLGN relay neurons that receive direct OFF RGC input[@b55].
DRN serotonergic neurons project in a topographic manner to widespread targets throughout the forebrain and lower brainstem. Serotonergic neurons in the DRND/L project to several structures of the central visual system including the dLGN, superior colliculus, and primary visual cortex where serotonin actively shapes the responses property of visual system networks[@b19]. In addition, serotonergic neurons in the DRND/L appear to exert inhibitory control over serotonergic projection neurons in the dorsal (DRD) and ventral (DRV) DRN subgroups, acting as serotonergic interneurons[@b11][@b56]. Thus, although the retinoraphe projection is strongest to the DRND/L, DRN-projecting RGCs may modulate the activity serotonergic neurons in other regions of the DRN either by direct input or indirectly via serotonergic and GABAergic interneurons thereby contributing to the serotonergic modulation of diverse processes.
Our understanding of the intra-DRN circuitry is far from complete and as new information is gathered our simple model of ON and OFF RGC input to the DRN will be refined. Recent technological advances such as optogenetic and *in vivo* fiber-optic recordings[@b57] will open a multitude of opportunities for functional studies of the neural basis underlying retinal regulation of DRN functions.
Additional Information
======================
**How to cite this article**: Zhang, T. *et al*. ON and OFF retinal ganglion cells differentially regulate serotonergic and GABAergic activity in the dorsal raphe nucleus. *Sci. Rep.* **6**, 26060; doi: 10.1038/srep26060 (2016).
This work was supported by grants from the National Natural Science Foundation of China (31400942), the National Program on Key Basic Research Project of China (973 Program, 2014CB542205), Guangdong Natural Science Foundation (2014A030313387 and S2013040014831), funds of Leading Talents of Guangdong (2013), Programme of Introducing Talents of Discipline to Universities (B14036), the Fundamental Research Funds for the Central Universities Grant (21609101).
**Author Contributions** Conceived and designed the experiments: C.R. and K.-F.S. Performed the experiments: T.Z., L.H. and M.J.T. Analyzed the data: T.Z., C.R., M.P., L.Z. and G.E.P. Contributed reagents/materials/analysis tools: C.R., K.-F.S. Contributed to the writing of the manuscript: C.R. and G.E.P.
{#f1}
{#f2}
{#f3}
{#f4}
{#f5}
{#f6}
###### Quantification of c-Fos-expressing cells in the DRND/L and retina.
Dark Light L-AP4-Light Saline-Dark Saline-Light MNU-Dark MNU-Light
---------------------------- -------------------- --------------------- ---------------------- ------------------- --------------------- --------------------- ---------------------
c-Fos+ 37.4 ± 3.6(n = 10) 201.6 ± 8.1(n = 10) 97.0 ± 8.6(n = 10) 44.8 ± 2.6(n = 8) 201.2 ± 8.2(n = 8) 566.5 ± 8.3(n = 10) 556.6 ± 6.8(n = 10)
c-Fos + /GABA+ 35.6 ± 4.1(n = 5) 171.2 ± 16.9(n = 5) 28.8 ± 3.6(n = 5) 40.6 ± 3.7(n = 4) 165.7 ± 7.8(n = 4) 205.8 ± 7.3(n = 5) 199.6 ± 5.3(n = 5)
c-Fos + /TPH+ 2.6 ± 0.9(n = 5) 33.0 ± 3.2(n = 5) 68.0 ± 5.6(n = 5) 2.4 ± 0.6(n = 4) 34.7 ± 3.8(n = 4) 331.2 ± 10.2(n = 5) 343.6 ± 12.5(n = 5)
c-Fos+ retinal cells/mm^2^ --- --- 121.2 ± 16.4(n = 10) 14.2 ± 2.8(n = 5) 935.8 ± 23.5(n = 5) 192.5 ± 14.6(n = 5) 197.6 ± 10.7(n = 5)
All values = mean ± SEM; n = animals or retinas/group.
[^1]: These authors contributed equally to this work.
| {
"pile_set_name": "PubMed Central"
} |
INTRODUCTION {#s1}
============
The 5-year survival rate of women with breast cancer has been improving steadily for the past 5 years, but metastasis to distant sites still limits survival for patients who have undergone radical surgery \[[@R1], [@R2]\]. Approximately one-third of women with early-stage breast cancer develops distant metastasis and suffers a tumor-related death \[[@R3]--[@R6]\].
Endocrine therapy as an adjuvant to prevent and treat metastatic breast cancer is generally well tolerated and efficient. However, the response to endocrine agents such as tamoxifen and aromatase inhibitors, whether partial or complete, is closely dependent on the hormone receptor (HR) status in cancer tissues \[[@R7], [@R8]\]. Patients whose primary breast tumors are positive for estrogen and progesterone receptors (ER and PR, respectively) are preferred candidates for salvage hormonal therapy, to attenuate progression of the disease.
In routine clinical practice, options for systemic therapy depend on the characteristics of the primary tumor, determined by routine histopathology, immunohistochemistry (IHC), or molecular analysis. However, the choice of treatment may better rest on features of the metastasized lesions rather than that of the primary tumors, in particular with regard to differences in HR status. Differences between the HR status of the primary tumors and metastatic lesions, termed receptor conversion, have been confirmed in 18%-54% of breast cancer patients \[[@R9]--[@R20]\]. These results support the necessity for biopsies of metastasized lesions, since the specific HR status of these may alter the choice of therapeutic regimen for these patients \[[@R21], [@R22]\].
ER and PR levels in primary breast cancers may be important indicators. For example, ER and PR positivity were associated with better treatment outcomes \[[@R23]\], and ER and PR negativity with poorer clinical outcomes \[[@R24]\]. There is some indication from retrospective studies that breast cancer cells that have undergone HR conversion may be more aggressive \[[@R25], [@R26]\], but a prospective study reported that a therapeutic regimen based on the receptor status of the metastasis did not contribute a survival benefit \[[@R10]\]. Thus the question remains whether identifying the HR status of metastasized breast cancer would aid the choice of therapy strategy.
The present retrospective study of a large cohort of breast cancer patients, compared the ER and PR statuses of primary breast tumors with that of paired metastatic lesions. In addition, we investigated whether HR conversion influenced the survival of these patients, and the effect of HR conversion on salvage hormonal therapy.
RESULTS {#s2}
=======
Patients\' clinicopathological characteristics {#s2_1}
----------------------------------------------
Six hundred and twenty-seven women with metastatic breast cancer were eligible and involved in our analysis (Table [1](#T1){ref-type="table"}). The median ages of the patients at primary tumor diagnosis and metastasis diagnosis were 44 years (range, 22-79 years) and 48 years (range, 25-80 years), respectively.
###### Patients\' clinicopathological characteristics, *n*/N (%)
ER PR
---------------------------------------------- ---------------------------------------------- ----------------- ----------------- ----------------- -----------------
Age, years \<35 76/346 (21.96) 43/281 (15.30) 75/315 (23.81) 44/312 (14.10)
35-60 252/346 (72.83) 211/281 (75.09) 223/315 (70.79) 240/312 (76.92)
\>60 18/346 (5.20) 27/281 (9.61) 17/315 (5.40) 28/312 (8.97)
Clinical stage I 70/346 (20.23) 40/281 (14.23) 61/315 (19.37) 49/312 (15.71)
II 236/346 (68.21) 197/281 (70.11) 212/315 (67.30) 221/312 (70.83)
III 40/346 (11.56) 44/281 (15.66) 42/315 (13.33) 42/312 (13.46)
Adjuvant therapy Chemotherapy 317/346 (91.62) 263/281 (93.59) 293/315 (93.02) 287/312 (91.99)
Hormone therapy 248/346 (71.68) 65/281 (23.13) 218/315 (69.21) 95/312 (30.45)
HER2 Positive[^a^](#tfn_001){ref-type="table-fn"} 44/346 (12.72) 71/281 (25.27) 47/315 (14.92) 68/312 (21.79)
Negative[^b^](#tfn_002){ref-type="table-fn"} 255/346 (73.70) 171/281 (60.85) 226/315 (71.75) 200/312 (64.10)
NA 47/346 (13.58) 39/281 (13.88) 42/315 (13.33) 44/312 (14.10)
Menstrual status Menopause 69/346 (19.94) 75/281 (26.69) 56/315 (17.78) 88/312 (28.21)
Pre-menopause 257/346 (74.28) 186/281 (66.19) 242/315 (76.83) 201/312 (64.42)
Unknown 20/346 (5.78) 20/281 (7.12) 17/315 (5.4) 23/312 (7.37)
Abbreviations: ER, estrogen receptor; PR, progesterone receptor; HER2, human epidermal growth factor receptor 2; IHC,immunohistochemistry; FISH, fluorescence in situ hybridization; NA, Not Available
3+ by IHC or FISH (+)
0-2+ by IHC or FISH (−)
In primary breast cancers, the positive rates for ER and PR were 55.18% and 50.24%, respectively. In accordance with the guidelines of the National Comprehensive Cancer Network, positive HR status was defined as positive signs of ER, PR, or both; and negative HR status was defined as negative signs of both ER and PR. In our study, HR positivity was detected in 62.84% (394/627) of all cases.
The biopsied metastatic sites were the following: soft tissues (*n* = 473), liver (*n* = 96), bone (*n* = 21), lung (*n* = 14), ovary (*n* = 9), pleura (*n* = 5), thyroid gland (*n* = 3), brain (*n* = 2), bladder (*n* = 1), stomach (n = 1), kidney (n = 1), and pancreas (*n* = 1). Soft tissues were the most frequent biopsy site of breast cancer metastasis, perhaps because biopsies of soft tissue are safer than of viscera. The next three most common biopsy sites of metastasis were, in descending order, liver, bone, and lung.
HR conversion {#s2_2}
-------------
A difference in ER status between the primary breast cancer and metastatic lesions was observed in 165 of 627 patients (26.31%). Specifically, in 106 of 346 patients (30.63%), the ER status had altered from positive to negative, while 59 of 281 patients (21.00%) had an alteration from negative to positive (McNemar\'s test, *P* \< 0.001).
A difference in PR status was observed in 213 of 627 cases (33.97%): in 158 of 315 patients (50.16%) the PR status had changed from positive in primary tumors to negative in the metastatic tissues, while 55 of 312 patients (17.63%) had changed from negative to positive (McNemar\'s test, *P* \< 0.001).
Thus, HR status conversion was detected in 170 of 627 cases (27.11%): 121 of 394 (30.71%) had changed from HR-positive in primary tumors to HR-negative in the metastatic tissues, and 49 of 233 (21.03%) from HR-negative to HR-positive (McNemar\'s test, *P* \< 0.001).
Reassessments were made from multiple biopsies from various organs during tumor progression in 82 patients with advanced breast cancers. The re-conversion rates for ER and PR were 26.83% and 35.36%, respectively (McNemar\'s test, *P* \< 0.05).
Unlike HR status, evidence of human epidermal growth factor receptor 2 (HER2) reflects a more aggressive cancer. The rate of discordance of HER2 between primary tumors and metastases was lower than that of ER and PR: 33 of 503 (6.56%) patients had HER2-negative primary tumors but HER2-positive metastases, and 22 of 503 (4.37%) had HER2-positive primary tumors but HER2-negative metastases (Table [2](#T2){ref-type="table"}).
###### HER2 status of primary tumor/metastases, by HR status of primary tumor/metastases
HR status [^b^](#tfn_004){ref-type="table-fn"} HER2 status [^a^](#tfn_003){ref-type="table-fn"}
------------------------------------------------ -------------------------------------------------- --------------- --------------- ----------------- -----
Prim^+^/Met^+^ 16/503 (3.18) 14/503 (2.78) 10/503 (1.99) 184/503 (36.58) 224
Prim^−^/Met^+^ 4/503 (0.80) 1/503 (0.20) 1/503 (0.20) 31/503 (6.16) 37
Prim^+^/Met^−^ 16/503 (3.18) 8/503 (1.59) 4/503 (0.80) 67/503 (13.32) 95
Prim^−^/Met^−^ 39/503 (7.75) 10/503 (1.99) 7/503 (1.39) 91/503 (18.09) 147
Sum 75 33 22 373 503
Abbreviations: HR, hormone receptor; HER2, human epidermal growth factor receptor 2; Prim, receptor status of the primary lesion; Met, receptor status of the metastatic lesion.
HER2 positive: 3+ by IHC or FISH (+); HER2 negative: 0-2+ by IHC or FISH (−)
HR positive: ER and/or PR positive; HR negative: ER and PR negative.
Survival analysis {#s2_3}
-----------------
The median follow-up time was 35 months (range, 4-220 months). The median OS (the time from primary breast cancer diagnosis to the date of death or the end of follow-up) and post-recurrent survival (the time from metastatic breast cancer diagnosis to the date of death or the end of follow-up) was calculated on the basis of ER status in the primary tumor and metastatic tissues (Table [3](#T3){ref-type="table"}).
###### Effects of HR status in primary tumors and metastatic lesions on OS, median months (95% CI)
*n* Deaths OS [^a^](#tfn_005){ref-type="table-fn"} *P* OS [^b^](#tfn_006){ref-type="table-fn"} *P*
----------------------------------------- ---------------- ----- -------------- ----------------------------------------------- --------------------------------------------- ----------------------------------------------- --------------------------------------------
ER Prim^+^/Met^+^ 240 69 135(88-NA) \<0.001 [\*](#tfn_008){ref-type="table-fn"} 68 (44-NA) \<0.001[\*](#tfn_008){ref-type="table-fn"}
Prim^−^/Met^+^ 59 27 85 (55-141) 0.125 [\*\*](#tfn_009){ref-type="table-fn"} 43 (27-90) 0.142 [\*\*](#tfn_009){ref-type="table-fn"}
Prim^+^/Met^−^ 106 45 107 (75-156) \<0.001 [\*\*](#tfn_009){ref-type="table-fn"} 56 (41-83) 0.001 [\*\*](#tfn_009){ref-type="table-fn"}
Prim^−^/Met^−^ 222 83 73 (42-107) \<0.001 [\*\*](#tfn_009){ref-type="table-fn"} 39 (28-71) \<0.001 [\*\*](#tfn_009){ref-type="table-fn"}
PR Prim^+^/Met^+^ 157 48 121 (90-NA) \<0.001[\*](#tfn_008){ref-type="table-fn"} 64 (44-NA) \<0.001[\*](#tfn_008){ref-type="table-fn"}
Prim^−^/Met^+^ 55 26 107 (64-NA) 0.608 [\*\*](#tfn_009){ref-type="table-fn"} 51 (34-NA) 0.212 [\*\*](#tfn_009){ref-type="table-fn"}
Prim^+^/Met^−^ 158 59 126 (73-156) 0.133 [\*\*](#tfn_009){ref-type="table-fn"} 64 (35-89) 0.086 [\*\*](#tfn_009){ref-type="table-fn"}
Prim^−^/Met^−^ 257 91 79 (45-142) \<0.001 [\*\*](#tfn_009){ref-type="table-fn"} 41(28-82) \<0.001 [\*\*](#tfn_009){ref-type="table-fn"}
HR [^c^](#tfn_007){ref-type="table-fn"} Prim^+^/Met^+^ 273 86 126 (86-NA) \<0.001 [\*](#tfn_008){ref-type="table-fn"} 64 (43-NA) \<0.001[\*](#tfn_008){ref-type="table-fn"}
Prim^−^/Met^+^ 49 24 105 (62-156) 0.091 [\*\*](#tfn_009){ref-type="table-fn"} 56 (34-83) 0.095 [\*\*](#tfn_009){ref-type="table-fn"}
Prim^+^/Met^−^ 121 50 85 (55-122) \<0.001 [\*\*](#tfn_009){ref-type="table-fn"} 44 (27-90) 0.002 [\*\*](#tfn_009){ref-type="table-fn"}
Prim^−^/Met^−^ 184 64 73 (42-111) \<0.001 [\*\*](#tfn_009){ref-type="table-fn"} 39 (28-82) \<0.001 [\*\*](#tfn_009){ref-type="table-fn"}
Abbreviations: ER, estrogen receptor; PR, progesterone receptor; Prim, receptor status of the primary lesion; Met, receptor status of the metastatic lesion; NA, not available; *n*, number of patients.
From breast cancer diagnosis to death or censoring.
From breast cancer metastases to death or censoring.
HR positive: ER, PR, or both positive; HR negative: ER and PR both negative.
Compared among 4 groups.
Compare with Prim^+^/Met^+^ group.
According to the ER status of the primary tumor and metastatic lesions, we divided all 627 breast cancer patients into 4 groups, as follows. Group A: ER-positive in primary tumor and ER-positive in metastatic tissues; Group B: ER-positive in primary tumor and ER-negative in metastatic tissues; Group C: ER-negative in primary tumor and ER-positive in metastatic tissues; and Group D: ER-negative in primary tumor and ER-negative in metastatic tissues.
There were significant differences in OS (log-rank, *P* \< 0.001, Figure [1A](#F1){ref-type="fig"}) and post-recurrent survival (log-rank, *P* \< 0.001, Figure [1B](#F1){ref-type="fig"}) among the 4 groups. The median OS was 135 (95% CI 88-NA) months in Group A, 85 (95% CI 55-141) months in Group B, 107 (95% CI 75-156) months in Group C, and 73 (95% CI 42-107) months in Group D. The median post-recurrent survival was 68 (95% CI 44-NA) months in Group A, 43 (95% CI 27-90) months in Group B, 56 (95% CI 41-83) months in Group C, and 39 (95% CI 28-71) months in Group D.
###### Kaplan-Meier survival curves in women of various ER status subtypes
**A.** OS associated with various ER statuses in primary breast cancer (Prim) or metastatic sites (Met). **B.** Post-recurrent survival associated with various ER statuses in primary breast cancer (Prim) and metastatic sites (Met).
According to the PR status of the primary tumor and metastatic lesions, we divided all 627 breast cancer patients into 4 groups, as follows. Group A: PR-positive in primary tumor and PR-positive in metastatic tissues; Group B: PR-positive in primary tumor and PR-negative in metastatic tissues; Group C: PR-negative in primary tumor and PR-positive in metastatic tissues; and Group D: PR-negative in primary tumor and PR-negative in metastatic tissues.
There were significant differences in OS (log-rank, *P* \< 0.001, Figure [2A](#F2){ref-type="fig"}) and post-recurrent survival (log-rank, *P* \< 0.001, Figure [2B](#F2){ref-type="fig"}) among the 4 groups. The median OS was 121 (95% CI 90-NA) months in Group A, 107 (95% CI 64-NA) months in Group B, 126 (95% CI 73-156) months in Group C, and 79 (95% CI 45-142) months in Group D. The median post-recurrent survival was 64 (95% CI 44-NA) months in Group A, 51 (95% CI 34-NA) months in Group B, 64 (95% CI 35-89) months in Group C, and 41 (95% CI 28-82) months in Group D.
###### Kaplan-Meier survival curves in women of various PR status subtypes
**A.** OS associated with various PR statuses in primary breast cancer (Prim) and/or metastatic sites (Met). **B.** Post-recurrent survival associated with various PR statuses in primary breast cancer (Prim) and metastatic sites (Met).
After adjusting for the age of patients with primary breast cancer, age of patients with metastatic breast cancer, PR status in metastatic sites, tumor stage, adjuvant endocrine therapy, adjuvant chemotherapy, and salvage endocrine therapy, the multivariate Cox proportional hazards model was applied. The results of pairwise analysis showed that patients with an ER status that changed from positive in the primary tumor to negative in the metastatic (Prim^+^/Met^−^ group) had a significantly increased hazard ratio for death compared with patients with no change in positive ER status (Prim^+^/Met^+^ group; hazard ratio: 1.74; 95% confidential interval \[CI\]: 1.19-2.56, *P* = 0.005, Table [4](#T4){ref-type="table"}).
###### Risks for patients with breast cancer, depending on ER status in primary tumor and metastases
OS [^a^](#tfn_010){ref-type="table-fn"} Trend test OS [^b^](#tfn_011){ref-type="table-fn"} Trend test
---------------- ---------------- ----- ----------------------------------------- ------------------------------------------ ----------------------------------------------- -------------------------------------------- ------- ------------------------------------------ ----------------------------------------------- ------------------------------------------- -------
ER Prim^+^/Met^+^ 240 69 1.0 [^d^](#tfn_013){ref-type="table-fn"} --- \<0.001[\*](#tfn_014){ref-type="table-fn"} 20.81 1.0 [^d^](#tfn_013){ref-type="table-fn"} --- 0.004[\*](#tfn_014){ref-type="table-fn"} 13.61
Prim^−^/Met^+^ 59 27 1.28 0.81 to 2.01 0.287 [\*\*](#tfn_015){ref-type="table-fn"} 1.18 0.75 to 1.86 0.470 [\*\*](#tfn_015){ref-type="table-fn"}
Prim^+^/Met^−^ 106 45 1.74 1.19 to 2.56 0.005 [\*\*](#tfn_015){ref-type="table-fn"} 1.66 1.13 to 2.44 \<0.001 [\*\*](#tfn_015){ref-type="table-fn"}
Prim^−^/Met^−^ 222 83 2.16 1.53 to 3.05 \<0.001 [\*\*](#tfn_015){ref-type="table-fn"} 1.80 1.28 to 2.53 \<0.001 [\*\*](#tfn_015){ref-type="table-fn"}
HR Prim^+^/Met^+^ 181 72 1.0 [^d^](#tfn_013){ref-type="table-fn"} --- 0.001 [\*](#tfn_014){ref-type="table-fn"} 20.72 1.0 [^d^](#tfn_013){ref-type="table-fn"} --- 0.007 [\*](#tfn_014){ref-type="table-fn"} 12.06
Prim^−^/Met^+^ 32 19 1.37 0.86 to 2.18 0.192 [\*\*](#tfn_015){ref-type="table-fn"} 1.22 0.77 to 1.93 0.404 [\*\*](#tfn_015){ref-type="table-fn"}
Prim^+^/Met^−^ 102 52 1.58 1.10 to 2.27 0.013 [\*\*](#tfn_015){ref-type="table-fn"} 1.48 1.03 to 2.11 0.033 [\*\*](#tfn_015){ref-type="table-fn"}
Prim^−^/Met^−^ 125 66 2.21 1.56 to 3.13 \<0.001 [\*\*](#tfn_015){ref-type="table-fn"} 1.78 1.27 to 2.50 \<0.001 [\*\*](#tfn_015){ref-type="table-fn"}
Abbreviations: ER, estrogen receptor; PR, progesterone receptor; HR, hazard ratio; OS, overall survival; Prim, receptor status of the primary lesion; Met, receptor status of the metastatic lesion; No., number of patients.
Adjusted for age and calendar year of primary breast cancer diagnosis, relapse diagnosis, progesterone receptor status, tumor stage, hormonal treatment, and chemotherapy.
From breast cancer diagnosis to death or censoring.
From breast cancer relapse to death or censoring.
HR positive: ER and/or PR positive; HR negative: ER and PR negative
Reference
Compare between 4 groups.
Compare with Prim^+^/Met^+^ group.
Salvage hormonal therapy {#s2_4}
------------------------
A subgroup analysis of this cohort was conducted to investigate the effects of salvage endocrine therapy on breast cancer patients whose HR status changed from positive to negative (121 patients). Of the 121 breast cancer patients in this group, 76 patients received both salvage endocrine therapy and chemotherapy, and 45 patients received salvage chemotherapy only. Univariate Kaplan-Meier analysis was performed to investigate the response to salvage endocrine therapy in terms of post-recurrent survival. We found that the median post-recurrent survival of patients who accepted both salvage endocrine therapy and chemotherapy (71, 95% CI 28-90 months) was better than that of the chemotherapy-only group (37, 95%C I 19-50 months; log rank; *P* = 0.030; Figure [3C](#F3){ref-type="fig"}).
###### Kaplan-Meier survival curves in women that are HR positive in primary breast cancer and negative in metastatic sites
Post-recurrent survival of the cohort receiving both salvage chemotherapy and endocrine therapy, or salvage chemotherapy only. **A.** Prim^+^/Met^+^ group; **B.** Prim^−^/Met^+^ group. **C.** Prim^+^/Met^−^ group; **D.** Prim^−^/Met^−^ group.
However, in the Prim^−^/Met^−^ group, the post-recurrent survival of the patients who accepted both salvage chemotherapy and endocrine therapy (*n* = 32) was comparable with the patients who accepted chemotherapy only (*n* = 152; log rank test, *P* = 0.266; Figure [3-D](#F3){ref-type="fig"}). In the Prim^+^/Met^+^ (Figure [3-A](#F3){ref-type="fig"}) or Prim^−^/Met^+^ (Figure [3-B](#F3){ref-type="fig"}) groups, the post-recurrent survival of the patients who accepted both salvage chemotherapy and endocrine therapy was also comparable with the patients who accepted chemotherapy only. However, it is worth noting that most patients with positive HR accepted endocrine therapy. Therefore, the sample size of the chemotherapy-only subgroup was very small (8 patients in the Prim^−^/Met^+^ group and 23 patients in Prim^+^/Met^+^ group). The small sample size restricts the power of these two subgroups to convince.
The multivariate Cox proportional hazards model was applied. After adjusting for the age of patients with primary breast cancer, age of patients with metastatic breast cancer, tumor stage, adjuvant endocrine therapy, adjuvant chemotherapy, clinical phase, and metastatic sites, the results showed that patients with salvage chemotherapy-only had a significantly higher hazard ratio for death compared with patients with both salvage endocrine therapy and chemotherapy (hazard ratio: 1.90; 95% confidential interval \[CI\]: 1.01-3.58, *P* = 0.045).
With regard to adjuvant endocrine therapy, the age of patients with metastatic breast cancer, tumor stage, adjuvant chemotherapy, clinical phase, and metastatic sites showed no statistically significant association with post-recurrent survival in this multivariate Cox proportional hazards model (Table [5](#T5){ref-type="table"}).
###### Multivariate Cox proportional hazards model for post-recurrent survival in 121 Prim^+^/Met^−^ patients
Parameter estimate Standard error χ^2^ P\* Hazard ratio 95% CI
---------------------------- -------------------- ---------------- ------- ------- -------------- ----------------
Salvage endocrine therapy 0.643 0.322 3.984 0.045 1.902 1.012 to 3.577
Adjuvant endocrine therapy −0.399 0.297 1.805 0.179 0.671 0.375 to 1.201
Metastatic sites 0.118 0.111 1.129 0.290 1.126 0.905 to 1.400
Adjuvant chemotherapy −0.732 0.561 1.704 0.192 0.481 0.160 to 1.444
Age at primary diagnosis 0.087 0.078 1.222 0.269 1.091 0.935 to 1.272
Age at relapse diagnosis −0.101 0.078 1.670 0.196 0.904 0.775 to 1.054
Clinical phase −0.288 0.203 2.005 0.158 0.750 0.504 to 1.117
Univariate Kaplan-Meier analyses of other parameters were also performed. In the Prim^+^/Met^−^ group, the post-recurrent survival of the patients who had accepted adjuvant endocrine therapy was comparable with that of patients who did not accept adjuvant endocrine therapy (log rank, *P* = 0.194, Figure [4-A](#F4){ref-type="fig"}). The post-recurrent survival of patients ≤35 years old at diagnosis of the primary tumor was comparable to that of patients aged 35-60 years or \>60 years (log rank, *P* = 0.567, Figure [4-B](#F4){ref-type="fig"}). The post-recurrent survival of patients aged ≤35 years at diagnosis of the metastatic tumor was comparable to that of patients aged 35-60 years old or \>60 years (log rank, *P* = 0.523, Figure [4-C](#F4){ref-type="fig"}). The post-recurrent survival of the patients who had accepted adjuvant chemotherapy was comparable to that of patients who did not receive adjuvant chemotherapy (log rank, *P* = 0.590, Figure [4-D](#F4){ref-type="fig"}). The differences among the 4 groups of metastatic sites (i.e., distant lymph node metastases; bone metastases; local relapse; and visceral metastases) was not statistically significant (log rank, *P* = 0.102, Figure [4-E](#F4){ref-type="fig"}). The differences among the clinical phase 1, 2, and 3 groups was not statistically significant (log rank, *P* = 0.409, Figure [4-F](#F4){ref-type="fig"}).
###### Univariate Kaplan-Meier analyses of post-recurrent survival of 121 patients in the Prim^+^/Met^−^ subgroup
**A.** adjuvant endocrine therapy; **B.** age at primary diagnosis. **C.** age at relapse diagnosis; **D.** adjuvant chemotherapy. **E.** metastatic sites; **F.** clinical phase.
DISCUSSION {#s3}
==========
In this retrospective clinical study, we compared the ER and PR status of primary breast tumors with matched metastases of 627 breast cancer patients, and found evidence that the ER or PR status may be different between primary tumor and metastatic tissues. Furthermore, in some metastatic sites, as the disease progresses, the ER and PR status may change again. We also observed that an ER or PR conversion from positive to negative was negatively associated with OS and post-recurrent survival in these patients. A subgroup analysis of patients with HR-positive status (ER, PR, or both positive) in the primary tumor but HR-negative (ER and PR both negative) in metastatic sites showed that patients who accepted both salvage endocrine therapy and chemotherapy had better post-recurrent survival than did those who accepted chemotherapy only. Systemic therapy prolonged the survival of patients with metastatic breast cancer. HR status in metastatic lesions that differs from that of the primary tumor may have clinical implications in salvage therapy and management.
Several studies have shown that ER and PR were not stable during carcinogenesis and tumor progression \[[@R12], [@R16], [@R21], [@R25], [@R27]--[@R38]\]. In our study, we investigated the conversion of ER and PR in a large cohort of women with breast cancer metastasis. The inclusion and exclusion criteria were chosen to omit bias. We found that the rates of receptor conversion were similar to several other retrospective studies \[[@R21], [@R36], [@R39]\]. It is likely that ER and PR conversion results from genetic mutation during tumor progression, intratumoral heterogeneity, and the selective pressure of therapies \[[@R29]\]. A suitable prospective study is needed to determine better the rates of receptor conversion.
According to previous studies, ER and PR conversion in both primary and metastatic lesions in metastatic breast cancer was a prognosticator of OS and post-recurrent survival time \[[@R29], [@R32], [@R39], [@R40]\]. We hypothesize that this correlation can be ascribed to inappropriate target therapy and selection of tumor cells with an unstable phenotype, which may result in more aggressive behaviors \[[@R41]\]. The results of a subsequent univariate analysis showed that a change from positive to negative ER or PR between the primary tumor and metastatic sites was significantly associated with shorter OS compared with no changes. A multi-factorial analysis revealed that only ER status correlated with OS, which was similar to the results of another retrospective study published previously in a peer-reviewed journal \[[@R42]\].
Prospective clinical trials have shown that 14% of patients with breast cancer had their therapeutic regimen modified as a result of changes in the HR/HER2 status in primary tumor tissues or metastatic cancerous tissue despite endocrine treatment before biopsy \[[@R10]\]. In our study, we investigated HR conversion in metastatic breast cancer as a prognostic biomarker during salvage hormonal therapy. Thus, it was of interest that patients whose HR status had altered from positive to negative still achieved a survival benefit from endocrine treatment, compared with those who did not receive endocrine agents. It is likely that HR status in breast cancer tissues will undergo changes several times during tumor progression. In this 627-patient cohort, for 82 patients whose ER and PR were reassessed 2-4 times during metastatic progression, ER and PR re-conversion occurred in 22 (26.83%) and 29 (35.36%), respectively. Thus, our study shows that subsequent and repeated evaluations are necessary for the metastatic breast cancer patient, and salvage hormonal therapy should not be abandoned because of a positive-to-negative HR conversion between the primary tumor and metastatic lesion.
Although we gained some useful insights in our study, its retrospective nature restricts its power to convince. Another shortcoming of our study is a relatively small sample size for the subgroup analysis, especially for the subgroup analysis of the salvage endocrine therapy in the Prim^+^/Met^−^ group. Therefore, a randomized controlled trial is necessary to verify our findings.
In conclusion, ER and PR conversion does occur in breast cancer metastases, and significantly influences survival. Furthermore, ER and PR status may change during tumor progression. Repeated evaluations of HR status are necessary in metastatic breast cancer. Salvage hormonal therapy is still worth trying in cases of positive-to-negative HR conversion.
MATERIALS AND METHODS {#s4}
=====================
Patients {#s4_1}
--------
We conducted a single-center retrospective clinical study, initially collecting the clinical data of 3674 patients with invasive breast cancer who had been hospitalized sometime between 1 January 2002 and 1 April 2016 at Affiliated Hospital of Academy of Military Medical Sciences, Beijing, China.
The inclusion criteria were: pathologically confirmed breast cancer; metastatic disease biopsy; obtainable status of ER, PR, or both of primary and metastatic tumors; and receiving treatment provided by an oncology team that included an oncosurgeon, a medical oncologist, and a radiologist. Patients\' clinicopathological information, including age, gender, date of invasive breast cancer diagnosis, date of metastatic diagnosis, location of distant metastasis, survival data, tumor pathological stage, adjuvant therapy regimen, and salvage therapy regimen was collected and recorded by the oncologists. Excluded from this study were patients without biopsy data on the metastasis with HR status, or patients with clinical stage IV breast cancer, bilateral primary breast cancer, contralateral breast metastasis, and those receiving neoadjuvant chemotherapy before biopsy.
The final analysis included 627 patients, all of whom had been biopsied at both the primary and paired metastatic sites. The academic and ethics committees of our hospital approved this retrospective study. All patients provided written informed consent before the biopsy.
ER, PR, and HER-2 testing {#s4_2}
-------------------------
The primary and metastatic ER and PR statuses of each patient were required for this study and were obtained from the pathology reports. The ER and PR status were both evaluated using immunohistochemistry (IHC) as described previously \[[@R43]\]. Immunohistochemical analysis was carried out on full 4-μm sections. For ER alpha (ERα) and PR, the percentage of positively stained nuclei was estimated. Appropriate negative and positive controls were used throughout. A threshold ≥1% of stained nuclei was considered a positive status. Scoring of IHC slides was performed by 2 independent pathologists, blinded to other data in the paired samples. In primary tumor samples, the adequacy of staining was checked by comparison with normal breast parenchyma of the same patient.
We used 2 methods to assess tissue HER2 status in our study: IHC and fluorescence *in situ* hybridization (FISH). To evaluate HER2 overexpression, IHC staining of specimens was conducted using paraffin-embedded breast cancer tissues and polyclonal rabbit anti-human HER2 oncoprotein. The latter targets the intracellular domain of HER2 protein. Tumors which scored 0 or 1+ were designated HER2-negative; those that scored 3+ were considered HER2-positive. The tumors that were scored via IHC as 2+, were analyzed further by FISH using a PathVysion HER-2 DNA probe kit (Abbott Laboratories, Abbott Park, IL), in accordance with the manufacturer\'s recommended protocol. A FISH result was defined as positive when the HER2/cep17 ratio was \>2.2.
Statistical analysis {#s4_3}
--------------------
Comparisons of ER and PR statuses between the primary tumor and paired metastatic sites were performed using McNemar\'s test. Overall survival (OS) was defined as the time from the date of the pathological diagnosis of primary breast cancer to the date of death or at the end of follow-up (1 May 2016). Post-recurrence survival was considered from the date of the pathological diagnosis of metastatic breast cancer to the date of death or at the end of follow-up (1 May 2016). The Kaplan-Meier analysis was applied to determine whether HR conversion could be a prognosticator of survival \[[@R44], [@R45]\]. The risk of tumor-related death associated with ER and PR status was modeled using a multivariable Cox proportional hazard model. An arbitrary level of 5% was used to indicate statistical significance. Our clinical data was analyzed using SAS version 9.2 statistical software.
**CONFLICTS OF INTEREST**
The authors have no conflicts of interest to declare.
| {
"pile_set_name": "PubMed Central"
} |
Introduction
============
Gastric cancer (GC) is the fourth most common cancer worldwide and the third leading cause of global cancer mortality, with an estimated 951,600 new cases and 723,100 deaths in 2012.[@b1-dddt-11-2595],[@b2-dddt-11-2595] Depending on GC classification and stage, clinical therapeutic regimens include surgery, systemic therapy, radiation therapy and multimodality treatments.[@b3-dddt-11-2595]--[@b5-dddt-11-2595] However, GC is often diagnosed at an advanced stage, at which surgical techniques are not suitable for these patients. It has been demonstrated that chemotherapy can provide both palliation and improved survival in patients with advanced and metastatic GC.[@b6-dddt-11-2595]
Older agents such as fluoropyrimidines, platinum compounds and, recently, taxanes and irinotecan (IRN) have shown the most activity as a single active ingredient and in combination regimens in patients with advanced GC.[@b7-dddt-11-2595]--[@b10-dddt-11-2595] Combination chemotherapy regimens have been widely applied in clinics and bring superior time-to-treatment failure (TTF), progression-free survival (PFS) and overall survival (OS) than single ones.[@b11-dddt-11-2595],[@b12-dddt-11-2595] These first-line combination regimens contain cisplatin and fluorouracil (CF); epirubicin, cisplatin and fluorouracil (ECF); epirubicin, oxaliplatin and capecitabine (ECX); fluorouracil, leucovorin and irinotecan (FOLFIRI), etc. The results of a randomized Phase III study comparing FOLFIRI to ECX in patients with advanced gastric or esophagogastric junction (EGJ) adenocarcinoma showed longer TTF and better toleration with FOLFIRI than with ECX.[@b12-dddt-11-2595] To optimize both or multi-drugs' synergistic therapeutic efficacies and reduce side effects, recent efforts have been devoted to developing novel combination nanomedicines. Therefore, we designed multifunctional nanoparticles (NPs) to co-deliver the hydrophilic drug (5-fluorouracil \[5-FU\]) and the hydrophobic drug (IRN) for gastric carcinoma therapy.
Layer-by-layer (LBL) assembly is a versatile technique to develop multilayer films by the electrostatic attraction of oppositely charged polyelectrolytes.[@b13-dddt-11-2595]--[@b15-dddt-11-2595] LBL techniques possess the ability to incorporate various agents with different physicochemical properties.[@b16-dddt-11-2595],[@b17-dddt-11-2595] In addition, LBL-based NPs could be engineered as the active targeting drug delivery system by the layering materials, which are targeting moieties.[@b18-dddt-11-2595],[@b19-dddt-11-2595] Moreover, the assembled polymer layers could carry a large amount of agents and control the release of them, thus having long blood circulation time.[@b13-dddt-11-2595]
The main purpose of this study was to specifically target and kill GC cells through co-delivery of 5-FU and IRN. As a result, we designed a novel polymer--chitosan (CH)--hyaluronic acid (HA) hybrid formulation (HA--CH--IRN/5-FU NPs) consisting of poly([d]{.smallcaps},[l]{.smallcaps}-lactide-*co*-glycolide) (PLGA) and IRN as the core, CH and 5-FU as a shell on the core and HA as the outmost layer to target GC cells. CH, the natural linear polysaccharide, is a biocompatible, biodegradable and mucoadhesive cationic polymer that exhibits increased and rapid uptake by cancer cells.[@b20-dddt-11-2595] Recently, CH NPs have been exploited extensively in the effective delivery of anticancer agents to the tumor area.[@b21-dddt-11-2595],[@b22-dddt-11-2595] HA is a nontoxic, biocompatible, biodegradable and negative polymer that could actively target the variant CD44 receptor commonly overexpressed on various cancers, including breast cancer and GC.[@b23-dddt-11-2595],[@b24-dddt-11-2595] HA-decorated NPs have been developed for CD44-targeted GC therapy.[@b25-dddt-11-2595]--[@b27-dddt-11-2595] Results showed that HA-coated NPs can be effectively used as a targeted antitumor drug delivery system to enhance therapeutic effects and overcome multi-drug resistance. In order to evaluate HA--CH--IRN/5-FU NPs, we further investigated the physical--chemical and biological characteristics, especially the in vitro and in vivo antitumor efficacy.
Materials and methods
=====================
Materials
---------
PLGA (50:50, molecular weight 5,000--15,000) was purchased from the Jinan Daigang Biomaterial Co., Ltd. (Jinan, China). HA (molecular weight 66--90 kDa) was obtained from Shandong Freda Biochem Co., Ltd. (Jinan, China). CH (degree of deacetylation: 80%; molecular weight 400 kDa), 5-FU, IRN, Pluronic F-68 and 3-(4,5-dimethyl-2-thiazolyl)-2,5-diphenyl-2H-tetrazolium bromide (MTT) were purchased from Sigma-Aldrich Co. (St Louis, MO, USA). Roswell Park Memorial Institute-1640 (RPMI-1640) and fetal bovine serum (FBS) were purchased from Thermo Fisher Scientific (Waltham, MA, USA). All chemicals and reagents used were of analytical grade or better.
Cells and culture
-----------------
Human gastric carcinoma cells (MGC803 cells) were purchased from American Type Culture Collection (Manassas, VA, USA) and were grown in an RPMI-1640 medium containing 10% FBS and 1% antibiotic antimycotic at 37°C and a humidified 5% CO~2~ atmosphere.
Animals and guidelines
----------------------
Balb/c nude mice (8 weeks old, 20 g weight) were purchased from Shanghai Slack Laboratory Animal Co., Ltd. (Shanghai, China) and housed in ventilated cages at a temperature of 20°C±2°C and a relative humidity of 60%±5%. All animal experiments comply with the National Institutes of Health Guide for the care and use of laboratory animals (NIH Publication No 8023, revised 1978), and the animal experiments was approved by the medical ethics committee of Shandong University (No 100120160189).
Preparation of IRN-loaded PLGA NPs
----------------------------------
IRN-loaded PLGA NPs (IRN NPs) was prepared by solvent displacement technique.[@b28-dddt-11-2595] Briefly, IRN (20 mg) and PLGA (100 mg) were dissolved in acetone (5 mL) to get the organic phase. In all, 1% F68 (w/v) was dissolved in Milli-Q water (50 mL) to obtain the aqueous solution. The organic phase was added drop by drop into the aqueous solution (50 mL) and stirred by a laboratory magnetic stirrer (400 rpm at room temperature) for 8 h to completely evaporate the organic solvent. The obtained NP suspensions were collected by centrifugation at 10,000 rpm for 15 min, resuspended in Milli-Q water, washed three times and filtered through a 0.45 μm membrane. The obtained NPs were stored at 2°C--8°C.
Preparation of 5-FU-loaded CH NPs
---------------------------------
5-FU-loaded CH NPs (5-FU NPs) were prepared by the self-assembly technique.[@b29-dddt-11-2595] Briefly, 5-FU (20 mg) was dissolved in Milli-Q water (5 mL) added drop by drop into the CH (100 mg) in acetic acid (1% v/v) solution. The obtained NP suspensions were collected by centrifugation at 10,000 rpm for 15 min, resuspended in Milli-Q water, washed three times and filtered through a 0.45 μm membrane. The obtained NPs were stored at 2°C--8°C.
Preparation of HA--CH--IRN/5-FU NPs
-----------------------------------
PLGA--CH hybrid NPs (CH--IRN/5-FU NPs) were prepared as follows: 5-FU (20 mg) was mixed with CH (100 mg) in acetic acid (1% v/v) solution and then added drop by drop into the IRN NP suspensions.
HA--CH--IRN/5-FU NPs ([Figure 1A](#f1-dddt-11-2595){ref-type="fig"}) was fabricated by adding HA solution drop by drop into the CH--IRN/5-FU NP suspensions.[@b30-dddt-11-2595] The obtained CH--IRN/5-FU NP and HA--CH--IRN/5-FU NP suspensions were collected by centrifugation at 10,000 rpm for 15 min, resuspended in Milli-Q water, washed three times and filtered through a 0.45 μm membrane. The obtained NPs were stored at 2°C--8°C.
Blank HA- and CH-containing NPs not including IRN and 5-FU (HA--CH NPs) were prepared by the same procedure without adding the two drugs.
Characterization of NPs
-----------------------
Transmission electron microscopy (TEM) was used to investigate the morphology of HA--CH--IRN/5-FU NPs, using JEM-1200EX transmission electron microscope (JEOL, Tokyo, Japan).[@b31-dddt-11-2595] One drop of HA--CH--IRN/5-FU NPs was placed onto a copper grid and air-dried. A drop of 2% aqueous solution of sodium phosphotungstate was applied as a negative stain, and the NPs were then examined under the microscope.
Mean particle size (volume mean diameter), polydispersity index (PDI) and zeta potential of NPs were characterized using photon correlation spectroscopy with a Zetasizer 3000 (Malvern Instruments, Malvern, UK) as described in our previous study.[@b32-dddt-11-2595]
The IRN and 5-FU entrapment efficiency (EE) in NPs was determined by high-performance liquid chromatography (HPLC; LC-20A; Shimadzu, Tokyo, Japan). 5-FU content was quantified using 250×4.6 mm C18 column.[@b29-dddt-11-2595] The mobile phase (flow rate 0.6 mL/min) was KH~2~PO~4~ 0.05 M triethylamine 0.1%, and ultraviolet detection was set at 266 nm. Quantitative analysis of IRN was carried out by spectrophotometric assay (UV-1800, UV--VIS spectrophotometer; Shimadzu) at 370 nm.[@b33-dddt-11-2595] The EE of the NPs was calculated according to the following equation: EE (%) = drug amount in NPs/drug feeding ×100.
Serum stability of HA--CH--IRN/5-FU NPs and single drug-loaded NPs was evaluated in phosphate-buffered saline (PBS) containing 10% FBS (v/v) at 37°C for 72 h. At times 0, 2, 4, 8, 24, 48 and 72 h, 1 mL of each sample was diluted with 2 mL of triethylamine and the mixture was bath sonicated for 5 min, followed by centrifugation at 10,000 rpm for 5 min. The variation trends of the particle size were evaluated.
In vitro drug release studies
-----------------------------
In vitro drug release studies were performed on IRN and/or 5-FU-co-loaded NPs.[@b34-dddt-11-2595] Briefly, 50 mg of NPs were resuspended in 10 mL of PBS and placed at 37°C in a shaker incubator shaking at the speed of 100 rpm. At predetemined times (1, 2, 4, 8, 12, 24, 48 and 72 h), the suspension was centrifuged at 12,000 rpm for 10 min, the supernatant was harvested and the amount of IRN and 5-FU was quantified by the method mentioned in the "characterization of NPs" section. The pellets of NPs were resuspended in fresh PBS in the original tube for further incubation.
Cellular uptake studies
-----------------------
The uptake rates of IRN and/or 5-FU-co-loaded NPs by MGC803 cells were tested using coumarin 6 as a model fluorescent molecule, which can be encapsulated into various NPs for quantitative investigation.[@b31-dddt-11-2595] Kinds of coumarin 6-loaded NPs were prepared by the same method as described in the "Preparation of IRN-loaded PLGA NPs", "Preparation of 5-FU-loaded CH NPs" and "Preparation of HA--CH--IRN/5-FU NPs" sections, with the adding of 20 mg coumarin 6 along with the drugs into each formula. Coumarin 6-loaded NPs were added at concentrations of 200 mg/mL into MGC803 cells equilibrated with Hank's buffered salt solution (HBSS) at 37°C. After incubation for 24 and 72 h, the medium was removed and the wells were washed three times with cold PBS solution. Then, cells were washed once with 1 mL of PBS; detached with trypsin/EDTA, centrifuged at 1,500 rpm, 4°C for 5 min; resuspended in 300 μL of PBS and directly introduced to a flow cytometer.
In vitro cytotoxicity assays
----------------------------
In vitro cytotoxicity of NPs was estimated in MGC803 cells using MTT assay.[@b35-dddt-11-2595] Briefly, MGC803 cells were seeded in 96-well plates at a density of 2,000 cells/well and incubated for 24 h. IRN- and/or 5-FU-contained solutions or NPs were added at various concentrations for 48 h. MTT was then added to the media at a concentration of 5 mg/mL and incubated for 4 h. The formazan crystals were dissolved in 100 μL dimethyl sulfoxide (DMSO), and optical absorbance was recorded at 570 nm with a Model 680 Microplate Reader (Bio-Rad Laboratories Inc., Hercules, CA, USA). Cell viability was calculated as percentage of the untreated control.
Synergistic effect evaluation and selection of the ratio of drugs
-----------------------------------------------------------------
To evaluate the synergistic effects and select the suitable ratio of drug co-loaded in the NPs, the combination index (CI) was measured according to the Chou and Talalay's method.[@b36-dddt-11-2595],[@b37-dddt-11-2595] Based on the cell viability results calculated in the "Cellular uptake studies" section, the median inhibitory concentration (IC~50~) values were calculated using Origin 8.0 (OriginLab, Northampton, MA, USA). CI was calculated by the equation: CI = \[(C)~IRN~/(C~50~)~IRN~\] + \[(C)~5-FU~/(C~50~)~5-FU~\]. (C)~IRN~ and (C)~5-FU~ represents the concentration of IRN and 5-FU in the combination system at the IC~50~ value. (C~50~)~IRN~ and (C~50~)~5-FU~ represent the IC~50~ value of IRN alone and 5-FU alone, respectively. CI \<1, CI =1 and CI \>1 indicate synergism, additive effect and antagonism, respectively.
Tumor induction
---------------
Tumors were induced in mice by subcutaneous injection of MGC803 cells (10^7^ cells suspended in 100 μL normal saline) into the right and left flanks on the dorsal side of the Balb/c nude mice.[@b38-dddt-11-2595] The measurements were taken in two perpendicular dimensions, and tumor volumes were calculated by the following equation: tumor volume (mm^3^) = (longest diameter × shortest diameter[@b2-dddt-11-2595])/2. When the volume of the tumor reached \~50 mm^3^ (\~1 day), the GC-bearing mice were divided into eight groups (with eight mice in each group) that were ready for the treatment.
In vivo tissue distribution study
---------------------------------
The GC-bearing mice model was used to investigate the in vivo tissue distribution of IRN- and/or 5-FU-contained solutions or NPs.[@b39-dddt-11-2595] The mice were divided into several groups and administered 1 mL of solutions or NPs through the tail vein separately. At predetermined time intervals, mice were sacrificed and the heart, liver, spleen, lung, kidney, stomach, colon and tumor of mice were collected. The tissues were cut into small pieces and homogenized with physiological saline. After appropriate dilution of supernatants, the content of IRN and 5-FU was quantified by the method mentioned in the "Characterization of NPs" section.
In vivo antitumor effect and system toxicity evaluation
-------------------------------------------------------
In vivo antitumor effect and system toxicity of the NPs were evaluated in terms of tumor volume and body weight changes.[@b40-dddt-11-2595] The mice were divided into several groups and administered IRN- and/or 5-FU-contained solutions or NPs every other day for 3 weeks. Tumor sizes and mouse body weights were measured. Tumor volumes were calculated using the equation presented in the "Synergistic effect evaluation and selection of the ratio of drugs" section.
Statistical analysis
--------------------
The level of significance in all statistical analyses was set at a probability of *P*\<0.05. Experiments were performed at least three times (n=3) and expressed as mean ± SD. Statistical differences were determined using Student's *t*-test for comparison of two groups and one-way ANOVA for multiple groups.
Results
=======
Characterization of HA--CH--IRN/5-FU NPs
----------------------------------------
Morphology of HA--CH--IRN/5-FU NPs exhibited an LBL type of spherical particle ([Figure 1B](#f1-dddt-11-2595){ref-type="fig"}). It appeared as a white core, a gray first layer over the core and a darker outside second layer. Morphology of other NPs is presented in [Figure 1C](#f1-dddt-11-2595){ref-type="fig"}. Mean particle size, PDI and zeta potential were also tested ([Table 1](#t1-dddt-11-2595){ref-type="table"}). The size of NPs was increased from 91 to 153 along with the LBL formation procedure, indicated that the coating of CH and HA layer enlarged the particles. Zeta potential of the NPs varied due to the positive or negative charges of the materials used. The EE of IRN- and/or 5-FU-loaded NPs was \>90%. Changes in size in the presence of serum are described in [Figure 2](#f2-dddt-11-2595){ref-type="fig"}. The NPs tested were stable up to 72 h without any significant size changes.
In vitro drug release
---------------------
In vitro release of IRN and 5-FU from NPs was in a sustained behavior ([Figure 3](#f3-dddt-11-2595){ref-type="fig"}). The release behavior of 5-FU ([Figure 3B](#f3-dddt-11-2595){ref-type="fig"}) from the NPs was faster than that of IRN ([Figure 3A](#f3-dddt-11-2595){ref-type="fig"}). The release rate of HA--CH--IRN/5-FU NPs was the slowest. The release rate of drugs from CH--IRN/5-FU NPs was slower than that of IRN NPs. 5-FU released from 5-FU NPs was more slower than from the CH--IRN/5-FU NPs.
Cellular uptake
---------------
Cellular uptake efficiency of the NPs was tested. As shown in [Figure 4](#f4-dddt-11-2595){ref-type="fig"}, cellular uptake efficiency of HA--CH--IRN/5-FU NPs and HA--CH NPs was significantly higher than other NPs, reached \>80% at 72 h post treatments (*P*\<0.05). Cellular uptake efficiency of CH--IRN/5-FU NPs, IRN NPs and 5-FU NPs showed no obvious difference compared with each other.
In vitro cytotoxicity
---------------------
In vitro cytotoxicity of IRN- and/or 5-FU-contained solutions or NPs was investigated using MGC803 cells. Following incubation with samples for 48 h, cell viability was summarized ([Figure 5](#f5-dddt-11-2595){ref-type="fig"}). HA--CH--IRN/5-FU NPs exhibited the most effective tumor cell inhibition ability among all samples tested. The results illustrated that the drug-loaded NPs were more efficient than free drug solutions (*P*\<0.05). Double drugs contained NPs, and the solution also had a higher cytotoxicity than single drug-contained NPs and solution (*P*\<0.05).
Synergistic effect and selection of the ratio of the two drugs
--------------------------------------------------------------
Dual drug-co-loaded NPs with various IRN to 5-FU weight ratios were applied to MGC803 cells using MTT assay to determine the synergistic effect and selection of the ratio of the two drugs ([Table 2](#t2-dddt-11-2595){ref-type="table"}). The combination therapy showed synergistic effect when IRN to 5-FU ratio was 2:1. The CI~50~ values for HA--CH--IRN/5-FU NPs, CH--IRN/5-FU NPs and IRN/5-FU solutions were 0.521, 0.884 and 0.787, respectively. The IRN to 5-FU weight ratio of 2:1 exhibited stronger synergism in both NPs and solution formulations; thus, this ratio was determined in this study. So the ration of 2:1 (IRN to 5-FU, w/w) was used for all the in vitro and in vivo studies.
In vivo tissue distribution
---------------------------
In vivo tissue distribution of HA--CH--IRN/5-FU NPs, CH--IRN/5-FU NPs and IRN/5-FU solution was investigated in GC-bearing mice model ([Figure 6](#f6-dddt-11-2595){ref-type="fig"}). IRN and 5-FU distributions of HA--CH--IRN/5-FU NPs were higher in the tumor tissue than those of CH--IRN/5-FU NPs and IRN/5-FU solution (*P*\<0.05). Drugs loaded in NPs had relatively lower accumulation in heart and kidney than drugs' solution (*P*\<0.05).
In vivo antitumor effect and system toxicity
--------------------------------------------
In vivo antitumor efficiency was evaluated in GC-bearing mice model ([Figure 7A](#f7-dddt-11-2595){ref-type="fig"}). The most obvious tumor inhibition was clearly observed in the HA--CH--IRN/5-FU NPs group; the tumor growth was prominently delayed, which attained \~217 mm^3^ on 21-day posttreatment. The 0.9% saline control group had a tumor volume of 1,530 mm^3^ on day 21. Tumor growth was more significantly inhibited by drug-loaded NPs than free drug solutions (*P*\<0.05). Dual drug-co-loaded NPs and solutions showed better tumor suppression efficacy than the single drug containing NPs and solutions (*P*\<0.05). HA--CH--IRN/5-FU NPs exhibited better tumor inhibition ability than non-modified CH--IRN/5-FU NPs (*P*\<0.05). The obvious emaciation could be observed in the free drug solution groups, with an obvious decrease in the body weight ([Figure 7B](#f7-dddt-11-2595){ref-type="fig"}). The NP groups did not cause a significant difference in body weight changes. A significant body weight increase was found in the 0.9% saline control group.
Discussion
==========
LBL assembly techniques are promising strategies for incorporating various agents with different physicochemical properties, and LBL-based NPs could be engineered as the targeted drug delivery system to provide enhanced stability, cellular uptake, regulation of drug release and targeting capabilities.[@b41-dddt-11-2595] In this study, we used the LBL technique to develop multilayer NPs by the electrostatic interaction of oppositely charged HA (negative) and CH (positive). Moreover, because of the specific binding between HA and CD44, our NPs can be used for targeted delivery of anticancer drugs into GC that are commonly overexpressed with CD44.
Electrostatic interactions are widely reported to be the main driving force for the multilayer formation process between two oppositely charged polyelectrolytes for the LBL technique.[@b42-dddt-11-2595],[@b43-dddt-11-2595] Morphology, size and zeta potential of NPs are often characterized to determine whether an additional layer could be successfully formulated onto the previous carriers. [Figure 1B](#f1-dddt-11-2595){ref-type="fig"} illustrates the layered core--shell structure of HA--CH--IRN/5-FU NPs; this could be the evidence that the system was successfully assembled. Moreover, the increased particle size indicated that the coating of CH and HA layers enlarged the particles. Zeta potential of the NPs was reversed due to the positive or negative charged materials used. The NPs tested were stable up to 72 h without any significant size changes in serum and suggest in a period of time that this formulation may be stable when intravenously administrated.
Sustained release behavior of IRN and 5-FU from NPs could significantly improve the therapeutic efficacy of the drugs loaded.[@b27-dddt-11-2595] Faster release behavior of 5-FU from the NPs than that from IRN could be explained by stating that the 5-FU was loaded in the outer layer of the NPs and the IRN was loaded in the inner core. More slowly release of 5-FU from 5-FU NPs than from the CH--IRN/5-FU NPs may be due to the preparation of 5-FU NPs using a large amount of CH that may stuck the 5-FU release.
The therapeutic effects of the NPs would depend on internalization of the NPs to the cancer cells. Coumarin 6, a fluorescent probe, was used to represent the drug in the NP formulation to analyze cellular uptake of the NPs. Cellular uptake efficiency of HA--CH--IRN/5-FU NPs was significantly higher than other NPs. This could be attributed to enhanced cancer cell-specific adherence of the HA ligands. This behavior may improve the activity and overcome drug resistance.[@b31-dddt-11-2595] In order to investigate the cytotoxicity and synergistic effect of dual drug-co-loaded NPs, we have performed a cell viability study using MGC803 cells in an in vitro MTT assay. Evaluation of drug--drug interaction is important in all areas of medicine and particularly in combination cancer chemotherapy.[@b44-dddt-11-2595] Chou and Talalay[@b45-dddt-11-2595] in 1983 introduced the term CI for quantification of synergism or antagonism for two drugs. If the CI value is \<1, synergism is indicated, and if the CI value is 1 to infinity, antagonism is indicated.[@b46-dddt-11-2595] When IRN to 5-FU weight ratio was 2:1, the CI~50~ values for HA--CH--IRN/5-FU NPs, CH--IRN/5-FU NPs and IRN/5-FU solution were \<1. This section confirmed the synergistic effect and the best IRN to 5-FU ratio of the two drug-containing systems.
In vivo drug distribution of NPs was higher in the tumor tissue and lower in heart and kidney, which could decrease the side effects during the tumor therapy.[@b47-dddt-11-2595] On the contrary, the drug solution samples mainly distributed in heart and kidney. This may lead to systemic toxicity. Higher drug distribution of NPs formulations in tumor might be due to the sustained release behavior and targeted ability of the NPs thus prolonged the blood circulation time and could better target the tumor site.
It was observed that in vivo antitumor efficiency of the dual drugs could be better after loading in the NPs because of the higher tumor volume suppression. The higher antitumor efficiency of drugs after co-loaded in HA--CH--IRN/5-FU NPs than non-modified CH--IRN/5-FU NPs is related to the targeted ability of HA.[@b48-dddt-11-2595] Based on the healthy body weight of NP-treated mice, the NPs constructed were proposed as safe carriers for the delivery of anticancer drugs. Minimizing the toxicity and side effects of drugs could prove the targeting efficiency of NPs. The in vivo antitumor results suggested the best anti-tumor effect of folate decorated double drugs contained NLCs due to the synergetic effect of the two drugs, and the least systemic toxic side effect of the NLC formulations for the head and neck cancer treatment.
Conclusion
==========
HA--CH--IRN/5-FU NPs consisting of PLGA and IRN as the core, CH and 5-FU as a shell on the core and HA as the outmost layer were prepared. HA--CH--IRN/5-FU NPs displayed enhanced antitumor activity in vitro and in vivo than non-modified NPs, single drug-loaded NPs and drug solutions. The results demonstrate that HA--CH--IRN/5-FU NPs can achieve impressive antitumor activity and the novel targeted drug delivery system offers a promising strategy for the treatment of GC.
**Disclosure**
The authors report no conflicts of interest in this work.
![Scheme graph and TEM images of HA-CH-IRN/5-FU NPs and other NPs.\
**Notes:** (**A**) Scheme graph of the structure of HA--CH--IRN/5-FU NPs. (**B**) TEM image of HA--CH--IRN/5-FU NPs. (**C**) TEM images of HA--CH--IRN/5-FU NPs (1), CH--IRN/5-FU NPs (2), IRN NPs (3), 5-FU NPs (4) and HA--CH NPs (5) (yellow scale bars represent 200 nm).\
**Abbreviations:** HA, hyaluronic acid; CH, chitosan; IRN, irinotecan; 5-FU, 5-fluorouracil; NP, nanoparticle; PLGA, poly([d]{.smallcaps},[l]{.smallcaps}-lactide-*co*-glycolide); TEM, transmission electron microscopy.](dddt-11-2595Fig1){#f1-dddt-11-2595}
{#f2-dddt-11-2595}
{#f3-dddt-11-2595}
{#f4-dddt-11-2595}
{#f5-dddt-11-2595}
{#f6-dddt-11-2595}
{#f7-dddt-11-2595}
######
Characterization of NPs
Formulations HA--CH--IRN/5-FU NPs CH--IRN/5-FU NPs IRN NPs 5-FU NPs HA--CH NPs
------------------------- ---------------------- ------------------ ------------- ------------- -------------
Particle size (nm) 153.8±5.6 127.3±4.8 91.4±3.6 86.9±3.9 155.1±5.2
Size distribution (PDI) 0.187±0.021 0.164±0.018 0.118±0.012 0.147±0.016 0.184±0.025
Zeta potential (mV) −13.7±1.9 +17.5±2.6 −16.9±2.2 +25.8±3.7 −15.2±2.4
EE of IRN (%) 94.5±2.3 93.6±2.1 92.8±2.8 N/A N/A
EE of 5-FU (%) 91.4±3.1 92.3±2.9 N/A 93.4±3.2 N/A
**Abbreviations:** NPs, nanoparticles; HA, hyaluronic acid; CH, chitosan; IRN, irinotecan; 5-FU, 5-fluorouracil; PDI, polydispersity index; EE, entrapment efficiency; N/A, not applicable.
######
Synergistic effect evaluation and selection of the ratio of drugs by CI calculation
Formulations IRN/5-FU ratio (w/w) IC~50~ IRN (μM) IC~50~ 5-FU (μM) CI~50~
---------------------- ---------------------- ----------------- ------------------ --------
IRN NPs N/A 0.261 N/A N/A
5-FU NPs N/A N/A 0.123 N/A
CH--IRN/5-FU NPs 4/1 0.242 0.061 1.423
CH--IRN/5-FU NPs 2/1 0.112 0.056 0.884
CH--IRN/5-FU NPs 1/1 0.096 0.096 1.148
HA--CH--IRN/5-FU NPs 2/1 0.066 0.033 0.521
IRN solution N/A 1.961 N/A N/A
5-FU solution N/A N/A 1.162 N/A
IRN/5-FU solution 2/1 0.837 0.419 0.787
**Abbreviations:** CI, combination index; IRN, irinotecan; 5-FU, 5-fluorouracil; IC~50~, median inhibitory concentration; NPs, nanoparticles; N/A, not applicable; CH, chitosan; HA, hyaluronic acid.
| {
"pile_set_name": "PubMed Central"
} |
1. Introduction {#sec1-jcm-09-00559}
===============
Wound healing follows a series of complex overlapping processes that leads to closure of the wound and restoration of the epithelial layer, including hemostasis, inflammation, cell proliferation and tissue remodeling \[[@B1-jcm-09-00559]\].
After an injury to skin, the exposed sub-endothelium, collagen and tissue factors activate platelet aggregation to form a clot (hemostasis). Next, neutrophils appear at the site of injury to remove debris and bacteria, promoting an environment for wound healing. Then, macrophages accumulate in the inflammatory phase facilitating phagocytosis of bacteria and damaged tissue. In acute wounds, inflammation is fast and short-lived (typically 3--5 days). The inflammatory cells orchestrate the inflammatory process and facilitate initiation of the reconstructive phase of healing. Lamentably, in chronic non-healing wounds the inflammatory cytokines are in excess, and the resultant pro-inflammatory environment causes degradation of growth factors and extracellular matrix (ECM) proteins \[[@B2-jcm-09-00559]\]. In acute wound, the proliferative phase follows the inflammatory phase, and is characterized by accumulation of fibroblasts and production of connective tissue. Fibroblasts play a critical role in supporting normal wound healing due to their involvement in several key processes, including breakdown of fibrin clot, creation of new ECM and collagen structures, and contraction of the wound \[[@B3-jcm-09-00559]\]. Finally, during the tissue remodeling phase, collagen bundles are reorganized, restoring epidermal barrier function and skin integrity. Overall, the healing process involves numerous cellular and biosynthetic processes, which all require energy in the form of adenosine triphosphate (ATP), as well as amino acids, and other precursor molecules to replace damaged tissue \[[@B4-jcm-09-00559],[@B5-jcm-09-00559],[@B6-jcm-09-00559]\].
Mitochondria are the key organelles responsible for ATP production in human cells, generating energy through oxidative phosphorylation \[[@B7-jcm-09-00559]\]. The mitochondrial respiratory chain consists of a five-subunit complex (I-V) through which electrons are exchanged at increasing reduction potentials, leading to production of ATP \[[@B8-jcm-09-00559]\]. Besides production of ATP for cellular energy, mitochondria also generate reactive oxygen species (ROS) as a by-product of ATP formation \[[@B9-jcm-09-00559]\]. The role of ROS in wound healing is complex. ROS contributes to the oxidative burst that kills bacteria, and acts as a cell signal to increase cell proliferation, which is vital in wound repair \[[@B10-jcm-09-00559]\]. However, ROS are also very potent molecules and can cause damage to DNA, lipids and proteins \[[@B11-jcm-09-00559]\]. Excessive levels of ROS have been associated with chronic wounds, resulting in tissue damage, excessive inflammation, and delayed healing \[[@B12-jcm-09-00559],[@B13-jcm-09-00559]\]. Furthermore, high levels of oxidative stress impact on mitochondrial morphology and positioning within cells, causing mitochondria to cluster around the nucleus as a protective strategy \[[@B14-jcm-09-00559],[@B15-jcm-09-00559]\].
Generally, mitochondria undergo a constant process of fission and fusion where they join and subsequently split back into separate entities. This process is thought to be a housekeeping effect to ensure that mitochondria stay as efficient as possible \[[@B16-jcm-09-00559]\]. The position of mitochondria within cells is dynamic based on localized signals or energy balance. For instance, during cell migration mitochondria will locate close to the migrating cell edges where rapid reorganization of the cytoskeleton filaments requires more ATP for energy \[[@B17-jcm-09-00559],[@B18-jcm-09-00559],[@B19-jcm-09-00559]\].
Photobiomodulation (PBM) is the term used to describe the application of low-level light energy to induce tissue regeneration or to protect tissue that is injured or degenerating \[[@B20-jcm-09-00559],[@B21-jcm-09-00559],[@B22-jcm-09-00559]\]. PBM has been shown to be effective in accelerating healing in chronic and acute wounds \[[@B23-jcm-09-00559],[@B24-jcm-09-00559]\], as well as reducing pain and inflammation in several conditions \[[@B25-jcm-09-00559],[@B26-jcm-09-00559],[@B27-jcm-09-00559]\]. Moreover, PBM reverses toxic effects of neurotoxins, stimulates stem cell proliferation, and displays therapeutic effects in reducing myocardial ischemia reperfusion related to myocardial injury \[[@B28-jcm-09-00559],[@B29-jcm-09-00559]\].
One of the most well-studied mechanisms of PBM is the ability to interact with endogenous chromophores in tissue that absorb photons (light). The most studied endogenous chromophore is complex IV of the mitochondrial respiratory chain, cytochrome c oxidase (CCO). Studies have shown that the activation spectrum of CCO runs from yellow through to red wavelengths (\~570--910 nm). Absorption of photons by CCO initiates a biochemical cascade that increases ATP and ROS generation within the electron transport chain \[[@B30-jcm-09-00559],[@B31-jcm-09-00559],[@B32-jcm-09-00559],[@B33-jcm-09-00559]\]. Additionally, evidence suggests that one possible mechanism of action of photons on the wound healing process is regulation of the homeostatic balance between mitochondrial fusion and fission \[[@B34-jcm-09-00559]\]. Mitochondria fusion helps reduce mitochondrial stress by mixing the contents of partially damaged mitochondria. Fission allows the creation of new mitochondria, but it also enables the removal of damaged mitochondria \[[@B35-jcm-09-00559]\].
A form of PBM is fluorescent light energy (FLE) that acts as a vehicle to induce biomodulation. To generate fluorescence, specialized chromophores (light absorbing molecules) are employed to translate light energy into a low-energy emission of fluorescence through a mechanism known as stokes shift \[[@B36-jcm-09-00559]\]. FLE is a unique form of photobiomodulation that has been demonstrated to advance healing of both acute and chronic wounds \[[@B37-jcm-09-00559],[@B38-jcm-09-00559]\]. Studies have demonstrated that acute incisional wounds have reduced inflammation, as well as more physiologic re-epithelization and collagen remodeling resulting in better quality and less visible scars \[[@B25-jcm-09-00559],[@B26-jcm-09-00559],[@B27-jcm-09-00559],[@B39-jcm-09-00559]\]. Additionally, patients with hard-to-heal chronic ulcers experienced accelerated healing and improved quality of life \[[@B23-jcm-09-00559]\]. These studies have documented the safety and efficacy of FLE in a variety of impaired skin conditions \[[@B40-jcm-09-00559]\]. The beneficial impact of FLE on inflammation has been well documented, with improved inflammatory profiles observed in clinical trials for wound healing \[[@B37-jcm-09-00559]\], acne treatment \[[@B41-jcm-09-00559]\], and management of rosacea \[[@B42-jcm-09-00559]\], as well as in canine pyoderma \[[@B43-jcm-09-00559]\] and mechanistic in vitro studies \[[@B36-jcm-09-00559]\]. These studies have demonstrated that FLE has the ability to reduce pro-inflammatory cytokines produced by human dermal fibroblasts (HDFs), as well as increase levels of some anti-inflammatory cytokines, which have a beneficial effect in the treatment of chronic wounds that are stuck in an inflammatory feedback loop \[[@B36-jcm-09-00559]\].
The effect of FLE on mitochondrial homeostasis is still largely unknown. We tested the effect of FLE on an in vitro model of inflammation. Cultures of normal HDFs were stimulated with a cocktail of pro-inflammatory cytokines to create inflamed cells. These inflamed cells were treated with either direct light (photobiomodulation) or FLE. Two forms of photoconverter substrates were tested: an amorphous gel (FLE-Gel) or sheet hydrogel matrix (FLE-Matrix). Variations in the mitochondrial network were analyzed with confocal microscopy. Expression profiles of genes related to mitochondrial dynamic, biogenesis and function were assessed by PCR gene array. Confocal microscopy showed the presence of mitochondrial fragmentation due to exposure to the inflammatory cytokines, but a restoration of the mitochondrial network 24 h post-treatments was observed, in particular in cultures treated with FLE. The gene expression analysis showed that treatment with FLE-Matrix upregulated UCP1 and CPT1B genes, which encode proteins that favor ATP production through oxidative phosphorylation and lipid β-oxidation, respectively. Treatment with FLE-Gel upregulated SLC25A31 gene, linked to cytosolic adenosine diphosphate (ADP) balance. These observations, together with the already known capacity of photobiomodulation to stimulate cytochrome c oxidase, show a beneficial effect of FLE in the treatment of inflamed wounds.
2. Experimental Section {#sec2-jcm-09-00559}
=======================
2.1. Cell Culture Preparation {#sec2dot1-jcm-09-00559}
-----------------------------
Normal human dermal fibroblasts (HDFs; PCS-201-012, American Type Culture Collection (ATCC, Manassas, VA, USA) were cultured at 37 °C and 5% CO~2~ in Fibroblast Basal Medium (phenol red-free; ATCC, Manassas, VA, USA) supplemented with Fibroblast Growth Kit-Low serum (ATCC, Manassas, VA, USA). The culture process was performed seeding HDFs in 6-well plastic plates at a density of 10 × 10^4^ cell/well. Healthy cells were incubated in basal medium for the duration of the experiment.
Inflamed cells (including all treatment groups) were incubated for 5--6 h in basal medium prior to inducing an in vitro inflammatory state with an 18-h incubation in an inflammatory cocktail comprising 20 ng/mL each of pro-inflammatory cytokines recombinant human tumor necrosis factor alpha (TNF-α; Miltenyi Biotec S.r.l., Bologna, Italy) and recombinant human Interleukin-1 beta (IL-1β; Miltenyi Biotec S.r.l.).
For all treatment groups (Light, FLE-Gel, and FLE-Matrix), after the 18-h incubation in the inflammatory cocktail, the media was replaced with Phosphate Buffer Saline (PBS; Euroclone S.p.A., Italy) for the illumination procedure (5-min) to minimize scattering or other interference with the light. After the illumination, the PBS was replaced with fresh media containing the TNFα/IL-1β inflammatory cocktail to continue with the inflammatory stimulus. Untreated inflamed HDFs were also placed in PBS for 5 min (to mimic the treatment conditions), and then were incubated in fresh media containing inflammatory cockail for the duration of the experiment.
Mitochondrial morphology was analyzed at 30-min and 24-h post-treatment for all groups, and gene expression profile was investigated at 6-h post-treatment.
2.2. Fluorescent Light Energy (FLE) Systems {#sec2dot2-jcm-09-00559}
-------------------------------------------
FLE Systems consist of a multi-LED lamp (KT-L lamp, Klox Technologies Inc., Laval, QC, Canada) and a topical photoconverter substrate in the form of an amorphous gel (FLE-Gel) or sheet hydrogel matrix (FLE-Matrix) (LumiHeal^TM^ Gel and LumiHeal^TM^ Matrix, Klox Technologies Inc., Laval, QC, Canada). The multi-LED lamp delivers non-coherent light between 400--520 nm with a peak at app. 447 nm and a power density between 110--150 mW/cm^2^ at a distance of 5 cm from the *light-emitting diodes* (LEDs). The lamp is equipped with a 5-min timer and a distance indicator. FLE photoconverters contain a chromophore, embedded within the gel or matrix, which can absorb some of the photons from the multi-LED lamp, and emit FLE in the range of approximately 510--700 nm. Thus cells treated with FLE receive a combination of direct light from the multi-LED lamp plus FLE emitted from the Gel or Matrix photoconverter, for delivery of a full spectral range between 400--700 nm. Of note, a dose response for FLE may be observed by assessing FLE-Gel compared with FLE-Matrix, as FLE-Gel generates 0.1--0.2 J/cm^2^ of fluorescence (\~510--700 nm) whereas FLE-Matrix generates 0.2--0.7 J/cm^2^.
2.3. Fluorescence Light Energy (FLE) Protocols {#sec2dot3-jcm-09-00559}
----------------------------------------------
Three treatment conditions were tested in order to study the impact of FLE on mitochondrial morphology and gene expression. Light-treated cells received a 5-min illumination with the multi-LED lamp placed 5 cm from the bottom of the plate, without the presence of a topical photoconverter. For FLE-treated cells (FLE-Gel or FLE-Matrix) the topical photoconverters were placed under the 6-well plate, not in direct contact with cells, and the multi-LED lamp was placed at 5 cm from the bottom of the plate. Light and FLE are transmitted unchanged through the plastic bottom of the plate, thus contact with the cells is not required to induce their effects. The illumination duration was 5-min for all treatment groups. Each of the following groups were tested:(a)Healthy: HDFs maintained in basal medium (no inflammatory cocktail or illumination).(b)Inflamed: HDFs incubated in TNFα/IL-1β inflammatory cocktail.(c)Light: Inflamed HDFs illuminated for 5-min with only the multi-LED lamp (no FLE).(d)Gel: Inflamed HDFs illuminated for 5-min with the FLE-Gel system consisting of the multi-LED lamp and topical photoconverter amorphous gel (LumiHeal Gel, Klox Technologies Inc., Laval, QC, Canada).(e)Matrix: Inflamed HDFs illuminated for 5-min with the FLE-Matrix system consisting of the multi-LED lamp and topical photoconverter sheet hydrogel matrix (LumiHeal Matrix, Klox Technologies Inc., Laval, QC, Canada).Healthy HDFs were considered as the control group of the experiment.
2.4. Mitochondrial Morphology {#sec2dot4-jcm-09-00559}
-----------------------------
Cells were seeded in coverslips 24-mm in diameter and allowed to grow to a confluence of 50--60%. After treatments, cells were fixed with 4% paraformaldehyde solution (Sigma-Aldrich, USA) and washed three-times. Next, cells were permeabilized with a solution of 0.1% triton x-100 (Sigma-Aldrich, USA), for 10 min at room-temperature (RT) on a plate-shaker. After three-washes, unspecific sites were blocked with a solution of 2% bovine serum albumin (Sigma-Aldrich, USA) supplemented of 0.01% triton x-100 for 45 min. at RT with agitation. Cells were next incubated with a primary antibody against TOM20 (mitochondrial marker of the inner membrane) (BD, USA) diluted 1:100 over-night at 4 °C. The next day, cells were washed with three washes of 10 min each with agitation at RT and next incubated with a specific secondary fluorescent antibody Alexa Fluor 488 (Thermo Fisher Scientific, Waltham, USA) diluted 1:1000 in the dark for 45 min at RT with agitation. Cells were acquired in z-stacks of 51 planes at 0.2 µm each at Nikon A1 confocal microscope equipped with a 63X objective. Images obtained were deconvolved to remove blurred signal and 3D reconstructed. The mitochondrial network was then quantified by using the 3D-object counter available in software Fiji (<http://fiji.sc/wiki/index.php/Fiji> accessed on 29 April 2017) that allow to measure the total object (mitochondria) volume and the number of total objects (mitochondria) per each cell. The mean volume of single mitochondria was calculated by divide the total mitochondria volume with the number of total mitochondria. For each condition, at least 20 cells were analyzed. Data are presented as mean ± SD. Multi comparison statistical analyses were performed by using one-way ANOVA. T test was to perform all pairwise comparisons between group means. Calculated mean ± SD are reported in figure legends.
2.5. Total RNA Isolation and PCR Array Profile {#sec2dot5-jcm-09-00559}
----------------------------------------------
As previously described \[[@B44-jcm-09-00559]\], total RNA was extracted by the RNeasy Mini Kit (Qiagen, Hilden Germany) which includes DNase digestion using the RNase-Free DNase Set (Qiagen). For each sample, 500 ng of total RNA were reverse transcribed with RT^2^ First Strand Kit (Qiagen) in SimpliAmp Thermal Cycler (Thermo Fisher Scientific) following the manufacture procedures. Then, the RT^2^ Profiler PCR Array Human Mitochondrial Energy Metabolism (Qiagen) and RT^2^ Profiler PCR Array Human Mitochondria (Qiagen) in StepOne Plus Real-Time PCR System (Thermo Fisher Scientific) were performed. The amplification protocol included the activation at 95 °C for 10 min, followed by 40 cycles of denaturation at 95 °C for 15 s, and elongation at 60 °C for 1 min. The 2ΔΔCT method was used to determine the relative expression of target genes. Cycle threshold (Ct) values of target genes were normalized to the geometric mean Ct values of five housekeeping genes (ACTB: actin, beta; B2M: beta-2-microglobulin; GAPDH: glyceraldehyde-3-phosphate dehydrogenase; HPRT1: hypoxanthine phosphoribosyl transferase 1; RPLP0: ribosomal protein, large, P0). For each target gene, the average of three normalized expression levels were calculated, and *p* values were calculated using Student's t-test based on 2ΔCT values for each gene in the test group compared to the control group. Statistical significance was set at *p* \< 0.05. Results were reported as fold regulation of target genes in test group compared with control group.
3. Results {#sec3-jcm-09-00559}
==========
3.1. Mitochondrial Morphology Analysis {#sec3dot1-jcm-09-00559}
--------------------------------------
Analysis of the mitochondrial morphology was performed in all HDF conditions: (a) healthy HDFs in normal media (Healthy), (b) inflamed HDFs in TNFα/IL-1β inflammatory cocktail (Inflamed), (c) inflamed HDFs treated with light alone (Light), (d) inflamed HDFs treated with FLE-Gel (Gel), and (e) inflamed HDFs treated with FLE-Matrix (Matrix). Importantly, incubation of HDFs with the TNFα/IL-1β inflammatory cocktail led to fragmentation of the mitochondrial network ([Figure 1](#jcm-09-00559-f001){ref-type="fig"}). A global reduction of mitochondrial volume per cell was observed in Inflamed HDFs ([Figure 1](#jcm-09-00559-f001){ref-type="fig"}a) accompanied by an increase in the number of mitochondria per cell, and a reduction of individual mitochondrion volume ([Figure 1](#jcm-09-00559-f001){ref-type="fig"}b,c, respectively). Predictably, greater exacerbation of the mitochondria was observed in Inflamed HDFs at 24 h compare to 30 min, due to the continued exposure to the inflammatory cocktail ([Figure 2](#jcm-09-00559-f002){ref-type="fig"}).
Interestingly, when inflamed cells were exposed to either direct light (Light) or FLE (Gel or Matrix) the mitochondrial network showed preliminary signs of improvement at 30-min post-treatment, with some statistical differences already observed in FLE-treated cells. This improvement occurred despite the continued exposure to the TNFα/IL-1β inflammatory cocktail. Mitochondrial number per cell in Inflamed HDFs was 51 ± 13, compared with 41 ± 11 in FLE-Gel HDFs (*p* = 0.019) and 41 ± 12 in FLE Matrix HDFs (*p* = 0.0148). Similarly, individual mitochondrion volume was 12.8 ± 4.13 µm^3^ in Inflamed HDFs, compared with 18.3 ± 6.5 µm^3^ (*p* = 0.0185) and 18.6 ± 6.3 µm^3^ (*p* = 0.0156) in FLE-Gel and FLE-Matrix HDFs, respectively. Taken together, this data indicates that mitochondria in FLE-treated cells have shifted their mitochondrial network into a more fused state, with larger mitochondria that are fewer in number. While the difference in the volume of the entire mitochondrial network per cell did not reach significance between Inflamed and FLE-treated cells at the 30 min time point, it is interesting to note that while both Inflamed and Light-treated HDFs are different from Healthy HDFs (648 ± 237 µm^3^ and 639 ± 273 µm^3^ vs. 840 ± 153 µm^3^, *p* = 0.0103 and 0.0093 respectively), neither of the FLE-treated groups were significantly different from Healthy HDFs, suggesting they are already starting to recover from the inflamed state.
In order to verify this possibility, the healthiness of the mitochondrial network was analyzed for a longer period of time. At 24-h post-treatment Inflamed HDFs treated with FLE had completely recovered to healthy-HDF mitochondrial morphology, while cells treated with Light were only partially recovered ([Figure 2](#jcm-09-00559-f002){ref-type="fig"}). All observed aspects of the mitochondrial networks were significantly improved in FLE-treated cells compared with Inflamed HDFs. Volume of the entire mitochondrial network per cell was 802 ± 133 µm^3^ in FLE-Gel HDFs, and 910 ± 159 µm^3^ in FLE-Matrix, compared with 523 ± 301 µm^3^ in Inflamed HDFs (*p* = 0.0018 and 0.0001, respectively). Values for the FLE-treated cells were not significantly different from Healthy HDFs (845 ± 205 µm^3^). Additionally, the count of mitochondria number per cell for Inflamed HDFs was 57 ± 16 compared with 48 ± 14 and 39 ± 9 for FLE-Gel and FLE-Matrix (*p* = 0.0016 and *p* = 0.0001, respectively), and individual mitochondrion volume was 9.5 ± 3.5 µm^3^ in Inflamed HDFs compared with 20 ± 5.6 µm^3^ and 24 ± 7.5 µm^3^ for FLE-Gel and FLE-Matrix (*p* = 0.0016 and *p* = 0.0001, respectively). By comparison, Light-treatment had induced some recovery, and was no longer significantly different from healthy HDFs in any parameter. However, it was only significantly improved compared with Inflamed cells in the individual mitochondrion volume (10 ± 3 vs. 16 ± 4 µm^3^ for Inflamed vs. Light respectively, *p* = 0.0104).
In [Figure 3](#jcm-09-00559-f003){ref-type="fig"}, radial graphs provide a visual representation of the ability of FLE to fully rescue the mitochondrial dynamics of inflamed HDFs. All values are indexed to the healthy values for volume of the mitochondrial network per cell (top, Mitochondria Volume per cell, MV/c), number of mitochondria per cell (bottom right, Mitochondria Number per cell, MN/c) and individual mitochondrion volume (bottom left, Individual Mitochondrion Volume, IMV). The green line represents the Healthy condition, and the red line indicates the Inflamed status. At both the 30-min ([Figure 2](#jcm-09-00559-f002){ref-type="fig"}a) and 24-h ([Figure 2](#jcm-09-00559-f002){ref-type="fig"}b) post-treatment time points, it is apparent that Light-treated cells (blue lines) are shifting away from the Inflamed state towards the Healthy condition. A noticeably greater recovery is visible in FLE-treated cells (FLE-Gel in yellow and FLE-Matrix in orange) at 30-min post-treatment, where both have already progressed beyond the Light cells toward the Healthy condition. Impressively, by 24-h post-treatment FLE-treated cells are completely recovered, even overlapping with the Healthy condition.
3.2. PCR Array Gene Expression Profile {#sec3dot2-jcm-09-00559}
--------------------------------------
In order to investigate the biological impact of the treatments on mitochondrial dynamic and function, mitochondrial gene expression was probed with two different real-time PCR array, measured 6-h post-treatment.
[Table 1](#jcm-09-00559-t001){ref-type="table"} presents data on the expression of 84 key genes associated with mitochondrial respiration, including genes encoding components of the electron transport chain and oxidative phosphorylation complexes. Oxidation of Nicotinamide adenine dinucleotide (NADH) and flavin adenine dinucleotide (FADH2), key metabolites in glycolysis and the citric acid cycle, occurs within a series of four protein complexes embedded in the inner mitochondrial membrane: NADH-coenzyme Q reductase, succinate-coenzyme Q reductase, coenzyme Q-cytochrome c reductase, and CCO. The free energy generated from these processes drives oxidative phosphorylation and ATP synthesis via a fifth protein complex (ATP synthase) \[[@B45-jcm-09-00559]\].
Of note, compared with Healthy HDFs, Inflamed HDFs had upregulated *ATP4A* and *ATP6V0A2* genes and downregulated *COX6C* gene. These genes encode enzymes belonging to ATP synthase and CCO, respectively, which are fundamental for energy production and oxidative phosphorylation. The remaining 81 investigated genes were expressed at similar levels in both conditions. Compared with Inflamed cells, no treatment altered the expression of these genes.
Compared with Inflamed HDFs, cells treated with Light overexpressed two genes belonging to CCO: *COX4I2* and *COX8C.* These genes remained unchanged in both cell group treated with FLE.
Genes belonging to complex I (NADH-Coenzyme Q Reductase), complex II (Succinate-Coenzyme Q Reductase), and complex III (Coenzyme Q-Cytochrome c Reductase) maintained the same expression profile in all the conditions investigated.
[Table 2](#jcm-09-00559-t002){ref-type="table"} reports the expression profiles of 84 genes related to biogenesis and function of mitochondria, including regulators and mediators of mitochondrial molecular transport of metabolites required for the electron transport chain and oxidative phosphorylation.
Compared to Healthy cells, Inflamed HDFs showed substantial upregulation of superoxide dismutase 2 (*SOD2*) and neurofilament (*NEFL*) genes. *SOD2* is an antioxidant enzyme that protects cells from oxidative damage, whereas *NEFL* influences the dynamics of mitochondria. Although to a lesser extent, an upregulation of BCL2-antagonist/killer 1 (*BAK1*) and solute carrier family 25, member 25 (*SLC25A25*) and a downregulation of BCL2 binding component 3 (*BBC3*)*,* Cyclin-dependent kinase inhibitor 2A (*CDKN2A*)*,* IMP1 inner mitochondrial membrane peptidase-like *(IMMP1L)*, stratifin *(SFN),* solute carrier family 25 (mitochondrial carrier; citrate transporter), member 1 *(SLC25A1)*, and tumor protein p53 (*PT53*) were also observed. The *BAK1* gene encodes a pro-apoptotic protein. *SLC25A25* and *SLC25A1* genes encode proteins belonging to the family of calcium-binding mitochondrial carriers in the inner membranes of the mitochondria. Their functions are to transport proteins, metabolites, nucleotides and cofactors through the mitochondrial membrane and thereby connect and/or regulate cytoplasm and matrix functions \[[@B46-jcm-09-00559]\]. The *SFN* gene encodes the stratifin protein, a cell cycle checkpoint protein that binds translation and initiation factors and functions as a regulator of mitotic translation. It also regulates signal transduction pathways and cellular trafficking \[[@B47-jcm-09-00559]\]. Cyclin-dependent kinase inhibitor 2A (CDKN2A) is an inhibitor of CDK4 kinase and arrests the cell cycle in G1 phase \[[@B48-jcm-09-00559]\]. The *BBC3* gene encodes the pro-apoptotic protein PUMA that is regulated by the protein tumor suppressor p53 \[[@B49-jcm-09-00559]\]. The *PT53* gene encodes p53 protein that exhibits diverse and global functions, including cell cycle arrest, senescence, and apoptosis. Through these pathways, p53 facilitates the repair and survival of damaged cells or eliminates severely injured cells from the replicative pool to protect the organism. One of the most dramatic responses to p53 activation is the induction of apoptosis \[[@B50-jcm-09-00559]\]. No treatment altered any of the above-mentioned genes compared with Inflamed HDFs.
Of interest, an upregulation of genes involved in the production of ATP and reduction of ROS was observed. Compared to Inflamed cells, Light and FLE-Matrix treated HDFs showed an increase in *UCP1* gene expression. This gene encodes uncoupling protein 1, a member of the family of mitochondrial anion carrier proteins. The uncoupling proteins (UCPs) separate oxidative phosphorylation from ATP synthesis with energy dissipated as heat, also referred to as mitochondrial proton leak. UCPs facilitate the transfer of anions from the inner to the outer mitochondrial membrane and the return transfer of protons from the outer to the inner mitochondrial membrane reducing the mitochondrial membrane potential. A minor decline in the mitochondrial membrane potential leads to a significant decrease in harmful levels of ROS production. UCPs proteins decrease mitochondrial membrane potential to a level still allowing both production of required amounts of ATP, and of lower ROS levels that would be relatively harmless to the cells \[[@B51-jcm-09-00559]\]. Light treated cells also had a reduction in the expression of *SLC25A27*, alias *UCP4*. Moreover, compared to Inflamed HDFs, FLE-Matrix treated cells had an upregulation of *CPT1B* gene. Carnitine palmitoyltransferase-1 (CPT1) is located in the inner aspect of the outer mitochondrial membrane and transports long-chain fatty acids into mitochondria for β-oxidation. The acetyl-coenzyme A (acetyl- CoA) produced by oxidative degradation of fatty acids enters the citric acid cycle for oxidation to carbon dioxide and water by the electron transport chain to yield ATP \[[@B52-jcm-09-00559]\]. Compared to Inflamed cells, the FLE-Gel treated HDFs showed an increase in *SLC25A31* gene expression. The protein encoded by this gene is a member of the ADP/ATP carrier family of proteins that exchange cytosolic ADP for matrix ATP in the mitochondria \[[@B53-jcm-09-00559]\].
4. Discussion {#sec4-jcm-09-00559}
=============
In recent years, FLE has been demonstrated to be a safe and effective non-pharmacological intervention for the treatment of acute and chronic wounds \[[@B23-jcm-09-00559],[@B37-jcm-09-00559],[@B39-jcm-09-00559]\] and other skin pathologies \[[@B41-jcm-09-00559],[@B42-jcm-09-00559]\], however, the exact mechanism through which FLE realizes an acceleration in the wound healing process is not yet clear.
The principle behind FLE is the absorption of photons by endogenous chromophores inside the treated tissues, resulting in various biological effects \[[@B36-jcm-09-00559]\]. It is widely known that mitochondria are an initial site of light action in cells, and the central molecule for this is CCO, the terminal enzyme of the mitochondrial respiratory chain. CCO transfers one electron (from each of four cytochrome c molecules), to a single oxygen molecule, producing two molecules of water. At the same time the four protons required are translocated across the mitochondrial membrane, producing a proton gradient that ATP synthase uses to synthesize ATP \[[@B54-jcm-09-00559]\]. CCO has two heme centers and two copper centers. Each of these metal centers can exist in an oxidized or a reduced state, and each are photo-acceptors with different absorption spectra in the red and near-infrared region (up to 950 nm). The absorption of photons by CCO leads to an increase in enzymatic activity, an increase in oxygen consumption, and an increase in ATP production thanks to the photodissociation of inhibitory nitric oxide (NO) \[[@B55-jcm-09-00559]\]. Since NO is non-covalently bound to the heme and copper centers and competitively blocks oxygen at a ratio of 1:10, a relatively low energy photon can move the NO, allowing a lot of respiration to take place \[[@B56-jcm-09-00559]\].
This theory describes the structural changes that CCO undergoes after exposure to photons. Conversely, the present work aims to investigate the mitochondrial morphology and the expression profile of genes related to the respiratory chain, and the mitochondrial dynamics and functions in inflamed cells after the exposure to photons as either direct light (PBM) or FLE. For this purpose, in vitro cultures of HDFs were stimulated overnight with a cocktail of pro-inflammatory cytokines (TNFα/IL-1β) to induce an inflamed state. Then, Inflamed HDFs were treated with one of three different photonic treatments: Light (a non-coherent light between 400 and 520 nm); FLE-Gel of FLE-Matrix (a mix of non-coherent light and FLE with a broad spectrum between 400 and 700 nm). All Inflamed and treated cells were maintained in the inflammatory cocktail environment post-treatment for the duration of the experiment.
Thirty minutes post-treatments, confocal microscopy analysis showed mitochondrial fragmentation in Inflamed HDFs and Light treated cells. Likewise, Motori and colleagues have previously shown that pro-inflammatory stimuli produced localized changes in mitochondrial dynamics, favoring fission over fusion \[[@B57-jcm-09-00559]\]. Indeed, the mitochondrial architecture of a cell results from movement, tethering, fusion and fission events. Owing to frequent fission and fusion, different shapes of mitochondria can be found within a cell, including small vesicles, short rods, and long reticular networks. Strong evidence has demonstrated that mitochondrial dynamics are important for cell viability, senescence, mitochondria health, bioenergetic function, quality control, and intracellular signaling. Shortening of mitochondria is a result of increased fission activity or decreased fusion activity, and is typical for states of reduced bioenergetic efficiency (increased respiratory leak) \[[@B58-jcm-09-00559]\].
Fragmentation of mitochondria may also indicate activation of the autophagic removal of mitochondria by a process known as mitophagy. This process is described as a pro-survival mechanism employed as an early response to cell stress, since it removes damaged mitochondria. Contrarily, if unfavorable condition persists, mitophagy results in cell death. Changes in mitochondrial morphology also occur in the early step of the cell death mechanisms apoptosis and necrosis \[[@B59-jcm-09-00559],[@B60-jcm-09-00559]\]. The progressive loss of mitochondrial network during exposure to inflammatory mediators affirms that chronic inflammation affects mitochondrial morphology and bioenergetic functions. Of note, even as early as 30-min post-treatment, FLE-treated cells were starting to progress toward a healthier state of mitochondria, with significant differences observed between Inflamed HDFs and FLE-treated HDFs in the number of mitochondria per cell and the individual mitochondrion volume.
A full recovery of mitochondrial network occurred by 24-h post-treatment with FLE, manifested by mitochondrial networks comparable to those of Healthy HDFs. A trend towards recovery was also observed with Light-treated cells (PBM), however, while there were no significant differences from Healthy HDFs for mitochondria morphology, only the individual mitochondrion volume was different between Inflamed and Light-treated HDFs. This improved biological effect of FLE compared with PBM has been previously observed \[[@B36-jcm-09-00559],[@B40-jcm-09-00559]\]. While not clearly understood, a possible explanation is biological sensitivity to changes in the light properties that arise during generation of FLE, including polarity, micro-pulsations, or coherency. Of note, opsin proteins are known to be sensitive to light polarity, thus it is feasible that other endogenous chromophores, including CCO, may be as well \[[@B61-jcm-09-00559]\]. This enhanced impact of FLE compared with Light treatment is clearly visible in the radial graphs ([Figure 3](#jcm-09-00559-f003){ref-type="fig"}), with FLE-treated HDFs demonstrating both a quicker and more potent recovery.
Interestingly, we may see a dose-response starting to occur between FLE-Gel and FLE-Matrix. At 30-min, both groups have almost identical mitochondrial dynamics, however, by 24 h it appears as if the FLE-Matrix may be shifting the mitochondria network balance further towards fusion, even beyond that of the Healthy HDFs. However, there is no statistical significance to this finding, so further studies would be needed to better understand any dose effects associated with FLE and its biological implication. It is known that processes associated with increased energy are characterized by mitochondrial elongation and by respiration coupled to ATP synthesis \[[@B58-jcm-09-00559]\]. We can hypothesize that the absorption of low energy photons by CCO has led to a restoration of the mitochondrial network due to an increase in enzymatic activity and respiration, and therefore an increase in ATP production.
Gene expression analysis did not report great alterations in expression of genes encoding enzymes of respiratory chain. In Light-treated cells, compared to Inflamed HDFs, upregulation of *COX4I2, COX8C* and *UCP1* genes was observed. Little is reported about the expression of these genes in the literature, and nothing about their expression during the inflammatory process. However, Suárez and workers linked the upregulation of *COX4I2* gene to the increase in *UCP1* expression in an obese animal model \[[@B62-jcm-09-00559]\]. They demonstrated that the overexpression of the thermogenic factor UCP1 improves metabolic phenotype concurrent with mitochondrial biogenesis trough the *COX4I2* gene overexpression. In Light-treated cells, the improvement in mitochondrial biogenesis due to the upregulation of CCO-coding genes could be the basis of the mitochondrial morphology recorded 24 h post-treatment. In fact, through an increase in the number of mitochondria, a restoration of the mitochondrial volume equal to healthy cells has been observed. Conversely, in Inflamed cells, where *COX6C* gene downregulation was recorded, drastic mitochondrial fragmentation and a decrease in whole mitochondrial volume were found.
It was demonstrated that TNF-α and IL-1β decrease ATP production by reducing the activity of complex I, as well as by reducing mitochondrial membrane potential and inducing mitochondrial DNA damage \[[@B63-jcm-09-00559]\]. The damage and mutations of mitochondrial DNA lead to the synthesis of functionally impaired respiratory chain subunits, promoting increased ROS production. In Inflamed cells, the increased production of ROS was confirmed by the upregulation of *SOD2*, the main mitochondrial antioxidant enzyme. It has been reported that when the inflammatory signal through the NF-κB pathway is strong and extended, the *SOD2* expression is upregulated to maintain ROS homeostasis. However, if the ROS level is not restrained, cell damage and death occur \[[@B64-jcm-09-00559]\]. In Light-treated cells, the levels of ROS are controlled by *UCP1* upregulation via a slight decline in mitochondrial membrane potential, however allowing the production of ATP \[[@B44-jcm-09-00559]\]. Since UCP1 was recently found deregulated during inflammation \[[@B65-jcm-09-00559]\], and essential to preserve mitochondrial structural integrity and function \[[@B66-jcm-09-00559],[@B67-jcm-09-00559]\], we can speculate that Light treatment may improve the mitochondrial function and recover mitochondrial network by the upregulation of this protein.
Instead, treatment with FLE-Matrix resulted in upregulation of *UCP1* as well as *CTP1B*, which is involved in mitochondrial β-oxidation of lipids long-chain fatty acids. Both genes are linked to an increase in ATP production because UCP separates oxidative phosphorylation from ATP synthesis allowing the production of ATP, also in the presence of ROS. Instead CTP1 favors β-oxidation of lipids ensuring the production of ATP \[[@B51-jcm-09-00559],[@B52-jcm-09-00559]\]. It has already been shown that an increase in ATP synthesis is connected with the mitochondrial elongation under different cell conditions \[[@B58-jcm-09-00559]\]. Compared to Light treatment, FLE-Matrix increased the volume of whole mitochondrial network through an increase in the volume of single mitochondria, favoring mitochondrial elongation. Having found the highest mitochondrial network recovery in this condition, we can assume that the gain in ATP production by mitochondrial β-oxidation may help in restoration of mitochondrial dynamics.
Conversely, an increase in *UCP1* and *CTP1B* expression was not found in FLE-Gel treated cells, but an increase in *SLC25A31* gene expression was observed. This gene encodes a nucleotide transporter which imports ADP into the mitochondrial matrix, where it can be converted to ATP by ATP synthase, and then exports the newly synthesized ATP to the cytosol \[[@B68-jcm-09-00559]\]. Therefore, this carrier protein plays an important role in spending the ATP on metabolic processes necessary for cell survival. Since both FLE-Gel and FLE-Matrix recover mitochondrial network, it can be hypothesized that upregulation of this mitochondrial carrier causes an increase in ATP production due to the activation of ATP synthase.
ATP is the cellular energy-carrying molecule essential for multiple cellular functions. Reduced energy levels threaten cellular homeostasis and integrity. Impaired energy metabolism may trigger pro-apoptotic signaling, oxidative damage, cytotoxicity and impede mitochondrial DNA repair \[[@B69-jcm-09-00559]\]. All the evaluated treatments led to an improvement in intracellular ATP levels, thus energizing the cells. However, it seems that FLE treatment rescued inflamed mitochondrial networks, returning them to a healthy condition despite the presence of oxidative stress caused by continued exposure to TNFα/IL-1β inflammatory cytokines. This return of mitochondrial homeostasis by FLE could be a key mechanism supporting accelerated healing and tissue regeneration in wounds and skin pathologies treated with FLE.
5. Conclusions {#sec5-jcm-09-00559}
==============
In conclusion, this work investigated the effect of FLE on mitochondrial homeostasis in an in vitro model of inflammation. FLE treatment leads to restoration of the mitochondrial network by 24-h post-treatment, as well as upregulation of *UCP1* and *CPT1B* genes, which encode proteins favoring the production of ATP through oxidative phosphorylation and lipid β-oxidation, respectively. These observations, coupled with the previously established capacity of PBM to stimulate CCO, show a beneficial effect of FLE in the treatment of inflamed wounds. However, these findings derive from an in vitro model and should be validated by an in vivo investigation.
The authors would like to acknowledge Emanuela Buonamici for her technical and graphical support.
Conceptualization, M.Z., S.E.C., L.H. and B.Z.; methodology, M.Z., L.F. and S.P.; formal analysis, M.Z., L.H., L.F. and S.P.; resources, M.Z., B.Z. and P.P.; data curation, L.F., M.Z., S.E.C and S.P.; writing---original draft preparation, M.Z., L.F. and S.P.; writing---review and editing, M.N., S.E.C, L.H., L.F. and S.P.; visualization, M.Z., C.S. and F.B.; supervision, B.Z. and P.P.; project administration, M.Z., S.E.C. and L.H. All authors have read and agreed to the published version of the manuscript.
This research was funded in part by Klox Technologies, Inc. Additionally, P.P. is supported by the Italian Association for Cancer Research (AIRC, IG-18624), Telethon (GGP11139B) and local funds from the University of Ferrara.
Drs. Zago, Campbell, Hébert, and Nielsen are employees of Klox Technologies. The other authors have no conflict of interest relevant to the content of this article.
{#jcm-09-00559-f001}
{#jcm-09-00559-f002}
{#jcm-09-00559-f003}
jcm-09-00559-t001_Table 1
######
Human Mitochondrial Energy Metabolism PCR array.
Gene Inflamed vs. Healthy *p* Value Light vs. Inflamed *p* Value FLE-Gel vs. Inflamed *p* Value FLE-Matrix vs. Inflamed *p* Value
------------ ---------------------- ----------- -------------------- ----------- ---------------------- ----------- ------------------------- -----------
*ATP12A* 1.08 0.3024 1.31 0.5219 −1.11 0.1645 1.02 0.1648
*ATP4A* **2.23** 0.7832 1.03 0.6392 −1.03 0.2884 −1.03 0.2967
*ATP4B* 1.45 0.3082 1.79 0.7798 −1.21 0.2096 1.52 0.2148
*ATP5A1* −1.08 0.9484 −1.14 0.9730 1.06 0.7168 1.03 0.8362
*ATP5B* −1.05 0.9663 −1.04 0.9969 1.00 0.9906 1.04 0.5569
*ATP5C1* −1.05 0.9517 −1.12 0.9898 −1.08 0.8672 −1.04 0.9454
*ATP5F1* −1.23 0.1785 1.01 0.2102 1.05 0.1455 1.07 0.1458
*ATP5G1* −1.02 0.8216 1.01 0.9565 −1.05 0.1904 1.02 0.2007
*ATP5G2* −1.15 0.8984 1.28 0.9757 1.28 0.2901 1.08 0.2901
*ATP5G3* −1.04 0.2888 −1.02 0.8484 −1.09 0.1986 −1.04 0.1983
*ATP5H* −1.13 0.4352 1.05 0.7275 1.14 0.2591 1.08 0.2589
*ATP5I* −1.08 0.9752 1.05 0.8689 −1.05 0.8167 0.99 0.9525
*ATP5J* 1.02 0.9968 −1.02 0.9737 −1.04 0.9612 0.99 0.9791
*ATP5J2* −1.23 0.9810 −1.07 0.9603 −1.01 0.2836 1.22 0.2834
*ATP5L* 1.68 0.5228 −1.36 0.8836 −1.53 0.2096 −1.27 0.2134
*ATP5O* −1.24 0.8745 1.05 0.7809 −1.03 0.8725 −1.04 0.5821
*ATP6V0A2* **2.59** 0.9270 1.03 0.3655 −1.79 0.2746 −1.09 0.1516
*ATP6V0D2* 1.91 0.6030 1.62 0.9800 1.36 0.2356 1.21 0.4197
*ATP6V1C2* 1.11 0.1784 1.12 0.6929 −1.18 0.2758 1.20 0.2789
*ATP6V1E2* 1.02 0.6367 1.01 0.8071 1.17 0.1381 1.04 0.1352
*ATP6V1G3* −1.70 0.4560 1.12 0.8299 −1.11 0.3228 −1.05 0.3231
*BCS1L* −1.46 0.5451 −1.07 0.8080 1.16 0.4274 1.24 0.4942
*COX4I1* 1.08 0.6312 −1.02 0.6719 1.03 0.7022 1.06 0.7194
*COX4I2* −1.85 0.9443 **3.24** 0.8211 1.40 0.1855 1.90 0.1854
*COX5A* −1.15 0.9779 −1.02 0.3636 −1.05 0.1026 1.10 0.0556
*COX5B* −1.08 0.8219 1.06 0.6803 1.01 0.6248 1.01 0.6164
*COX6A1* 1.05 0.9912 −1.02 0.6452 −1.05 0.9001 −1.06 0.8699
*COX6A2* 1.29 0.7882 1.55 0.4056 −1.03 0.5157 −1.02 0.3269
*COX6B1* 1.09 0.8830 1.06 0.3851 −1.12 0.3845 −1.06 0.1568
*COX6B2* 1.54 0.6464 1.36 0.9643 −1.06 0.2774 1.21 0.2775
*COX6C* **−2.23** 0.2374 1.12 0.4507 −1.11 0.2241 1.18 0.2246
*COX7A2* 1.31 0.8657 1.00 0.7080 1.00 0.2980 1.02 0.3576
*COX7A2L* 1.14 0.9597 1.18 0.3701 −1.02 0.2149 1.02 0.2810
*COX7B* −1.23 0.1399 1.03 0.9086 −1.11 0.8089 −1.07 0.8180
*COX8A* −1.02 0.6213 1.04 0.7314 −1.01 0.3849 1.03 0.4075
*COX8C* 1.11 0.3632 **3.60** 0.3340 −1.20 0.2921 1.67 0.2925
*CYC1* 1.03 0.9564 1.02 0.9543 −1.01 0.8904 1.00 0.8846
*LHPP* −1.12 0.7322 1.14 0.6447 1.10 0.9146 1.02 0.9953
*NDUFA1* 1.07 0.6609 −1.06 0.7084 1.06 0.2131 −1.16 0.2167
*NDUFA10* −1.06 0.9250 0.99 0.7389 1.01 0.6267 0.99 0.7675
*NDUFA11* −1.03 0.9466 −1.02 0.9339 1.04 0.8978 1.09 0.9636
*NDUFA2* 1.07 0.9914 1.02 0.3434 1.11 0.1150 1.01 0.0749
*NDUFA3* −1.29 0.8155 1.05 0.9753 −1.04 0.8924 −1.03 0.8893
*NDUFA4* 1.09 0.5712 1.03 0.7243 −1.04 0.0812 −1.04 0.0819
*NDUFA5* −1.97 0.5816 1.28 0.9752 −1.24 0.2903 1.15 0.3374
*NDUFA6* 1.03 0.1884 −1.07 0.5686 1.18 0.1108 1.12 0.1032
*NDUFA7* −1.03 0.9399 −1.03 0.9026 1.04 0.3395 1.00 0.2997
*NDUFA8* −1.23 0.8994 −1.08 0.7818 1.23 0.8069 0.99 0.4317
*NDUFAB1* 1.08 0.5897 1.03 0.9813 −1.03 0.7660 −1.04 0.7574
*NDUFB10* −1.03 0.2064 −1.03 0.6846 1.00 0.7707 −1.02 0.7615
*NDUFB2* −1.15 0.9577 1.05 0.2898 1.05 0.1720 −1.05 0.3253
*NDUFB3* −1.14 0.8127 1.03 0.8891 1.07 0.3763 −1.03 0.3061
*NDUFB4* 1.24 0.3245 −1.06 0.7046 1.09 0.0957 1.04 0.0861
*NDUFB5* −1.07 0.5122 1.04 0.1030 1.05 0.0993 1.03 0.0712
*NDUFB6* −1.08 0.6834 1.09 0.3958 −1.07 0.4169 −1.21 0.1946
*NDUFB7* −1.25 0.8837 1.03 0.9139 1.07 0.9146 1.11 0.9221
*NDUFB8* −1.29 0.9782 −1.08 0.4260 1.04 0.0860 1.04 0.0867
*NDUFB9* −1.07 0.8154 1.04 0.6321 −1.03 0.3974 1.00 0.3031
*NDUFC1* −1.13 0.6376 1.06 0.4091 1.10 0.1097 −1.03 0.1077
*NDUFC2* 1.03 0.6633 1.03 0.5582 1.26 0.4491 1.21 0.6712
*NDUFS1* −1.05 0.8937 1.03 0.8077 1.06 0.2821 1.06 0.2782
*NDUFS2* −1.05 0.6046 1.08 0.8433 1.08 0.5569 1.05 0.5270
*NDUFS3* 1.14 0.7629 −1.09 0.3173 1.05 0.1314 1.05 0.0777
*NDUFS4* −1.11 0.3228 1.05 0.6511 −1.01 0.6848 −1.04 0.5687
*NDUFS5* −1.13 0.9234 1.05 0.8181 1.01 0.3283 −1.03 0.4957
*NDUFS6* −1.10 0.7346 1.09 0.9146 1.05 0.6572 1.11 0.6882
*NDUFS7* −1.29 0.7148 −1.32 0.7148 1.05 0.3395 1.02 0.2788
*NDUFS8* −1.23 0.8670 1.03 0.9886 1.18 0.9448 1.23 0.9275
*NDUFV1* −1.21 0.8922 0.99 0.6639 1.04 0.1205 1.06 0.1277
*NDUFV2* −1.21 0.8671 1.08 0.5305 1.06 0.1460 −1.02 0.2872
*NDUFV3* 1.13 0.7272 −1.10 0.3081 1.26 0.1167 1.24 0.1145
*OXA1L* −1.13 0.5362 1.14 0.5885 1.27 0.1040 1.22 0.1026
*PPA1* −1.71 0.2951 −1.03 0.5926 1.04 0.8757 −1.02 0.8405
*PPA2* −1.19 0.5189 1.04 0.4868 −1.02 0.2007 −1.04 0.2018
*SDHA* −1.42 0.3281 1.09 0.5285 1.05 0.1407 1.02 0.1423
*SDHB* 1.04 0.5917 1.01 0.8479 1.04 0.6252 1.03 0.7692
*SDHC* −1.19 0.7466 1.22 0.6577 −1.05 0.1798 −1.02 0.1799
*SDHD* −1.22 0.8702 −1.23 0.9003 1.00 0.5197 −1.07 0.7193
*UQCR11* 1.26 0.4176 −1.16 0.8663 −1.04 0.1969 −1.06 0.1947
*UQCRC1* −1.03 0.9857 −1.13 0.8575 −1.02 0.9116 −1.04 0.8919
*UQCRC2* −1.29 0.8090 −1.21 0.1460 1.02 0.1357 1.04 0.1405
*UQCRFS1* −1.03 0.9645 −1.17 0.8910 −1.03 0.9580 −1.03 0.9512
*UQCRH* −1.03 0.9254 −1.07 0.9792 −1.04 0.9931 −1.04 0.9962
*UQCRQ* 1.00 0.5753 −1.11 0.7945 1.06 0.8875 1.03 0.8636
ATP12A: ATPase, H+/K+ transporting, nongastric, alpha polypeptide; ATP4A: ATPase, H+/K+ exchanging, alpha polypeptide; ATP4B: ATPase, H+/K+ exchanging, beta polypeptide; ATP5A1: ATP synthase, H+ transporting, mitochondrial F1 complex, alpha subunit 1, cardiac muscle; ATP5B: ATP synthase, H+ transporting, mitochondrial F1 complex, beta polypeptide; ATP5C1: ATP synthase, H+ transporting, mitochondrial F1 complex, gamma polypeptide 1; ATP5F1: ATP synthase, H+ transporting, mitochondrial Fo complex, subunit B1; ATP5G1: ATP synthase, H+ transporting, mitochondrial Fo complex, subunit C1; ATP5G2: ATP synthase, H+ transporting, mitochondrial Fo complex, subunit C2; ATP5G3: ATP synthase, H+ transporting, mitochondrial Fo complex, subunit C3; ATP5H: ATP synthase, H+ transporting, mitochondrial Fo complex, subunit d; ATP5I: ATP synthase, H+ transporting, mitochondrial Fo complex, subunit E; ATP5J: ATP synthase, H+ transporting, mitochondrial Fo complex, subunit F6; ATP5J2: ATP synthase, H+ transporting, mitochondrial Fo complex, subunit F2; ATP5L: ATP synthase, H+ transporting, mitochondrial Fo complex, subunit G; ATP5O: ATP synthase, H+ transporting, mitochondrial F1 complex, O subunit; ATP6V0A2: ATPase, H+ transporting, lysosomal V0 subunit a2; ATP6V0D2: ATPase, H+ transporting, lysosomal 38 kDa, V0 subunit d2; ATP6 V1C2: ATPase, H+ transporting, lysosomal 42 kDa, V1 subunit C2; ATP6V1E2: ATPase, H+ transporting, lysosomal 31 kDa, V1 subunit E2; ATP6V1G3:ATPase, H+ transporting, lysosomal 13 kDa, V1 subunit G3; BCS1L: BCS1-like (S. cerevisiae); COX4I1: Cytochrome c oxidase subunit IV isoform 1; COX4I2: Cytochrome c oxidase subunit IV isoform 2; COX5A: Cytochrome c oxidase subunit Va; COX5B: Cytochrome c oxidase subunit Vb; COX6A1: Cytochrome c oxidase subunit VIa polypeptide 1; COX6A2: Cytochrome c oxidase subunit VIa polypeptide 2; COX6B1: Cytochrome c oxidase subunit Vib polypeptide 1; COX6B2: Cytochrome c oxidase subunit VIb polypeptide 2; COX6C: Cytochrome c oxidase subunit Vic; COX7A2: Cytochrome c oxidase subunit VIIa polypeptide 2; COX7A2L: Cytochrome c oxidase subunit VIIa polypeptide 2 like; COX7B: Cytochrome c oxidase subunit VIIb; COX8A: Cytochrome c oxidase subunit VIIIA; COX8C: Cytochrome c oxidase subunit VIIIC; CYC1: Cytochrome c-1; LHPP: Phospholysine phosphohistidine inorganic pyrophosphate phosphatase; NDUFA1: NADH dehydrogenase 1 alpha subcomplex, 1, 7.5 kDa; NDUFA10: NADH dehydrogenase 1 alpha subcomplex, 10, 42 kDa; NDUFA11: NADH dehydrogenase 1 alpha subcomplex, 11, 14.7 kDa; NDUFA2: NADH dehydrogenase 1 alpha subcomplex, 2, 8 kDa; NDUFA3: NADH dehydrogenase 1 alpha subcomplex, 3, 9 kDa; NDUFA4: NADH dehydrogenase 1 alpha subcomplex, 4, 9 kDa; NDUFA5: NADH dehydrogenase 1 alpha subcomplex, 5, 13 kDa; NDUFA6: NADH dehydrogenase 1 alpha subcomplex, 6, 14 kDa; NDUFA7: NADH dehydrogenase 1 alpha subcomplex, 7, 14.5 kDa; NDUFA8: NADH dehydrogenase 1 alpha subcomplex, 8, 19 kDa; NDUFAB1: NADH dehydrogenase 1, alpha/beta subcomplex, 1, 8 kDa; NDUFB10: NADH dehydrogenase 1 beta subcomplex, 10, 22 kDa; NDUFB2: NADH dehydrogenase 1 beta subcomplex, 2, 8 kDa; NDUFB3: NADH dehydrogenase 1 beta subcomplex, 3, 12 kDa; NDUFB4: NADH dehydrogenase 1 beta subcomplex, 4, 15 kDa; NDUFB5: NADH dehydrogenase 1 beta subcomplex, 5, 16 kDa; NDUFB6: NADH dehydrogenase 1 beta subcomplex, 6, 17 kDa; NDUFB7: NADH dehydrogenase 1 beta subcomplex, 7, 18 kDa; NDUFB8: NADH dehydrogenase 1 beta subcomplex, 8, 19 kDa; NDUFB9: NADH dehydrogenase 1 beta subcomplex, 9, 22 kDa; NDUFC1: NADH dehydrogenase 1, subcomplex unknown, 1, 6 kDa; NDUFC2: NADH dehydrogenase 1, subcomplex unknown, 2, 14.5 kDa; NDUFS1: NADH dehydrogenase Fe-S protein 1, 75 kDa; NDUFS2: NADH dehydrogenase Fe-S protein 2, 49 kDa; NDUFS3: NADH dehydrogenase Fe-S protein 3, 30 kDa; NDUFS4: NADH dehydrogenase Fe-S protein 4, 18 kDa; NDUFS5: NADH dehydrogenase Fe-S protein 5, 15 kDa; NDUFS6: NADH dehydrogenase Fe-S protein 6, 13 kDa; NDUFS7: NADH dehydrogenase Fe-S protein 7, 20 kDa; NDUFS8: NADH dehydrogenase Fe-S protein 8, 23 kDa; NDUFV1: NADH dehydrogenase flavoprotein 1, 51 kDa; NDUFV2: NADH dehydrogenase flavoprotein 2, 24 kDa; NDUFV3: NADH dehydrogenase flavoprotein 3, 10 kDa; OXA1 L: Oxidase (cytochrome c) assembly 1-like; PPA1: Pyrophosphatase 1; PPA2: Pyrophosphatase 2; SDHA: Succinate dehydrogenase complex, subunit A, flavoprotein (Fp); SDHB: Succinate dehydrogenase complex, subunit B, iron sulfur (Ip); SDHC: Succinate dehydrogenase complex, subunit C, integral membrane protein, 15 kDa; SDHD: Succinate dehydrogenase complex, subunit D, integral membrane protein; UQCR11: Ubiquinol-cytochrome c reductase, complex III subunit XI; UQCRC1: Ubiquinol-cytochrome c reductase core protein I; UQCRC2: Ubiquinol-cytochrome c reductase core protein II; UQCRFS1: Ubiquinol-cytochrome c reductase, Rieske iron-sulfur polypeptide 1; UQCRH: Ubiquinol-cytochrome c reductase hinge protein; UQCRQ: Ubiquinol-cytochrome c reductase, complex III subunit VII, 9.5 kDa.
jcm-09-00559-t002_Table 2
######
Human Mitochondria PCR array.
Gene Inflamed vs. Healthy *p* Value Light vs Inflamed *p* Value FLE-Gel vs. Inflamed *p* Value FLE-Matrix vs. Inflamed *p* Value
------------ ---------------------- ----------- ------------------- ----------- ---------------------- ----------- ------------------------- -----------
*AIFM2* 1.17 0.9186 1.05 0.9694 1.15 0.8808 1.05 0.3564
*AIP* −1.12 0.9070 1.16 0.8602 1.21 0.9389 1.13 0.8818
*BAK1* **2.18** 0.4666 −1.09 0.8772 1.02 0.7111 −1.05 0.1289
*BBC3* **−2.91** 0.9816 1.31 0.9722 1.08 0.9145 1.04 0.6868
*BCL2* 1.93 0.4330 1.13 0.4945 1.14 0.4947 1.16 0.5045
*BCL2L1* −1.37 0.9929 1.07 0.4951 −1.07 0.3020 −1.07 0.3250
*BID* 1.51 0.6882 1.01 0.8719 1.06 0.9178 −1.35 0.2987
*BNIP3* −1.36 0.8372 1.06 0.9668 −1.23 0.9024 −1.16 0.9280
*CDKN2A* **−2.99** 0.7208 −1.03 0.5949 1.13 0.3788 −1.23 0.3111
*COX10* −1.03 0.6641 1.01 0.4646 −1.10 0.5937 −1.08 0.3185
*COX18* −1.36 0.7253 1.32 0.6170 1.30 0.4723 1.09 0.3871
*CPT1B* −1.33 0.8146 0.99 0.9582 1.82 0.9127 **2.04** 0.4682
*CPT2* 1.11 0.6167 1.11 0.5475 1.04 0.4144 −1.03 0.3732
*DNM1L* 1.04 0.6504 1.02 0.4551 −1.02 0.3857 1.01 0.3536
*FIS1* −1.29 0.7077 1.12 0.4843 1.15 0.4490 1.07 0.3857
*GRPEL1* 1.40 0.9575 −1.15 0.8785 −1.12 0.3880 −1.01 0.8202
*HSP90AA1* −1.19 0.8602 1.06 0.5892 1.13 0.1296 1.27 0.4059
*HSPD1* −1.17 0.8993 −1.01 0.9746 1.00 0.9350 1.07 0.8728
*IMMP1L* **−2.51** 0.9134 −1.02 0.9908 1.06 0.9928 1.10 0.9780
*IMMP2L* −1.31 0.5724 1.08 0.9491 1.08 0.9692 1.06 0.4700
*LRPPRC* −1.20 0.9990 1.19 0.7798 1.07 0.6555 1.10 0.6068
*MFN1* −1.07 0.7624 1.05 0.9604 1.11 0.8119 1.10 0.5266
*MFN2* −1.09 0.9807 −1.11 0.7044 −1.15 0.5278 −1.06 0.5023
*MIPEP* −1.84 0.9939 −1.03 0.8895 1.23 0.9541 1.26 0.4258
*MPV17* −1.56 0.9050 1.05 0.9991 1.06 0.9772 1.04 0.5992
*MSTO1* 1.04 0.9795 1.10 0.9732 1.26 0.6936 1.17 0.3500
*MTX2* 1.01 0.4066 1.07 0.4383 1.33 0.2515 1.28 0.1868
*NEFL* **16.93** 0.6088 −1.36 0.9775 −1.65 0.4401 −1.51 0.3560
*OPA1* −1.20 0.5516 1.15 0.9987 1.10 0.6269 1.15 0.9292
*PMAIP1* −1.06 0.8934 −1.01 0.9052 −1.24 0.9505 −1.03 0.9594
*RHOT1* −1.42 0.8552 1.13 0.6221 1.20 0.4302 1.17 0.3495
*RHOT2* 1.07 0.7768 1.03 0.5066 1.00 0.5196 1.00 0.9821
*SFN* **−2.38** 0.4837 1.66 0.8181 1.90 0.4771 1.75 0.4248
*SH3GLB1* −1.22 0.9383 1.28 0.9551 1.52 0.8607 1.52 0.3748
*SLC25A1* **−2.07** 0.8876 1.13 0.8759 1.05 0.7808 1.04 0.7874
*SLC25A10* 1.38 0.6742 −1.62 0.9591 −1.85 0.9908 −1.60 0.9927
*SLC25A12* −1.26 0.8391 −1.36 0.9808 −1.22 0.9588 −1.12 0.3523
*SLC25A13* 1.51 0.8949 1.11 0.9549 −1.05 0.9313 1.09 0.3061
*SLC25A14* 1.13 0.5778 1.06 0.9728 1.04 0.9826 1.03 0.2791
*SLC25A15* 1.20 0.6720 −1.02 0.5093 1.18 0.4814 1.21 0.3807
*SLC25A16* −1.12 0.9002 1.13 0.8996 1.03 0.8461 1.16 0.6215
*SLC25A17* 1.26 0.8423 −1.05 0.9918 −1.10 0.8879 −1.06 0.5068
*SLC25A19* 1.44 0.8739 −1.08 0.9144 −1.07 0.8900 −1.06 0.9309
*SLC25A2* 1.14 0.6760 1.17 0.4763 1.15 0.4239 1.48 0.3471
*SLC25A20* −1.19 0.8596 −1.14 0.7744 1.05 0.8105 1.05 0.5398
*SLC25A21* −1.42 0.8432 0.99 0.5065 1.30 0.5826 1.43 0.3696
*SLC25A22* 1.21 0.7463 1.07 0.6007 1.11 0.6098 1.17 0.3715
*SLC25A23* −1.43 0.7167 −1.09 0.4906 −1.52 0.4234 −1.27 0.3587
*SLC25A24* −1.13 0.8747 1.05 0.8545 1.02 0.9513 1.03 0.3730
*SLC25A25* **2.30** 0.9570 −1.06 0.9910 −1.11 0.8293 −1.09 0.6173
*SLC25A27* −1.27 0.3919 **−2.25** 0.9605 −1.10 0.4895 −1.55 0.7273
*SLC25A3* −1.20 0.3646 1.02 0.7269 1.03 0.4790 1.01 0.1843
*SLC25A30* −1.36 0.9004 1.12 0.9875 1.08 0.9820 1.10 0.9916
*SLC25A31* −1.12 0.8581 1.28 0.9795 **2.47** 0.9464 −1.06 0.4897
*SLC25A37* 1.92 0.6504 1.07 0.9261 1.21 0.8353 1.24 0.8423
*SLC25A4* −1.86 0.6059 1.06 0.8584 −1.03 0.6999 1.02 0.8379
*SLC25A5* 1.00 0.9160 −1.09 0.5923 1.01 0.5904 1.05 0.3863
*SOD1* −1.14 0.6756 1.09 0.4430 1.08 0.4504 1.14 0.3994
*SOD2* **17.28** 0.8853 1.28 0.7959 1.19 0.7786 1.13 0.7637
*STARD3* −1.04 0.3845 1.00 0.8834 −1.02 0.9115 1.01 0.9488
*TAZ* −1.26 0.8577 1.02 0.3116 1.27 0.2549 1.22 0.9748
*TIMM10* 1.22 0.8033 −1.08 0.5207 −1.09 0.4573 1.03 0.3611
*TIMM10B* 1.14 0.7175 1.06 0.9674 1.01 0.6375 −1.01 0.7582
*TIMM17A* 1.22 0.7552 −1.08 0.7767 1.05 0.7586 1.01 0.7073
*TIMM17B* −1.33 0.7433 1.29 0.5761 1.15 0.5304 1.23 0.4434
*TIMM22* −1.06 0.7358 −1.08 0.4982 1.00 0.4019 1.17 0.3696
*TIMM23* 1.07 0.9080 −1.05 0.9481 1.00 0.9475 −1.01 0.7871
*TIMM44* 1.13 0.7086 −1.06 0.5628 −1.14 0.7906 −1.13 0.3392
*TIMM50* −1.06 0.9866 1.09 0.8853 1.00 0.9968 1.09 0.8634
*TIMM8A* 1.28 0.7994 −1.11 0.1517 −1.05 0.2349 −1.01 0.2730
*TIMM8B* −1.11 0.9864 −1.06 0.9095 −1.09 0.8707 1.02 0.9847
*TIMM9* 1.37 0.6803 −1.14 0.6204 −1.29 0.4490 −1.33 0.3385
*TOMM20* 1.07 0.9647 1.00 0.9962 −1.05 0.9785 −1.03 0.9860
*TOMM22* −1.03 0.9395 −1.03 0.9564 −1.05 0.9226 1.00 0.8254
*TOMM34* 1.16 0.9339 1.04 0.9646 0.99 0.9878 1.08 0.9486
*TOMM40* 1.29 0.5140 −1.16 0.6759 −1.09 0.8690 0.99 0.7455
*TOMM40L* −1.01 0.9861 1.13 0.8819 1.05 0.6619 1.15 0.3629
*TOMM70A* −1.16 0.9512 1.07 0.9631 1.19 0.8725 1.17 0.3903
*TP53* **−2.04** 0.7434 1.44 0.8865 1.27 0.5986 −1.05 0.3399
*TSPO* −1.30 0.8269 1.10 0.5675 1.02 0.4945 1.00 0.3463
*UCP1* −1.52 0.8095 **3.89** 0.8001 −1.09 0.6418 **2.70** 0.3743
*UCP2* −1.52 0.9517 1.10 0.9361 1.08 0.8185 1.09 0.3625
*UCP3* 1.59 0.7862 1.64 0.6184 −1.31 0.5903 1.48 0.4868
*UXT* 1.07 0.9794 1.09 0.6893 1.05 0.8536 1.02 0.4300
AIFM2: Apoptosis-inducing factor, mitochondrion-associated, 2; AIP: Aryl hydrocarbon receptor interacting protein; BAK1: BCL2-antagonist/killer 1; BBC3: BCL2 binding component 3; BCL2: B-cell CLL/lymphoma 2; BCL2L1: BCL2-like 1; BID: BH3 interacting domain death agonist; BNIP3: BCL2/adenovirus E1B 19 kDa interacting protein 3; CDKN2A: Cyclin-dependent kinase inhibitor 2A; COX10: COX10 homolog, cytochrome c oxidase assembly protein; COX18: COX18 cytochrome c oxidase assembly homolog; CPT1B: Carnitine palmitoyltransferase 1B; CPT2: Carnitine palmitoyltransferase 2; DNM1L: Dynamin 1-like; FIS1: Fission 1; GRPEL1: GrpE-like 1, mitochondrial; HSP90AA1: Heat shock protein 90 kDa alpha, class A member 1; HSPD1: Heat shock 60 kDa protein 1; IMMP1L: IMP1 inner mitochondrial membrane peptidase-like; IMMP2L: IMP2 inner mitochondrial membrane peptidase-like; LRPPRC: Leucine-rich PPR-motif containing; MFN1: Mitofusin 1; MFN2: Mitofusin 2; MIPEP: Mitochondrial intermediate peptidase; MPV17: MpV17 mitochondrial inner membrane protein; MSTO1: Misato homolog 1; MTX2: Metaxin 2; NEFL: Neurofilament, light polypeptide; OPA1: Optic atrophy 1; PMAIP1: Phorbol-12-myristate-13-acetate-induced protein 1; RHOT1: Ras homolog gene family, member T1; RHOT2: Ras homolog gene family, member T2; SFN: Stratifin; SH3GLB1: SH3-domain GRB2-like endophilin B1; SLC25A1: Solute carrier family 25 (mitochondrial carrier; citrate transporter), member 1; SLC25A10: Solute carrier family 25, member 10; SLC25A12: Solute carrier family 25, member 12; SLC25A13: Solute carrier family 25, member 13; SLC25A14: Solute carrier family 25, member 14; SLC25A15: Solute carrier family 25, member 15; SLC25A16: Solute carrier family 25, member 16; SLC25A17: Solute carrier family 25, member 17; SLC25A19: Solute carrier family 25, member 19; SLC25A2: Solute carrier family 25, member 2; SLC25A20: Solute carrier family 25, member 20; SLC25A21: Solute carrier family 25, member 21; SLC25A22: Solute carrier family 25, member 22; SLC25A23: Solute carrier family 25, member 23; SLC25A24: Solute carrier family 25, member 24; SLC25A25: Solute carrier family 25, member 25; SLC25A27: Solute carrier family 25, member 27; SLC25A3: Solute carrier family 25, member 3; SLC25A30: Solute carrier family 25, member 30; SLC25A31: Solute carrier family 25, member 31; SLC25A37: Solute carrier family 25, member 37; SLC25A4: Solute carrier family 25, member 4; SLC25A5: Solute carrier family 25, member 5; SOD1: Superoxide dismutase 1; SOD2: Superoxide dismutase 2; STARD3: StAR-related lipid transfer (START) domain containing 3; TAZ: Tafazzin; TIMM10: Translocase of inner mitochondrial membrane 10 homolog; TIMM10B: Fracture callus 1 homolog; TIMM17A: Translocase of inner mitochondrial membrane 17 homolog A; TIMM17B: Translocase of inner mitochondrial membrane 17 homolog B; TIMM22: Translocase of inner mitochondrial membrane 22 homolog; TIMM23: Translocase of inner mitochondrial membrane 23 homolog; TIMM44:Translocase of inner mitochondrial membrane 44 homolog; TIMM50: Translocase of inner mitochondrial membrane 50 homolog; TIMM8A: Translocase of inner mitochondrial membrane 8 homolog A; TIMM8B: Translocase of inner mitochondrial membrane 8 homolog B; TIMM9: Translocase of inner mitochondrial membrane 9 homolog; TOMM20: Translocase of outer mitochondrial membrane 20 homolog; TOMM22: Translocase of outer mitochondrial membrane 22 homolog; TOMM34: Translocase of outer mitochondrial membrane 34; TOMM40: Translocase of outer mitochondrial membrane 40 homolog; TOMM40L: Translocase of outer mitochondrial membrane 40 homolog-like; TOMM70A: Translocase of outer mitochondrial membrane 70 homolog A; TP53: Tumor protein p53; TSPO: Translocator protein (18 kDa); UCP1: Uncoupling protein 1; UCP2: Uncoupling protein 2; UCP3: Uncoupling protein 3; UXT: Ubiquitously-expressed transcript.
[^1]: co first author.
| {
"pile_set_name": "PubMed Central"
} |
Introduction {#Sec1}
============
Burn injury is associated with a complex pathophysiological response with rapidly cascading influences impacting on the whole of the patient with multisystem disruption.\[[@CR1]\] The injury results in physical and psychological sequel such that every very intervention from the point of injury will influence the scar worn for life. The patient embarks upon a path which may be widely variable in terms of first aid, wound cleaning, prehospital care and preparation for transfer, pain management, resuscitation, surgery, wound care, nutrition, scar management, and functional and psychological rehabilitation.\[[@CR2]\] Clearly the clinical problem faced on a daily basis is complex and research is essential in continuing to developing innovative solutions to solve the clinical problems.\[[@CR3]\]Access this article online**Quick Response Code:** {#d30e202}**Website:** [www.burnstrauma.com](http://www.burnstrauma.com)**DOI:** 10.4103/2321-3868.123071
Looking back at the accumulation of knowledge over the past decades demonstrates significant progress has been made with ongoing research being core to the development of burn care. With improvement in survival from burn injury the focus has shifted to the continuous improvement in the quality of survival.\[[@CR4]\]
In order to give the patient the best evidence based treatment at every point in time from the time of injury, we need to consider three aspects;The patient needs based on the clinical assessment at that point in time.The experience/knowledge level of the treating clinician underpinned by ongoing education and training.The environment of operation optimized based on prior knowledge.The concept of the "triangle of care" has been used to understand the capacity to treat the patient at a given time and drive the triage process.\[[@CR5]\]
Bringing these elements together will dictate the outcome for that patient at that time guiding each intervention.\[[@CR6]\] That is the starting point, bringing together what we know and delivering that knowledge to the bedside.\[[@CR7]\]
Challenges in the research of burn care {#Sec2}
=======================================
So where is the challenge?**Utilizing what we know effectively:** It has been stated that if current knowledge was implemented we would see a significant reduction in complications and concurrent improvement in outcomes. The translation of evidence into practice is an area in need of research and audit. The investigation of drivers and barriers to the implementation of evidence is vital to action timely translation into clinical practice. Pivotal to the continuous improvement in care is a systematic approach to education and training in the broadest sense. Community education ensuring appropriate use of first aid can have a significant impact on outcome.\[[@CR8]\] There is no substitution for meticulous attention to detail; using what we know in the best capacity possible, then we can push the boundaries.\[[@CR9]\]**Harnessing the potential opportunities from the ever expanding knowledge base in science and technology:** Pushing the boundaries, engaging in collaborations with basic science, population health, and clinical research to provide innovative solutions to complex clinical problems. Collaboration between disciplines provides real opportunities for improvements in clinical care, translating to improved outcomes for patients. We need to understand how the experimental situation, with control of all but one variable, can be extrapolated into the clinical situation. The design of clinical trials is an ongoing challenge due to the complex nature of the responses. A clear focus on burn injury is essential for a targeted problem solving approach and has facilitated great advances in care of the burn injured patient. However, this should not be at the expense of a broad general knowledge gaining insight into potential links and facilitating cross-fertilization.
Clinical practice is a fusion of experience and knowledge based on the observations of the natural history and the impact of interventions guiding advancements.
An essential element in observation is development of an appropriate measurement tool for each variable to facilitate the research process.\[[@CR10]\] How do we measure the extent of the injury, the impact of the injury on the given individual, and the outcome post injury?
How do we measure and correct for events, both intrinsic and extrinsic to the patient, along such a complex pathway?
The initial measurement of the injury severity and scoring systems is vexed as complex factors are weighted to give scores to facilitate comparison.\[[@CR11]\]
We have many examples linking wide ranging aspects of the individual demonstrating the complexity, such as the impact of premorbid personality on injury outcome the measurement of which is challenging and stimulates much debate.\[[@CR12]\]
We also know that knowledge of the etiology is essential as we observe routinely the progression of a scald injury differing from flame injury. Understanding and documenting the detail in data collection is key to understanding the validity of comparison of either a single variable, or of a network of interacting variables.
Such tools as the laser Doppler scanner have the potential to assist in depth assessment, but it is influenced by many factors such as resuscitation and of course availability.\[[@CR13]\] The use of laser body surface mapping has potential to percentage total body surface area, but relies on accurate definition of the edge of the injury and availability. The other known drivers to outcome such as age may be easily recordable but chronological age does not equate with physiological age in many individuals, with respect to concurrent pathologies and use of alcohol and drugs.\[[@CR14]\]
Considering the measurement the impact of the multiple interventions post injury, attempts have been made to breakdown the measurement into aspects such as the domains of the Burn Specific Health Scale (BSHS) physical and psychological? Increasing efforts in the area of outcome measurement related to the burn injury have validated a combination of subjective and objective measures.
Progress has been made in the area of assessment and measurement, but is still a work in progress and an area of great potential as new technologies have the potential to shift the subjective to the objective. For example, the accurate reliable wound assessment and scar outcome is fundamental to clinical research and areas were many technologies have been explored.**Developing rigorous systems of validated measurement and data collection with transparent analysis published for the benefit of all the burn population:** As we develop and increasingly invest in robust research systems based on sound research governance, we have to understand the responsibility of research and in particular the publication of results with transparency. Steven E. Wolf, editor of BURNS has publicly stated on many occasions the responsibility to see the work through to the finish with publication is vital. The funding of research could be put forward as the greatest challenge for the future. Yes it is a challenge, but all of us engaged in the endeavor need to share by publication to be realistic regarding the worth of our work. Nothing should be wasted.\[[@CR15]\]
The 50^th^ anniversary edition of the Medical Journal of Australia published the vision of clinical care in a number of disciplines in 50 years' time. With respect to burn injury repair the aim is healing by regeneration and restoration of function. The vision was described;
"Assessment is key in understanding the extent of injury.
Debridement is focused on tissue salvage.
Reconstruction balances repair with regeneration.
Investigation of multimodality, multiscale characterization, including confocal microscopy and synchrotron technology will quantify assessment.
Debridement using autolytic inflammatory control techniques with image-guided physical methods will ensure the vital tissue frameworks are retained.
Tissue-guided regeneration afforded by self-assembly nanoparticles will provide the framework to guide cells to express the appropriate phenotype in reconstruction.
To solve the clinical problem a multidisciplinary scientific approach is needed to ensure the quality of the scar is worth the pain of survival."
Within a decade many of the technologies highlighted are available and in need for research to move along the innovation pathway to ensure safe implementation into healthcare systems. Progress requires collaboration at all stages from basic science, clinical trial design to population health research with a link to health economics, driven by improved clinical outcomes.
We need to challenge our thinking and in particular revisit our preconceived ideas and "gold standards". For example, we should set our sights higher than wound healing achieved by split thickness skin grafting which commits the patient to a lifelong scar. Rather, we should aim for regeneration of the skin to the preinjury state matched specifically for that body site.\[[@CR16]\] There is an increasing understanding of variation in individual responses to therapeutic interventions; what is the role of *n* = 1 clinical trials?\[[@CR17]\] Is such case reporting anecdotal or an opportunity to engage in individualized medicine, linking genetic knowledge with potential therapies to focus on the outcome for that single individual?
Understanding that every intervention from the time of injury influences the scar worn for life has driven research by the multidisciplinary burns teams in a multitude of directions; from first aid and prevention to stem cells to rehabilitation strategies to name a few.
We live in exciting times with tools which have the potential to identify novel therapeutic targets as we strive towards regenerative healing. The exploration of regeneration and the interplay between genes, cells and tissues is possible with advanced bioinformatics systems developing to understand the network interactions.\[[@CR18]\]
There has been great progress in burn care driven by dedicated individuals focused on improving the outcome of burn patients over the past decades. The challenge we face now is to capitalize on that tradition and link with the opportunities afforded by the exciting and unprecedented growth in science and technology.\[[@CR19]\]
As we drive to "One World One Standard of Burn Care" as championed by International Society for Burn Injuries (ISBI) at the 2012 international meeting, we need to understand the key drivers of outcome, standardize the care around the key elements, and then we may make progress towards meaningful comparisons.\[[@CR20]\] The key to improvement in global outcomes is in the translation of research into practice. The essential element in the equation is the education at all levels from community injury prevention and first aid, to the dedicated burns team setting and all in between to ensure the quality of the outcome is worth the pain of survival.\[[@CR21]\]
**How to cite this article:** Wood FM. Burn care: The challenges of research. Burn Trauma 2013;1:105--8.
**Source of Support:** Nil, **Conflict of Interest:** None declared.
| {
"pile_set_name": "PubMed Central"
} |
Plain-language summary
======================
An important part of managing COPD is self-management by the patient. With IT-based self-management programs, patients can do this from their own home, which is easier for them and more cost-effective for the health care provider. Our goal was to explore how a generic Internet-based health-promotion program, The Preventive Plan (TPP), containing programs to support behavior change and health information, combined with support from a specialist nurse, could help patients to manage their COPD. We collected data from and interviewed 19 patients. We found that TPP and nurse support improved patient activation (the belief that one has a role in self-managing care and possession of knowledge and skills to do so), physical function, mood and confidence, and reduced patients' information needs. Patients with more severe COPD gained the most benefit. Support from the nurse was considered to be more important than the Internet-based program by those with more severe COPD, and lack of motivation to use IT was a barrier to use, but poor IT skills and disease severity were not. Our findings demonstrate that a low-intensity intervention can produce meaningful benefits, and provides learning about the challenges of this approach and the importance of the nurse coach role.
Introduction
============
COPD is a progressive lung disease causing breathlessness, cough, fatigue, reduced exercise capacity, and frequent infections, with high societal burden.[@b1-copd-12-2955] Medication optimizes airway function and reduces symptoms, but cannot address the psychosocial impact of the disease, including anxiety and depression,[@b2-copd-12-2955] social isolation, and loss of independence and self-esteem.[@b3-copd-12-2955]
Self-management support interventions in COPD have been shown to improve health-related quality of life,[@b4-copd-12-2955],[@b5-copd-12-2955] exercise capacity,[@b5-copd-12-2955] and self-efficacy[@b5-copd-12-2955] and reduce COPD-related hospital admissions.[@b4-copd-12-2955] However, for those unable or unwilling to attend group classes, IT-based approaches can provide support.[@b6-copd-12-2955]--[@b11-copd-12-2955] IT-based self-management and lifestyle-change interventions for COPD have reported improved health-related quality of life,[@b9-copd-12-2955],[@b12-copd-12-2955]--[@b16-copd-12-2955] physical activity,[@b13-copd-12-2955],[@b14-copd-12-2955],[@b17-copd-12-2955] exercise capacity,[@b12-copd-12-2955],[@b16-copd-12-2955] and functional capacity[@b18-copd-12-2955],[@b19-copd-12-2955] and reduced number and duration of hospital admissions.[@b9-copd-12-2955],[@b18-copd-12-2955],[@b20-copd-12-2955]--[@b23-copd-12-2955] Participants have shown improved knowledge and awareness of their symptoms and health status,[@b19-copd-12-2955],[@b24-copd-12-2955] and qualitative analyses indicate positive effects on self-management, psychological condition, and coping ability.[@b18-copd-12-2955],[@b25-copd-12-2955] However, other studies have failed to detect positive changes in quality of life,[@b7-copd-12-2955],[@b17-copd-12-2955],[@b22-copd-12-2955] smoking cessation,[@b6-copd-12-2955] physical activity,[@b6-copd-12-2955],[@b7-copd-12-2955],[@b11-copd-12-2955] or dyspnea,[@b6-copd-12-2955],[@b7-copd-12-2955],[@b11-copd-12-2955],[@b14-copd-12-2955] and results for self-efficacy are mixed,[@b7-copd-12-2955],[@b17-copd-12-2955],[@b25-copd-12-2955] which may reflect heterogeneity among interventions: some were complemented with nurse or allied health-professional support,[@b7-copd-12-2955],[@b8-copd-12-2955],[@b15-copd-12-2955]--[@b18-copd-12-2955],[@b20-copd-12-2955],[@b22-copd-12-2955] while others offered a stand-alone technological platform.[@b6-copd-12-2955],[@b12-copd-12-2955]--[@b14-copd-12-2955],[@b21-copd-12-2955],[@b24-copd-12-2955],[@b26-copd-12-2955]
Patient activation is the belief that one has a role in self-managing care and the possession of knowledge and skills to manage one's condition, maintain functioning, collaborate with health care providers, and access appropriate care.[@b27-copd-12-2955] Activation impacts on health behavior, clinical outcomes, health care costs, and patient experiences.[@b28-copd-12-2955] However, little is known about the impact of IT-based interventions on patient activation in COPD.
Using a case-series design, we explored the feasibility of an off-the-shelf Internet-based health-promotion program, the Preventive Plan (TPP), coupled with nurse-coach support for home-based self-management of COPD with a focus on patient activation. We also aimed to identify patient-reported self-management benefits and factors influencing self-management goal achievement, and to generate generic learning to benefit users of ehealth interventions.
Materials and methods
=====================
Participants
------------
Patients were recruited between May 2012 and January 2013. Inclusion criteria at the outset were: COPD diagnosis (forced expiratory volume in 1 second \[FEV~1~\]/forced vital capacity \[FVC\] \<0.7), FEV~1~ \<50% predicted (GOLD stage 3 and 4), two or more COPD exacerbation-related hospital admissions within the last year, had not attended pulmonary rehabilitation within the last year, and age \>18 years. Exclusion criteria were diagnosis of lung cancer within the last year and being under the care of a psychologist for management of anxiety/depression. Eleven patients fulfilling these criteria were recruited (group 1). It was challenging to engage patients who met these criteria, and so to increase the number of participants, we adjusted the inclusion criteria for a second group (group 2) to include: FEV~1~ \<80% predicted (GOLD stage 2--4) and troubled by breathlessness. We imposed no restriction on the number of exacerbation-related hospital admissions for this group. This enabled us to recruit a further eight patients.
Design
------
This case series employed mixed methods of quantitative outcome assessment and qualitative interviews. The intervention comprised an off-the-shelf Internet-based program, TPP, and nurse-coach support. TPP encompassed primary prevention, secondary prevention, and chronic-condition management. It was a generic tool personalized by uploading personal details, completing a health-risk assessment, and generation of a personalized prevention plan. Users accessed health-information resources and action programs to support behavior change. There was a messaging facility for email contact with the nurse. TPP could be accessed ad libitum by participants. The nurse coach (same for all participants) was a respiratory nurse specialist. She was trained to use TPP by the application developers and received training from a clinical communication skills specialist to use a coproduction consultation model to support self-management.[@b29-copd-12-2955] Through home visits, telephone, and email she provided self-management support, disease education, assisted participants to use TPP and IT hardware, and "signposted" complementary self-management and COPD-specific resources (Supplementary material). For group 2, she added further disease-specific content through weekly emails of material that replicated the educational content of pulmonary rehabilitation. This addition was made in the context of a pragmatic exploration of how the online platform could be built on for the benefit of patients. We provided laptop computers and Internet connectivity for patients with no existing access. No changes were made to participants' usual care.
Data collection
---------------
Baseline measures, including spirometry, were collected by the nurse. Outcomes were measured at 3-month follow-up. Patient activation for self-management was assessed by the Patient Activation Measure.[@b30-copd-12-2955] Scores of 0--100 represent four activation levels: level 1 (\<47) -- disengaged, overwhelmed, and may not believe patient role is importantlevel 2 (47.1--55.1) -- becoming aware, but lacks confidence and knowledgelevel 3 (55.2--67) -- beginning to take actionlevel 4 (\>67.1) -- making changes, but may have difficulty maintaining behaviors over time.
We also assessed patient health-related quality of life using the Chronic Respiratory Questionnaire (CRQ),[@b31-copd-12-2955] anxiety and depression using the Hospital Anxiety and Depression Scale (HADS),[@b32-copd-12-2955] and information needs using the Lung Information Needs Questionnaire (LINQ).[@b33-copd-12-2955]
All participants were interviewed by KH at 1 month after baseline to explore interim progress and at 3-month follow-up. Semistructured interviews explored participants' self-management aims, benefits achieved, and challenges or enablers to achieving their self-management goals ([Table S1](#SD1-ijn-8-2589){ref-type="supplementary-material"}). Website activity was logged and time-stamped to identify the number of days on which participants logged in and the number of pages loaded. The nurse recorded the number and duration of contacts by email, phone, and home visit.
Data analysis
-------------
Baseline characteristics for groups 1 and 2 were compared using Fisher's exact, independent *t*-, and Mann--Whitney *U* tests. Baseline and follow-up outcome data for each group (14 patients in total) were analyzed with paired *t*- and Wilcoxon signed-rank tests. Statistical significance was set at *P*≤0.05. All statistical analysis was performed in SPSS version 23 (IBM, Armonk, NY, USA).
Transcribed interviews were imported into NVivo software and analyzed thematically.[@b34-copd-12-2955] Transcripts were initially read by FE to gain an overview of patients' experiences. A deductive coding framework was developed by FE, consisting of codes from the interview-topic areas, and transcripts were coded by FE and ER. During coding, deductive codes were supplemented by inductive codes to capture unexpected findings. Codes were checked for duplication and redundancy and organized into categories. Categories were compared across transcripts and between the groups and summarized in matrices, on which the final interpretation was based.
Ethical considerations
----------------------
This case series was categorized as a service evaluation by the Research and Development Department at Cambridge University Hospitals NHS Foundation Trust and was subsequently approved and registered as a service evaluation with the Safety and Quality Support Department at the same trust (project registration 570). Patients gave written informed consent to participate in the study to the nurse coach at the first visit. Specific written consent to participate in qualitative interviews was taken by KH at the beginning of the first interview.
Results
=======
Participants
------------
A total of 46 patients were invited and 19 responded (eleven in group 1 and eight in group 2) ([Figure 1](#f1-copd-12-2955){ref-type="fig"}). Commonest reasons for declining were not wishing to use the Internet and not needing help. A total of 14 patients completed 3-month follow-up (eight in group 1 and six in group 2). Patients lost to follow-up were older (mean 73.2±11.84 vs 60.64±9.44 years, *P*=0.028), but did not differ on any other baseline characteristic. Nine patients from group 1 and six from group 2 were interviewed, with interviews lasting 20--58 minutes.
Baseline characteristics
------------------------
Several indicators of disease severity were more severe in group 1 compared to group 2: lower FEV~1~, FEV~1~% predicted and FEV~1~/FVC, and higher GOLD stage, Medical Research Council dyspnea scale, supplementary oxygen use, and hospital admissions in the previous year ([Table 1](#t1-copd-12-2955){ref-type="table"}). The mean number of exacerbations in the past year for group 2 was seven, higher than for group 1, despite few hospital admissions. These were self-reported exacerbations that required antibiotics or steroid use for worsening symptoms, and the data for group 2 were skewed by two patients who had 20 and 14 exacerbations (but only zero and one hospital admissions, respectively).
Patient activation
------------------
Fourteen patients provided baseline and follow-up data ([Table 2](#t2-copd-12-2955){ref-type="table"}). Baseline activation scores were significantly lower (*P*=0.043) for group 1 (51.98±7.36) than group 2 (61.22±7.9). At follow-up, group 1 showed a significant increase (*P*=0.022) in activation from level 2 (51.98±7.36) to 3 (64.11±8.09). Group 2 increased insignificantly (*P*=0.111) from level 3 (61.22±7.89) to 4 (72.73±16.11).
Other outcomes
--------------
The only significant difference at baseline between groups 1 and 2 was in mastery (mean 3.44±1.38 vs 5.17±0.61, *P*=0.015). At follow-up, group 1 showed a clinically meaningful and statistically significant improvement in mastery (mean 3.44±1.38 to 4.38±1.4, *P*=0.04) and improvement in anxiety (mean 11.38±3.81 to 8.88±4.32, *P*=0.028). Both groups had reduced information needs (group 1, median \[IQR\] 5.5 \[4.25--6\] to 3 \[1.25--3.75\], *P*=0.008; group 2, 5.5 \[2.26\] to 3.33 \[1.75\], *P*=0.027).
Interview data
--------------
[Tables 3](#t3-copd-12-2955){ref-type="table"} and [4](#t4-copd-12-2955){ref-type="table"} describe the categories into which the codes were organized. In some cases, common category headings apply across both groups, but the way in which the category is manifested differs between the groups.
Aims
----
Both groups wished to improve their physical functioning and better manage their health and COPD ([Table 3](#t3-copd-12-2955){ref-type="table"}). In this regard, group 1 were focused on getting fitter and staying out of hospital, while group 2, with less severe COPD, focused on daily activities and slowing COPD progression. Group 1 also wanted more independence, and group 2 wanted increased motivation.
Benefits
--------
Both groups benefited from improved physical functioning ([Table 3](#t3-copd-12-2955){ref-type="table"}). For group 1 this was the most commonly reported benefit and enabled more independence. Both groups felt better able to manage their COPD and gained understanding of healthy behaviors, such as healthy eating. Social and emotional functioning improved for both groups. While group 2 gained in self-belief, for group 1 the benefits were wider-ranging, including being able to connect better with others and improved mood and confidence: "I felt good enough to get out the door under me own steam, carrying my oxygen ... I walked round the garden ... feeling normal for, you know, it could be years actually. \[115, group 1\]"
Some in group 2 benefited from being able to communicate better with health care professionals. One patient was able to be more assertive in explaining why they wanted a hospital referral, and this helped them to get the referral they had long wanted: "I've never been referred to the hospital, so \[nurse coach\] said, "What's your goal?" Well, I went to \[the doctor\] and said, "I feel different and I want to know if there's anything else going on", so he sent me for chest X-rays ... and took my blood ... He phoned up \[community respiratory nurse\] and said about referring me to hospital ... I achieved my goals. \[122, group 2\]"
Factors facilitating goal achievement
-------------------------------------
Nurse-coach support was important for both groups and the benefits similar, although group 1 derived more in terms of emotional support. For this group, it had a greater impact than TPP ([Table 4](#t4-copd-12-2955){ref-type="table"}): "I was getting depressed and down, so \[the nurse\] ... she suggested about the models ... and that was an ideal thing ... every time she spoke to me by phone or sent me an email or come round, she's always asked how it's going with the models. So yeah, that's made a difference. \[110, group 1\]The most helpful thing is, like, talking to \[the nurse\] about different things. Computers don't really answer you back when you ask it a question. \[108, group 1\]"
TPP provided additional support for both groups through information and action plans, eg, healthy eating and smoking cessation, but group 2 found TPP relatively more useful: "\[TPP\] gives you different ideas, what they class as one serving ... it's made me think about what I'm eating ... I would have just thrown steak into a frying pan ... whereas now I'll grill it. \[109, group 2\]"
Both groups used TPP most frequently to contact the nurse. Other facilitating factors relevant to both groups were social support from family and friends, psychological factors relating to the individual's own feelings of confidence and motivation, and the possession of self-management skills, which in turn were supported by the nurse coach and the use of TPP.
Challenges to achieving goals
-----------------------------
Poor physical well-being affected progress and prevented activity for both groups, but particularly so for group 1 ([Table 4](#t4-copd-12-2955){ref-type="table"}): "I've had a couple of chest infections, so I've been finding it quite hard. I've got a DVD that \[the nurse\] gave me. I haven't been able to get into it because of the last few months. My tummy's quite painful. \[101, group 1\]"
For some in group 1, this reduced their confidence and was compounded by being in hospital. Psychological factors were a challenge for both groups, with anxiety a particular challenge for group 1: "I just get so out of breath that I just sit here because ... I'm scared to get out of breath. Even though I know it's not going to hurt me, it's just very unpleasant. \[105, group 1\]"
Both groups reported other priorities that made it more difficult to work toward goals. For group 1, there were more health-related commitments, and for group 2 more domestic tasks. Bad weather limited outside activity for both groups. In addition, some in group 1 reported lack of social support and resources for activities, while for some in group 2 behavioral factors, such as trying to do too much and poor housing, presented challenges.
Challenges in using TPP
-----------------------
Orientation toward IT and the design of TPP were challenges for some participants in both groups. In group 1, lack of interest in computers (one patient), poor IT skills and confidence (five) and difficulty using the site (three) made patients less willing or able to use TPP. Two who reported no benefits at all from TPP found it difficult to use, and one had poor IT skills. However, two of the highest users had poor IT skills, and one had found TPP difficult to use. In group 2, two participants reported no benefits from TPP. One found it impersonal and complicated: "I don't feel that I'm mastering it ... perhaps there's too much in there. \[118, group 2\]"
However, similarly to group 1, of the two highest users in group 2, one had poor IT skills and both found TPP difficult to use. Barriers related to individual circumstances were ill health among group 1 and pain and poor memory for passwords among group 2.
Contact and usage
-----------------
On average, participants in group 1 accessed TPP 1.21 days/week and had 12.51 minutes of face-to-face contact time/week, 3.53 minutes of phone time/week, and 0.71 emails per week from the nurse coach ([Table 5](#t5-copd-12-2955){ref-type="table"}). Group 2 accessed TPP 1.32 days per week and had 18.54 minutes of face-to-face contact time/week, 2.67 minutes of phone time/week, and 1.27 emails per week. Both groups used TPP most frequently for messaging the nurse and secondly for action programs. The most frequent users were from both groups.
Discussion
==========
In this case series of a self-management support intervention for COPD patients comprising an Internet-based health-promotion program coupled with nurse-coach support, patients with more advanced COPD (group 1) improved significantly in activation, mastery, and anxiety and reported increased independence and confidence. Both this group and a second group with less advanced disease (group 2) felt better informed about COPD and reported improved physical functioning, though benefits overall were greater for patients with more advanced COPD. This occurred despite group 1 not receiving disease-specific content that was made available to group 2 and having fewer emails from the nurse coach. For group 1, the motivational, emotional, and goal support from the nurse was more important than TPP in building confidence to self-manage; some patients with clear goals used the action plans and information resources in TPP as additional support. Group 2 participants valued the nurse support, but emphasized it less than TPP.
Group 1, with more advanced disease, had lower activation at baseline and achieved more positive outcomes than group 2, despite reporting more challenges, eg, being physically unwell, anxious, and low in confidence. This is consistent with other studies where participants with poorer baseline scores achieved greater gains from self-management and pulmonary rehabilitation programs[@b35-copd-12-2955] and improvement in self-efficacy following an expert-patients program.[@b36-copd-12-2955]
Our work found benefits in CRQ mastery in group 1, in contrast to a previous study of a self-management program consisting of tailored sessions and telephone support from a nurse, which found no significant change in any CRQ domains or the COPD self-efficacy scale;[@b37-copd-12-2955] this may have been a result of greater disease severity in our cohort. We also observed significant improvements in HADS -- anxiety in group 1, consistent with a supported self-management intervention of initial training sessions followed by nurse home visits in a similar cohort.[@b38-copd-12-2955] Other studies have found that patient-education programs have the potential to improve health-related quality of life within short time frames (4 months), although results are mixed.[@b39-copd-12-2955],[@b40-copd-12-2955]
Both groups benefited from the nurse input (which was more important than TPP for group 1) and used TPP most frequently for contacting the nurse. Nurses add value to IT-based interventions by raising awareness of the need for behavior change,[@b17-copd-12-2955] providing positive feedback and goal support[@b41-copd-12-2955] and social and motivational support to persevere with interventions.[@b17-copd-12-2955] In our intervention, the nurse was reassuring, and talking to her about COPD was supportive for patients. Relationships, including supportive and empathic relationships with healthcare professionals, are one of three core mechanisms for successful ehealth interventions.[@b42-copd-12-2955]
Participants varied in their use of TPP. The most frequent users were from both groups and two of the highest users had poor IT skills, so disease severity and low IT confidence and skills were not necessarily barriers where participants were motivated. Williams et al[@b43-copd-12-2955] found that patients could use a tablet-based mobile-health application regardless of previous experience, and Cummings et al[@b44-copd-12-2955] reported that negative previous computer experience did not prevent adoption of an online self-monitoring diary.
Patient preference may be important for engagement in Internet-based interventions.[@b7-copd-12-2955] Not all participants were enthusiastic about using TPP: some later declared little interest in computers, and it may have differed from their expectations. Common reasons for low usage of ehealth applications are unfamiliarity with technology among elderly populations, insufficiently stimulating applications, lack of change of content over time, and low use by health care professionals.[@b6-copd-12-2955],[@b19-copd-12-2955],[@b20-copd-12-2955] Factors that increase user engagement, such as refreshing content on a regular basis, sending prompts, being able to choose elective components,[@b45-copd-12-2955] or tailoring health information to personal needs,[@b46-copd-12-2955] were present in TPP, but still not all participants were engaged. Some found it complex and difficult to navigate, which was a barrier to use. High-quality user-centered design is important for uptake and use of ehealth interventions.[@b43-copd-12-2955],[@b47-copd-12-2955] TPP was an off-the-shelf application, previously evaluated in workplace settings in the US,[@b48-copd-12-2955] and not validated in COPD patients. More engagement may be achieved with a dedicated application designed collaboratively with the target group and technology-design approaches combined with health models.[@b49-copd-12-2955]
Implications for future research, policy, and practice
------------------------------------------------------
We found that participants with the lowest levels of activation gained the most. Nurse support was particularly important for those with greater emotional and self-management support needs. This suggests the need for careful assessment of patient needs to enable efficient targeting of specialist nurse support.
Disease severity and lack of IT skills were not necessarily barriers to using the Internet-based program, but lack of motivation to use computers was. It may be helpful for patients to be introduced to an internet-based package before an intervention, in order to assess motivation and manage expectations. Costs of the intervention were nurse-coach training in use of TPP and self-management support skills and nurse--patient contact time. The mean time spent by the nurse per patient was 15.25 minutes/week for group 1 and 21 minutes/week for group 2. However, self-management interventions have the potential to be cost-saving relative to usual care[@b50-copd-12-2955] and reduce respiratory-related hospitalisations.[@b4-copd-12-2955]
Effing et al[@b51-copd-12-2955] provided a detailed definition of the requirements for a COPD self-management intervention. Key aspects of the definition include the need for structured personalized support, with an emphasis on motivating and engaging patients. The process requires interactions between patients and health care professionals that focus on identifying needs and motivations, eliciting goals, formulating strategies, and building confidence and competence. The definition emphasizes the importance of the relationship with the health care professional. The importance of this relationship was also apparent in our study, where the input from the nurse coach was key. There were no patients who reported that the online platform was helpful and the nurse coach was not; however, there were some patients who made little use of TPP or used it primarily to contact the nurse coach. The nurse coach was important, and for some the content and functionality of TPP provided additional benefits. Further research is needed on how and under what circumstances Internet-based support can best complement personal relationships with health care providers and how nonmedical factors impact the potential to benefit from such interventions.
Limitations
-----------
The total sample size was small, rendering numbers in each of the two groups very small. However, our aim was to explore the potential of combined ehealth support with nurse coaching in COPD, rather than to determine efficacy or establish generalizability. Results are thus presented as a contribution to generic learning in this field. For group 2, a pragmatic decision was made to provide additional structured pulmonary rehabilitation materials, and this could have impacted on outcomes. It is notable, however, that the lack of this extra resource for group 1 did not appear to be detrimental and that they achieved relatively greater benefits despite this. When measuring outcomes over time, regression to the mean can occur. Patients may join because they feel vulnerable and then naturally improve over time; however, improvement was different for the two groups. Data on patients' treatment regimens and comorbidities were not available for analysis, and these factors may have an impact on patient activation and responses to this type of intervention. A wider study incorporating factors beyond COPD severity would be needed to establish this. Resources only enabled a follow-up period of 3 months, hence it was only possible to explore immediate impact. Sustainability of outcomes from self-management support interventions is an important issue, and a longer-term follow-up would be needed to address this.
Conclusion
==========
This case series demonstrates the feasibility of combining nurse-coach support aligned to an Internet-based health resource, TPP, in COPD. Patients with more advanced disease and lower baseline activation had the largest change in activation. The case series provides knowledge about the challenges of such an approach and the importance of the nurse-coach role.
Supplementary materials
=======================
Further detailed description of the intervention and interview-topic guide are available from the corresponding author.
The Preventive Plan
-------------------
The Preventive Plan (TPP) was a web-based program encompassing primary prevention (health promotion), secondary prevention (biometric and lab screening and early detection/diagnosis), and tertiary prevention (chronic condition management). TPP was personalized through the uploading of personal details, including medical history, medication, family history, weight, waist measurement, current lifestyle, dietary intake, and activity levels, via a 77 item questionnaire. The application incorporated four elements, described in the following paragraphs.
Following the uploading of personal details, a personalized health-risk assessment was generated, which indicated the individual's highest risks of developing a range of conditions if no action were taken on current behavior. Risks were graded high, severe, and moderate and calculated using a variety of risk models, such as Framingham heart risk.[@b52-copd-12-2955] Based on this a personalized prevention plan was generated to address the health risks with a range of pre-programmed recommended action programs. Focusing on health-related behaviors, such as healthy eating, alcohol consumption, exercise, and smoking cessation, these guided the user through behavior change and offered step-by-step recommendations to achieve health benefits, while encouraging users to upload individual goal achievements. The duration of these action programs was around 8 weeks. This process was personalized to the extent that the action plan addressed an identified risk for the individual, but the plans themselves were not patient-led or -formed.
TPP also incorporated information resources, including daily health-news bulletins, which could be preselected relevant to identified health risks and action plans, and a health library/tutorial facility. In addition, local uploads were installed including links to the British Lung Foundation, carer support, local resources, and a COPD self-management personal health plan that included condition-related information, symptom-monitoring, and self-management action-plan templates. Participants also received a handheld copy of this personal health plan. Access was available to health information aimed at enhancing health literacy, an important aspect of self-management. TPP enabled email communication with the nurse coach, who could also monitor when users accessed the program.
Nurse-coach support
-------------------
Use of the package was integrated with support from a trained respiratory nurse via home visits, telephone, and email contact. The nurse coach assisted participants to use the website and hardware (if provided), and supported patient self-management through individualized patient-led goal-setting. The role comprised user support for TPP, encouragement to use TPP, and "signposting" to other online resources for self-management through supportive email messages.
Patient-centered coaching to enhance confidence to self-manage was also a component. This included agenda-setting, patient-led goal-setting, support for and action planning to complement TPP-generated action plans, problem-solving, goal review, verbal encouragement, and focus on past successes. Patient-led goals could range from specific health-related behaviors, such as healthy eating, to broader lifestyle issues, such as meeting friends. Specifically, the communication style of this role incorporated coproduction, and was distinct from that of the nurse coach's usual role as a respiratory nurse specialist, in that there was a strong emphasis on support relative to expert medical advice. For example, if a patient were to say that they thought they were having an exacerbation and ask if they should start their rescue pack, the role of the nurse coach was to educate through asking: "Why do you think you are having an exacerbation? What are your symptoms? What advice have you been given about these symptoms previously? Do you have any written instructions, such as a self-management plan?"
This required specific skills distinct from medical knowledge. Education to complement information was available through TPP, with condition-specific education as appropriate to each participant's needs.
When a participant joined the project, the nurse made an initial home visit. During this visit, she introduced the participant to TPP, provided contact details and written information about the program, collected baseline assessment data, supported the patient in completing the health-risk assessment, generated the personal prevention plan, and discussed self-management priorities, patient-led goal-setting, and action plan, and agreed on follow-up contact.
During each subsequent contact, the nurse would support the participant in problem-solving and working toward their goals, agreeing on methods and timing of goal follow-up on each occasion according to patient preference. Participants were asked whether they would like the next contact to be a visit, phone call, or text. If more than 14 days had elapsed since the previous contact, then the nurse coach would text the patient inquiring what progress they were making. If no reply had been received within 3--4 days, then she would telephone. The level of support provided by the nurse coach was flexible and responsive to the patient's level of engagement with the intervention, eg, more frequent phone calls or emails to encourage contact with patients who were not readily engaging.
######
Interview-topic guide
------------------------------------------------------------
Opening discussion
• What patient hoped to gain from taking part
• Experience date
• Benefits to date
Self-management
• What self-management means to the patient
• What goals the patient has been working toward
• Goal progress and what has helped or hindered progress
TPP and nurse coach
• How TPP has contributed to goal progress
• Frequency and experience of contact with the nurse coach
• Aspects of TPP used
• Use and benefits of health-risk assessment in TPP
• Role of family/carers in supporting use of TPP
• Ease of use of TPP and hardware
Any other benefits from taking part
------------------------------------------------------------
**Abbreviation:** TPP, The Preventive Plan.
The Preventive Plan was supplied by UK Preventive Medicine Ltd (Acre House, 11/15 William Road, London NW1 3ER, UK), and we are grateful to Stefan Wisbauer from UK Preventive Medicine for technical support in its use. We thank Kate Homan who conducted the qualitative interviews, Petrea Fagan who supported the nurse coach in developing the self-management support skills necessary for the study, and the patients who participated in this study. The study was funded by NHS East of England Regional Innovation Funding, and delivery of this work was supported by the Cambridge Biomedical Research Centre.
**Author contributions**
JF and FE conceived, sought funding for, and designed the study. FE and ER conducted the qualitative data analysis. FE, Ella M, and JF conducted the statistical analysis. JY and Emma M were key contributors to the interpretation of the data. All authors have been involved in drafting and revision of the manuscript for important intellectual content, and have approved the final version to be published. All authors agree to be accountable for aspects of the work. JF is the guarantor of the paper.
**Disclosure**
The authors report no conflicts of interest in this work.
{#f1-copd-12-2955}
######
Participant characteristics at baseline
Group 1 (n=11) Group 2 (n=8) Statistic *P*-value
--------------------------------------------------------------------------------- ------------------- ------------------ ------------ -------------------------------------------------------
Age, mean (SD) 66.36 (12.33) 60.63 (9.47) *t*=1.099 0.287
Sex, n (%)
Male 6 (54.55) 3 (37.5) 0.65
Female 5 (45.45) 5 (62.5)
Educational level, median (IQR)[a](#tfn1-copd-12-2955){ref-type="table-fn"} 1 (1--1) 1 (1--1) *U*=43 1
Employment, n (%)
Employed 0 1 (12.5) 0.421
Unemployed/retired 11 (100) 7 (87.5)
Living arrangements, n (%)
Alone 1 (9.09) 3 (37.5) 0.262
With others 10 (90.91) 5 (62.5)
Current smoking, n (%)
Yes 3 (27.27) 5 (62.5) 0.181
No 8 (72.73) 3 (37.5)
Pack-years, median (IQR) 40 (33--80) 40 (16.5--49.1) *U*=36 0.527
**Disease severity**
FEV~1~ (L), mean (SD) 0.72 (0.25) 1.36 (0.67) *t*=−2.561 0.032[\*](#tfn4-copd-12-2955){ref-type="table-fn"}
FEV~1~ % predicted, median (IQR) 25.4 (18.4--36.2) 48.2 (33--64.8) *U*=12 0.006[\*](#tfn4-copd-12-2955){ref-type="table-fn"}
FEV~1~/FVC, median (IQR) 0.31 (0.27--0.43) 0.54 (0.4--0.68) *U*=16.5 0.022[\*](#tfn4-copd-12-2955){ref-type="table-fn"}
GOLD stage, median (IQR) 4 (3--4) 2.5 (2--3) *U*=22.5 0.046[\*](#tfn4-copd-12-2955){ref-type="table-fn"}
GOLD ABCD classification, median D D 0.421
MRC dyspnea scale, median (IQR) 5 (4--5) 4 (4--4) *U*=17.5 0.012[\*](#tfn4-copd-12-2955){ref-type="table-fn"}
Supplemental oxygen, n (%)
Yes 5 (45.45) 0 0.045[\*](#tfn4-copd-12-2955){ref-type="table-fn"}
No 6 (54.55) 8 (100)
Duration of COPD, median (IQR) 10.3 (7.39) 6.43 (3.91) *t*=1.259 0.227
Exacerbations in past year, mean (SD) 3.7 (2.75) 7 (6.78) *t*=-1.409 0.178
Hospital admissions in past year, median (IQR) 4 (2--6) 0 (0--0.75) *U*=0 \<0.0005[\*](#tfn4-copd-12-2955){ref-type="table-fn"}
A&E attendances in past year, median (IQR) 0 (0--1) 0 (0--0.75) *U*=38 0.716
**Other**
Confidence managing COPD, mean (SD)[b](#tfn2-copd-12-2955){ref-type="table-fn"} 6.18 (1.83) 7.75 (1.91) *t*=−1.809 0.088
Previous PR, n (%)
Yes 7 (63.64) 5 (62.5) 1
No 4 (36.36) 3 (37.5)
**Computer/Internet use**
Internet access at home, n (%)
Yes 9 (81.82) 6 (75) 1
No 2 (18.18) 2 (25)
Internet user, n (%)
Yes 10 (90.91) 5 (62.5) 0.262
No 1 (9.09) 3 (37.5)
Email user, n (%)
Yes 5 (45.45) 4 (50) 1
No 6 (54.55) 4 (50)
Frequency of use, median (IQR)[c](#tfn3-copd-12-2955){ref-type="table-fn"} 3 (2--4) 2.5 (0--4) *U*=34 0.425
Confidence using IT, mean (SD)[b](#tfn2-copd-12-2955){ref-type="table-fn"} 6.8 (2.82) 6.0 (3.07) *t*=0.575 0.573
**Notes:**
0 -- none, 1 -- secondary education, 2 -- college/higher education;
0 --10 scale;
0 -- never, 1 -- occasional, 2 -- once a week, 3 -- several times a week, 4 -- daily.
significant at *P*≤0.05.
**Abbreviations:** FEV~1~, forced expiratory volume in 1 second; FVC, forced vital capacity; GOLD, Global Initiative for Chronic Obstructive Lung Disease; MRC, Medical Research Council; PR, pulmonary rehabilitation.
######
Change in outcome measures
Group 1 (n=8) Group 2 (n=6)
---------------------------- ------------------ ------------------- ------------ ---------------------------------------------------- ------------------- ------------------- ------------ ----------------------------------------------------
Patient Activation Measure 51.98 (7.36) 64.11 (8.09) *t*=−2.939 0.022[\*](#tfn7-copd-12-2955){ref-type="table-fn"} 61.22 (7.89) 72.73 (16.11) *t*=−1.931 0.111
CRQ -- dyspnea 1.8 (1.65--2.15) 1.90 (1.4--2.75) *W*=13 0.906 2.8 (1.36) 2.93 (1.51) *t*=−0.791 0.465
CRQ -- emotion 3.51 (1.41) 4.08 (1.53) *t*=−1.725 0.128 4.65 (2.04) 4.15 (1.83) *t*=1.739 0.143
CRQ -- fatigue 2.13 (1.75--3) 2.88 (1.81--4.5) *W*=6 0.133 2.25 (1.75--3.38) 3.13 (1.44--5.31) *W*=5.5 0.344
CRQ -- mastery 3.44 (1.38) 4.38 (1.4) *t*=−2.525 0.04[\*](#tfn7-copd-12-2955){ref-type="table-fn"} 5.17 (0.61) 4.92 (1.47) *t*=0.518 0.627
HADS -- depression 8.75 (3.69) 7.38 (3.42) *t*=1.553 0.164 5.67 (1.86) 6.17 (6.08) *t*=−0.209 0.843
HADS -- anxiety 11.38 (3.81) 8.88 (4.32) *t*=2.758 0.028[\*](#tfn7-copd-12-2955){ref-type="table-fn"} 8.83 (3.76) 8.33 (3.44) *t*=0.436 0.681
LINQ 5.5 (4.25--6) 3.00 (1.25--3.75) *W*=0 0.008[\*](#tfn7-copd-12-2955){ref-type="table-fn"} 5.5 (2.26) 3.33 (1.75) *t*=3.081 0.027[\*](#tfn7-copd-12-2955){ref-type="table-fn"}
**Notes:** Mean (SD) for all *t*-statistics; medians (IQR) for all *W*-statistics.
significant at *P*≤0.05.
**Abbreviations:** CRQ, Chronic Respiratory Questionnaire; HADS, Hospital Anxiety and Depression Scale; LINQ, Lung Information Needs Questionnaire.
######
Aims and benefits
Category Group 1 Group 2
---------------------------------- ------------------------------------------------------------------------------------------------------------------------------------------------------------------- --------------------------------------------------------------------------------
**Aims**
Physical functioning To be physically fitter and more active To do more and have more stamina and a better lifestyle
Manage health and COPD To stay out of hospital and better manage COPD and infections To help myself more by preventing worsening of symptoms and avoiding hospital
Independence To be more independent and depend less on family and friends NA
Increase motivation NA To gain motivation for more daily activities
**Benefits**
Physical functioning More able to walk, exercise and do household activities; more energy; able to be more independent More physically able, better sleep
Managing health and COPD Increased confidence to manage symptoms, more aware of own health behaviors, eg, diet and activity level Better understanding and managing of symptoms, more aware of healthy behaviors
Social and emotional functioning More confident, empowered, improved mood, positive sense of self, more connected to others, not alone, more likely to talk about problems with family and friends Enhanced self-belief
Communicating with HCPs NA More assertive, goal-oriented consultations
**Abbreviations:** HCPs, health care practitioners; NA, not applicable.
######
Facilitators and challenges in achieving self-management goals and using TPP
Category Group 1 Group 2
----------------------------------------------------- -------------------------------------------------------------------------------------------------------------------------------------------------------------- ---------------------------------------------------------------------------------------------------------------------------------------------------------------------
**Facilitators in achieving self-management goals**
Nurse-coach support Motivational, emotional, and goal support; links to relevant information, understanding and acceptance of COPD, talking through questions; help in using TPP Motivational, goal support; understanding COPD; talking through questions; support for communicating with health care professionals
TPP Action plans, information about COPD and health, inspiration, reminders, security, means of contacting nurse coach Action plans, information about COPD and health, reminders, advice on self-management skills, motivational, advice for communicating with health care professionals
Social support Support from family and friends for goal achievement, health care, company for activities, and help with IT Support from family and friends for goal achievement, company for activities, sharing experiences, and problem-solving
Psychological factors Feeling confident and motivated Feeling confident and motivated
Self-management skills The ability to use self-management skills and engage in health behaviors The ability to use self-management skills and engage in health behaviors
**Challenges in achieving self-management goals**
Physical well-being Being ill, in pain, low energy, poor eyesight Low energy, pain, being ill
Psychological factors Anxiety, low confidence, demotivation, difficulty accepting limitations of COPD and so doing too much Low mood, safety concerns, demotivation
Other priorities Health-related commitments and appointments, social activities Domestic jobs taking time and energy
Weather Bad weather limited outside activities Bad weather
Social support Lack of social support NA
Access to resources Lack of exercise facilities and equipment, finance, money for fares NA
Behavioral factors NA Trying to do too much, unhelpful triggers in the environment, eg, chocolates in the house
Housing NA Poor housing conditions
**Challenges to using TPP**
IT orientation Poor IT skills, lack of interest in computers Poor IT skills, lack of interest in computers
TPP design TPP was difficult to use TPP was difficult to use
Individual barriers Illness Poor memory for passwords, pain inhibiting computer use
**Abbreviations:** NA, not applicable; TPP, The Preventive Plan.
######
The Preventive Plan usage and nurse-contact time
Group 1 (n=8) Group 2 (n=6) Statistic *P*-value
------------------------------------------------------- ---------------------- --------------------- ------------ --------------------------------------------------------
Days TPP accessed/weeks in program, mean (SD) 1.21 (0.77) 1.32 (0.8) *t*=−0.244 0.812
Nurse phone time/weeks in program, mean (SD) 3.53 (0.95) 2.67 (1.58) *t*=1.28 0.225
Home visits/weeks in program, median (IQR) 0.19 (0.15--0.22) 0.29 (0.19--0.38) *U*=9 0.052
Face-to-face contact time/weeks in program, mean (SD) 12.51 (3.91) 18.54 (6.37) *t*=−2.044 0.076
Emails sent by nurse/weeks in program, mean (SD) 0.71 (0.21) 1.27 (0.23) *t*=−4.732 \<0.0005[\*](#tfn11-copd-12-2955){ref-type="table-fn"}
Emails received by nurse/weeks in program, mean (SD) 0.32 (0.22) 0.39 (0.27) *t*=−0.462 0.653
Page loads (%): messages, mean (SD) 44.33 (4.65) 47.99 (10.31) *t*=−0.81 0.447
Page loads (%): action programs, median (IQR) 12.19 (11.63--13.48) 10.48 (9.75--18.58) *U*=15 0.282
Page loads (%): health-risk assessment, mean (SD) 3.46 (1.56) 3.9 (1.64) *t*=−0.508 0.621
Page loads (%): personal health record, mean (SD) 2.57 (1.24) 2.9 (2.48) *t*=−0.329 0.748
Page loads (%): health news, mean (SD) 1.52 (1.3) 1.25 (0.92) *t*=0.43 0.674
Page loads (%): challenge programs, mean (SD) 0.69 (0.5) 0.51 (0.71) *t*=0.543 0.597
**Note:**
significant at *P*≤0.05.
**Abbreviation:** TPP, The Preventive Plan.
| {
"pile_set_name": "PubMed Central"
} |
Introduction {#s1}
============
Understanding racial disparities in breast cancer is complicated by the fact disparities vary among the different breast cancer subtypes. Characterizing the social inequalities that give rise to each of these disparities is necessary to inform preventive interventions and work toward health equity ([@B1]--[@B4]). This is especially true of triple negative breast cancer (TNBC) due to the aggressive nature of this type of cancer and relative lack of treatment modalities available to patients.
TNBC is an aggressive, heterogeneous subtype of breast cancer, immunohistochemically negative for estrogen receptor α (ER^−^) and progesterone receptor (PR^−^), as well as lacking amplification of the human epidermal growth factor receptor 2 locus (HER2^−^). While TNBC accounts for an estimated 12--15% of all breast cancers, the unique epidemiology and biology of TNBC draw substantial research attention. Numerous studies have shown that relative to European Americans (EA/White), TNBC is significantly more prevalent among African American (AA/Black) women ([@B5]--[@B8]), and is associated with a poorer prognosis than other subtypes of breast cancer ([@B9], [@B10]). TNBC is characterized by diagnosis at a younger age, higher tumor grade, and larger tumor size ([@B11], [@B12]).
Distinct racial disparities exist in TNBC outcomes (i.e., incidence, stage at diagnosis, and survival). For this reason, there is a vital need to understand the role of biological and non-biological factors in disparities ([@B4], [@B13]--[@B15]). Some studies have begun to describe the nature of these racial disparities in terms of the risk due to non-biological social determinants ([@B4], [@B7], [@B16], [@B17]). However, the findings are not yet at the point they can establish the translational impact necessary for developing policy interventions.
Several issues concerning TNBC disparities research need to be addressed to draw translationally impactful conclusions. First, there needs to be an appreciation of the fact that the various outcomes (e.g., incidence, stage at diagnosis, survival) involve different stages in cancer carcinogenesis (i.e., initiation, promotion and progression). At each of these stages unique factors are likely to contribute to racial disparities, which may or may not include social determinants ([@B18]). Consequently, results indicating a role for social determinants in explaining a TNBC disparity may vary by the stage of carcinogenesis. Second, characterizing the role of modifiable factors in the physical and social environment requires the integration of secondary data to characterize the local environment (i.e., neighborhood) in which the patient lives. This type of integration requires an analytic approach involving large population-based datasets using multilevel models at the appropriate scale to characterize social determinants ([@B3], [@B19]).
In both national and state level registry based studies, AA women have a 2-fold increase in risk of being diagnosed with TNBC compared with EA ([@B13], [@B15]). Several studies have incorporated measures of socioeconomic risk in individual level models and found these factors appeared unrelated to incidence and therefore did not explain any of the racial disparity ([@B7], [@B16], [@B17]). Racial disparities in TNBC stage at diagnosis and survival also exist, with AA women more often diagnosed at later stages and experiencing lower survival ([@B20]). In contrast to TNBC incidence, racial disparities in TNBC stage at diagnosis and survival do appear to be partially, if not completely, socially determined. Tao et al. ([@B4]) found measures of neighborhood SES to fully explain the disparity in stage-specific survival for AA women diagnosed with TNBC. These findings suggest the etiologic pathways for TNBC disparities vary across the continuum of the disease.
Besides the District of Columbia, Louisiana ranks first in breast cancer mortality ([@B21]). Unfortunately, the excess breast cancer mortality can be almost entirely attributed to the high mortality rate for AA women. Among AA women, the mortality rate is 15% above the national average, while among EA women, the mortality rate is comparable to the national average ([@B21]). Racial disparities in breast cancer mortality are complex and originate from a multitude of factors including greater incidence of more aggressive subtypes, such as TNBC among AAs, later stages at diagnosis, higher levels of comorbidities, as well as inequality in treatment, access to care, and adherence ([@B22]--[@B25]). In this study, we conducted an analysis of TNBC incidence, stage at diagnosis, and survival in Louisiana to better understand the origins of racial disparities in this disease by assessing the role of a robust measure of risk associated with the neighborhood environment in explaining racial disparities across TNBC outcomes in the same population. In addition, we introduce a multilevel approach to the analysis of tumor registry data that permits inferences regarding the potential role of neighborhood level social determinants with respect to individual level risk ([@B26], [@B27]).
Methods {#s2}
=======
Case Definition
---------------
This study used 2010--2012 data from the Louisiana Tumor Registry (LTR), a participant of the National Cancer Institute\'s (NCI) Surveillance, Epidemiology and End Results (SEER) Program and the Centers of Disease Control and Prevention\'s National Program of Cancer Registries (CDC-NPCR). All study variables regarding demographics (age, sex, race, ethnicity, address) and tumor characteristics (site, behavior, histology, stage, grade) were collected from hospital and medical records in accordance with registry guidelines. Primary invasive breast cancer cases were identified by International Classification of Diseases for Oncology, Third Edition (ICDO-3) site codes C500-C509. ICD-O-3 histology codes 9050-9055, 9140, and 9590-9989 were excluded. *In-situ* cases were also excluded. Breast cancer cases were classified into subtypes by estrogen receptor, progesterone receptor and human epidermal growth factor 2 (HER2) status, which have been routinely collected by SEER registries since 2010 ([@B28]). Hormone receptor (HR) status was considered negative if the tumor lacked both estrogen and progesterone reactivity. The four molecular breast cancer subtypes were defined as HR+/HER--, HR+/HER+, HR--/HER2-- (Triple Negative), and HR--/HER2+. The Louisiana State University Health Sciences Center Institutional Review Board approved this research.
Geocoding Cases
---------------
TNBC cases were geocoded to 2010 census tracts using the Automated Geospatial Geocoding Interface Environment system, which was developed as a uniform geocoding platform for open use by cancer registries[^1^](#fn0001){ref-type="fn"}. To ensure a high certainty of patient location, we restricted matched cases to those geocoded using street address at time of diagnosis. We used 2010 US Census population for demographic groups within census tracts as the population at-risk in determining incidence of disease. Census tract socioeconomic measures were obtained from the US Census American Community Survey (ACS).
Measuring Disadvantage
----------------------
Neighborhood concentrated disadvantage index (CDI) scores for census tracts were calculated based on the PhenX Toolkit Protocol[^2^](#fn0002){ref-type="fn"}. The toolkit is an established resource of consensus measures of phenotypes and exposures for biomedical research. CDI is a sample-based composite score derived from a principal components analysis of 6 measures at the census tract level (given as percentages): (1) individuals below the poverty line; (2) households receiving public assistance income; (3) female-headed families; (4) individuals that are unemployed; (5) individuals below the age of 18; and (6) individuals that are Black. The construct operationalizes urban theory regarding the overconcentration of Blacks, children and female-headed families in economically disadvantaged neighborhoods ([@B29]). This measure of disadvantage has been associated with poor outcomes in breast and ovarian cancer ([@B30], [@B31]). We derived CDI using 2008--2012 5-year estimates of American Community Survey (ACS), to best align with the study period[^3^](#fn0003){ref-type="fn"}. All measures follow US Census definitions provided by the ACS. Factor scores for study census tracts follow a standard normal distribution with a mean of zero and standard deviation of 1. Factor loadings are provided as [Supplementary Table 1](#SM1){ref-type="supplementary-material"}.
Census Tract Exclusions
-----------------------
From the 2010 US Census, Louisiana has 1,148 census tracts. Standard US census tracts typically contain between 2,500 and 8,000 residents and are designed to be relatively homogenous in terms of socioeconomic characteristics[^4^](#fn0004){ref-type="fn"}. Census tracts with zero population per the 2010 US census were not included (*n* = 19). We excluded eight low-population (population of 500 or less) census tracts and the census tract constituting the Orleans Parish Prison. After these exclusions, the study included 1,120 census tracts. Cancer outcomes were assessed for women aged 30 or older, of Black or White race, who resided in an eligible Louisiana census tract during the study period. Cases in women younger than 30 years old and cases in races other than black or white were not included in these analyses due to small age and race group numbers. Inclusion of Hispanic women was based on identified race regardless of ethnicity.
Outcome Determinations
----------------------
Incidence of disease was a Binomial response constructed from the number of incident TNBC cases out of the population-years at risk for age and race specific groups within census tracts. Stage at diagnosis was a Multinomial response based on the coded SEER summary 2000 system, with 3 stages, local, regional and distant[^5^](#fn0005){ref-type="fn"}. Patient survival was calculated as the number of months elapsed from date of diagnosis to date of last contact or death, as recorded by the LTR. The latest date of contact in the patient dataset was August 2016. The event of interest was breast-cancer specific survival and was determined using patient vital status and cause of death, based on SEER cause-specific death coding system[^6^](#fn0006){ref-type="fn"}.
Statistical Analysis
--------------------
To examine the role of neighborhood disadvantage in TNBC outcomes we used multivariable multilevel models of individuals (level 1) nested within census tracts (level 2). A random intercept at the census tract level was used to account for correlation among individuals within census tracts, or neighborhood clustering, for all models. Cancer incidence was modeled with binomial regression. Stage at diagnosis was analyzed using proportional odds logistic regression. We compared stage-specific survival for Black and White races using Kaplan-Meier plots of time to breast cancer death. Stage-specific survival was analyzed using multilevel Cox proportional hazards regression (frailty model) and included fixed effects for stage at diagnosis. For the study, we assumed that racial disparities in TNBC are independent of age ([@B4], [@B7]). Initial multilevel models were executed as mixed effects models with fixed effects for age, as 10-year age groups, and race. This model was used to estimate racial disparities in TNBC incidence and outcomes after adjusting for the effects of age. To test the hypothesis that neighborhood concentrated disadvantage is associated with adverse outcomes among both races, a subsequent model included a fixed effect for CDI, in addition to main effects for age and race. All statistical analyses were performed in SAS version 9.4 software (SAS Institute, Cary, NC). Generalized mixed linear models were executed using the GLIMMIX Procedure, with maximum likelihood estimation via adaptive quadrature method. Cox proportional hazards regression was executed in the PHREG Procedure. We considered two-sided *p* \< 0.05 as statistically significant.
Results {#s3}
=======
Breast cancer cases diagnosed in Louisiana from 2010 to 2012 with a histologically confirmed subtype are summarized in [Table 1](#T1){ref-type="table"}. TNBC cases that met the study criteria accounted for 14.79% (*n* = 1,216) of cases. The demographic distribution of cases varied by subtype. An earlier age of onset for TNBC was observed in the patient sample. More than 25% of TNBC cases were diagnosed in women younger than 50 years old. Racial distribution of breast cancer cases also differed by subtypes. AA women constituted 47% of TNBC cases, the highest proportion for any subtype. There were stark differences in tumor grade and stage at diagnosis by subtype, with HR+/HER2-- tumors having a greater share of low grade and localized tumors compared to other subtypes.
######
Demographics and tumor stage characteristics of breast cancer patients by molecular subtype, Louisiana 2010--2012.
**HR+/HER2−** **HR+/HER2+** **HR−/HER2+** **TNBC**
--------------------------------------------------------------------- --------------- --------------- --------------- -------------
N 5,650 944 410 1,216
**AGE, YEARS (%)**
30--39 2.81 5.51 8.05 7.32
40--49 13.27 17.16 17.80 18.59
50--59 23.86 24.79 27.07 29.28
60--69 27.65 25.53 26.34 23.52
70+ 32.41 27.01 20.73 21.30
**RACE (%)**
White 73.82 69.70 61.46 51.81
Black 25.33 29.77 37.56 47.45
Other 0.85 0.53 0.98 0.74
**SEER SUMMARY STAGE 2000 (%)**[^**a**^](#TN1){ref-type="table-fn"}
Local 64.12 56.04 51.95 57.81
Regional 30.48 36.12 37.32 33.31
Distant 5.12 7.73 10.49 8.39
Unknown 0.28 0.11 0.24 0.49
**CDI**
Mean (Std Dev) −0.21 (0.88) −0.16 (0.9) −0.04 (0.91) 0.00 (0.97)
*(N = 8,220). HR, Hormone Receptor; HER2, Human Epidermal Growth Factor 2; TNBC, Triple Negative Breast Cancer; SEER, Surveillance, Epidemiology and End Results Program; CDI, Concentrated Disadvantage Index*.
<https://www.registrypartners.com/seer-summary-staging-manual-2000/>
Incidence
---------
Our analysis of TNBC incidence included 1,308,564 female residents aged 30 years and older. AA women comprised 30.9% of the study population. AA women had significantly higher age-specific incidence of TNBC, given as cases per 100,000, compared to EA women ([Figure 1](#F1){ref-type="fig"}). There was also a notable difference in the exposure to CDI for Louisiana women by race. [Figure 2](#F2){ref-type="fig"} illustrates that AA women were disproportionately represented in more disadvantaged areas, with an average CDI of 0.56 (standard deviation *SD* = 6.06) compared to −0.47 (*SD* = 5.64) among EA women. [Table 2](#T2){ref-type="table"} provides adjusted risk ratios (RR) and 95% confidence intervals (CI) from multilevel binomial regression for triple negative breast cancer incidence. In our initial model, controlling for age, AA women in Louisiana had 2.21 times the risk of TNBC compared to EA \[*RR* = 2.21, 95% CI (1.96, 2.48)\]. In model 2, neighborhood concentrated disadvantage was not significantly associated with incidence of the disease \[*RR* = 0.96, 95% CI (0.89, 1.03)\] and the racial disparity appeared to be independent of CDI \[*RR* = 2.30, 95% CI 2.01, 2.64)\].
{#F1}
{#F2}
######
Estimated risk ratios (RR) and 95% confidence intervals (CI) from multilevel binomial regression of triple negative breast cancer, Louisiana 2010--2012.
**Model 1** **Model 2**
-------------------- ------------------- -------------------
**AGE, YEARS**
30--39 (reference) 1.00 1.00
40--49 2.33 (1.82, 2.99) 2.33 (1.82, 2.99)
50--59 3.68 (2.91, 4.66) 3.69 (2.92, 4.66)
60--69 4.44 (3.49, 5.65) 4.45 (3.50, 5.66)
70+ 4.06 (3.18, 5.18) 4.07 (3.19, 5.20)
**RACE**
White (reference) 1.00 1.00
Black 2.21 (1.96, 2.48) 2.30 (2.01, 2.64)
**CDI**
1 SD Increase -- 0.96 (0.89, 1.03)
*RR, risk ratio; CI, confidence interval; CDI, concentrated disadvantage index; SD, standard deviation*.
*Model 1 is adjusted for age and race. Model 2 is adjusted for age, race and CDI*.
*All models include random intercept for US census tracts*.
Stage at Diagnosis
------------------
Our analysis of stage at diagnosis (local, regional, distant) included all TNBC patients in the sample with known stage (*n* = 1,201). Results from multilevel proportional odds logistic regression models are presented as [Table 3](#T3){ref-type="table"}. In our initial model, controlling for age, AA women had 42% greater odds of being diagnosed with TNBC at a later stage \[*OR* = 1.42, 95% CI (1.11, 1.81)\]. Results from model 2 showed a single standard deviation increase in CDI was associated with a 20% increase in the odds of diagnosis at a more advanced stage \[*OR* = 1.20, 95% CI (1.03, 1.39)\]. After adjusting for CDI, the racial disparity was no longer statistically significant \[*OR* = 1.17, 95% CI (0.88, 1.56)\].
######
Estimated odds ratios (OR) and 95% confidence intervals (CI) from multilevel proportional odds logistic regression models of SEER stage at diagnosis in triple negative breast cancer, Louisiana 2010--2012.
**Model 1** **Model 2**
-------------------- ------------------- -------------------
**AGE, YEARS**
30--39 (reference) 1.00 1.00
40--49 0.62 (0.38, 1.02) 0.62 (0.38, 1.01)
50--59 0.56 (0.35, 0.90) 0.56 (0.35, 0.89)
60--69 0.74 (0.46, 1.19) 0.72 (0.45, 1.15)
70+ 0.66 (0.40, 1.07) 0.64 (0.39, 1.04)
**RACE**
White (reference) 1.00 1.00
Black 1.42 (1.11, 1.81) 1.17 (0.88, 1.56)
**CDI**
1 SD Increase 1.20 (1.03, 1.39)
*OR, odds ratio; CI, confidence interval; SEER, Surveillance, Epidemiology and End Results Program; CDI, concentrated disadvantage index; SD, standard deviation*.
*Model 1 is adjusted for age and race. Model 2 is adjusted for age, race and CDI*.
*All models include random intercept for US census tracts*.
Survival
--------
Our analysis of survival included patients for which TNBC was their first cancer diagnosis and the tumor was of known stage (*n* = 989). In this sample, 191 patients experienced breast cancer-related death during follow up. Stage-specific survival curves ([Figure 3](#F3){ref-type="fig"}) suggested similar survival among Black and White patients. Multilevel Cox proportional hazard model results are provided as [Table 4](#T4){ref-type="table"}. In our initial model, controlling for stage and age at diagnosis, we found no statistically significant difference in stage-specific survival by race \[*HR* = 1.08, 95% CI (0.80, 1.46)\]. However, when introduced into the model, CDI was significantly associated with poorer survival \[*HR* = 1.19, 95% CI (1.01, 1.39)\].
{#F3}
######
Estimated hazard ratios (HR) and 95% confidence intervals (CI) from Cox proportional hazards model of cancer-related death in triple negative breast cancer, Louisiana 2010--2012.
**Model 1** **Model 2**
----------------------------------------------------------------- --------------------- ----------------------
**SEER SUMMARY STAGE 2000**[^**a**^](#TN2){ref-type="table-fn"}
Local (reference) 1.00 1.00
Regional 4.03 (2.76, 5.91) 4.01 (2.75, 5.86)
Distant 26.30 (17.4, 39.76) 24.70 (16.47, 37.04)
**AGE, YEARS**
30--39 (reference) 1.00 1.00
40--49 0.59 (0.33, 1.07) 0.58 (0.32, 1.04)
50--59 0.78 (0.46, 1.33) 0.77 (0.46, 1.30)
60--69 1.01 (0.59, 1.72) 0.99 (0.59, 1.68)
70+ 1.25 (0.72, 2.16) 1.16 (0.67, 1.99)
**RACE**
White (reference) 1.00 1.00
Black 1.08 (0.80, 1.46) 0.88 (0.62, 1.25)
**CDI**
1 SD Increase -- 1.19 (1.01, 1.39)
*The event of interest is breast cancer-related death among triple negative breast cancer patients in Louisiana, diagnosed 2010--2012. HR, hazard ratio; CI, confidence interval; SEER, Surveillance, Epidemiology and End Results Program; CDI, concentrated disadvantage index; SD, standard deviation*.
*Model 1 is adjusted for SEER stage, age and race. Model 2 is adjusted for SEER stage, age, race and CDI*.
*Models include random intercept (frailty term) for US census tracts*.
<https://www.registrypartners.com/seer-summary-staging-manual-2000/>
Discussion {#s4}
==========
In the US, the incidence of TNBC in Black women is approximately 2-fold that of White women across all ages ([@B6]) and it has been shown to have the worst stage-specific survival of all breast cancer subtypes ([@B32]). The poor prognosis among TNBC patients is thought to be one of the major contributors to racial disparities in breast cancer mortality. While some investigators attribute it to differences in income and social status, which affect access to and compliance with treatment at disproportionate rates among minorities, others credit it to racial/ethnic differences in tumor biology and responsiveness to treatment ([@B33]--[@B38]). Our study supports the views expressed by other groups that both tumor biology and socioeconomic factors are major driving forces behind racial disparities, but their impact varies along the continuum of the disease. We found that CDI was not associated with disparities in incidence of TNBC but was associated with diagnosis at more advanced stages and poorer survival.
The role of social determinants in TNBC is of interest because low socioeconomic status is associated with many of the shared characteristics of breast tumors that occur in AA women, including high grade, high clinical stage at diagnosis, and ER-negative status ([@B39], [@B40]). In the US, socioeconomic status is intrinsically linked with race and lifestyle behaviors, such as physical activity, obesity, diet, reproductive experiences such as having more children, and screening behaviors, which vary in prevalence across different populations of women ([@B41]). The poverty rate in Louisiana is one of the highest in the nation, with socioeconomic conditions worse among AA. One in three Black Louisianans lives below the federal poverty level and has limited access to the health system ([@B42]). This study assessed the effects of neighborhood CDI, which is a robust measure of neighborhood environment. Our results support the notion that adverse living environments correlates with unfavorable stage and survival outcomes. It is therefore likely that socioeconomic disadvantages that are more prevalent in the AA community coupled with the increased risk for TNBC found for AA contribute to the high breast cancer mortality gap in Louisiana.
In view of the higher incidence of TNBC in women of African than of European ancestry, African ancestry might be associated with inherited genetic variants that predispose carriers to TNBC. One potential contributor to the higher incidence of early-onset aggressive breast cancer among African American patients may be a previously undefined higher burden of inherited breast cancer in this population. International studies have revealed that breast cancer frequencies of TNBC compared to other populations in the world ([@B43]). Triple negative breast cancer is known to be a marker of hereditary breast cancer susceptibility syndromes, such as BRCA1 mutations. High frequencies of mutations in *BRCA1* and *BRCA2* have already been observed in breast cancer patients of African ancestry from Nigeria ([@B44]) and from the Bahamas ([@B45]).
Our results indicate that disparities in neighborhood environment do contribute to racial disparities in TNBC outcomes between AA and EA in stage and survival, but not incidence. It is important to note that we lacked the case numbers needed to address disparities among other racial/ethnic groups. Additional limitations include a relatively limited study time period and cross-sectional design. SEER registries began collecting HER2 status in 2010; therefore only 3 years of data were available at the time of the study. Despite these limitations, we were able to identify indications of significant neighborhood level effects of concentrated disadvantage in these analyses. Due to the cross-sectional design, the duration of exposure or the risk associated with neighborhood environment over time was not established. An additional limitation is that we have assessed the effects of neighborhood living environment based on a census-defined spatial unit (tract), which is designed to be relatively homogeneous in terms of social characteristics but does lack a subjective definition of "neighborhood." This study does not attempt to characterize other potentially influential social determinants of health, such as individual income or education. Finally, the study lacked information on individual level risk factors such as family history, reproductive history, comorbidities and health behaviors, which are needed to fully specify individual risk of this disease. The models did not adjust for cancer treatment, which also plays an important role in survival.
We have presented results from a population-based study in a single state regarding the impact of neighborhood concentrated disadvantage on disparities in TNBC outcomes. Our results support the need for a multipronged approach to address racial disparities that originate from the environment in which people live. The incidence of TNBC appears to be largely biologically determined, thus disparities will likely persist in terms of individual risk and incidence. However, disparities in TNBC outcomes are not due to biology alone, and these results highlight the need for targeted interventions and effective therapies among high-risk populations. Future efforts in breast cancer health equity should have a particular focus on individuals identified through known genetic risk and AA women in general.
Author Contributions {#s5}
====================
FH and DD contributed to the study concept, design, methods, and contributed to manuscript development. OP and AG made considerable contributions of intellectual content and contributed to manuscript development. TF, NS, CL, and QY contributed to the study concept and revised the article for intellectual content. X-CW contributed to documentation of data and methods and revised the article for intellectual content. LM and RS made substantial contributions to the study concept, design and intellectual content. All authors have approved the content and agree to be accountable for the work with regard to accuracy and integrity.
Conflict of Interest Statement
------------------------------
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
^1^NAACCR Geocoder <https://www.naaccr.org/gis-resources/>
^2^PhenX Toolkit -- Neighborhood Concentrated Disadvantage Index <https://www.phenxtoolkit.org/index.php?pageLink=browse.protocoldetails&id=211302>
^3^American FactFinder <https://factfinder.census.gov/faces/nav/jsf/pages/index.xhtml>
^4^US Census Geographic Areas Reference Manual <https://www.census.gov/geo/reference/garm.html>
^5^SEER Summary Staging Manual 2000 <https://seer.cancer.gov/tools/ssm/ssm2000/>
^6^SEER Cause of Death <https://seer.cancer.gov/codrecode>
**Funding.** This work was supported by the National Institutes of Health (NIH 5R24MD008121-03, U54MD008176, U54GM104940, and 1P20CA233374-01).
Supplementary Material {#s6}
======================
The Supplementary Material for this article can be found online at: <https://www.frontiersin.org/articles/10.3389/fpubh.2019.00018/full#supplementary-material>
######
Click here for additional data file.
[^1]: Edited by: André F. S. Amaral, Imperial College London, United Kingdom
[^2]: Reviewed by: Salman M. Tajuddin, National Institute on Aging (NIA), United States; Huabin Luo, East Carolina University, United States
[^3]: This article was submitted to Epidemiology, a section of the journal Frontiers in Public Health
[^4]: †Present Address: Neal Simonsen, Consultant, New Orleans, LA, United States
| {
"pile_set_name": "PubMed Central"
} |
Introduction
============
Pancreatic cancer (PC) is one of the most aggressive types of human malignancy, exhibiting an overall 5-year survival rate of \<2%, and is the fourth most common cause of cancer-associated mortality in the western world ([@b1-mmr-12-04-4843]). PC is characterized by rapid disease development and the absence of specific symptoms, thus limiting early diagnosis and the success of curative treatment ([@b2-mmr-12-04-4843],[@b3-mmr-12-04-4843]). Surgical resection is currently the only curative treatment for PC; however, due to late diagnosis, the majority of patients are diagnosed with advanced stage PC and only a minority (10--20%) respond well to surgery ([@b4-mmr-12-04-4843]). Owing to the high recurrence rate, patients with PC who have undergone surgery require adjuvant chemotherapy with or without radiotherapy, resulting in 5-year survival rates of between 15 and 25% ([@b5-mmr-12-04-4843]--[@b7-mmr-12-04-4843]). Since the majority of cases of PC are inoperable, the majority of patients rely on palliative treatment using conventional chemotherapy. Gemcitabine and 5-fluorouracil (5-FU) are the standard chemotherapeutic drugs used to treat PC, which offer mild improvement of tumor-associated symptoms and minimal improvements in survival rates. Despite providing improvements in quality of life, current standard treatment with gemcitabine or 5-FU results in a median survival rate of just a few months ([@b8-mmr-12-04-4843],[@b9-mmr-12-04-4843]). The limitations of conventional chemotherapy are due to the profound resistance of PC cells towards anticancer drugs, which results from efficient protection against chemotherapeutic drugs due to an altered balance of pro- and anti-apoptotic proteins, which results in significantly reduced susceptibly to apoptosis ([@b10-mmr-12-04-4843],[@b11-mmr-12-04-4843]). Since the majority of established anticancer therapeutic strategies depend on the elimination of tumor cells by apoptosis, the capability of tumor cells to escape apoptosis is a major hurdle in treatment. As with other cancer cells, PC cells have developed resistance mechanisms, which enable them to resist chemotherapy ([@b12-mmr-12-04-4843]). Among these mechanisms, protection from apoptosis appears to be the most relevant. With such poor response rates to current chemotherapeutics, there is an immediate requirement to identify novel and effective therapeutic strategies to treat PC ([@b13-mmr-12-04-4843]--[@b15-mmr-12-04-4843]). The present study aimed to determine the cytotoxic potential of the polyphenol-rich extract of *Salvia chinensis*, and to investigate its role in cell cycle arrest, mitochondrial membrane potential loss and apoptosis in pancreatic cancer cells.
*Salvia chinensis* Benth, also referred to as Shijianchuan (Chinese Sage)\] is a plant belonging to the Labiatae plant family. *S. chinensis* is an annual plant that is native to several provinces in China, including Hubei, Sichuan, Guangxi, Guangdong and Hunan, and grows in forests and in clusters of grass on hillsides or plains at 100 and 500 m elevation. *S. chinensis* grows on stems, which are erect or prostrate, up to a height of 20--60 cm ([@b16-mmr-12-04-4843]). *S. chinensis* was primarily recorded in the Compendium of Materia Medica (Ming Dynasty, A.D. 1590), in which it was recorded as a treatment for ostealgia and swollen carbuncles ([@b17-mmr-12-04-4843]). In addition, ethnopharmacological investigation revealed that this herbal medicine has been used to treat breast, liver and stomach cancer, and hepatitis ([@b18-mmr-12-04-4843]). Phytochemical investigation of *S. chinensis* has resulted in the detection of \>50 chemical constituents, in four classes of compounds: Terpenoids (monoterpenoids, sesquiterpenes and triperpenoids), phenolic acids, flavonoids, and dibenzylcyclooctadiene lignans ([@b19-mmr-12-04-4843]). In addition, boswellic acids, blumenol A, pinafaenoic acid, salvianolic acid B, salvianolic acid D, 5,7,4′-trihydroxydihydroflavonol, protocatechuic acid, 3,5,7-trihydroxychromone and kaempferol have been reported to be present in *S. chinensis* ([@b20-mmr-12-04-4843]--[@b27-mmr-12-04-4843]).
Previous pharmacological investigations have demonstrated that water extract of *S. chinensis* markedly inhibits the proliferation of CNE human nasopharynx cancer cells and MGC-803 human gastric cancer cells ([@b28-mmr-12-04-4843]). In addition, polysaccharides isolated from *S. chinensis* exhibit marked antitumor activity ([@b29-mmr-12-04-4843],[@b30-mmr-12-04-4843]), B-lymphocyte stimulation and, at a concentration of 20 mg/l, protection of PC-12 cells against H~2~O~2~-induced injury ([@b31-mmr-12-04-4843],[@b32-mmr-12-04-4843]). Furthermore, *S. chinensis* has been reported to protect against CCl~4~-induced acute liver injury in mice, possibly due to the antioxidant activity of the phenolic acids present ([@b33-mmr-12-04-4843]).
In view of the reported use of *S. chinensis* in traditional medicine, in combination with reports of its use against various types of cancer, the present study aimed to determine the phytochemical composition and anticancer activity of the polyphenol-rich extract of *S. chinensis*. In addition, the mechanism of action of this extract was evaluated by investigating its effects on cell cycle phase distribution, apoptosis and mitochondrial membrane potential using flow cytometry and fluorescence microscopy.
Materials and methods
=====================
Plant material and extraction procedure
---------------------------------------
*S. chinensis* was collected between June and July 2013 from a local site in Jianguo, China, and the plant material was confirmed by Professor JW Chen (College of Pharmaceutical Science, Nanjing University of Chinese Medicine, Nanjing, China). The aerial parts of *S. chinensis* were washed thoroughly with tap water, air dried and then sectioned into small pieces. Methanol (95%) was used for the hot extraction, which was performed after 4 h using a Soxhlet extraction apparatus (BSXT-02; Shanghai Bilon Instrument Co., Ltd. Shanghai, China). In this method, the finely ground crude drug is placed in a porous bag made of strong filter paper, which is placed in chamber E of the Soxhlet apparatus The extract was concentrated under reduced pressure in a rotary evaporator at 45°C, and was maintained at in a refrigerator at 4°C prior to use.
Liquid chromatography-electrospray ionization/multi-stage mass spectrometry (LC-ESI-MSMS)/high performance liquid chromatography (HPLC) analyses
------------------------------------------------------------------------------------------------------------------------------------------------
The LC-MS equipment consisted of a chromatographic system (LC-MS Infinity; Agilent Technologies, Inc., Santa Clara, CA, USA) coupled with an Agilent 1100 Series LC system (Agilent Technologies, Inc.), which was equipped with a binary solvent delivery system, auto-sampler, column temperature controller, photo diode array detector and Finnigan LCQ Deca XP Plus ion trap mass spectrometer (Thermo Finnigan; Thermo Fisher Scientific, Waltham, MA, USA) via an ESI interface. MS spectra were obtained using positive and negative modes; nebulizer gas, 45 Psi; capillary voltage, 4,000 V. The operating parameters for MS were as follows: Collision gas, ultrahigh-purity helium (He); nebulizing gas, high purity nitrogen (N~2~); ion spray voltage, −5.5 kV; sheath gas (N~2~) at a flow rate of 70 arbitrary units; auxiliary gas (N~2~) at a flow rate of 30 arbitrary units; capillary temperature, 360°C; capillary voltage, −15 V; and tube lens offset voltage, −30 V. Full scan data acquisition was performed between 80 and 1,800 m/z in MS scan mode.
HPLC analysis was performed on an Agilent 1260 Infinity series (Agilent Technologies, Inc.) using a Chromolith RP-18e column (4.6 mm ID, 60 mm length). The mobile phase consisted of (A) 0.5% aqueous acetic acid and (B) methanol. Mobile phase gradient: 0--10 min, linear gradient between 10 and 20% of B; 10--15 min, isocratic conditions at 25% of B; 15--20 min, linear gradient between 25 and 40% of B; 20--40 min, linear gradient between 40 and 50% of B; 40--50 min; linear gradient between 50 and 100% of B. Flow rate: 1.5 ml/min.
Chemicals and reagents
----------------------
RPMI-1640 growth medium was purchased from Hangzhou Sijiqing Biological Products Co., Ltd. (Hangzhou, China). Fetal calf serum (Gibco Life Technologies, Carlsbad, CA, USA), trypsin, penicillin, MTT, streptomycin, dimethyl sulfoxide and phosphate-buffered saline (PBS) were used in the present study (all purchased from Sigma-Aldrich, St. Louis, MO, USA). The MTT kit was obtained from Roche Diagnostics (Indianapolis, IN, USA). Camptothecin was used as a positive control for the mitochondrial membrane loss and was purchased from Sigma-Aldrich. An Annexin V-Fluorescein Isothiocyanate (FITC)-Propidium Iodide (PI) Apoptosis Detection kit was purchased from Sigma-Aldrich. All other chemicals and solvents used were of the highest purity grade. Cell culture plasticware was purchased from Falcon^®^ (Corning Life Sciences, Tewkesbury, MA, USA).
Cell lines
----------
MCF-7 human breast cancer cells, A549 human lung cancer cells, HCT-116 human colon cancer cells, COLO-205 human colon cancer cells, MiapaCa-2 human pancreatic cancer cells, and the normal cell line, NIH-3T3 mouse embryonic fibroblasts, were obtained from the Shanghai Institute of Cell Resource Center of Life Science (Shanghai, China). All the cell lines were cultured in a humidified atmosphere of 5% CO~2~ at 37°C in RPMI-1640 medium supplemented with 10% heat-inactivated fetal calf serum, 100 IU/ml penicillin and 100 *µ*g/ml streptomycin.
MTT cell viability assay
------------------------
The inhibition of cell proliferation following treatment with the extract was measured using an MTT assay. Briefly, the MCF-7, A-549, HCT-116, COLO-205, MiapaCa-2 and NIH-3T3 cells were plated in separate 96-well culture plates (1×10^5^ cells/well). Following 24 h incubation at 37°C, the cells were treated with the polyphenol-rich extract (10, 20, 40, 60, 80 and 100 *µ*g/ml; eight wells per concentration) for 12, 24 or 48 h. MTT solution (5 mg/ml) was subsequently added to each well. Following incubation for 4 h, the formazan precipitate was dissolved with 100 *µ*l dimethyl sulfoxide, and the absorbance was measured using an ELISA reader (SpectraMax Plus 384 microplate reader; Molecular Devices, LLC, Sunnyvale, CA, USA) at a wavelength of 570 nm. The cell viability ratio was calculated using the following formula: Inhibitory ratio (%) = \[(ODcontrol − ODtreated) / (ODcontrol)\] × 100%; OD, optical density. Cytotoxicity was expressed as the half maximal inhibitory concentration.
Investigation of apoptosis using fluorescence microscopy
--------------------------------------------------------
Fluorescence microscopy was performed to evaluate morphological alterations in the MiapaCa-2 cancer cells, following treatment with the extract. The cells (1×10^6^ cells/ml) were seeded in 6-well plates and treated with the extract (20, 40, 60 and 80 *µ*g/ml concentrations). Following 24 h incubation at 37°C, the cells were centrifuged at 112 × g for 5 min at 4°C. The resuspended pellet was then dissolved in PBS. The air-dried smears were then fixed in methanol at −20°C, stained with 4′,6-diamidino-2-phenylindole (DAPI; 2 *µ*g/ml) and incubated at 37°C for 20 min. The culture plates were subsequently observed under an inverted light microscope (Eclipse Ti-E; Nikon Corporation, Tokyo, Japan) for morphological analysis.
Annexin V binding assay and quantification of apoptotic cell death
------------------------------------------------------------------
To establish and confirm that the cells were undergoing apoptosis, an annexin V binding assay was performed using flow cytometry. Briefly, the MiapaCa-2 pancreatic cancer cells (2×10^6^ cells/ml) were treated with the polyphenol-rich extract at 20, 40, 60 and 80 *µ*g/ml for 24 h at 37°C. Subsequently, the treated and untreated cells were harvested by trypsinization. The harvested cells were then incubated with annexin V-FITC (80 ng/ml) and PI (50 *µ*g/ml), at room temperature in the dark for 20 min, and analyzed using a FACSCalibur flow cytometer (BD Biosciences, San Jose, CA, USA). A minimum of 2×10^4^ cells were measured in each sample.
Agarose gel electrophoresis for the detection of DNA frag-mentation
-------------------------------------------------------------------
For the detection of DNA fragmentation, the cells were lysed in a solution containing 10 mM Tris-HCl (pH 7.4), 150 mM NaCl, 5 mM EDTA and 0.5% Triton X-100) at room temperature for 30 min. The lysates were then vortexed and cleared by centrifugation at 15,000 rpm for 15 min. DNA was extracted from the supernatant using a 20:20:10 (v/v/v) equal volume of neutral phenol:chloroform:isoamyl alcohol. The DNA was then separated by electrophoresis on 1.0% agarose gels containing 0.1 *µ*g/ml ethidium bromide (Sigma-Aldrich), and DNA fragmentation was detected under UV illumination.
Cell cycle analysis
-------------------
MiapaCa-2 cells (5×10^6^) were seeded into 60 mm dishes and subjected to various concentrations (20, 40, 60 and 80 *µ*g/ml) of polyphenol-rich extract for 48 h at 37°C. The floating and adherent cells were collected by trypsinization and washed twice with PBS. The remaining cells were then incubated in 70% ethanol at −20°C overnight, treated with 10 *µ*g/ml RNase A, and stained with 2.0 *µ*g/ml PI. The stained cells were subsequently analyzed using flow cytometry at a wavelength of 488 nm (FACSCalibur; BD Biosciences), equipped with CellQuest 3.3 software.
Measurement of mitochondrial membrane potential (ΛΨm) loss
----------------------------------------------------------
Mitochondrial membrane potential (ΛΨm) was measured using 1 mM Rhodamine-123 (Rh-123) dye. Rhodamine fluorescence can be used as a measure of membrane polarization in live cell assays within mitochondria. Briefly, 5×10^5^ MiapaCa-2 cells were treated with different concentrations (20, 40, 60 and 80 *µ*g/ml) of the polyphenol-rich extract for 48 h at 37°C. Subsequently, ΛΨm was measured using flow cytometry (FACSCalibur; BD Biosciences). Rh-123 (2 mM) was added 1.5 h prior to termination of the experiment. The cells were then collected, washed in PBS and incubated with PI (10 *µ*g/ml) for 15 min at room temperature. The reduction in fluorescence intensity, due to loss of ΛΨm, was analyzed using flow cytometry. The mean fluorescence intensity was detected using the FL1 channel of the FACSCalibur.
Statistical analysis
--------------------
All the data were analyzed using a one-way analysis of variance, followed by Dunnett\'s test for pair wise comparison with GraphPad Prism 4.0 (GraphPad Software, Inc., La Jolla, CA, USA). Values are presented as the mean ± standard deviation. P\<0.05 was considered to indicate a statistically significant difference.
Results
=======
Evaluation of the antitumor activity of the polyphenol-rich extract using cytotoxicity assays
---------------------------------------------------------------------------------------------
The polyphenol-rich extract was evaluated for antiproliferative activity using an MTT assay. The MCF-7 human breast cancer cells, A549 human lung cancer cells, HCT-116 human colon cancer cells, COLO-205 human colon cancer cells, MiapaCa-2 human pancreatic cancer cells and NIH-3T3 mouse embryonic fibroblast normal cell line were treated with the extract for 24 h ([Fig. 1](#f1-mmr-12-04-4843){ref-type="fig"}). The extract exhibited potent dose-dependent cytotoxic activity against the different cancer cell lines. The COLO-205 and MCF-7 cancer cells were the most susceptible to treatment with the extract, and exhibited increased growth inhibition. The HCT-116 and MiapaCa-2 cells exhibited higher levels of growth inhibition only following treatment with higher concentrations of the extract. In order to examine the toxic effects of the extract on normal cells, the cytotoxic effects of the extract were assessed against the NIH-3T3 mouse embryonic fibroblast cell line. The extract demonstrated reduced cytotoxicity towards the normal cell line, compared with the cancer cell lines, suggesting that its effects are specific to cancer cells. The effect of the polyphenol-rich extract on the growth of MiapaCa-2 pancreatic cancer cells was evaluated using an MTT assay at three different time intervals (12, 24 and 48 h). The cytotoxic effect of the extract on the cells was dose- and time-dependent. At increased time intervals, high levels of growth inhibition were observed ([Fig. 2](#f2-mmr-12-04-4843){ref-type="fig"}).
Evaluation of apoptotic morphological changes in MiapaCa-2 cells
----------------------------------------------------------------
In order to establish whether the polyphenol-rich extract of *S.chinensis* induced apoptosis in MiapaCa-2 cells, the cells were treated with various concentrations of the extract (0, 20, 40,60 and 80 *µ*g/ml) for 48 h. Subsequently, the representative morphological features of apoptosis were examined under an inverted light fluorescence microscope, using DAPI as a staining agent. As shown in [Fig. 3](#f3-mmr-12-04-4843){ref-type="fig"}, compared with the untreated viable cells, treatment with the extract resulted in the appearance of cell contraction and membrane blebbing, both of which are distinguishing features of apoptosis. When treated with a higher concentration of the extract (100 *µ*g/ml), the majority of the cancer cells had shrunk considerably, and no cells exhibiting normal morphological features were detected.
Quantification of apoptotic cell death using an annexin V binding assay
-----------------------------------------------------------------------
The translocation of phosphatidylserine to the exterior surfaces of the plasma membrane is a distinguishing feature of early apoptosis, which can be identified and detected by annexin V-FITC binding. If cell death occurs, fragmented and damaged DNA becomes permeable for binding with PI ([@b34-mmr-12-04-4843]). Following staining of cells with annexin V in combination with PI, this reagent enters the cells only when the plasma cell membrane has deteriorated. In the present study, flow cytometry revealed that, in the extract-treated cells, a higher number of annexin V-positive cells were detected, compared with the untreated control cells ([Fig. 4A and B](#f4-mmr-12-04-4843){ref-type="fig"}). The percentage of apoptotic cells was low following treatment with lower concentrations of the extract. However, at higher extract concentrations (60 and 80 *µ*g/ml), the total number of apoptotic cells increased considerably. This assay provided a quantitative estimation of the rate of apoptotic cell death following drug exposure.
Effects of the polyphenol-rich extract on cell cycle distribution
-----------------------------------------------------------------
Apoptosis and cell cycle dysfunction are closely associated biochemical processes, and any disturbance in cell cycle progression may lead to apoptotic cell death ([@b35-mmr-12-04-4843]). In order to determine the mechanism underlying the growth inhibitory effect of the extract on MiapaCa-2 cancer cells, flow cytometric analysis was performed to detect whether the extract induced cell cycle arrest. Treatment with different concentrations of the extract for 48 h induced G~0~/G~1~-phase growth arrest in the MiapaCa-2 cells. As shown in [Fig. 5](#f5-mmr-12-04-4843){ref-type="fig"}, following treatment of the MiapaCa-2 cells with different concentrations of the extract (20, 40, 60 and 80 *µ*g/ml), considerable G~0~/G~1~ cell cycle growth arrest was observed. The apoptotic cells were observed as shrunken cells with degraded chromatin, increased side scatter and decreased forward scatter properties. The increase in the sub-G~1~ cell population (hypodiploid DNA content) may be due to DNA fragmentation, which eventually results in apoptotic cell death. Inhibition of cell cycle progression may be one of the molecular events associated with the cytotoxic activities of the extract. Following treatment of the cells with low concentrations of the extract, no significant differences were observed in the levels of apoptosis; however, following treatment with extract concentrations of 60 and 80 *µ*g/ml, the percentage of cells undergoing apoptosis (G~0~/G~1~ arrest) increased significantly.
Polyphenol-rich extract induces ΛΨm loss
----------------------------------------
Depolarization of the mitochondrial membrane and subsequent seepage of the outer membrane is a key step in the intrinsic apoptotic pathway. This is usually followed by the release of cytochrome *c* and pro-apoptotic molecules ([@b36-mmr-12-04-4843]). The present study used the fluorescent probe, Rh-123, to detect the ΛΨm in living cells. Treatment with the extract induced a substantial reduction in the number of cells with intact membrane potential, and increased the number of cells with a low ΛΨm after 48 h ([Fig. 6](#f6-mmr-12-04-4843){ref-type="fig"}). Loss of ΛΨm is a crucial event in the mitochondrial apoptotic pathway. The loss of ΛΨm was found to exhibit dose-dependence, and the number of cells with reduced ΛΨm increased with increasing concentrations of the extract.
DNA fragmentation is induced by treatment with polyphenol-rich extract
----------------------------------------------------------------------
A DNA fragmentation assay also revealed that treatment with the extract resulted in DNA laddering, which is indicative of apoptosis ([Fig. 7A](#f7-mmr-12-04-4843){ref-type="fig"}). DNA fragmentation in the polypheno-rich extract-treated cells was confirmed using agarose gel electrophoresis, which detected the presence of DNA laddering, a marker of apoptosis, in the extract-treated MiapaCa-2 cells. By contrast, the untreated control cells demonstrated no evidence of DNA laddering. As shown in [Fig. 7B](#f7-mmr-12-04-4843){ref-type="fig"}, the number of cells exhibiting degraded DNA (sub-G~1~ DNA content) increased in a dose-dependent manner.
LC-ESI-MSMS/HPLC analysis
-------------------------
In the present study, analysis of the *S. chinensis* methanol extract was performed using LC-ESI-MS as well as HPLC techniques. The extract was run under positive and negative ESI-MS conditions, and demonstrated several major and minor peaks. The total ion MS chromatogram, and the structure of the identified molecules are shown in [Figs. 8](#f8-mmr-12-04-4843){ref-type="fig"} and [9](#f9-mmr-12-04-4843){ref-type="fig"}, respectively. Fragmentation of the major peaks was used for identification of the compounds. Identification of the chemical compounds was also achieved by comparing the molecular ion peaks and MS fragmentation patterns with those described in the literature. The six chemical constituents identified in the extract were as follows: Protocatechuic acid, salvianolic acid B, salvianolic acid D, xeractinol, kaempferol and apigenin. All these compounds belong to the polyphenol class of natural products.
Discussion
==========
Dysregulation of cell division and apoptosis are associated with the development of the majority of types of cancer. Therefore, the ability of cancer cells to induce apoptosis has been recognized as one of the major mechanisms that may assist in the development of novel anticancer treatment strategies. Of the two apoptotic pathways, the intrinsic pathway is primarily controlled by members of the B-cell lymphoma (Bcl-2) protein family ([@b35-mmr-12-04-4843],[@b37-mmr-12-04-4843]). The anti-apoptotic Bcl-2 proteins, including Bcl-2 and Bcl-extra large, stimulate cell survival by inhibiting mitochondrial permeability and the release of cytochrome *c*, therefore, effectively inhibiting apoptosis. Pro-apoptotic proteins, including Bcl-2-associated X protein and Bcl-2-associated death promoter, stimulate cell death through a reduction in ΛΨm ([@b38-mmr-12-04-4843],[@b39-mmr-12-04-4843]). Therefore, the proportion of pro-apoptotic to anti-apoptotic molecules is considered to be a determining factor for mitochondria-associated apoptosis.
In the intrinsic apoptotic pathway, mitochondria have a vital role ([@b36-mmr-12-04-4843]). Disruption of the ΛΨm results in the release of cytochrome *c* into the cytosol. Release of cytochrome *c*, along with apoptotic protease-activating factor-1, allows formation of the apoptosome complex, which activates caspase-9 ([@b40-mmr-12-04-4843],[@b41-mmr-12-04-4843]). Activated caspase-9 then cleaves and activates effector caspases, including caspase-3, which results in the apoptotic process. Release of cytochrome *c* from the mitochondria into the cytosol is regulated by pro- and anti-apoptotic Bcl-2 family proteins, which regulate mitochondrial membrane permeability and polarization. In the present study, the results of the flow cytometric analyses demonstrated that the mitochondrial membranes were depolarized following treatment with the extract, particularly following exposure to higher concentrations.
Although a number of studies have reported the anticancer activity of *S. chinensis* against certain malignancies ([@b28-mmr-12-04-4843]--[@b32-mmr-12-04-4843]), the mechanism of action has not been investigated in detail. The aim of the present study was to evaluate the anticancer effects of the polyphenol-rich extract of *S. chinensis* in various human cancer cell lines, and to determine the mechanism underlying the anticancer action against MiapaCa-2 pancreatic cancer cells by evaluating its effects on cell viability, cell cycle phase distribution, apoptosis, DNA fragmentation, and ΛΨm. This is the first study, to the best of our knowledge, to confirm the reported benefits of this plant in Chinese medicine folklore.
The results of the present study demonstrated that the extract from *S. chinensis* exhibited potent cytotoxic effects against cancer cell lines, including MCF-7 human breast cancer cells, A549 human lung cancer cells, HCT-116 and COLO-205 human colon cancer cells, MiapaCa-2 human pancreatic cancer cells and NIH-3T3 mouse embryonic fibroblast normal cells, following exposure for 24 h. Notably, the extract exhibited lower cytotoxicity towards the normal cell line. These results are encouraging, since cancer cell specificity is important for the production of novel anticancer agents. In addition, in order to identify the anticancer action of the extract, its effects on the apoptosis of pancreatic cancer cells were evaluated using DAPI staining, flow cytometry and gel electrophoresis. The results demonstrated that the extract induced apoptosis by inducing DNA damage and cell cycle arrest at G~0~/G~1~ phase. The annexin V binding assay revealed the extent of the apoptosis induced by the extract. The present study also evaluated the role of the extract in disrupting the mitochondrial membrane in pancreatic cancer cells using flow cytometry. The results demonstrated that treatment with the extract induced potent ΛΨm loss in the cells.
In conclusion, polypheno-rich extract from *S. chinensis* was found to induce significant growth inhibition of MCF-7 human breast cancer cells, A549 human lung cancer cells, HCT-116 and COLO-205 human colon cancer cells, and MiapaCa-2 human pancreatic cancer cells. The extract was demonstrated to be selective, as it exhibited lower cytotoxicity towards the NIH-3T3 normal cell line, compared with the cancer cell lines. The extract induced apoptosis in the pancreatic cancer cells, determined using flow cytometry, fluorescence microscopy and agarose gel electrophoresis. In addition, the extract induced G~0~/G~1~ cell cycle arrest and loss of ΛΨm. This requires further elucidation of the exact mechanism of action in order to render it an effective therapeutic strategy against different types of cancer.
{#f1-mmr-12-04-4843}
{#f2-mmr-12-04-4843}
{#f3-mmr-12-04-4843}
{#f4-mmr-12-04-4843}
{#f5-mmr-12-04-4843}
{#f6-mmr-12-04-4843}
{#f7-mmr-12-04-4843}
{#f8-mmr-12-04-4843}
{#f9-mmr-12-04-4843}
[^1]: Contributed equally
| {
"pile_set_name": "PubMed Central"
} |
Background
==========
Biology is replete with periodically patterned elements, from the sensory bristles of a fruit fly to the hair and teeth of mammals. Models of periodic patterning seek to explain the developmental origin of boundaries separating adjacent repetitive structures and the maintenance of cellular compartments once formed \[[@B1]-[@B8]\]. For example, the formation of feather tracts on the dorsal surface of chick embryos \[[@B9]\] and mammary (milk) lines on the ventral surface of embryonic mammals \[[@B10]\] serve to pre-pattern regions competent for the initiation of these structures. Similarly, a functionally equivalent field is established along the axis of the oral jaws in most vertebrates, competent to form tooth bud primordia \[[@B11]-[@B15]\]. In teleost fish this initial field is known as the primary odontogenic band (OB) \[[@B12],[@B13],[@B15]\] and in mammals it is termed the dental lamina \[[@B4],[@B14]\]. This band or lamina sets the regionally restricted \'field\' along the jaw axis from which tooth induction is triggered.
As with all periodically patterned systems, an initial \'field of competence\' is set from a once-homogeneous cellular region, followed by the establishment of positional information throughout the restricted \'field\' \[[@B16],[@B17]\]. The initial field may be set up by a number of diffusing molecules such as morphogens that allow regionalization to occur, from which cellular differentiation responds along a gradient \[[@B5],[@B18]\], probably by means of a reaction-diffusion-type mechanism \[[@B19],[@B20]\]. Positional information determines cell differentiation, cellular compartmentalization and subsequent unit placode initiation, the first of which is imperative for iterative initiation of adjacent placodes via activator-inhibitor mechanisms \[[@B2]\]. Placode initiation is thought to be triggered by cellular accumulation (self-organization) over a given threshold that reacts to a number of positional cues within the competent field \[[@B1]\]. Within the placode itself, additional activators and inhibitors determine the boundaries of the placode unit and the spacing between units. Studies of periodically patterned systems such as the developing vertebrate dentition and developing chick feathers have led to the identification of a number of molecules that have been modelled as activators or inhibitors within the specific developing system \[[@B2],[@B21],[@B22]\]. In feather placode patterning, Shh and members of the Eda pathway have activator roles, while Bmp2 and Bmp4 are thought to act as inhibitors \[[@B2],[@B23],[@B24]\]. During mouse odontogenesis the same molecules are involved in patterning the molar cusps. Attempts have been made to model cusps according to activator-inhibitor patterning mechanisms; however, whether individual candidates can be classed as activators or inhibitors during tooth development is largely stage dependent \[[@B21],[@B25]-[@B27]\].
Molecules involved in the establishment of vertebrate dentition have been well characterized from studies of the mouse \[[@B11],[@B21],[@B28]\]. A number of these molecules are known to have detrimental effects on the murine dentition when removed/inhibited from the dental network early in tooth development; *Shh*\[[@B29],[@B30]\], *Pitx2*\[[@B31]\] and *Pax9*\[[@B32]\] are among those with severe dental phenotypes \[[@B33]\]. For example, inhibition of *Shh*in mandibular explants during the transition of dental competence to initiation (E10.5) leads to tooth arrest at the bud stage \[[@B29],[@B30]\]. Thus, it is clear that this gene is essential for the correct establishment of the global dental programme. However, these studies are specific to the mouse experimental model, which develops a single set of teeth with no replacements. We therefore know nothing of the resulting phenotypes when modifications occur to these networks, for example the hedgehog pathway, in vertebrates with numerous functional tooth rows and continuous replacement cycles.
The morphogenesis of teeth, like that of other periodically patterned vertebrate organs (for example, hair, mammary glands, feathers), is regulated both by sequential and reciprocal molecular interactions between two adjacent cell layers, the epithelium and the directly underlying mesenchyme \[[@B11],[@B28]\]. During early stages, these distinct organs share a number of features and express a familiar suite of genes with common roles \[[@B28],[@B34]\]. Many studies have attempted to identify the morphodynamic control of iterative organization and how such patterning mechanisms change during development to generate evolutionary novelty \[[@B6],[@B7],[@B21],[@B35]-[@B37]\]. We sought to characterize the expression of a set of these molecules in the dentitions of Lake Malawi cichlids to tackle an unanswered and fundamental biological question: how is the diversity of periodically patterned elements generated in nature? Malawi cichlids are exemplars of natural craniofacial diversity. In essence, natural selection has conducted an experiment in micro-evolutionary diversification, and we want to know how development works to produce variation in phenotype \[[@B37]\]. The range of dental variety in Malawi is tremendous given a common ancestor in the last 500,000 to 1 million years \[[@B26]\]; species possess about 10 teeth in a single row (per jaw), or as many as 700 teeth in up to 20 rows. Species differ in tooth size, spacing and shape in coordinated fashion \[[@B26],[@B36]\]. We focus on three closely related Lake Malawi cichlids with alternative dental phenotypes (Figure [1](#F1){ref-type="fig"}): *Cynotilapia afra*(CA, Figure [1a](#F1){ref-type="fig"}), a unicuspid species with two tooth rows of large, widely spaced teeth; *Metriaclima zebra*(MZ, Figure [1b](#F1){ref-type="fig"}), a bicuspid and tricuspid species with five or six tooth rows of intermediately sized and spaced teeth; and *Labeotropheus fuelleborni*(LF, Figure [1c](#F1){ref-type="fig"}), a uniformly tricuspid species with 10 or more tooth rows of small, tightly packed teeth \[[@B26],[@B36]\]. Previously, we have used these species to identify the chromosomal basis of divergent tooth shapes among species \[[@B26],[@B36]\]. Here we address a different question: we ask how conserved gene networks are deployed to influence the diversity in the size, number, spacing within rows and number of rows of teeth.
{#F1}
Results and Discussion
======================
Variation in developmental gene networks prefigures differences in adult cichlid dentitions
-------------------------------------------------------------------------------------------
We cloned cichlid orthologues of genes required during oral epithelial organization and tooth germ initiation (*bmp2*, *fgf8*, *pitx2*, *shh*) \[[@B11],[@B38]\] as well as mesenchymal markers (*bmp2*, *bmp4*\[[@B36]\], *pax9*, *runx2*) involved in reciprocal signalling to the epithelium \[[@B11]\]. Teleost tooth development has been well characterized in the zebrafish and thus our nomenclature for the early stages of tooth development will follow that model. Two stages of early odontogenesis are relevant: the thickened epithelium stage and the bell-shaped epithelium stage \[[@B39]-[@B41]\]. Developing teeth beyond this point will be referred to as tooth germs, spanning the progression of the tooth from a bell-shaped unit to various stages of functional maturity, characterized by cytodifferentiation.
The transcription factor *pitx2*, described as a putative odontogenic-commissioning gene \[[@B12],[@B13],[@B15]\], has a broad expression pattern that encompasses both the developing tooth unit and the inter-tooth region, marking the extent of the dental-competent oral epithelium, including regions of future tooth rows (Figures [2a](#F2){ref-type="fig"} and [3A--C](#F3){ref-type="fig"}). *pitx2*is one of the earliest dental epithelial markers (Figure [3C](#F3){ref-type="fig"}) with expression in the thickened dental epithelium (Figure [3C](#F3){ref-type="fig"}) and both the inner dental epithelium (IDE) and outer dental epithelium (ODE) of the maturing tooth (Figure [3A](#F3){ref-type="fig"} and [3B](#F3){ref-type="fig"}). Interestingly, the early pattern of *pitx2*expression differs across the three species prior to and during morphogenesis of the first tooth, and reflects the future organization of these distinct dentitions (Figure [2a](#F2){ref-type="fig"}). *L. fuelleborni*shows the greatest region of dental competence (expression of *pitx2*), consistent with the later elaboration of teeth and tooth rows (Figures [1c](#F1){ref-type="fig"} and [2a](#F2){ref-type="fig"}). Future tooth rows also show expression of *pitx2*, labelling the lingual progression of the subsequent OB (Figure [3A](#F3){ref-type="fig"} and [3B](#F3){ref-type="fig"}). Similar to studies in other fishes \[[@B38],[@B41]\], we found that *fgf8*expression is not associated with initiating tooth germs in Malawi cichlids (not shown).
{#F2}
{#F3}
*pax9*, one of the earliest mesenchymal markers of odontogenesis in the mouse, is either absent from or weakly expressed in the dentitions of zebrafish and Mexican tetra \[[@B38]\]. By contrast, in Malawi cichlids *pax9*is expressed initially in oral mesenchymal cells as a dental field along the mesiodistal jaw axis (Figure [2b](#F2){ref-type="fig"}, OB stage, and Figure [3D](#F3){ref-type="fig"}), then it is strongly up-regulated in the underlying mesenchyme at the epithelial thickening stage of the first tooth (Figure [2b](#F2){ref-type="fig"}, first-tooth stage, and Figure [3D](#F3){ref-type="fig"}). Expression of *pax9*is then restricted to cells of the dental mesenchyme enveloping the tooth during morphogenesis and is absent from the cells of the dental papilla (Figure [2b](#F2){ref-type="fig"}, 3--4-teeth stage, and Figure [3E](#F3){ref-type="fig"} and [3F](#F3){ref-type="fig"}). The expression of *runx2*essentially replicates that of *pax9*for the stages examined (data not shown).
Expression of *shh*is up-regulated from the primary OB into the individual tooth germs (Figure [2c](#F2){ref-type="fig"} and [3G--I](#F3){ref-type="fig"}). *shh*continues to be expressed during tooth morphogenesis, marking the bell-shaped dental epithelium and later the IDE of the tooth during cytodifferentiation (Figure [3G--I](#F3){ref-type="fig"}). Notably, *shh*is never present in regions between or around teeth, called the zone of inhibition (ZOI) \[[@B15]-[@B17]\]. In Malawi cichlids *shh*expression continues to label OBs, marking subsequent initiation of more lingual tooth rows, one at a time (Figures [2c](#F2){ref-type="fig"} and [3G--I](#F3){ref-type="fig"}). *C. afra*does not develop a third OB (nor a third row), while *M. zebra*and *L. fuelleborni*initiate an OB for each future tooth row (see Figure [4](#F4){ref-type="fig"}). *bmp2*is co-expressed in the competent epithelial OB with *shh*and *pitx2*. From the initial epithelial OB, *bmp2*expression is up-regulated in the cells of the contorted bell-shaped epithelial germ and continues to be expressed during differentiation (Figure [3L](#F3){ref-type="fig"}) before becoming localized to cells of the mesenchymal dental papilla (Figure [3J--L](#F3){ref-type="fig"}). In addition, *bmp2*is restricted to epithelial cells at the developing tooth tip (Figure [3K](#F3){ref-type="fig"}), which will differentiate to ameloblasts, partially responsible for the secretion of enameloid, the first mineralized tissue of the teleost tooth \[[@B42]\]. These are an equivalent set of cells to the mammalian \'enamel knot\'. *bmp4*is expressed initially in the mesenchymal field along the mesiodistal axis prior to tooth germ initiation, much like *pax9*and at the 3--4-teeth stage *bmp4*is restricted to the dental papilla (data not shown \[[@B36]\]).
{#F4}
These data, in conjunction with data reported earlier for *bmp4*\[[@B36]\], suggest that early patterns of gene expression differentiate among Malawi cichlid dentition types. *L. fuelleborni*, *M. zebra*and *C. afra*exhibit marked variation in the field of odontogenic competence both in the epithelium (*pitx2, shh*) and mesenchyme (*pax9*), in the spacing (*bmp2*, *bmp4*, *pax9*, *runx2*, *shh*) and in the size (*bmp4*, *pax9*, *runx2*, *shh*) of the initial tooth germs. These differences are readily quantifiable (Table [1](#T1){ref-type="table"}), arise before teeth acquire their functional adult shape, and correspond directly to the size and spacing observed in the adult dentitions \[[@B26]\]. For example, *C. afra*embryos have the largest initial dental germs and surrounding ZOI (Table [1](#T1){ref-type="table"}) while adults have the largest, fewest teeth (mean ± SD of 1.3 ± 0.20 per millimetre of jaw width) with the greatest inter-unit spacing compared with the other two species, *M. zebra*(3.3 ± 0.52) and *L. fuelleborni*(4.8 ± 0.74) \[[@B26]\]. The measurements obtained from *shh*expression in Table [1](#T1){ref-type="table"} highlight an early developmental origin of dental diversity among these three species. Streelman and Albertson noted a similar pattern from *bmp4*expression, with a comparable range of cichlid species \[[@B36]\]. While some measurements shown here seem counterintuitive (for example, Table [1b](#T1){ref-type="table"}, columns 3 and 4), it is because species also differ in more subtle aspects of tooth initiation. *C. afra*never initiates a tooth between the first three teeth to develop; *M. zebra*has an initiation order that places tooth 5 between teeth 1 and 3, whereas *L. fuelleborni*initiates tooth 5 between teeth 1 and 2 (Figure [2c](#F2){ref-type="fig"}). These early differences in tooth size, spacing and organization are perhaps surprising because cichlids (and all teleost fish) continuously replace their teeth; therefore, we might have expected inter-specific variety to develop gradually, over multiple rounds of tooth replacement, from a common dental \'ground\' state, as is the case for tooth shape \[[@B26]\].
######
Size and spacing of tooth germs in three species of Malawi cichlids
\(a\)
------------- ----------------------------- ---------------------------------------------------------- ---------------------------------------------------------- --------------------------------------- --------------------------------------- --------------------------------- ----------------------------------------- ---------------------------------
**Species** **Diameter of first tooth** **ZOI including first tooth** **ZOI medial to first tooth** **ZOI distal to first tooth** **ZOI from First tooth to second OB** **MD length of OB** **Width of OB at level of first tooth** **Area (μm^2^) of first tooth**
*CA* 22.91 ± 0.839 51.80 ± 1.298 18.67 ± 0.344 17.99 ± 0.482 8.80 ± 0.56 110.72 ± 7.337 22.21 ± 1.135 528.08 ± 31.348
*MZ* 17.62 ± 0.149 44.05 ± 0.898 16.50 ± 0.213 12.87 ± 0.322 6.84 ± 0.084 116.62 ± 5.013 23.02 ± 0.382 384.06 ± 4.754
*LF* 14.87 ± 0.349 39.9 ± 1.697 11.95 ± 0.285 15.24 ± 0.392 7.28 ± 0.545 120.91 ± 3.834 19.41 ± 0.782 268.47 ± 8.084
\(b\)
**Species** **Diameter of first tooth** **Inter-germ space between first and next medial tooth** **Inter-germ space between first and next distal tooth** **ZOI from First tooth to second OB** **MD length of second row OB** **Area (μm^2^) of first tooth**
*CA* 25.68 ± 0.359 16.64 ± 0.128 14.02 ± 0.096 8.29 ± 0.266 140.02 ± 3.669 684.57 ± 19.248
*MZ* 18.19 ± 0.231 19.63 ± 0.222 11.96 ± 0.318 10.18 ± 0.434 138.47 ± 2.897 435.09 ± 19.98
*LF* 18.37 ± 0.437 13.06 ± 1.112 20.33 ± 1.065 8.02 ± 0.295 168.39 ± 5.432 404.15 ± 23.182
\(a\) We report mean values (three to five individuals) of measurements taken from the expression of *shh*at the first-tooth stage in the three cichlid species: CA, *Cynotilapia afra*; MZ, *Metriaclima zebra*; LF, *Labeotropheus fuelleborni*. MD, mesiodistal; OB, odontogenic band; ZOI, zone of inhibition. (b) Mean values (three to five individuals) from the 3--4-teeth stage. All values are in micrometres ± SE.
Organizing the periodic pattern with molecular \'spacers\'
----------------------------------------------------------
Given the set of molecules localized to the first tooth germs (notably *shh*, but also *pitx2*and *bmp2*), we hypothesized that other factors expressed within the ZOI surrounding these germs might guide the size and spacing of early cichlid tooth units. We therefore analyzed the expression of genes involved in a putative spacing mechanism. We cloned three genes, *eda*, *edar*and *wnt7b*, with antagonistic effects on *shh*in the initiation of mouse teeth, mouse hair follicles and chick feathers \[[@B9],[@B30],[@B43]-[@B50]\]. *edar*(data not shown) is expressed within the germs themselves in a pattern similar to *shh*(Figures [2c](#F2){ref-type="fig"} and [3G--I](#F3){ref-type="fig"}). We observed the expression of both *wnt7b*and *eda*surrounding the initial *shh*-positive tooth germs (Figure [5a--c](#F5){ref-type="fig"}) within the ZOI/inter-germ regions across the three species.
{#F5}
*eda*is expressed locally and strongly surrounding the first-tooth germ (expressing *shh*and *edar*) in all three species (first-tooth stage data not shown). By the 3--4-teeth stage, *eda*remains expressed in the mesenchyme locally and heavily at the lingual margin of the first tooth in *C. afra*and *M. zebra*(Figures [5a](#F5){ref-type="fig"} and [5b](#F5){ref-type="fig"} and [6A--C](#F6){ref-type="fig"}), but its lingual expression is broader and more diffuse in *L. fuelleborni*(Figure [5c](#F5){ref-type="fig"}). Notably, there appears to be a lateral bias in the expression of *eda*in both *L. fuelleborni*and *M. zebra*, which may reflect the influence of (or may influence) the initiation of tooth 5 between existing germs 1 and 2 or 1 and 3, respectively. An apparent *eda*-negative region exists in both *L. fuelleborni*and *M. zebra*where the fifth tooth in the series will appear, a different position in each species (Figure [5](#F5){ref-type="fig"}, middle column). *C. afra*lacks tooth initiation between these first three positions, an arrangement that continues into the adult dentition.*wnt7b*expression coincides with the ZOI surrounding the first teeth in all species (Figures [5](#F5){ref-type="fig"} and [6D--F](#F6){ref-type="fig"}); this is best illustrated comparatively with image overlays with *shh*as depicted in Figure [5a--c](#F5){ref-type="fig"}.
{#F6}
Remarkably, these genes seem also to be employed in the initiation and spacing of future tooth rows, an iterated expression pattern similar to tooth germ organization within each row. *shh*labels each OB for subsequent tooth rows (Figures [2c](#F2){ref-type="fig"} and [4](#F4){ref-type="fig"}); *eda*and *wnt7b*are expressed between the first tooth row and the OB of the second (Figures [5](#F5){ref-type="fig"} and [6](#F6){ref-type="fig"}). Specifically, *eda*expression partly overlaps that of *shh*in the lingual OB, while *wnt7b*is expressed either side of *shh*. Thus, eda from the enveloping mesenchyme (Figure [6A--C](#F6){ref-type="fig"}) may induce and maintain *shh*expression in tooth germs as well as in future tooth rows, and planar epithelial wnt7b (Figure [6D--F](#F6){ref-type="fig"}) may inhibit dental competence in these regions, similar to the role of these molecules in other systems \[[@B9],[@B30],[@B44]-[@B48]\].
Hedgehog signalling is required for initiation of periodic dental patterning
----------------------------------------------------------------------------
Our data suggest that the ZOI has an important role in patterning the size (and spacing) of the *shh*-positive tooth germs, especially the first unit to initiate. We speculated that the first tooth might possess unique regulatory properties as a source for continued induction and patterning of the dental program. We tested the role of the first tooth as a source of communicative signal for the organization of the dentition using targeted chemical inhibition of the hedgehog pathway at the first epithelial thickening stage. *C. afra*embryos at the first-tooth stage (five days post-fertilization (dpf); Figure [2c](#F2){ref-type="fig"}) were treated for 24 hours in 50 μM cyclopamine in 1% DMSO (based on protocols in \[[@B52]\]), thus spanning the time from the initiation of the first tooth to the three-teeth stage (6 dpf), by which time the second row OB had established territory. A subset of treated embryos further developed for an additional 24 hours under standard conditions (7 dpf; Figure [7a](#F7){ref-type="fig"} and [7b](#F7){ref-type="fig"}); the remaining embryos from the same brood were allowed to develop for an additional six days (12 dpf; Figure [7c](#F7){ref-type="fig"} and [7d](#F7){ref-type="fig"}) to span the period of both first row eruption and development of the lingual tooth row (Figures [4](#F4){ref-type="fig"}, [7](#F7){ref-type="fig"} and [8](#F8){ref-type="fig"}). Treated *C. afra*fixed at 7 dpf showed varying low levels of *shh*expression localized to a reduced number (one or two) of tooth germs on each side of the dentary and the OB for the second row (Figure [7b](#F7){ref-type="fig"}). *shh*expression appears within the region allocated for the ZOI (seen as *shh*-negative inter-germ regions in all controls; Figures [2c](#F2){ref-type="fig"} and [7a](#F7){ref-type="fig"}) normally expressing *eda and wnt7b*(Figures [5](#F5){ref-type="fig"} and [6](#F6){ref-type="fig"}), suggesting a breakdown in both the initiation and spacing mechanisms (Figure [7b](#F7){ref-type="fig"}). Expression of *eda*in treated *C. afra*was absent in regions of tooth development (data not shown), implying that *eda*may respond to signals from, or downstream of, the hedgehog pathway. 1% DMSO control *C. afra*showed patterns of *shh*expression identical to standard controls (untreated).
{#F7}
![**A periodic pattern generator for diversity in Malawi cichlid dentitions**. (A) A cross-section through three thickened tooth germs showing planar signalling in the zone of inhibition (ZOI)/inter-germ space (grey, *wnt7b*) demarcating the location of teeth (green, *shh*and *edar*), with contribution from dental mesenchyme (blue, *eda*, *bmp4*, *pax9*and *runx2*). Non-dental mesenchyme, purple. (B) The proposed interactions of wnt7b and eda (around the thickened tooth germs) with shh and edar within the thickened epithelial germ \[[@B9],[@B30],[@B44]-[@B51]\] in the context of our observations. wnt7b acts as an inhibitor of *shh*during tooth patterning and eda from the mesenchyme acts as an activator maintaining *shh*, perhaps via the eda-receptor *edar*, within the tooth germ. Wnt signals (possibly including wnt7b) may induce *eda*in the mesenchyme. DE, dental epithelium; DM, dental mesenchyme; OE, oral epithelium; OM, oral mesenchyme. (C)-(K) the induction of each tooth row acts in a \'copy-and-paste\' mechanism utilizing the same genes for the pattern and organization of teeth and new tooth rows. Differential regulation of the genes during periodic patterning lead to the species diversityRedeployment of genes from OB1 establishes new tooth rows periodically in a lingual progression until the \'copy-and-paste\' mechanism fails, due lack of combinatorial *pitx2*and *shh*, essential for sequential row addition. Upper panel (C)-(E) around 5--6 days post-fertilization (dpf); middle panel (F)-(H) around 6--7 dpf; lower panel (I)-(K) around 12 dpf; each panel represents the left half of the dentary in dorsal view. (C)-(E the first tooth initiates from OB, a combinatorial expression of *pitx2*(white) and *shh*(green). The initial *pitx2*positive region is greatest in LF, with more rows. Up-regulation of *shh*and *edar*in the epithelial germs plus the establishment of a ZOI (*eda*and *wnt7b*) set the size and space restriction for the epithelial tooth germs. Local *eda*surrounds the first germ for each species (darker blue denotes stronger local expression of *eda*, which later becomes a broader domain and less intense, lighter blue). (F)-(H) *eda*remains strong and local in CA and at this stage (3--4-teeth stage) is broader and less intense in LF. CA does not initiate more than two tooth rows because *pitx2*is constrained along the oral-aboral axis (lingual to the first teeth). The size, spacing and positioning of teeth and tooth rows is regulated by the interactions of *wnt7b*(grey), *eda*(blue) and *shh*(green) within the *pitx2*-positive field (white; see the main text and Figure 5). Future tooth row initiation in (J) MZ (OB3) and (K) LF (OB4) depends on the lingual co-expression of *pitx2*and *shh*. (I) CA does not initiate a third row of teeth, and lacks co-expression of *shh*and *pitx2*in a third OB (see Figure 4). Arrows mark the direction of tooth initiation for continued tooth addition on older rows. (I)-(K) *eda*and *wnt7b*expression domains are predicted from earlier stages. A more diffuse expression of *eda*(light blue) will continue to regulate size, spacing and row initiation of first-generation teeth. Numbers indicate the order of appearance for tooth rows. S, jaw symphysis.](1741-7007-6-32-8){#F8}
In all treated individuals that were allowed to develop for a further six days (fixed at 12 dpf), we found that the first tooth continues partial development and shows signs of mineralization, although it does not complete development or attachment (Figure [7d](#F7){ref-type="fig"}). With the exception of a mineralized remnant of the first tooth, all other teeth, adjacent to the first and in subsequent rows, failed to develop (Figure [7d](#F7){ref-type="fig"}). Knockdown of the hedgehog pathway at the 3--4-teeth stage resulted in a functional, patterned and replacing dental system (data not shown). These observations demonstrate that when perturbed (via the hedgehog pathway) at the first-tooth stage, the dental programme cannot recover, despite continued cycles of periodic patterning past this stage in untreated individuals.
The periodic pattern generator for dental diversity
---------------------------------------------------
The comparison of gene expression across Malawi cichlid species with divergent dentitions suggests a simple model implicating *pitx2*, *eda*and *wnt7b*, and their interaction with *shh*and *edar*, as primary features of a periodic pattern generator for diversity in Lake Malawi cichlid dentitions (Figure [8](#F8){ref-type="fig"}). The model accounts for two aspects of dental patterning: how to put tooth rows in jaws, and how to put teeth in tooth rows. Our data suggest that the combination of *pitx2*and *shh*is required for a competent field of tooth initiation (the OB, Figures [2](#F2){ref-type="fig"}, [3](#F3){ref-type="fig"}, [4](#F4){ref-type="fig"} and [8](#F8){ref-type="fig"}). *M. zebra*and *L. fuelleborni*exhibit expanded expression of *pitx2*lingually on the embryonic lower jaw; *C. afra*does not (Figure [2](#F2){ref-type="fig"}). *pitx2*and *shh*are also co-expressed in each subsequent OB for *M. zebra*and *L. fuelleborni*(Figure [4](#F4){ref-type="fig"}); *C. afra*does not initiate a third OB. Therefore, the lack of lingual/oral co-expression of *pitx2*and *shh*in *C. afra*(Figures [4](#F4){ref-type="fig"}, [8F](#F8){ref-type="fig"} and [8I](#F8){ref-type="fig"}) may account for the reduction in row number compared with the other species (Figures [4](#F4){ref-type="fig"}, [8G](#F8){ref-type="fig"} and [8J, H](#F8){ref-type="fig"} and [8K](#F8){ref-type="fig"}, respectively). The lack of combinatorial expression of *shh*and *pitx2*in the oral region of zebrafish may partially explain the lack of teeth \[[@B38]\]. Here we show that this mechanism likely accounts for variation in tooth row number among Malawi cichlids. Thus, molecular mechanisms used to pattern the first row of teeth (the only row of teeth in mammals and most vertebrates) are redeployed as \'triggers\' of dental competence and initiation in each subsequent row. We suggest that the initiation of new tooth rows follows a \'copy and paste\' mechanism wherein the dental expression network is redeployed for each new tooth row. Therefore, our model posits that preceding tooth rows are required as a source of signal to initiate the next lingual row during sequential addition.
The combination of comparative gene expression data and perturbation of the hedgehog pathway suggests that the correct initiation and maintenance of the first-tooth germ, via activation of *shh*, is necessary for the periodically patterned dental programme in Malawi cichlids (Figure [8](#F8){ref-type="fig"}). Comparison of the cyclopamine phenotype at the first-tooth to the 3--4-teeth stages shows that disturbing the development of the first-tooth germ has an effect on the entire dentition, whereas disrupting the dentition at later stages results in a mildly reduced phenotype with additional teeth forming and completing development. We do not yet understand the molecular mechanisms (for example, decreased epithelial proliferation and/or increased cell death) of this severe dental phenotype at the first-tooth stage.
Our data imply that *eda*and *wnt7b*, expressed in the ZOI, regulate initial tooth germ size and position within rows, through interactions with *shh*; wnt7b inhibits the germ through planar epithelial signals (Figures [5](#F5){ref-type="fig"} and [6D--F](#F6){ref-type="fig"}) and eda maintains the tooth germ (*shh*and *edar*) from within the surrounding mesenchyme (Figures [5](#F5){ref-type="fig"} and [6A--C](#F6){ref-type="fig"}). The ZOI may not lie solely within the layers of the epithelium and we suggest that inhibitor/activator controls signal from within the underlying mesenchyme that envelops the thickened dental epithelium \[[@B34]\]. Once the periodic pattern is established, other molecules may act as inhibitors from within the developing tooth unit, for example bmp2, which is present both in the early epithelial thickening and within the dental papilla (mesenchyme) during maturation (Figure [3J--L](#F3){ref-type="fig"}), and bmp4, which is restricted to the dental papilla (data not shown; \[[@B36]\]).
The expression of *eda*in the mesenchyme surrounding the developing dental germs of cichlids (Figures [5](#F5){ref-type="fig"} and [6A--C](#F6){ref-type="fig"}) is more similar to that deployed during the patterning of feather placodes and salivary primordia \[[@B9]\] than that observed in mammalian dentitions, where it is restricted to epithelium \[[@B44],[@B46]\]. In our model, a large initial tooth germ in *C. afra*results from sustained local and intense *eda*expression on a comparatively similar inhibitory background of *wnt7b*(Figure [5a](#F5){ref-type="fig"}). The size of this tooth germ is reduced in *M. zebra*(Figure [5b](#F5){ref-type="fig"}) and *L. fuelleborni*(Figure [5c](#F5){ref-type="fig"}) because the *eda*expression broadens earlier (especially for *L. fuelleborni*), a heterochronic imbalance setting the stage for more, closely packed *shh*-positive tooth germs (Table [1](#T1){ref-type="table"} and Figure [8](#F8){ref-type="fig"}). Consistent with our results, transgenic mice (K14-*eda*) with increased levels of *Ectodysplasin*expression exhibit larger tooth germs \[[@B48],[@B52]\]. Furthermore, Eda null mutant mice have reduced tooth germs \[[@B48],[@B53]-[@B55]\]. However, in the mouse, effects of Eda on tooth size correlate positively with effects on tooth number; for example, higher levels of *Eda*lead to a single extra molar \[[@B49],[@B52]\]. Our data and model point to an important distinction between overall levels of *eda*and its spatial expression over time. An earlier dispersion of *eda*expression after initiation of the first tooth (as in *L. fuelleborni*), rather than continued localized expression around that first-tooth germ (as in *C. afra*), may in fact lead to the production of more, smaller tooth germs (Figure [8](#F8){ref-type="fig"}).
The position of subsequent tooth rows is also specified in part by the expression of *wnt7b*and *eda*in our model. Mesenchymal *eda*plays a permissive role in the positioning of the lingual OB (Figures [5](#F5){ref-type="fig"} and [6A](#F6){ref-type="fig"}). In *C. afra*, its expression is strongest medial to the first tooth, while in *M. zebra*and *L. fuelleborni*it appears more as a band along the mesiodistal axis (Figures [5](#F5){ref-type="fig"} and [6A--C](#F6){ref-type="fig"}; also see the second row tooth positions in Figure [1](#F1){ref-type="fig"}). *wnt7b*also appears to demarcate the location of the second row, as its expression is either side of the *shh*-positive second OB (Figures [5](#F5){ref-type="fig"} and [6D--F](#F6){ref-type="fig"}) and, in a similar iterative manner to the patterning of individual tooth units, *wnt7b*is restricted to the inter-row space (Figure [5a--c](#F5){ref-type="fig"}). Once the initiation of the primary dental pattern for each row is established, the essential nature of *shh*and genes that occupy the ZOI is lost; although they likely continue to be expressed during further morphogenesis (Figures [3](#F3){ref-type="fig"} and [6](#F6){ref-type="fig"}), these molecules are probably no longer required for initiation of the secondary, replacement dentition \[[@B13]\].
Conclusion
==========
Periodically patterned phenotypes such as the dentitions of Lake Malawi cichlids present important exemplars for evolutionary developmental biology. The discipline has heretofore focused on the molecular basis of evolutionary novelty among distantly related organisms \[[@B35],[@B56]\] or the genetic/transcriptional basis of discrete trait loss among closely related groups \[[@B51],[@B57]\]. Trait elaboration (for example, bigger, longer, stronger \[[@B58],[@B59]\]) is more difficult to study because phenotypes are subtler, but this remains the more common type of evolutionary change \[[@B37]\]. Dental diversity is an intermediate case; quantitative elaboration takes the form of gain or loss of discrete units. Our results support the general model that old genes, and entire developmental modules, are deployed anew to generate micro-evolutionary novelty in iterative structures.
Methods
=======
Fish husbandry
--------------
Embryos and fry of three species of Lake Malawi cichlids (*C. afra*, *M. zebra*and *L. fuelleborni*) were raised to the required stage in a recirculating aquarium system at 28°C. Embryo ages (in dpf) were set after the identification of mouth brooding females (day 0). Embryos were then removed from the mouths of brooding females and, if required, were maintained for further development in separate culture tanks at 28°C.
Sequences
---------
Cloned sequences used to generate digoxigenin-labelled antisense riboprobes from Malawi cichlid species have been deposited in GenBank (accession numbers: [EU867210](EU867210) -- [EU867219](EU867219)). Many of the genes were identified through partial genome assemblies of *L. fuelleborni*and *M. zebra*\[[@B60]\] and cloned from *M. zebra*and *L. fuelleborni*cDNA libraries. Sequences of cDNA used to generate the probes are identical across the three species. Overall, these species exhibit almost no sequence divergence; the average nucleotide diversity for comparisons across the Malawi assemblage is 0.2%, less than among laboratory strains of the zebrafish \[[@B60]\].
*In situ*hybridization
----------------------
To ensure the embryos of the three species were of equivalent stages (especially during gene expression comparisons), specimens were stage-matched based on external features, including pectoral and caudal fin development and eye development and maturity. Specimens for *in situ*hybridization were anaesthetized in tricaine methanesulfonate (MS222, Argent) and fixed overnight in 4% paraformaldehyde (PFA) in 0.1% phosphate-buffered saline (PBS) at 4°C. Whole-mount *in situ*hybridization experiments were based on protocols from \[[@B12]\] and modified as follows: embryos were transferred to methanol for dehydration and stored at -20°C. Specimens were rehydrated through to PBS with Tween-20 and digested with 4--10 μg/ml proteinase K (PK); the final concentration was based on the specific stage of embryo/fry (for example, embryos at approximately 5 dpf were digested with 5 μg/ml PK). Following hybridization, embryos were washed in TST (10 mM NaCl, 10 mM Tris-HCl, Tween-20 in depc-H~2~O). During the colour reaction stage of the protocol, all embryos were allowed to fully develop the colour. Thus, embryos were continuously transferred into fresh NBT/BCIP solution (Roche) in NTMT until full staining had ensued; this was determined after multiple regions of known expression became positive. Specimens were stage-matched based on external features, including pectoral and caudal fin development and eye development and maturity. All *in situ*hybridization experiments were performed with multiple specimens (multiple individuals were fixed at regular intervals, within single broods, then repeated at least twice with alternative broods) to fully characterize the expression patterns within and across the three species. After colour reaction (NBT/BCIP, Roche) embryos were washed in PBS and fixed again in 4% PFA, before whole-mount imaging using a Leica Microsystems stereomicroscope (MZ16). Embryos were embedded in gelatin and chick albumin with 2.5% gluteraldehyde. The gelatin-albumin blocks were post-fixed in 4% PFA before sectioning. Thin sections were cut at 15--25 μm using a Leica Microsystems VT1000 vibratome.
Cyclopamine manipulation of the hedgehog pathway
------------------------------------------------
From a single brood of 24 individuals, 14 *C. afra*embryos were treated with cyclopamine (LC Laboratories) compound (50 μM) from a stock (5 mM cyclopamine in DMSO) to make up a final 1% DMSO solution in fish water. Five *C. afra*individuals were used as a 1% DMSO control, under the same incubation conditions as the treated embryos (Figure [7a](#F7){ref-type="fig"} and [7c](#F7){ref-type="fig"}). A further five individuals were kept as standard controls (wild-type), developing in the Georgia Institute of Technology aquarium. Treatment and control experiments were performed in ventilated Petri dishes spinning at 28°C in an oscillating platform culture incubator (Barnstead Lab-Line Max 4000). Following the treatment experiments and for the controls with DMSO, fishes were washed 10 times in fresh fish water to remove any remnant of cyclopamine compound or DMSO before transferring to culture vessels containing at least 300 ml of fish water, changed daily until ready for analysis. Although initial experiments with 50 μM cyclopamine using 1% (of 95%) ethanol as the solvent (suggested by the manufacturer, LC Laboratories and previous reports \[[@B51],[@B61]\]) showed differential expression patterns of *shh*to the 1% ethanol control experiments, alizarin red preparation of embryos raised to 12 dpf showed gross phenotypic effects on the ethanol-administered controls. Therefore, we substituted 1% DMSO for ethanol solvent, after which controls could not be distinguished from standard controls (untreated). While DMSO is not the best solvent for cyclopamine because of limited solubility above concentrations of 4 mg/ml, at the low concentrations used for enhanced viability of treated embryos, DMSO proved to be a better solvent than ethanol because of lower solvation temperatures and faster solvation times from -20°C storage temperatures.
Authors\' contributions
=======================
GJF and JTS designed the study. GJF carried out the analyses. RFB performed the cyclopamine treatment experiments. All authors contributed to the preparation of the manuscript and read and approved the final version.
Acknowledgements
================
We thank Craig Albertson, Marty Cohn, Anthony Graham, Darrin Hulsey, Moya Smith, and three anonymous reviewers for comments on previous drafts of the manuscript. Alizarin red-prepared fish (Figure [1](#F1){ref-type="fig"}) were collected from Lake Malawi and stained by Darrin Hulsey. Cichlid *pitx2*was cloned by Keen Wilson, University of Georgia, Athens, GA, USA. Research is supported by grants, from the Petit Institute for Bioengineering and Bioscience (IBB 1241318), the NIH (DE 017182), and the Alfred P Sloan Foundation (BR-4499), to JTS. RFB was a GIT Presidential Undergraduate Research Fellow.
| {
"pile_set_name": "PubMed Central"
} |
1. Introduction {#sec1-ijerph-16-03584}
===============
Today's adolescents are generally considered as the digital generation \[[@B1-ijerph-16-03584]\]. They have grown up using mobile phones, which have become an important part of their life and have reshaped their social life and behavior. Although mobile phones have many advantages such as the convenience of searching information, researchers have expressed concern regarding the potential negative effects of problematic mobile phone use, such as depression, anxiety, sleep disturbance, technostress, and poor academic performance \[[@B2-ijerph-16-03584],[@B3-ijerph-16-03584]\]. In this regard, Korean adolescents have been found to be at considerable risk of mobile phone addiction. A recent national survey in South Korea in 2018 revealed that 29.3% of adolescents were dependent on mobile phones \[[@B4-ijerph-16-03584]\].
Various terms have been used to describe varying degrees of potential problems related to uncontrolled mobile phone use; the terms of problematic mobile phone use \[[@B5-ijerph-16-03584]\], mobile phone dependence \[[@B6-ijerph-16-03584]\], and mobile phone addiction \[[@B7-ijerph-16-03584]\] are used interchangeably. Although researchers have not reached a consensus regarding the definition of mobile phone addiction, potential indicators of mobile phone addiction include preoccupation with one's mobile phone, conflicts with one's family members resulting from the excessive use of the mobile phone, use of the mobile phone to handle changes in mood, and feeling of unease when mobile phone use is inhibited \[[@B7-ijerph-16-03584]\].
Researchers have identified gender differences in mobile phone use and mobile phone addiction \[[@B1-ijerph-16-03584],[@B6-ijerph-16-03584],[@B8-ijerph-16-03584]\]. Several studies have reported that female youth are likely to spend more time on their mobile phones than male youth \[[@B6-ijerph-16-03584],[@B8-ijerph-16-03584]\], suggesting that females are potentially more vulnerable to mobile phone addiction than males \[[@B8-ijerph-16-03584]\]. In addition, gender differences have been found in the patterns of mobile phone usage. For example, female adolescents tend to use their mobile phones for text messaging, social media, for playing online games and for other forms of entertainment \[[@B6-ijerph-16-03584],[@B8-ijerph-16-03584]\].
As regards the association between problematic mobile phone addiction and mental health issues, mobile phone addiction is consistently linked to depression \[[@B9-ijerph-16-03584]\]. Behavioral addiction such as internet addiction and mobile phone addiction during adolescence can be understood with a developmental psychopathology framework \[[@B10-ijerph-16-03584]\]. Adolescents go through the rapid development of emotional, social, and psychological changes, and, in particular, impulsive tendency is demonstrated during this time. Such impulsive behavior of the adolescent is intensified if the self-control is limited. This adolescent's limited ability for self-control may be more vulnerable and susceptible to addictive behavior \[[@B11-ijerph-16-03584]\].
Most cross-sectional studies on mobile phone addiction have yielded inconclusive findings, but a few longitudinal studies have shown a bidirectional association between mobile phone addiction and depression \[[@B12-ijerph-16-03584],[@B13-ijerph-16-03584]\]. For instance, Jun \[[@B12-ijerph-16-03584]\] investigated reciprocal effects between mobile phone addiction and depression using autoregressive cross-lagged modeling and determined the bidirectional associations between mobile phone addiction and depression over time. Specifically, significant reciprocal relationships were found over a period of three years, and both mobile phone addiction and depression became severe with time. On the other hand, Coyne Stockdale, and Summers \[[@B14-ijerph-16-03584]\] tested a bidirectional longitudinal model representing the relationship between depression and problematic mobile phone use and did not find any bidirectional relationship between them.
Considering these mixed findings on the directional relationships between mobile phone addiction and depression, we aimed to fill the gaps found in previous studies in a number of ways. In particular, the present study expanded on the earlier study by Jun \[[@B12-ijerph-16-03584]\], which showed the reciprocal association between mobile phone addiction and depression using the same Korean Children and Youth Panel Survey (KCYPS) data. First, this study included mobile phone use data, which were not included in Jun's study \[[@B12-ijerph-16-03584]\], and examined the longitudinal associations among mobile phone use, mobile phone addiction, and depression. Several studies have found relationships between the frequency or duration of mobile phone use and mobile phone addiction \[[@B15-ijerph-16-03584],[@B16-ijerph-16-03584]\], suggesting that an increase in mobile phone use causes mobile phone addiction. Second, the present study incorporated important factors related to family (i.e., parents' education and income) and school (i.e., academic activity and relationships with peers), which were not controlled in Jun's study \[[@B12-ijerph-16-03584]\]. Jun \[[@B12-ijerph-16-03584]\] mentioned that one of the limitations of his study was that it did not include relevant variables associated with mobile phone addiction and depressive symptoms. Third, considering that several studies found gender differences in various relationships, this study further investigated the role of gender in the longitudinal relationships among mobile phone use, mobile phone addiction, and depression in Korean youth.
Therefore, the purposes of this study were as follows: (1) to compare differences between Korean boys and girls regarding the mean scores of mobile phone use, mobile phone addiction, and depression across three time points, and (2) to investigate gender differences in the longitudinal relationships among mobile phone use, mobile phone addiction, and depression across different time periods.
2. Present Study {#sec2-ijerph-16-03584}
================
The structural model used in this study is displayed in [Figure 1](#ijerph-16-03584-f001){ref-type="fig"}. The hypotheses are as follow:
*The mean values of mobile phone use, mobile phone addiction, and depression for Korean girls will be higher than those for Korean boys in the second year of middle school (Time 1), first year of high school (Time 2), and third year of high school (Time 3).*
*The longitudinal relationships among mobile phone use, mobile phone addiction, and depression will change from Time 1 to Time 3, and the strengths of these changes will differ between Korean boys and girls.*
H2a. There will be contemporaneous effects of mobile phone use on mobile phone addiction at each time point (Path a, Path b, and Path c).
H2b. There will be first-order autoregressive effects of mobile phone use (Paths d and e), mobile phone addiction (Path f and Path g), and depression (Path h and Path i) across time periods.
H2c. There will be second-order autoregressive effects of mobile phone use, mobile phone addiction, and depression (Path j, Path k, and Path l) from Time 1 to Time 3.
H2d. There will be cross-lagged effects between mobile phone use and mobile phone addiction (Path m, Path n, Path o, and Path p) and between mobile phone addiction and depression (Path q, Path r, Path s, and Path t) across time periods.
H2e. There will be bidirectional relationships between mobile phone addiction and depression at each time point (Path u, Path v, and Path w).
H2f. There will be gender differences in the hypotheses (H2a\~H2e) stated previously.
3. Materials and Methods {#sec3-ijerph-16-03584}
========================
3.1. Data Source and Sample {#sec3dot1-ijerph-16-03584}
---------------------------
This study used the second panel of the KCYPS data collected by the National Youth Policy Institute ([www.nypi.re.kr/archive](www.nypi.re.kr/archive)) from 2010 to 2016. The KCYPS is a nationally representative longitudinal panel survey that repeatedly collects data of three cohorts of Korean children and adolescents. The sampling frame of the KCYPS was based on stratified multistage cluster sampling. Specific data collection procedures used in the KCYPS have been described in the literature \[[@B17-ijerph-16-03584]\].
The present study used the middle school student cohort data from the second wave (the year of 2011, 2nd grade in middle school, 14-year-old, Time 1), fourth wave (the year of 2013, 16-year-old, 1st grade in high school, Time 2), and sixth wave (the year of 2015, 3rd grade in high school, 18-year-old, Time 3) of the KCYPS. The observed sample at Time 1 consisted of a total of 2280 adolescents (1152 boys and 1128 girls). We included only those adolescents who had owned mobile phones in each time period when the data were collected. A final sample was 1794 adolescents (897 boys and 897 girls) \[[@B17-ijerph-16-03584]\].
This study was exempt from the Institutional Review Board (IRB) for the protection of human subjects (IRB No: 1040647-201810-HR-006).
3.2. Measures {#sec3dot2-ijerph-16-03584}
-------------
Mobile Phone Use. The KCYPS assessed mobile phone use using a set of nine items \[[@B17-ijerph-16-03584]\]. Items assessed the frequency of mobile phone use by asking the respondents how often they used their mobile phones, such as 'How often do you use your mobile phone to talk to your family?' 'How often do you use your mobile phone to text or message your friends?' 'How often do you use your phone to use game applications?' Items were rated on a four-point Likert scale ranging from 1 (not at all) to 4 (very frequently). All the items required reverse coding, higher scores indicated higher levels of mobile phone use. The alpha coefficients for this measure in this study were 0.690 (Time 1), 0.703 (Time 2), and 0.706 (Time 3).
Mobile Phone Dependence. Mobile phone addiction was assessed according to a seven-item version of the Inventory for Mobile Phone Dependency (IMPD) \[[@B18-ijerph-16-03584]\]. It examines how much the respondents were addicted to their mobile phones, for instance: 'I feel anxious without my mobile phone when I go out of the house,' 'I lose track of time when using my mobile phone,' and 'I cannot live without my mobile phone.' The responses were rated on a four-point Likert scale ranging from 1 (never true) to 4 (always true), and all the items required reverse coding. Higher scores indicated greater levels of mobile phone dependency. Cronbach's alpha coefficients in this study were 0.898 (Time 1), 0.884 (Time 2), and 0.867 (Time 3).
Depression. Depression was measured according to a shorter 10-item version of the Korean Manual of Symptom Checklist (KMSC) \[[@B19-ijerph-16-03584]\] such as 'I have little energy,' 'I feel unhappy and depressed,' 'I have many worries,' and 'Everything is overwhelming for me.' These items were rated on a four-point Likert scale ranging from 1 (never true) to 4 (always true), and all the items were reverse-scored. Higher scores indicated more severe depression. The alpha coefficients in this study were 0.901 (Time 1), 0.888 (Time 2), and 0.880 (Time 3).
Covariates. Covariates included factors related to family (i.e., two-income family, both parents living in the household, father's education, and mother's education, annual household income) and school (i.e., academic activity, compliance with school rules, and youth relationships with peers and teachers). Gender was used as a moderator.
The factor "two-income family" was dichotomized into two categories: both parents were working (1) and at least one parent was not working (0). The factor "both parents living in the household" was assessed by asking respondents whether or not they were living with both their parents, the responses were scored as either 1 (yes) or 0 (no). The annual household income was categorized into four categories as follow: 1 = less than \$30,000, 2 = \$30,000 to \$40,000, 3 = \$40,000 to \$55,000, and 4 = greater than \$55,000. The educational level of parents was scored as follows: 1 = less than high school, 2 = high school graduate, 3 = two-year college graduate, and 4 = four-year college graduate or above.
The School Adjustment Inventory was composed of 19 items with four subscales: academic activity, compliance with school rules, relationships with peers, and relationships with teachers \[[@B17-ijerph-16-03584]\]. The subscale "relationships with peers" consisted of four items, whereas the other three subscales included five items. The responses were rated on a four-point Likert scale ranging from 1 (never true) to 4 (always true). Higher scores indicated a greater level of school adjustment. Cronbach's alpha coefficients for the four abovementioned subscales in this study were 0.696, 0.783, 0.677, and 0.835, respectively.
3.3. Strategy for Data Analyses {#sec3dot3-ijerph-16-03584}
-------------------------------
The family and school factors were descriptively examined for Korean boys and girls and compared using *t*-tests and chi-square tests. The *t*-tests were used to compute group mean comparisons of mobile phone use, mobile phone addiction, and depression in boys and girls at each time point.
A measurement invariance model was tested to ensure the metric equivalence of the three factors between boys and girls. Then, a multiple group structural equation modeling (MGSEM) approach was performed to compare gender group differences in unstandardized coefficients for the paths in [Figure 1](#ijerph-16-03584-f001){ref-type="fig"} \[[@B20-ijerph-16-03584]\]. First, an unconstrained model without equality constraints was estimated across the entire set of boys and girls. Second, a constrained model with equality constraints was separately estimated across groups. Then, a nested chi-square difference test was performed to compare the constrained and unconstrained models. Model fit was estimated using the following indices: the overall chi-square test (*p* \> 0.05), the comparative fit index (CFI \> 0.95), the root mean square error of approximation (RMSEA \< 0.05), and the standardized root mean residual (SRMR ≤ 0.08) \[[@B21-ijerph-16-03584]\]. Missing data were dealt with using a full-information maximum likelihood approach \[[@B22-ijerph-16-03584]\]. Mplus 8.1 \[[@B23-ijerph-16-03584]\] and SPSS version 24.0 \[[@B24-ijerph-16-03584]\] were used for data analyses.
4. Results {#sec4-ijerph-16-03584}
==========
4.1. Descriptive Analyses {#sec4dot1-ijerph-16-03584}
-------------------------
[Table 1](#ijerph-16-03584-t001){ref-type="table"} presents the group comparisons of family and school factors between Korean boys and girls in the second year of middle school (Time 1). Some significant differences between boys and girls were found for the school factor but not for the family factor. Among the school factors, girls reported higher levels of compliance with school rules (*t* = 2.94, *p* \< 0.01) and relationships with peers (*t* = 3.80, *p* \< 0.001).
4.2. Group Comparisons {#sec4dot2-ijerph-16-03584}
----------------------
[Table 2](#ijerph-16-03584-t002){ref-type="table"} lists the group mean comparisons of mobile phone use, mobile phone addiction, and depression between boys and girls. Statistically significant group differences were found in the means of mobile phone use, mobile phone addiction, and depression at Times 1, 2, and 3. Mobile phone use for girls was higher than that for boys across time periods (Time 1: *t* = 7.25, *p* \< 0.001, Time 2: *t* = 5.44, *p* \< 0.001, Time 3: *t* = 4.94, *p* \< 0.001). Similarly, girls showed higher mobile phone addiction than boys at each time point (Time 1: *t* = 8.20, *p* \< 0.001, Time 2: *t* = 6.49, *p* \< 0.001, Time 3: *t* = 5.80, *p* \< 0.001). The means of the depression for girls were also higher than those for boys over time (Time 1: *t* = 6.12, *p* \< 0.001, Time 2: *t* = 7.71, *p* \< 0.001, Time 3: *t* = 7.50, *p* \< 0.001). Therefore, hypothesis H1 was supported.
4.3. Structural Equation Model {#sec4dot3-ijerph-16-03584}
------------------------------
Measurement invariance tests were performed for mobile phone use, mobile phone addiction, and depression among boys and girls using multiple-group confirmatory factor analyses. Among the 69 group comparisons, 15 paths linking from the unobserved variables to the observed indicators were statistically significant. These links represented 21.7% of all the group contrasts, which indicated that the path coefficients for the three factors largely measured the mobile phone use, mobile phone addiction, and depression in a similar manner between boys and girls. Thus, this study used latent composite variables to examine the structural model shown in [Figure 1](#ijerph-16-03584-f001){ref-type="fig"}.
To test hypothesis 2, the structural model of mobile phone use, mobile phone addiction, and depression in boys and girls were tested using an MGSEM strategy with a Huber--White maximum likelihood estimator. The model fit was good (χ^2^(df = 26) = 33.826, *p* \> 0.05, CFI = 0.998, RMSEA = 0.018, SRMR = 0.008, 90% C.I. = (0.000, 0.034)). Focused fit tests revealed no theoretically meaningful significant points of ill fit. [Figure 2](#ijerph-16-03584-f002){ref-type="fig"} shows the unstandardized path coefficient comparisons between boys and girls. For each of the two groups, statistically significant path coefficients were found for the contemporaneous effects of mobile phone use on mobile phone addiction at Times 1, 2, and 3 for boys (path coefficients = 0.397, 0.314, and 0.245, respectively) and girls (path coefficients = 0.448, 0.353, and 0.318, respectively). Therefore, hypothesis H2a was supported.
For the first-order autoregressive effects of mobile phone use, statistically significant effects were found for boys (Time 1 to Time 2: path coefficient = 0.221, Time 2 to Time 3: path coefficient = 0.286) and girls (Time 1 to Time 2: path coefficient = 0.299, Time 2 to Time 3: path coefficient = 0.353). Similarly, the first-order autoregressive effects of mobile phone addiction were observed for boys (Time 1 to Time 2: path coefficient = 0.318, Time 2 to Time 3: path coefficient = 0.356) and girls (Time 1 to Time 2: path coefficient = 0.345, Time 2 to Time 3: path coefficient = 0.370). For depression, the predicted first-order autoregressive effects from Time 1 to Time 2 and from Time 2 to Time 3 were observed for boys (path coefficients = 0.292 and 0.449) and Korean girls (path coefficients = 0.338 and 0.397). Additionally, the second-order autoregressive effects of mobile phone use, mobile phone addiction, and depression were found from Time 1 to Time 3 for boys (path coefficients = 0.183, 0.167, and 0.120) and girls (path coefficients = 0.096, 0.147, and 0.164). Therefore, hypotheses H2b and H2c were supported.
As for cross-lagged effects, statistically significant path coefficients were found between mobile phone use at Time 1 and mobile phone addiction at Time 2 for boys (path coefficient = 0.072) and between mobile phone use at Time 2 and mobile phone addiction at Time 3 for boys (path coefficient = −0.126) and girls (path coefficient = −0.139). However, no reciprocal causal relationships were found between mobile phone use and mobile phone addiction across time periods for both groups. Statistically significant path coefficients were found from mobile phone addiction at Time 1 to depression at Time 2 for girls (path coefficient = 0.079) and from mobile phone addiction at Time 2 to depression at Time 3 for boys (path coefficient = 0.106). Reciprocal causal relationships linking depression at Time 1 to mobile phone addiction at Time 2 were not observed, but reciprocal causal relationships linking depression at Time 2 to mobile phone addiction at Time 3 were found for boys (path coefficient = 0.060) and girls (path coefficient = 0.078). Therefore, hypothesis H2d was partially supported.
For bidirectional relationships, statistically significant path coefficients were found between mobile phone addiction and depression at each time point for boys (path coefficients = 6.917, 6.003, and 3.942) and girls (path coefficients = 6.396, 6.036, and 3.612). Therefore, hypothesis H2e was supported.
Finally, there were no gender differences for each path in the structural model. Thus, hypothesis H2f was not supported.
5. Discussion {#sec5-ijerph-16-03584}
=============
This study aimed to investigate the gender differences in the mean scores of mobile phone use, mobile phone addiction, and depression among Korean youth and to compare differences in the longitudinal relationships of these three variables among boys and girls for a four-year period. The following four main points are discussed: (1) gender differences in the mean values of mobile phone use, mobile phone addiction, and depression, (2) longitudinal effects of mobile phone use, mobile phone addiction, and depression, (3) bidirectional relationships between mobile phone addiction and depression, and (4) gender differences in the structural model.
First, the mean values of mobile phone use, mobile phone addiction, and depression were differed by gender. Compared to boys, Korean girls were likely to use their mobile phones more and were at a higher risk of mobile phone addiction and depression during the second year of middle school (Time 1), the first year of high school (Time 2), and the third year of high school (Time3). These findings were consistent with those of a previous study by Sánchez-Martínez and Otero \[[@B6-ijerph-16-03584]\], which reported that female adolescents tended to use their mobile phones more often and were more likely to depend on mobile phone devices than their male counterparts. Similarly, girls were more likely to have depression than boys \[[@B25-ijerph-16-03584]\].
With regard to the autoregressive effects of mobile phone use, mobile phone addiction, and depression over time, the present study found that early mobile phone use, mobile phone addiction, and depression during middle school were associated with later mobile phone use, mobile phone addiction, and depression during high school. Frequent mobile phone users at middle school showed the tendency to use their mobile phones frequently even in high school. Similarly, Korean adolescents who showed higher levels of mobile phone addiction during the junior high school period were more likely to depend on their mobile phones during the high school period. This finding was consistent with previous studies \[[@B14-ijerph-16-03584],[@B26-ijerph-16-03584]\], which indicated that mobile phone addiction tended to increase over time. As for depression, Korean adolescents with higher depression during the early adolescence period were at a higher risk of developing depression during the late adolescence period, which was consistent with the results of a previous study \[[@B25-ijerph-16-03584]\].
Regarding the cross-lagged effects of mobile phone use, mobile phone addiction, and depression, the present study found that the frequency of mobile phone use could not predict mobile phone addiction at a later time period. However, mobile phone addiction and depression concurrently predicted each other across time periods. First, frequent mobile phone users in middle school were less likely to addict to their mobile phones in high school. This finding was contrary to our hypothesis. The inconsistent finding might be attributable to the use of different measurement tools. Haug et al. \[[@B27-ijerph-16-03584]\] found that smartphone addiction was more strongly associated with the total time spent per day using smartphones than the frequency of smartphone use. In contrast, Lin et al. \[[@B28-ijerph-16-03584]\] reported that the frequency of mobile phone use was an indicator of smartphone addiction rather than the total time spent using mobile phones. These two studies \[[@B27-ijerph-16-03584],[@B28-ijerph-16-03584]\] used both the frequency of mobile phone use and the total time spent using mobile phones and mainly carried out cross-sectional analysis. The present study did not include the variable of the total time spent using mobile phones but longitudinally analyzed the relationship between the frequency of mobile phone use and mobile phone addiction. Therefore, further longitudinal study is warranted to examine the relationships among the total time spent using mobile phones, the frequency of mobile phone use, and mobile phone addiction. Second, earlier depression in the first year of high school (Time 2) was associated with later mobile phone addiction in the third year of high school (Time 3). This finding supports that of a previous study, which indicated that depression was a predictor of mobile phone addiction \[[@B29-ijerph-16-03584]\]. Third, this present study found that earlier mobile phone addiction was associated with later depression. This finding, that supported the study by Coyne et al. \[[@B14-ijerph-16-03584]\], indicated that early problematic mobile phone use was a significant predictor of depression.
This study confirmed that the bidirectional relationship between mobile phone addiction and depression persisted even after the other variables were controlled. This finding is consistent with Jun's study \[[@B12-ijerph-16-03584]\]. There are some possible explanations for this causal pathway. Adolescents with depression are more likely to be addicted to their mobile phones because mobile phones may provide an environment where such adolescents can relate to others in a safer and less demanding environment \[[@B6-ijerph-16-03584]\]. Inversely, a possible explanation of why mobile phone addiction may lead to depression is that adolescents with higher levels of mobile phone addiction might be at an increased risk of interpersonal problems, resulting in depression later \[[@B30-ijerph-16-03584]\]. This finding highlights the importance of providing simultaneous intervention for reducing both depression and mobile phone addiction.
Finally, contrary to our expectation, the present study found no gender differences for each path in the structural model. A plausible explanation for this finding is that the sample considered in this study consisted of middle and high school Korean students. In particular, high school students have less time to use their mobile phones than young adults because high school students need to prepare for university entrance exams. This commonality of adolescent life in South Korea may result in there being no gender differences. Although the present study analyzed data longitudinally, four years of research might not be sufficient to investigate gender differences in the relationships among mobile phone use, mobile phone addiction, and depression. Prior literature reported that gender was an important predictor in the relationship between the internet or mobile phone use and mental health outcomes but these studies were cross-sectional \[[@B1-ijerph-16-03584],[@B31-ijerph-16-03584]\]. Few studies have explored the causal effects of gender on this issue, thus, more studies are needed to conduct with longitudinal data.
This study has some limitations. First, considering the nature of the secondary analysis, the present study was limited by the measurement of mobile phone use. The KCYPS did not measure the actual time and frequency of mobile phone use, which may yield different results regarding the relationship between mobile phone use and mobile phone addiction. A further longitudinal study is warranted to examine the relationships among the total time spent using mobile phones, the frequency of mobile phone use, and mobile phone addiction. Second, our study did not distinguish between different types of mobile phone devices but treated them (from the most advanced smartphones to standard cell phones) as the same. Thus, the findings of this study should be interpreted with caution. Future research is warranted to investigate whether the types of mobile phone devices are related to adolescents' patterns of mobile phone use. Third, although we found significant associations among mobile phone use, mobile phone addiction, and depression over time in this study, causal inferences among these variables are limited due to the inconsistent results on this topic. Lastly, although this study included parents' socio-economic variables such as income and education, it did not include parents' psychological status such as mobile phone use of parents, which may affect their children's mobile phone use. Future research is needed to examine the effect of parents' psychological status on their children's mobile phone use and mobile phone addiction.
6. Conclusions {#sec6-ijerph-16-03584}
==============
In summary, excessive mobile phone use and mobile phone addiction have been increasing concerns among Korean adolescents. Findings of this study revealed that Korean girls were more exposed to mobile phone use and they were at higher risk of mobile phone addiction and depressive symptoms. In addition, the longitudinal relationships among mobile phone use, mobile phone addiction, and depressive symptoms were observed in Korean adolescents. This study implies that it is necessary to acknowledge the negative effects of mobile phone use and to design adequate intervention strategies to prevent mobile phone addiction and depressive symptoms among adolescents.
Conceptualization, S.-Y.P. (So-Young Park) and S.-Y.P. (So-Youn Park); methodology, S.-Y.P. (So-Young Park) and H.J.; software, S.-Y.P. (So-Young Park) and S.-Y.P. (So-Youn Park); validation, S.Y., H.J., and C.-S.S.; formal analysis, S.-Y.P. (So-Young Park) and S.-Y.P. (So-Youn Park); investigation, C.-S.S.; resources, S.-Y.P. (So-Young Park), S.Y., H.J., and S.-Y.P. (So-Youn Park); data curation, S.-Y.P. (So-Young Park) and S.-Y.P. (So-Youn Park); writing-original draft preparation, S.-Y.P. (So-Young Park), S.Y. and S.-Y.P. (So-Youn Park); writing-review and editing, C.-S.S., and H.J.; visualization, S.-Y.P. (So-Young Park); supervision, S.Y., C.-S.S., and S.-Y.P. (So-Youn Park); project administration, S.-Y.P. (So-Youn Park); funding acquisition, S.-Y.P. (So-Youn Park).
This study was supported by National Research Foundation of Korea (2017 Grant Number NRF-2017S1A5B8057479.
The authors declare no conflict of interest.
{#ijerph-16-03584-f001}
{#ijerph-16-03584-f002}
ijerph-16-03584-t001_Table 1
######
Group comparisons of family and school factors between boys and girls at Time 1 (*N* = 1794).
Variable Boys (n = 897) Girls (n = 897) *t*/χ^2^
------------------------------------- ---------------- ----------------- -------------- ------------- ------
**Family factor**
Two-income Family
Yes 492 (54.8) 509 (56.7) 0.65
No 405 (45.2) 388 (43.3)
Both parents in household
Yes 766 (87.7) 772 (89.9) 1.97
No 107 (12.3) 87 (10.1)
Annual household income (\$)
0--30,000 283 (33.4) 258 (30.7) 5.31
30,000--40,000 193 (22.8) 177 (21.1)
40,000--55,000 168 (19.8) 204 (24.3)
≥55,000 203 (24.0) 201 (23.9)
Father's education
Less than high school 28 (3.5) 29 (3.6) 1.06
High school graduate 356 (44.2) 338 (42.1)
Two-year college graduate 76 (9.5) 85 (10.6)
Four-year college graduate or Above 344 (42.8) 351 (43.7)
Mother's education
Less than high School 22 (2.7) 29 (3.5) 2.01
High school graduate 463 (57.7) 455 (55.3)
Two-year college graduate 83 (10.3) 80 (9.7)
Four-year college graduate or Above 235 (29.3) 259 (31.5)
**School factor**
Learning activity 13.64 ± 2.64 13.72 ± 2.46 0.68
Compliance with school rules 13.81 ± 2.87 14.18 ± 2.54 2.94 \*\*
Youth relationships with peers 12.11 ± 2.04 12.45 ± 1.73 3.80 \*\*\*
Youth relationships with teachers 14.12 ± 3.34 13.95 ± 3.24 1.06
Note: \*\*\* *p* \< 0.001, \*\* *p* \< 0.01. *M* = mean, *SD* = standard deviation.
ijerph-16-03584-t002_Table 2
######
Group mean comparisons of major variables between boys and girls (*N* = 1794).
Variable Boys (n = 897) Girls (n = 897) *t*
---------------------------------------------------- ---------------- ----------------- -------------
2nd Grade of middle school at Time 1 (14-year-old)
Mobile phone use 27.68 ± 4.27 29.04 ± 3.63 7.25 \*\*\*
Mobile phone addiction 15.07 ± 4.97 17.06 ± 5.32 8.20 \*\*\*
Depression 18.36 ± 5.94 20.09 ± 6.02 6.12 \*\*\*
1st Grade of high school at Time 2 (16-year-old)
Mobile phone use 29.43 ± 4.08 30.40 ± 3.50 5.44 \*\*\*
Mobile phone addiction 15.98 ± 4.53 17.42 ± 4.89 6.49 \*\*\*
Depression 17.79 ± 5.34 19.78 ± 5.60 7.71 \*\*\*
3rd Grade of high school at Time 3 (18-year-old)
Mobile phone use 29.57 ± 4.01 30.47 ± 3.70 4.94 \*\*\*
Mobile phone addiction 15.68 ± 4.51 16.91 ± 4.46 5.80 \*\*\*
Depression 17.75 ± 5.41 19.65 ± 5.34 7.50 \*\*\*
Note: \*\*\* *p* \< 0.001. *M* = mean, *SD* = standard deviation.
| {
"pile_set_name": "PubMed Central"
} |
See related research by Phillips *et al.*, <http://ccforum.com/content/13/2/R30>
In the present issue of *Critical Care*an experimental model of hemorrhagic shock in pigs describes a significant difference in the extent of lung injury as assessed by the extravascular lung water (EVLW) between two crystalloid solutions -- normal saline (NS) or Ringer\'s lactate (RL) -- when administered in volumes \>250 ml/kg \[[@B1]\]. In general, which type of fluid to use in the resuscitation from hemorrhagic shock is still a matter of debate. In this context, with respect to developing organ dysfunction, early detection of lung injury during resuscitation from hemorrhagic shock is widely considered of particular clinical importance. For these purposes, the transpulmonary thermodilution technique -- which enables one to assess the extent of fluid in the interstitial space of the lungs (EVLW) by describing the relation between intravascular fluid and extravascular fluid (that is, pulmonary capillary permeability) -- is clinically applied. Especially in the scenario of fluid shift during resuscitation in hemorrhagic shock, the type of fluid may have particular influence on endothelial function in the lungs.
The study of Phillips and colleagues attempted to mimic severe hemorrhagic shock in both the prehospital and early-hospital periods in humans \[[@B1]\]. Studies examining the effects of NS versus RL on hemodynamic response have so far been conducted in controlled hemorrhage models, and many studies reinfused shed blood with the resuscitation fluid. In contrast, this study used a more clinically relevant model by adding tissue injury to uncontrolled hemorrhage and initiating early resuscitation with crystalloids alone and resuscitating to a goal blood pressure. As the authors themselves mention, however, their study has several limitations that make the interpretation of the effects of the fluid type independent of the volume difficult. Nevertheless, the results probably allow more reliable extrapolation to the human clinical scenario than previous studies.
First, total blood loss was greater in the NS group than in the RL group, which may impact on the study findings. Furthermore, although animals in the NS received significantly more fluid, the mean arterial pressure was significantly lower in the resuscitation phase when compared with the RL group. At a time point during resuscitation when differences between both groups in EVLW first became significant, however, a higher capillary permeability as assessed by the index parameter pulmonary capillary permeability was already observed. Significantly higher pulmonary capillary permeability in the presence of a comparable filling volume implies either greater extravasation of fluid into the lung due to increases in permeability, or changes in the transcapillary oncotic pressure gradient, or an impairment of fluid clearance, or a combination of the three.
In general, the various factors for transcapillary fluid flux have been described in the Starling equation. Since this occurred at a time when there were no differences in the volumes of resuscitation administered or in the central filling volumes, however, this finding suggests that NS may have caused a pulmonary capillary endothelial permeability injury relative to the RL group to explain the differences in EVLW. It should be mentioned that nobody has previously examined EVLW, oxygenation, and hemodynamic effects in an uncontrolled trauma-related hemorrhagic shock model.
With respect to the monitoring technology used, a clinical study compared NS with different colloids and found that pulmonary edema (that is, EVLW as measured by trans-pulmonary thermodilution) and the lung injury score were not affected by the type of fluid loading in fluid responsiveness in both septic patients and nonseptic patients \[[@B2]\]. van der Heijden and colleagues, however, did not compare different crystalloids and did not involve individuals with a different underlying pathophysiology \[[@B2]\]. Phillips and colleagues present evidence that the difference in EVLW between NS versus RL may have been due to increased pulmonary vascular permeability \[[@B1]\] -- surprisingly, there is little in the literature describing this effect. Further pathophysiological mechanisms to explain the authors\' findings are peripheral vasodilatation by NS and acidosis, which itself may cause vasodilation and impair endothelial integrity \[[@B3]\].
Interestingly, oxygenation (PaO~2~to FiO~2~ratio) was not significantly different between both groups while EVLW was. Previous clinical and experimental studies showed that transpulmonary thermodilution-derived EVLW is sufficiently accurate when compared with gravimetry or with the double-indicator dilution technique \[[@B4]-[@B6]\]. Consequently, to explain this difference and to exclude limitations by the technique itself, a reference technique -- such as gravimetry, which would have been easy to apply in an animal study -- would have been useful. The authors adequately mentioned in their discussion the following possible explanations for the differences observed between NS and RL: inflammation, oncotic pressure (which unfortunately was not measured) and different influences on coagulation and platelet activation. Histological examinations that were obtained by the same group in the same model \[[@B7]\], however, revealed that increased numbers of sequestered neutrophils in the lung between NS-resuscitated and RL-resuscitated animals were not different. Finally, ventilator settings that may have influenced the study results were discussed. Although clinically not recommended and hopefully not applied, an identical tidal volume of 12 ml/kg was used in both groups. According to the authors, this tidal volume may have caused some lung injury and may have played a role in the development of increased EVLWI; however, using the same tidal volume made it unlikely as the cause of the difference between the groups.
Further, more sophisticated studies are required to validate and explain these findings. The authors are to be congratulated for completing this first phase of a promising line of investigation. Future research should further clarify the clinical significance of different fluids on pulmonary capillary function in patients with hemorrhagic shock.
Abbreviations
=============
EVLW: extravascular lung water; NS: normal saline; RL: Ringer\'s lactate.
Competing interests
===================
SGS is a member of the medical advisory board of Pulsion Medical Systems AG, Munich, Germany, and received honoraria for presenting lectures.
| {
"pile_set_name": "PubMed Central"
} |
1. Introduction {#sec1-molecules-25-01319}
===============
The Thomsen-Friedenreich (TF, β-[d]{.smallcaps}-Gal*p*-(1,3)-α-[d]{.smallcaps}-GalNAc*p*) antigen is a tumor associated carbohydrate antigen (TACA) that has significant roles in the progression of carcinomas of the breast, colon, prostate, liver, and more \[[@B1-molecules-25-01319]\]. High expression levels of the TF antigen on tumor cells is positively correlated to poor prognosis, and an increased ability to metastasize \[[@B2-molecules-25-01319],[@B3-molecules-25-01319]\]. The relationship between TF expression and metastasis has been definitively shown to be mediated through galectin-3 on endothelial cells, where selective pressure to non-TF expressing tumor populations can decrease metastasis \[[@B4-molecules-25-01319],[@B5-molecules-25-01319]\]. Furthermore, patients that were able to develop endogenous anti-TF antibodies have significantly improved prognosis \[[@B6-molecules-25-01319],[@B7-molecules-25-01319]\]. These characteristics of the TF antigen coupled to the fact that the TF antigen is not exposed on normal, healthy tissue makes the TF antigen a viable target for immunotherapy \[[@B8-molecules-25-01319],[@B9-molecules-25-01319]\]. Regardless, strategies to develop potent immune responses to the TF antigen have been thwarted by its seemingly low immunogenicity \[[@B10-molecules-25-01319],[@B11-molecules-25-01319],[@B12-molecules-25-01319],[@B13-molecules-25-01319],[@B14-molecules-25-01319],[@B15-molecules-25-01319],[@B16-molecules-25-01319]\].
Our research efforts focus on the issues of developing carbohydrate specific immunity by utilizing a semi-synthetic entirely carbohydrate vaccine platform. This two-component platform consists of an anomeric aminooxy TACA derivative conjugated to Polysaccharide A1 (PS A1) via oxime bond and has been demonstrated with the TF antigen, the Thomsen-nouveau (Tn, α-[d]{.smallcaps}-GalNAc*p*) antigen, and the sialyl-Tn (STn, α-[d]{.smallcaps}-Neu5Ac-(2,6)-α-[d]{.smallcaps}-GalNAc*p*) antigen \[[@B17-molecules-25-01319],[@B18-molecules-25-01319],[@B19-molecules-25-01319]\]. PS A1 is a zwitterionic polysaccharide (ZPS) isolated from the commensal bacteria *Bacteroides fragilis* (ATCC 25285/NCTC 9343) capable of activating CD4+ T cells in a major histocompatibility complex-II (MHC-II) dependent mechanism \[[@B20-molecules-25-01319],[@B21-molecules-25-01319]\]. PS A1 is an effective immunogenic carrier for TACAs and is evidenced by our previous work with Tn-PS A1 and STn-PS A1 \[[@B17-molecules-25-01319],[@B18-molecules-25-01319],[@B22-molecules-25-01319]\]. These PS A1 conjugates were able to induce protective immune responses that consisted of potent antibody production, cellular immunity, and the production of proinflammatory markers such as T helper 17 (Th17) cells and IL-17, which are essential in the protection against cancer \[[@B23-molecules-25-01319],[@B24-molecules-25-01319]\].
Another elegant strategy to improve carbohydrate immunogenicity involves targeting innate immune receptors on the surface of antigen presenting cells (APCs) \[[@B25-molecules-25-01319],[@B26-molecules-25-01319],[@B27-molecules-25-01319]\]. C-type lectin receptors (CLRs), such as the macrophage galactose *N*-acetylgalactosamine specific lectin 2 (MGL2, CD301b) have been investigated for their potential to influence immune responses \[[@B28-molecules-25-01319],[@B29-molecules-25-01319],[@B30-molecules-25-01319],[@B31-molecules-25-01319],[@B32-molecules-25-01319]\]. The major ligand for MGL2 is the Tn glycan, and reports suggest the Tn glycan has direct positive influence on antigen uptake mediated in an MGL2 dependent mechanism by dendritic cells (DCs) that lead to CD4+ T cell activation \[[@B33-molecules-25-01319],[@B34-molecules-25-01319],[@B35-molecules-25-01319],[@B36-molecules-25-01319],[@B37-molecules-25-01319]\]. Studies by Leclerc et al. have validated this phenomenon, demonstrating a correlation between an increased Tn density on MUC6 and other peptides to enhance antigen internalization by APCs \[[@B38-molecules-25-01319]\]. Antigen internalization was markedly increased when compared to the non-glycosylated peptides, which was believed to be a consequence of MGL2 interaction with Tn. These responses were also associated with an increase in Th2 related cytokines, Th17 related cytokines, and an increased expansion of B cell populations within germinal centers \[[@B38-molecules-25-01319],[@B39-molecules-25-01319]\]. Van Kooyk et al. have also reported increased MGL2 dependent antigen internalization and presentation of Tn containing antigens. These responses lead to Th1 polarization and increased IFN-γ production coupled with enhanced cross-presentation ultimately activating antigen specific CD4+ T cells and CD8+ T cell responses \[[@B40-molecules-25-01319],[@B41-molecules-25-01319]\].
Based on our current understanding of carbohydrate immunity, we sought to create an immunogen with the ability to increase an immune response to the TF antigen without the need to generate unnatural epitopes on the TF structure which requires additional synthetic steps and antibody cross-reactivity to the natural TF epitope. The inclusion of the Tn antigen into the TF-PS A1 conjugate may induce an "adjuvant-like" affect by recruiting key components in the development of an immune response, such as MGL2, and the subsequent activation of the adaptive immune response involving both T and B cells. Herein, we report the synthesis of the unimolecular, bivalent Tn-TF-PS A1 conjugate and its immune efficacy as compared to the TF-PS A1 and Tn-PS A1 conjugates.
2. Results and Discussion {#sec2-molecules-25-01319}
=========================
2.1. Synthesis and Characterization of TACA-PS A1 Conjugates {#sec2dot1-molecules-25-01319}
------------------------------------------------------------
The synthesis of TACA-PS A1 conjugates Tn-PS A1 (**4a**) \[[@B18-molecules-25-01319]\], TF-PS A1 (**4b**) \[[@B19-molecules-25-01319]\], and Tn-TF-PS A1 (**4c**) were achieved when sodium periodate was used to regioselectively oxidize the terminal vicinal diol on the [d]{.smallcaps}-galactofuranose moiety on the repeating unit of PS A1 using a 0.1 M acetate buffering system at a pH of 5. Periodate oxidations form a cyclic, five-membered ring intermediate in which terminal vicinal diols will undergo oxidative cleavage faster than cyclic trans-diols \[[@B42-molecules-25-01319]\]. To avoid oxidative cleavage of the furanose ring, 0.5 equivalents of sodium periodate was used per repeating unit of PS A1 to ensure regioselectivity. The resulting aldehyde was then exposed to the anomeric aminooxy derivatives of Tn (**2**), TF (**3**), and a 1:1 molar ratio of **2** and **3** to form the oxime conjugates in Tn-PS A1 (**4a**), TF-PS A1 (**4b**), and Tn-TF-PS A1 (**4c**), respectively as shown in [Scheme 1](#molecules-25-01319-sch001){ref-type="scheme"}. Purification of these conjugates was achieved through dialysis with a 10 kDa molecular weight cut off followed by lyophilization of the dialyzed material. The resulting white solid was analyzed by proton and COSY NMR spectroscopy to confirm chemical transformations which are evidenced by: (1) oxime proton signal (7.99--8.00 ppm), (2) new anomeric proton signals (\~5.77 ppm), and (3) additional *N*-acetyl peaks (2.29--2.31 ppm) (Doc. S1). An estimated percent loading can be obtained from the proton spectra of Tn-PS A1 (**4a**), TF-PS A1 (**4b**), and Tn-TF-PS A1 (**4c**) by integration of the *N*-acetyl peaks. Although this method may introduce potential error due to peak overlap, we have demonstrated this method previously \[[@B17-molecules-25-01319],[@B43-molecules-25-01319]\]. Based on the integration of *N*-acetyl peaks, percent loadings were calculated to be 13% for Tn-PS A1 (**4a**), 19% for TF-PS A1 (**4b**), and a combined loading of 18% for Tn-TF-PS A1 (**4c**) ([Table S1](#app1-molecules-25-01319){ref-type="app"}).
2.2. ELISA Reveals Enhanced Antibody Production Against the TF antigen with the Tn-TF-PS A1 Conjugate {#sec2dot2-molecules-25-01319}
-----------------------------------------------------------------------------------------------------
Conjugates Tn-PS A1, TF-PS A1, and Tn-TF-PS A1 were evaluated for their immunological potency in conjunction with either the depot adjuvant TiterMax^®^ Gold (TMG) \[[@B44-molecules-25-01319],[@B45-molecules-25-01319]\] or the monophosphoryl lipid A (MPLA) containing adjuvant Sigma Adjuvant System^®^ (SAS) \[[@B44-molecules-25-01319],[@B46-molecules-25-01319]\]. Groups of five Jax C57BL/6 mice were immunized intraperitoneally four times in biweekly intervals when using TMG or were immunized intraperitoneally three times in triweekly intervals when SAS was used. To determine antibody production against the native antigens, antibody titers were determined by enzyme-linked immunosorbent assay (ELISA) with 96-well plates coated with either glycoconjugate Tn-bovine serum albumin (Tn-BSA) or TF-BSA ([Scheme S1](#app1-molecules-25-01319){ref-type="app"}).
[Figure 1](#molecules-25-01319-f001){ref-type="fig"}a displays IgG titers as an average from individual mice against Tn-BSA. Antiserum from monovalent Tn-PS A1 immunized mice exhibited high IgG titers when SAS (red bar) and TMG (blue bar) was used as an adjuvant. The antibody response to Tn-PS A1 had increased IgG isotype populations as compared to the IgM isotype ([Figure 1](#molecules-25-01319-f001){ref-type="fig"}b). This response is an indication of T cell activation, immunological memory, and affinity maturation which are desirable outcomes in cancer immunotherapies. [Figure 1](#molecules-25-01319-f001){ref-type="fig"}c depicts IgG titers against TF-BSA and most notably, monovalent TF-PS A1 immunized mice did not showcase significant IgG binding events. The humoral response against TF-PS A1 remained exclusively within the IgM isotype ([Figure 1](#molecules-25-01319-f001){ref-type="fig"}d) and was independent of adjuvant used. The contrast in these ELISA data between conjugates Tn-PS A1 and TF-PS A1 correspond to the low immunogenicity of the native TF antigen. Individual mouse titer values obtained from immunizations in conjunction with SAS are illustrated in [Figure S1](#app1-molecules-25-01319){ref-type="app"}.
As shown in [Figure 1](#molecules-25-01319-f001){ref-type="fig"}a--d mice immunized with Tn-TF-PS A1 developed high IgG and IgM titers for both Tn-BSA and TF-BSA. The most significant result shown in [Figure 1](#molecules-25-01319-f001){ref-type="fig"}c provides evidence of an enhanced antibody production and maturation against the TF antigen. Mice immunized against TF-PS A1 did not generate significant IgG titers to the TF antigen, but high IgG titers were developed when mice were immunized with the bivalent conjugate Tn-TF-PS A1. We also detected minimal antibody cross reactivity from mice immunized with the monovalent conjugates Tn-PS A1 and TF-PS A1 which suggested that is not the likely source for the development of the IgG titers as observed in [Figure 1](#molecules-25-01319-f001){ref-type="fig"}c. These ELISA data indicated that the addition of Tn onto the TF-PS A1 construct augmented the immune response towards the TF antigen. To rule out immune activation from differential particle sizes by individual aggregation characteristics, PS A1 (1) and conjugates Tn-PS A1, TF-PS A1, and Tn-TF-PS A1 were examined with dynamic light scattering (DLS) \[[@B47-molecules-25-01319]\]. These experiments showed that conjugates Tn-PS A1, TF-PS A1, and Tn-TF-PS A1 have similar hydrodynamic radii and polydispersity indexes ([Figure S2](#app1-molecules-25-01319){ref-type="app"}).
2.3. Polyclonal Antibodies Bind Human Tumor Cell Lines {#sec2dot3-molecules-25-01319}
------------------------------------------------------
Antiserum was further evaluated with flow cytometry to determine polyclonal antibody binding to human tumor cells MCF-7 \[[@B48-molecules-25-01319]\] ([Figure 2](#molecules-25-01319-f002){ref-type="fig"}a), OVCAR-5 \[[@B49-molecules-25-01319],[@B50-molecules-25-01319]\] ([Figure 2](#molecules-25-01319-f002){ref-type="fig"}b), and normal human breast tissue MCF-10A \[[@B51-molecules-25-01319]\] ([Figure S3](#app1-molecules-25-01319){ref-type="app"}). The anti-serum of Tn-TF-PS A1 demonstrated a 97% gated-shift in fluorescently sorted cell populations compared to MCF-7 cells alone. This result is expected as the MCF-7 cell line is known to express both the Tn and the TF antigens \[[@B48-molecules-25-01319]\]. However, similar binding events were seen using human ovarian tumor cell line OVCAR-5 with antiserum from Tn-TF-PS A1 giving a 98% shift in fluorescently sorted cell populations. We also note in [Figure S3](#app1-molecules-25-01319){ref-type="app"} that there is negligible binding of antiserum from TACA-PS A1 constructs to MCF-10A.
2.4. Polyclonal Antibodies Mediate Tumor Cell Killing with Complement {#sec2dot4-molecules-25-01319}
---------------------------------------------------------------------
Antibody function was then assessed using a complement dependent cytotoxicity (CDC) assay where antiserum derived from native PS A1, Tn-PS A1, TF-PS A1, or Tn-TF-PS A1 were compared. This experiment also utilized antiserum derived from PBS to evaluate nonspecific antibody interactions with the target cells, as well as a complement control to evaluate nonspecific killing of tumor cells. In [Figure 3](#molecules-25-01319-f003){ref-type="fig"}a, antiserum from Tn-TF-PS A1 exhibited 59% cytotoxicity towards the MCF-7 cell line which was statistically significant in comparison with antiserum from TF-PS A1. Additionally, antiserum from Tn-TF-PS A1 had 53% cytotoxicity towards OVCAR-5 ([Figure 3](#molecules-25-01319-f003){ref-type="fig"}b) and was also statistically significant when compared to antiserum from TF-PS A1. Interestingly, there is no significant difference between the observed cytotoxicity from antisera derived from monovalent constructs Tn-PS A1 and TF-PS A1, which seems to be in contrast to the data in [Figure 1](#molecules-25-01319-f001){ref-type="fig"}. We hypothesized these small variations in cytotoxicity to be a result of the similar IgM responses of Tn-PS A1 and TF-PS A1, as the C1q protein in complement binds IgM Fc portions 1000-fold more than the IgG Fc portions \[[@B52-molecules-25-01319]\]. In all cases, we observed negligible cytotoxicity to the non-carcinoma MCF-10A cell line as a control ([Figure 3](#molecules-25-01319-f003){ref-type="fig"}c). Collectively, we observed that antiserum from Tn-TF-PS A1 had significant cytotoxicity over the antiserum derived from the monovalent counterpart TF-PS A1.
2.5. Quantification of Cytokines Released from Splenocytes In Vitro {#sec2dot5-molecules-25-01319}
-------------------------------------------------------------------
[Figure 4](#molecules-25-01319-f004){ref-type="fig"} illustrates an investigation between Th17, Treg, and Th1 related cytokines activated by native PS A1, Tn-PS A1, TF-PS A1, or Tn-TF-PS A1. PS A1 and its conjugates have been documented to influence the activation of Th17 cells, which are known to produce IL-17 \[[@B22-molecules-25-01319]\] and can assist in antibody production \[[@B53-molecules-25-01319]\]. Notably, co-culture of construct Tn-TF-PS A1 and the splenocytes harvested from mice immunized with Tn-TF-PS A1, revealed an increase in IL-17 by 59% when compared to PS A1. Similarly, immunization with Tn-PS A1 resulted in an increase in IL-17 by 57%. However, IL-17 production resulting from immunization with TF-PS A1 was decreased. The pattern of IL-10 quantification was opposite to that of IL-17, peaking with the construct TF-PS A1 and minimized with Tn-TF-PS A1. IL-10 is a major regulatory cytokine associated with Tregs \[[@B54-molecules-25-01319]\]. The observed high level of IL-10 matched with a low level of IL-17 for TF-PS A1 is a possible explanation for the poor IgG development seen when immunizations were done with TF-PS A1 in conjunction with either TMG or SAS. Finally, we observed IFN-γ production in splenocytes from mice immunized with Tn-PS A1, TF-PS A1, and Tn-TF-PS A1 were increased when compared to **1** alone. The increase in IFN-γ may be attributed to MPLA, which is known to stimulate Th1 polarization \[[@B55-molecules-25-01319]\].
2.6. Quantification of Cytokine Producing Cells with ELISpot {#sec2dot6-molecules-25-01319}
------------------------------------------------------------
In conjunction with cytokine quantification via sandwich ELISA, we performed a cellular based ELISpot assay to quantify cytokine producing cells. The cytokine IL-17 is solely produced by Th17 cells, and this relationship was exploited to show differences in CD4+ T cell activation between constructs PS A1, Tn-PS A1, TF-PS A1, and Tn-TF-PS A1 \[[@B22-molecules-25-01319],[@B23-molecules-25-01319],[@B24-molecules-25-01319]\]. [Figure 5](#molecules-25-01319-f005){ref-type="fig"} summarizes the number of Th17 producing cells developed from a splenocyte culture derived from mice immunized against the respective stimuli. The largest development of IL-17 producing cells was observed with Tn-PS A1, which showed a five-fold increase in IL-17 producing cells when compared to splenocytes derived from native PS A1. Conversely, TF-PS A1 developed even less IL-17 producing cells than PS A1 although it was not a significant change. To determine the influence conjugate Tn-TF-PS A1 had on IL-17 producing cells, anti-Tn-TF-PS A1 splenocytes were co-cultured with either Tn-PS A1, TF-PS A1, or Tn-TF-PS A1. The highest count of IL-17 producing cells in this group was observed when Tn-PS A1 was used as the stimulus. Alternatively, when TF-PS A1 conjugate TF-PS A1 was used to stimulate anti-Tn-TF-PS A1 splenocytes a three-fold increase of IL-17 producing cells was observed as compared to anti-TF-PS A1 splenocytes stimulated with the same construct. These data support the self-adjuvanting effect Tn can produce when incorporated as a bivalent construct, increasing immunity towards the TF antigen. Specificity was also supported by data presented in [Figure S4](#app1-molecules-25-01319){ref-type="app"} where the stimulation of anti-Tn-PS A1 splenocytes with TF-PS A1 did not increase IL-17 production.
IFN-γ is a canonical cytokine produced by Th1 cells, and the amount of IFN-γ producing cells in an ELISpot assay can provide insight into T cell activation \[[@B17-molecules-25-01319],[@B56-molecules-25-01319],[@B57-molecules-25-01319]\]. Displayed in [Figure S5](#app1-molecules-25-01319){ref-type="app"}, anti-Tn-TF-PS A1 splenocytes were able to generate increased levels of IFN-γ producing cells than anti-TF-PS A1 splenocytes when both were co-cultured with TF-PS A1. This result further supports an increased cellular response against the TF antigen with the bivalent Tn-TF-PS A1 conjugate. In addition to quantifying IL-17 and IFN-γ producing cells, we also quantified IL-10 producing cells ([Figure S6](#app1-molecules-25-01319){ref-type="app"}). Interestingly, the increase in IL-10 producing cells when anti-Tn-TF-PS A1 splenocytes were stimulated with TF-PS A1 was not as a significant increase when compared to the previous cytokine producing cells.
2.7. Biotinylated Conjugate Probes Bind to Recombinant MGL2 {#sec2dot7-molecules-25-01319}
-----------------------------------------------------------
To identify MGL2 as a potential mediator for the observed increase in immune activation by construct Tn-TF-PS A1, four biotinylated conjugate probes (**5a**, **5b**, **5c**, and **5d**, [Scheme S2](#app1-molecules-25-01319){ref-type="app"}) were synthesized. These probes were constructed by reacting the primary amine of the 2,4-dideoxy-4-amino-[d]{.smallcaps}-*N*-acetyl-fucose (AAT) constituent saccharide on the PS A1 repeating unit with sulfo-NHS-biotin. The probes (**5a**, **5b**, **5c**, and **5d**) were used in an assay where recombinant MGL2 was coated on 96-micro-well plates and streptavidin-alkaline phosphatase was used to detect binding interactions ([Figure 6](#molecules-25-01319-f006){ref-type="fig"}). The results indicated that probes **5a** and **5c** showed sufficient binding to MGL2 ([Figure 6](#molecules-25-01319-f006){ref-type="fig"}a). **5b** also gave a positive response but to a lesser extent than probes **5a** and **5c**. Constructs **5a**, **5b**, **5c**, and **5d** (10 μg mL^−1^) were also observed to be competitively inhibited by Tn-BSA (10 μg mL^−1^) where **5b** experienced the greatest inhibition ([Figure 6](#molecules-25-01319-f006){ref-type="fig"}b). MGL2 binding was negligible for the negative control **5d**.
3. Materials and Methods {#sec3-molecules-25-01319}
========================
3.1. Vaccinations with TiterMax^®^ Gold or Sigma Adjuvant System^®^ {#sec3dot1-molecules-25-01319}
-------------------------------------------------------------------
Jax C57BL/6 male mice (6 weeks) were obtained from Jackson Laboratories and maintained by the Department of Laboratory Animal Resources (DLAR) at the University of Toledo. All animal protocols were approved and performed in compliance with the relevant laws and institutional guidelines set forth by the Institutional Animal Care and Use Committee (IACUC) of the University of Toledo (protocol number 107956). Individual Tn-, TF-, and Tn-TF-PS A1 constructs were mixed in a 1:1 ratio of 50 μL of TiterMax^®^ Gold or 1:1 ratio of 100 μL SAS to achieve a final concentration of 20 μg for TACA-PS A1 constructs and injected into 7-week-old C57BL/6 mice. Groups of mice (n = 5) were immunized by intraperitoneal injections (IP) on day 0, 14, 28, 42 for TMG or on day 0, 21, 42 for SAS. Blood sera were obtained using a cardiac puncture technique on day 52.
3.2. Enzyme Linked Immunosorbent Assay (ELISA) {#sec3dot2-molecules-25-01319}
----------------------------------------------
Either Tn- or TF-BSA was coated on Immulon^®^ Microtiter™ 4 HBX 96 well-plates using 3 μg mL^−1^ in 0.1 M carbonate buffer (pH 9.2) and incubated for 18 h at 4 °C. Plates were washed three times with 200 μL of washing buffer (1× PBS buffer with 0.05% Tween^®^ 20 (*v*/*v*)) and blocked with 200 μL of 3% BSA (*w*/*v*) for 1 h at room temperature. Serum from mice were initially diluted at 1:100 and then serially half-log~10~ diluted for a final volume of 100 μL in each well, and then incubated for 2 h at 37 °C. After incubation, the plates were washed three times with 200 μL of washing buffer. Alkaline phosphatase-linked secondary antibodies (anti-IgM and anti-IgG) were used to detect primary antibodies bound to either Tn- or TF-BSA. The procedure for the secondary anti-IgM (Southern Biotech) antibodies called for a 1:1000 dilution and anti-IgG antibodies (Jackson ImmunoResearch) were diluted (1:5000). A total of 100 μL of secondary antibody was placed in wells and incubated for 1 h at 37 °C. The plates were washed three times with 200 μL of washing buffer and *p*-nitrophenyl phosphate (PNPP) (1 mg mL^−1^) in diethanolamine buffer (pH 9.8) was added at 100 μL per well and incubated for 30 min. The optical density was read at 405 nm using a BioTek PowerWave HT Microplate Spectrophotometer. All assays were performed in triplicate. Titers were determined by regression analysis with dilutions plotted against absorbance. The titer cutoff value was set at 0.2 which was two times the control PBS anti-serum for titer determination. Statistical analysis from ELISAs for experimental groups were compared with the controls using paired *t* test and GraphPad Prism 6.
3.3. Flow Cytometry {#sec3dot3-molecules-25-01319}
-------------------
MCF-7, OVCAR-5, and MCF-10A were cultured in 10% FBS RPMI 1640. 1.0 × 10^6^ cells of each cell line were incubated at 4 °C for 1 h in the dark with 1:50 dilution of the following separate antiserums: 1× PBS control, **1**, **4a**--**4c**. The cells were washed three times in 250 μL of FACS buffer (2% FBS (*v*/*v*) in 1× PBS, 0.001% sodium azide (*w*/*v*)) by centrifuging at 1000 rpm. 100 μL Anti-IgG Alexa Fluor^®^ 488 (1:50 dilution) was added to the cells and incubated at 4 °C in the dark for 1 h followed by three washes with 250 μL of FACS staining buffer. The cells were fixed with freshly prepared 1% paraformaldehyde (*v*/*v*) and analyzed using BD Biosciences FACS Calibur at the University of Toledo Core Flow Cytometry Facility. FlowJo analysis software was used to process flow cytometry data.
3.4. Complement Dependent Cytotoxicity Assay {#sec3dot4-molecules-25-01319}
--------------------------------------------
MCF-7 cells (1.0 × 10^4^) and OVCAR-5 cells (1.0 × 10^4^) were seeded in 96 well plates and incubated overnight in a 5% CO~2~ incubator at 37 °C. The plates were then washed with 2% BSA (*w*/*v*) in DPBS. After washing, 100 μL of experimental antiserum solutions, diluted 1:20 in DPBS, were added to the corresponding wells and incubated for 1 h. The experimental wells were washed and incubated with 10% rabbit complement (Pel-Freez) for 1 h at 37 °C. The control values of the LDH assay kit (Cytotoxicity Detection Kit; Roche, Mannheim, Germany) were determined from spontaneous LDH release (low control) and 1% Triton X-100 (*v*/*v*) (high control) and incubated for 1 h at 37 °C. 50 μL of cell supernatant was transferred to a new 96 well-plate containing 50 μL of DPBS. According to manufacturer protocol, 100 μL of the colorimetric LDH detection reagent was added to each well and the O.D. was read at 490 nm. The percentage cellular cytotoxicity was calculated using the following equation: Cell cytotoxicity % = (experimental values -- low control values)/(high control values -- low control values) × 100.
3.5. MGL2 Binding Assay {#sec3dot5-molecules-25-01319}
-----------------------
Mouse recombinant MGL2 (R&D systems) 2.5 μg mL^−1^ was used to coat Immulon^®^ Microtiter™ 4 HBX 96 well-plates in 1× PBS buffer (with CaCl~2~/MgCl~2~) pH 7.2 for 18 h at 4 °C. The plates were then washed with 200 μL of 1× PBS washing buffer (with CaCl~2~/MgCl~2~ and 0.05% Tween 20 (*v*/*v*)) three times. Biotin-PS A1 and respective biotinylated conjugates **4a**--**4c** were serially diluted from 40--0.625 μg mL^−1^ and incubated for 2 h at 37 °C in 1× DPBS with CaCl~2~/MgCl~2~. Plates were then washed with 200 μL of 1× PBS washing buffer three times. A streptavidin-alkaline phosphatase (Sigma Aldrich) solution was diluted (1:1000) and 100 μL was added to each well and incubated for 1 h at 37 °C. The plates were washed three times with 200 μL of 1× PBS washing buffer and then PNPP (1 mg mL^−1^) in diethanolamine buffer (pH 9.8) was added at a 100 μL per well and incubated for 30 min. The optical density was determined at 405 nm. Percent inhibition by Tn-BSA followed the same procedure, however, 10 μg mL^−1^ of Tn-BSA was co-incubated with **4a**--**4c** before binding competition to MGL2 was attempted. Percent inhibition was calculated using the equation: \[(O.D. of **4a**--**4c** binding to MGL2) − (O.D. of co-incubation of 4a--c with Tn-BSA)/(O.D. of **4a**--**4c** binding to MGL2)\] × 100.
3.6. Other Methods {#sec3dot6-molecules-25-01319}
------------------
Additional methodology is included in Doc. S5-Materials and Methods.
4. Conclusions {#sec4-molecules-25-01319}
==============
Tn-PS A1 was observed to be consistent in mounting an IgG immune response towards the Tn antigen whether TMG or SAS was used. However, using TF-PS A1 to elicit immunity towards the TF antigen provided unsatisfactory results. [Figure 1](#molecules-25-01319-f001){ref-type="fig"} indicated that the use of either TMG or SAS in conjunction with TF-PS A1 had little effect on IgG development, which is a critical intimation of T cell help and B cell maturation. When PS A1 was conjugated with both Tn and TF, there was a profound difference in proliferation of anti-TF IgGs when compared to the monovalent TF-PS A1 construct. Notably, polyclonal antibodies obtained from Tn-PS A1, TF-PS A1, and Tn-TF-PS A1 immunizations were assessed to bind Tn-BSA or TF-BSA. These BSA conjugates permitted the focus of antibody specificity towards either Tn or TF and not PS A1 as the carrier.
To further suggest an adjuvant-like effect with the Tn glycan, we observed an increase in pro-inflammatory cytokines and pro-inflammatory cytokine producing cells noted in [Figure 4](#molecules-25-01319-f004){ref-type="fig"} and [Figure 5](#molecules-25-01319-f005){ref-type="fig"} respectively. The increased pro-inflammatory environment is also suggested by the enhanced IgG response from Tn-TF-PS A1, enhanced tumor cell recognition in FACS ([Figure 2](#molecules-25-01319-f002){ref-type="fig"}), and increased tumor cell killing as noted in the CDC assay ([Figure 3](#molecules-25-01319-f003){ref-type="fig"}).
To determine if MGL2 was a potential target, four biotinylated probes (**5a**, **5b**, **5c**, and **5d**) were evaluated in a colorimetric assay. It was determined that Tn-biotin-PS A1 (**5a**) and Tn-TF-biotin-PS A1 (**5c**) had similar binding profiles, which in part, support the claim that the inclusion of the Tn antigen promotes a greater interaction with MGL2. MGL2 has been documented to have a lower affinity towards TF, which was confirmed in the observations noted in [Figure 6](#molecules-25-01319-f006){ref-type="fig"} \[[@B29-molecules-25-01319],[@B31-molecules-25-01319]\]. These experiments, in conjunction with competitive binding, suggested the affinity of the TF epitope is less for MGL2. These data support our hypothesis and the inclusion of the Tn epitope in Tn-TF-PS A1 permitted greater interactions with MGL2 which led to enhanced antigen uptake and antigen presentation by APCs. However, further experiments are required to definitively determine the mechanisms of immune activation.
The development of a semi-synthetic, bivalent Tn-TF-PS A1 construct has led to an increase in immunogenicity for the TF antigen as observed with ELISA, flow cytometry, CDC, and cytokine production. Literature precedent suggests that the increase in immune response is attributed to the MGL2 receptor which led to a more efficient uptake of the Tn-TF-PS A1 construct \[[@B33-molecules-25-01319],[@B38-molecules-25-01319],[@B39-molecules-25-01319],[@B40-molecules-25-01319]\]. This stands in contrast to other unimolecular multivalent constructs for which there was no enhanced response towards an individual TACA. \[[@B58-molecules-25-01319]\]. However, we did observe a similarity to unimolecular multivalent constructs where the seemingly immunodominant epitope was suppressed. This simple model seems applicable to multiple vaccine platforms which may include peptides, proteins, nanoparticles, and lipids leading to an increase in the therapeutic potential of carbohydrate-based vaccines.
We acknowledge Fred Valeriote, Joe Media, and Halina Pietraszkiewicz (Henry Ford Health System) for insightful discussions regarding immunotherapies and for their gracious donation of cell lines. We also acknowledge Yong-Wah Kim in the Department of Chemistry and Biochemistry at the University of Toledo for providing assistance with the NMR instrument.
**Sample Availability:** Samples of the compounds are not available from the authors.
The following are available online, Doc.S1--S5.
######
Click here for additional data file.
Conceptualization, P.R.A.; data curation, K.A.K. and K.R.T.; formal analysis, K.A.K., K.R.T. and P.R.A.; funding acquisition, P.R.A.; methodology, K.A.K. and K.R.T.; resources, M.S. and J.-P.B.; supervision, P.R.A.; visualization, K.R.T., K.A.K. and M.S.; writing---original draft, K.R.T.; writing---review and editing, K.A.K. and P.R.A. All authors have read and agreed to the published version of the manuscript.
This research was funded by the National Institutes of Health (NIH NCI R01 CA156661)
The authors declare no conflict of interest.
Figures and Scheme
==================
{#molecules-25-01319-sch001}
{#molecules-25-01319-f001}
{#molecules-25-01319-f002}
{#molecules-25-01319-f003}
{#molecules-25-01319-f004}
{#molecules-25-01319-f005}
{ref-type="app"}). (**b**) Percent inhibition by Tn-BSA in presence of **5a**, **5b**, **5c**, and **5d**. Experiments were performed in triplicate. Error bars are shown as mean ± standard error of mean. \* denotes % inhibition by Tn-BSA.](molecules-25-01319-g006){#molecules-25-01319-f006}
[^1]: These authors contributed equally to this work.
| {
"pile_set_name": "PubMed Central"
} |
The four circulating serotypes of dengue virus (DENV1-4) are mosquito-borne flaviviruses that cause 390 million human infections annually [@bib0001]. Approximately 25% of these infections result in symptoms ranging from a mild fever to a potentially fatal disease characterized by hemorrhagic fever or shock syndrome [@bib0001]. Although antibodies are a critical component for flavivirus immunity, complex antibody responses to DENV1-4 hinder the design of effective vaccines. Epidemiological studies have established that intermediate levels of antibodies from a prior DENV infection with one serotype are associated with the development of severe disease following subsequent infection with a different DENV serotype [@bib0002]. The prevailing theory suggests that during secondary exposure, pre-existing cross-reactive antibodies of the IgG isotype can bind to DENV particles from a different serotype without neutralizing infectivity. Instead, these antibodies can promote viral uptake into target cells expressing Fc gamma receptors in a process called antibody-dependent enhancement (ADE). Understanding the components and functions of the antibody response to DENV is important for informing the design of safe and effective antibody-based vaccines and therapies.
Unlike previous work, which was mostly restricted to analysis of B cells expressing the IgG antibody isotype, in the April 2020 issue of *EBioMedicine*, Waickman, Gromowski, et al. used single-cell RNA sequencing to obtain a more unbiased profile of the overall B cell repertoire of six individuals who had experienced primary or secondary DENV infection [@bib0003]. The authors examined paired heavy- and light-chain antibody sequences from over 9000 B cells, including short-lived plasmablasts generated early after infection, as well as memory B cells that persist long after the infection has resolved. Among memory B cells, there were no appreciable differences in isotype distribution in primary versus secondary infection. Consistent with a previous study, there was a low prevalence of IgA- relative to IgG-expressing plasmablasts upon secondary infection [@bib0004]. In contrast, following primary infection, this new study found an unexpectedly high proportion of plasmablasts expressing IgA antibodies, many of which were extensively hypermutated, suggesting a recall response despite no known prior DENV exposure. Further studies are warranted to investigate the origin of these hypermutated IgA plasmablasts detected following primary infection.
While IgA antibodies have recently been shown to contribute to the overall serum neutralizing activity against HIV [@bib0005], the functional significance of IgA antibodies in the context of DENV infection remains to be determined. Waickmann, Gromowski, et al. posited that plasmablast-derived DENV-specific IgA antibodies may be protective: as they appeared to recognize epitopes commonly targeted by IgG antibodies, by virtue of their inability to bind to Fc gamma receptors on relevant target cells, IgA antibodies may compete for binding to DENV with their IgG counterparts to abrogate the Fc gamma receptor-mediated ADE pathway. However, it is difficult to reconcile this proposed protective mechanism given the fact that except in the case of infants born to DENV-immune mothers [@bib0007], ADE is mostly implicated following secondary infection [@bib0002], during which IgA antibodies are apparently less prevalent. Indeed, given the reported characteristics of DENV-specific IgA antibodies in the current study, including 1) abundance during primary infection, 2) overlap in epitope specificity with IgG, and 3) limited neutralizing capacity (at least when tested with IgG Fc), it is equally possible that IgA antibodies could inhibit IgG-mediated neutralization of DENV. Additional studies to deconvolute the contribution of different antibody isotypes [@bib0005] in the humoral response to DENV infection or vaccination will help define their functional significance.
A limitation of the study by Waickman, Growmowski, et al. is its restricted sample size and population: six pediatric patients of whom five were infected with DENV1. Studies with a larger sample size that includes multiple age groups infected with other DENV serotypes will be needed to confirm the findings reported here and to ultimately draw correlations with disease outcomes. Nevertheless, this study highlights the utility of single-cell RNA sequencing to efficiently profile a large number of B cells in an unbiased manner, capturing diverse antibody isotypes and cellular states. Other recent studies further demonstrated this technology\'s capacity to link B cell receptor sequences [@bib0008] or transcriptional profiles [@bib0011] to antigen specificity. In the future, it would be interesting to investigate whether particular B cell transcriptomic signatures can predict antibody functions such as direct neutralization and Fc-dependent effector mechanisms. In addition to single-cell genomics, a 'systems serology' approach [@bib0009] to probe the biochemical and biophysical modifications to antibodies will also be important, given the emerging role of Fc glycoforms in regulating dengue disease [@bib0010]. Rapid advances in profiling humoral immunity at high throughput and resolution hold promise for comprehensively identifying the correlates of antibody-mediated protection and pathogenesis in DENV and other infections.
Declaration of Competing Interests
==================================
Authors have no conflicts of interest to disclose.
| {
"pile_set_name": "PubMed Central"
} |
Introduction {#s1}
============
*Pseudomonas aeruginosa* is considered one of the main Gram-negative bacteria causing health-care associated infections ([@B1]). In the hospital setting, *P. aeruginosa* is widely present in the environment and can be retrieved from different sources, such as respiratory therapy equipment, antiseptics, soap, sinks, and hydrotherapy pools ([@B2]). This pathogen was also found to be part of the endogenous microbiota of 2.6--24% of hospitalized patients ([@B3], [@B4]). Patients with compromised host defense mechanisms, such as neutropenia, severe burns, or cystic fibrosis, are particularly affected by this pathogen whose infections lead to high morbidity and mortality ([@B5], [@B6]). *P. aeruginosa* has been previously described as the second most common organism responsible for infections acquired in the intensive care units (ICUs) ([@B7]).
*P. aeruginosa* population structure is consensually believed to be panmictic-epidemic ([@B8]--[@B10]), i.e., a superficially clonal structure with frequent recombination that creates new strains with unique genetic characteristics, in which occasionally highly successful epidemic clones arise. In addition, clinical isolates are indistinguishable from environmental isolates; and there are no specific clones related to a specific habitat ([@B10]).
Molecular epidemiological investigations have become essential for active surveillance of infection and detection of outbreaks*. P. aeruginosa* possesses a very complex ecology. For that reason, only powerful typing methods can give insight on the relatedness of strains, and consequently on the routes of colonization and/or infection ([@B11]). Pulsed-field gel electrophoresis (PFGE) has been considered the "gold standard" for DNA fingerprinting of *P. aeruginosa* ([@B12]--[@B14]). However, this method as several disadvantages, such as long analysis time, low intra- and inter-laboratory reproducibility and is labor-intensive, which make it not the optimal method for large investigations ([@B15]--[@B17]). To overcome these limitations, alternative amplification-based molecular methods have been implemented such as multi-locus sequence typing (MLST), that showed to be efficient in the study of the global population structure of *P. aeruginosa* ([@B18]). Another sequence-based method, double locus sequence typing (DLST), based on partial sequencing of two highly variable loci has been successfully used to investigate the epidemiology of *Staphylococcus aureus* and *Pseudomonas aeruginosa* ([@B1], [@B19]--[@B21]). With the advance of new technologies, whole genome sequencing (WGS) has been used in recent studies on *P. aeruginosa* evolution and epidemiological investigations in the hospital settings ([@B22]--[@B24]).
Following an increase in *P. aeruginosa* incidence in the ICUs of the Lausanne University Hospital, clinical and environmental isolates were typed using DLST ([@B1]). Three major DLST types were identified, the larger being previously reported as the cause of an outbreak in the burn unit ([@B25]). The other two types showed sporadic occurrence with only few cases of possible transmission between patients. The discriminatory power of whole genome sequencing (WGS) was used to further investigate these three major DLST types.
Materials and Methods {#s2}
=====================
Bacterial Isolates and Molecular Typing
---------------------------------------
From 2010 to 2014, *P. aeruginosa* isolates were collected from patients and environment from the five ICUs of the University Hospital of Lausanne. All consecutive patients hospitalized in the ICU with a clinical sample growing *P. aeruginosa* at any site were considered. No routine screening of *P. aeruginosa* carriage was performed. Based on colony morphology, one or several *P. aeruginosa* isolates per clinical sample were chosen for further typing analysis. For patients with prolonged ICU stays, multiple samples were considered for isolate recovery. In 2012, the ICU environment was investigated for the presence of *P. aeruginosa*. Tap water samples and environmental swabs obtained from taps and sink traps of all ICU rooms, as well as from the environment of the hydrotherapy room (including shower trolleys and shower mattresses), were analyzed. Thereafter, sink traps were investigated twice a year.
All isolates were typed by the double locus sequence typing ([www.dlst.org](http://www.dlst.org)) method as previously described ([@B1]). Three major DLST types, i.e., types with the highest number of patients, were further analyzed in this study: DLST 1--18 (24 patients), 6--7 (21 patients), and 1--21 (16 patients). For WGS, at least one isolate was selected per patient. If several isolates were collected from one patient, only isolates sampled 15 days apart were selected, unless they belonged from different sample sites. All environmental isolates from the three genotypes (mainly from sink traps) were included. A total of 74 DLST 1--18 isolates (56 clinical and 18 environmental), 50 DLST 6--7 isolates (35 clinical and 15 environmental), and 31 DLST 1--21 isolates (18 clinical and 13 environmental) were selected for WGS. Epidemiologic and genetic data of all clinical and environmental isolates are listed in [Table S1](#SM1){ref-type="supplementary-material"}.
Epidemiological Investigation
-----------------------------
Epidemiological data (dates of ICU admission and discharge, unit and room of hospitalization and clinical data) were retrieved from the hospital databases and used to construct the timeline of patient\'s hospital stay and annotate the phylogenetic trees. Epidemiological links between patients or environment were considered in the following situations: (i) patients hospitalized during overlapping periods in the same ICU, or (ii) patients showing an identical DLST type with an environmental sample isolated in the same unit during the period of the study. An outbreak was defined as two or more cases with epidemiological links. A case with no epidemiological links was considered as a single case.
DNA Extraction and Whole Genome Sequencing
------------------------------------------
A single colony was picked and incubated into 5 ml Lysogenic Broth (LB) to reach an early exponential phase. We extracted genomic DNA using the GenElute bacterial genomic DNA kit (Sigma-Aldrich, St. Louis, MO, USA). Whole genome sequencing was performed at the Lausanne Genomic Technologies Facility (GTF, University of Lausanne). The sequencing libraries were prepared using the Nextera DNA Library Preparation Kit (Illumina, San Diego, CA, USA) for 100-bp paired-end sequencing runs on Illumina HiSeq 2500, aiming for a 100-fold coverage. All reads data have been deposited with the National Center for Biotechnology Information (<https://www.ncbi.nlm.nih.gov/>) under the accession project number [PRJNA503802](PRJNA503802).
SNPs and Phylogenetic Analysis
------------------------------
Isolates\' sequence types (ST) were assigned from the short reads data by the Short Read Sequence Typing 2 (SRST2) software ([@B26]). A first step of downsampling the raw reads to a given threshold coverage was added. Its purpose was to avoid the occurrence of excessive depth which creates non-informative data and higher computational costs. The depth threshold was set to 70× since this was the minimum read depth observed in our dataset. Complete reference genomes were created by sequencing the first collected clinical isolate of each ST with both PacBio and Illumina HiSeq technologies ([@B27]). The subsampled reads were then mapped against their respective complete reference genome with BWA-MEM (<https://arxiv.org/abs/1303.3997>). Variant calling was performed with FreeBayes (<https://arxiv.org/abs/1207.3907>) with a minimum mapping quality of 60. Putative phages, repeat regions and potential recombination regions were excluded from the genome alignment. Putative phages were found with PHASTER ([@B28]). Repeat regions were detected using an in-house script based on NUCmer ([@B29]) with a minimum percentage of identity of 80, and a minimum length of 80 to report a repeat region. Recombination search was performed with an in-house script on the VCF file acquired with FreeBayes (<https://arxiv.org/abs/1207.3907>) using a probability of 0.001 to remove a region of high SNPs density and a window size of 2000 for SNPs counting.
A maximum likelihood (ML) tree was constructed from the final SNPs alignment using the PhyML algorithm implemented in Seaview version 4.6.1 ([@B30]). Tree visualization was done with FigTree version 1.4.3 (<http://tree.bio.ed.ac.uk/software/figtree>).
*In silico* Identification of Resistance and Virulence Genes
------------------------------------------------------------
Acquired antibiotic-resistance and putative virulence genes were identified with SRST2 ([@B26]) by mapping the short reads against the ARG-ANNOT ([@B31]) and VFDB ([@B32]) curated databases, respectively. Additionally, the Antimicrobial Resistance Identification By Assembly (ARIBA) tool ([@B33]), which combines mapping/alignment and targeted local assembly of the paired sequencing reads, was applied for detection of antimicrobial resistance genes using the Comprehensive Antibiotic Resistance Database (CARD) ([@B34], [@B35]) as reference.
Results {#s3}
=======
Three DLST Types With Different Epidemiology
--------------------------------------------
Epidemiological data of patients and environmental isolates included in this study is schematically represented in [Figure 1](#F1){ref-type="fig"}. DLST 1--18 was previously considered responsible for an outbreak in the burn unit from 2010 to 2012 ([@B25]). From the 24 patients harboring this DLST type, 19 were hospitalized in the burn unit (ICU 3), and five in other ICUs. Epidemiological links were found between these patients, either through environmental contamination (especially the hydrotherapy shower room and sink traps) or through patient-to-patient transmission. However, Patient 1 was first isolated in the medical ward and afterwards hospitalized in ICU 5, without obvious epidemiological link with other patients.
{#F1}
Within DLST 1--21, epidemiological links were identified only between two patients hospitalized in ICU 2 and an environmental sample retrieved from a sink trap in the same ICU. The remaining patients were dispersed throughout the six ICUs during the study period ([Figure 1](#F1){ref-type="fig"}), suggesting they were not involved in an outbreak. One patient (Patient 6) was infected in the traumatology ward, before its stay in the ICU.
Regarding DLST 6--7, epidemiological links were found between five patients hospitalized in the burn unit in 2010. For the other 16 patients, no epidemiological link was suspected, as patients were not hospitalized in the same ICU during overlapping periods and had no link with environmental samples. Similarly to DLST 1--21, this DLST type occurred mainly sporadically throughout the study period and was not responsible for a major outbreak.
Congruence Between DLST and MLST
--------------------------------
Although DLST allows inter laboratory comparison of genotypes, the universal standard of multilocus sequence typing (MLST) is still widely used for strain comparison and identification. Therefore, the Illumina HiSeq raw reads were used to identify the STs of each isolate. Results showed that all DLST 1--18 isolates belonged to ST1076, DLST 1--21 to ST253 and DLST 6--7 to ST17, except for one DLST 6--7 isolate, which was found to be of ST845, a single-locus variant from ST17 at the *nuoD* locus. This confirms the previously documented congruence between both methods ([@B36]).
Contribution of WGS to Decipher the Epidemiology of the Three DLST Genotypes
----------------------------------------------------------------------------
The genetic similarities between isolates of DLST 1--18 are shown in the maximum likelihood tree in [Figure 2](#F2){ref-type="fig"}. Patient 1, who had no epidemiological link with the outbreak, clustered apart from the remaining isolates with 108--120 SNPs differences. Most of the outbreak isolates were closely related with 0--14 SNP differences, confirming the clonal origin of the outbreak. Isolates retrieved from the same patient were different by \<10 SNPs, e.g., Patient 4 (0--7 SNPs) and Patient 24 (0--2 SNPs).
{#F2}
The ML trees of DLST 1--21 and 6--7 are shown in [Figures 3](#F3){ref-type="fig"}, [4](#F4){ref-type="fig"}, respectively. A great diversity was found between isolates of DLST 1--21, reinforcing the premise that they were not all part of a single chain of transmission. Nevertheless, several clades of highly similar isolates were observed. Two isolates from two patients and one isolate from the environment showed 0--1 SNP differences (Patients 10, 11 and Env. 5 in [Figure 3](#F3){ref-type="fig"}), confirming the suspected epidemiological links. Isolates belonging to the same patient were closely related with a maximum of 4 SNPs (Patients 2 and 6). Highly similar isolates were retrieved from the same environment over a 1-year period (sink traps in ICU3), and a 10-year period (pediatric ICU and ICU5 which are adjacent units). Conversely, WGS results also showed high similarities (6 SNP differences) between isolates with no suspected epidemiological link (Patients 5 and 12, hospitalized 2 years apart in two different ICUs).
{#F3}
{#F4}
Similarly, the ML tree of DLST 6--7 showed a large diversity between isolates ([Figure 4](#F4){ref-type="fig"}). Two clades of highly similar isolates were observed. The first was composed of isolates with 0--13 SNPs differences, retrieved from patients with suspected epidemiological links (Patients 1, 2, 3, 4, and 5). The second clade includes isolates from Patient 11 and environmental isolates from the same ICU (0--7 SNPs differences). All isolates recovered from a single patient (Patients 7, 13, and 15) were highly similar with only 0--6 SNPs differences.
Resistance and Virulence Profiles of *P. aeruginosa* Isolates
-------------------------------------------------------------
The presence/absence of resistance and virulence genes found in the 153 *P. aeruginosa* genomes is listed in [Table S2](#SM2){ref-type="supplementary-material"}. Although the detected antimicrobial resistance and virulence profiles were very similar among all DLST types, isolates of DLST 1--21 harbored additional genes that were not present in isolates of DLST 1--18 or DLST 6--7. However, specific clusters of closely related isolates observed for each DLST type were not similar in gene content. Several genes encoding resistance to aminoglycosides, fosfomycin, β-lactams, chloramphenicol, polymyxins, bicyclomycin, fluoroquinolones, and inherent multidrug resistance, mainly associated with efflux pump complexes, were found for all isolates ([Table S2](#SM2){ref-type="supplementary-material"}). A plasmid-encoded ciprofloxacin resistance protein, CrpP, which inactivates the antibiotic by phosphorylation ([@B37]), was lacking in all DLST 1--18 isolates but detected in 93% (29/31) and 100% of DLST 1--21 and DLST 6--7 isolates, respectively. The regulatory protein, PvrR, responsible for the conversion between antibiotic-resistant and antibiotic-susceptible forms of *P. aeruginosa* biofilms ([@B38]) was only detected in 45% (14/31) of DLST 1--21 isolates. In addition, DLST 1--21 isolates comprised two integron-encoded aminoglycoside adenylyltransferase genes: *aadA6* in 6 isolates from two patients and *aadA7* in 23% (7/31). Resistance to sulfonamides was likewise unique for this genotype due to the presence of the integron-encoded sulfonamide dihydropteroate synthase, *sul1* ([@B39]), in 29% (9/31) of the isolates. One isolate belonging to Patient 13 from DLST 6--7 harbored the plasmid-associated *strB* gene, which encodes an aminoglycoside O-phosphotransferase, APH(6)-Id, responsible for streptomycin resistance ([@B40]).
Interestingly, *in silico* screening of the presence/absence of putative virulence genes demonstrated that all DLST 1--21 isolates comprising the gene encoding the regulatory protein PvrR, also contained a set of fimbriae genes, *cupD*. These genes are exclusive of the *P. aeruginosa* PA14 strain, which belongs to the same ST253 as the DLST 1--21 isolates, and are regulated by the Rcs/Pvr network ([@B41]). As previously suggested by Peña et al. ([@B42]), the more frequent type III secretion system (T3SS) exotoxin genes, *exoT* and *exoY*, were common to all DLST types. Nonetheless, the presence of *exoS* was only observed in DLST 6--7 isolates, and ExoU was solely encoded in DLST 1--18 and 1--21 genomes.
Discussion {#s4}
==========
In this study, we provide insight into the epidemiology of *P. aeruginosa* in the ICUs of a tertiary care hospital by combining a fast and cheap molecular typing method (DLST) with a more sophisticated and discriminatory method (WGS).
Identifying the isolates\' ST by MLST check is one of the advantages of WGS. We were able to determine that DLST 1--21 isolates (ST253) belong to the clinical and international well described clonal complex (CC) PA14, and DLST 6--7 (ST845 and 17) was previously reported as part of the clonal complex C, both CCs being the worldwide most abundant clonal complexes in the *P. aeruginosa* population ([@B43], [@B44]). Little is known so far on ST1076 genomes (DLST 1--18).
The combination of DLST and epidemiological data revealed three genotypes with different epidemiological behaviors. Most of DLST 1--18 patients were hospitalized in the burn unit during overlapping periods. The high number of epidemiological links between patients, along with the wide presence of this DLST type in the environment of the burn unit, previously allowed us to consider this genotype was responsible for an outbreak with an environmental source ([@B25]). The discriminatory power of WGS confirmed the high similarity between all DLST 1--18 outbreak isolates (0--10 SNP differences) except for Patient 1. This patient was initially considered as the index case based on DLST, but he was hospitalized in a distant unit without geographical link to the outbreak. According to WGS, this patient could be excluded from the outbreak.
While congruence between epidemiological data and DLST typing was good in this DLST 1--18, this was not the case for the other two DLST types for which only few epidemiological links were recorded. The high diversity between these isolates revealed by WGS confirmed they are not part of a single chain of transmission. Nevertheless, both suspected outbreaks (one in DLST 1--21 and one in DLST 6--7) were confirmed by WGS results and most isolates from a single patient were also found to be highly similar.
The epidemiology of *P. aeruginosa* nosocomial infections is intricate due to the ubiquitous presence of this pathogen in the environment. Since *P. aeruginosa* is capable to survive on wet surfaces such as sinks, sink traps, pipes, and hydrotherapy equipment, several nosocomial outbreaks have been associated with these specific reservoirs. DLST 1--18 environmental isolates retrieved from shower mattresses and sink traps from the hydrotherapy room support the assumption of an environmental source of infection. This assumption was confirmed by results of WGS showing that environmental isolates were highly similar to patients isolates (\<10 SNPs).
Apart from contamination in the hydrotherapy room, which was resolved by infection control measures, sink traps were the main reservoir of *P. aeruginosa* in our ICUs during the study period. Our study showed the added value of WGS for the confirmation of epidemiological links between patients and this environment in several occasions. The constant presence of both DLST 1--21 and 6--7 in the environment might have led to sporadic infections, which would explain infections in patients with no epidemiological links (e.g., DLST 1--21, Patients 5 and 12). We also showed that these sink traps of the same or adjacent units can be contaminated with highly similar isolates. More intriguing is the fact that highly similar isolates (\<14 SNPs differences) were recovered 10 years apart from sink traps of adjacent units. This low diversity is unexpected considering the long time between isolate sampling, and contrast with isolates retrieved from a single patient, weeks apart, exhibiting already several SNPs differences. One explanation can be the slower evolution of *P. aeruginosa* isolates in the environment of ICUs, as opposed to locations with high selective pressure such as inflamed cystic fibrosis lungs ([@B45]).
One limitation of our study relies on the environmental sampling. Extensive sampling was performed in 2012 but not prior to 2012, and only twice a year thereafter. A more frequent and regular sampling might have helped to discover possible epidemiological links between infected patients and environmental sources. It would have also allowed us to thoroughly investigate the diversity of isolates within this ecological niche. In addition, our study focus on the ICU without considering transmission or presence of *P. aeruginosa* in the environment of other hospital units. This limitation is exemplified by one patient infected with one of the major genotypes (Patient 6, DLST 1--21) before its admission to the ICU.
Although we initially hypothesized that DLST 1--18 isolates would have a different resistant/virulent gene content due to its association with a 2.5 years outbreak, the genetic characterization of all genotypes revealed differences primarily between DLST 1--21 and the remaining DLST types (DLST 1--18 and 6--7). Antibiotic resistance in *P. aeruginosa* is commonly associated with the presence of plasmids ([@B46]) and several resistance genes are included in mobile genetic elements such as transposons, integron, and IS elements ([@B47]--[@B49]). The additional resistance observed in DLST 1--21 isolates was associated with plasmid/integron encoded antimicrobial resistance genes. Nevertheless, these genes were only present in an unpatterned subset of DLST 1--21 isolates and were not representative of the entire DLST 1--21 isolate\'s collection.
A significantly higher number of virulence genes was detected for DLST 1--21 (ST253) isolates. This large set of additional putative additional virulence factors may represent an increased virulence indiscrimination of this DLST type, which is suspected for other highly virulent strains of the same ST, e.g., PA14 ([@B50]). *P. aeruginosa* isolates can be differentiated into cytotoxic or invasive, according to their interaction with epithelial cells. The cytotoxic phenotype, as opposed to invasive isolates, encodes for the ExoU toxin ([@B51]). *In silico* virulence analysis revealed this toxin is present in both DLST 1--18 and DLST 1--21 isolates, characterizing these genotypes as potentially cytotoxic. However, the expression of this toxin would have to be confirmed by *in vitro* testing.
Being fast and cheap, DLST is performed as a routine surveillance of *P. aeruginosa* in our ICUs by typing all patient and environmental isolates on a quarterly basis. Only if several patients harbor a similar genotype a field epidemiological investigation is undertaken and WGS performed.
In conclusion, we showed the high value of WGS for the investigation of nosocomial *P. aeruginosa* infections. Although WGS costs are decreasing, its implementation as a routine surveillance method for *P. aeruginosa* still comes at a higher price per isolate than the currently used DLST. Additionally, analysis of WGS data requires a certain level of bioinformatic expertise that is not available in all laboratories ([@B52]). Thus, using DLST as a first-line molecular typing tool for surveillance and WGS to solve problematic clusters would culminate in an accurate and cost-efficient typing strategy.
Data Availability Statement {#s5}
===========================
All datasets generated for this study are included in the article/[Supplementary Material](#s8){ref-type="sec"}.
Ethics Statement {#s6}
================
All procedures performed in this study were in accordance with the requirement of the local ethics committee (Commission cantonale d\'éthique de la recherche sur l\'être humain, Lausanne, Switzerland).
Author Contributions {#s7}
====================
BM is the principal investigator. DB is the principal supervisor. BV and MA are bioinformatician supervisors. GP and GG are microbiologist responsible for clinical analysis. LS is the epidemiologist and infection control supervisor.
Conflict of Interest
--------------------
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
We thank the medical staff of the IUCs for their fruitful collaboration, M. J. Thevenin for all environmental sampling, and the laboratory technicians for processing the samples and DLST typing. Thanks to Didier Hocquet for the discussion on the resistome and virulome of *P. aeruginosa*. We also thank the Ultra-High-Throughput Sequencing (UHTS) unit of the Lausanne Genomic Technologies Facility (LGTF) at the University of Lausanne for the Illumina HiSeq and PacBio sequencing services.
Supplementary Material {#s8}
======================
The Supplementary Material for this article can be found online at: <https://www.frontiersin.org/articles/10.3389/fpubh.2020.00003/full#supplementary-material>
######
Epidemiological data of each clinical and environmental isolate.
######
Click here for additional data file.
######
Summary of the detected resistance and virulence genes in the 153 *P. aeruginosa* isolates.
######
Click here for additional data file.
[^1]: Edited by: Richard V. Goering, Creighton University, United States
[^2]: Reviewed by: Jie Zheng, United States Food and Drug Administration, United States; Dinesh Sriramulu, Independent Researcher, Chennai, India
[^3]: This article was submitted to Infectious Diseases - Surveillance, Prevention and Treatment, a section of the journal Frontiers in Public Health
[^4]: †Present address: Mohamed M. H. Abdelbary, Division of Oral Microbiology and Immunology, Department of Operative and Preventive Dentistry and Periodontology, RWTH Aachen University Hospital, Aachen, Germany
| {
"pile_set_name": "PubMed Central"
} |
To the Editor:
==============
In 2018, I embarked on my residency interviews along with thousands of other hopeful residents and crisscrossed the country exploring opportunities for my education. As a former natural resource major with a thesis on climate change, I could not help but feel guilt for the hefty contribution to carbon dioxide emissions that I was making through air travel. Was walking the halls of recruiting hospitals worth the detriment to the environment? As a researcher, I knew the anthropogenic contribution of climate change was scientifically established. As a medical student, I knew that the changing climate had implications for human disease. But what choice did I have? For residency interviews, travel was the name of the game.
Then I discovered tools to quantify my concerns.^[@R1],[@R2]^ I calculated that, in the 12 interviews I had attended (4 with my spouse), we had accumulated 25,733 air miles resulting in 6,049 kg of CO~2~ emissions. Accounting for my seats alone, that was 329.4 square feet of sea ice, just smaller than the average American studio apartment footprint. Now, consider the environmental cost of 44,600 residency applicants and their partners, crossing the country and sometimes the globe, to attend an ever-increasing number of interviews.
It is well established that climate change threatens global health.^[@R3]^ Furthermore, the suffering brought on by climate change is inequitable, disproportionately shouldered by our most vulnerable---the elderly, unsheltered, and those with poor access to weather refuge. As a profession dedicated to the health of humanity, we must confront the effects of climate change on health and align our actions with our values by recognizing the carbon footprint of residency interview travel and taking steps to reduce it.
In the wake of the disruptions to medical education brought on by the COVID-19 pandemic, we have the opportunity to reexamine and redesign systems to better fit our professional values. As training programs adopt virtual interviews out of social-distancing necessity, we urge them, for the sake of the environment, to consider investing permanently in a virtual interview system. In doing so, we honor our commitment as physicians to global health and social justice.
*Disclosures:* None reported.
| {
"pile_set_name": "PubMed Central"
} |
In collaboration with [C&EN](http://cen.acs.org/index.html).
[Natia Frank](http://franklabuvic.org/) wants to reduce the energy consumption of future computers. The University of Victoria chemist designs and studies novel organic electronic materials, and she recently developed [a material](https://patentscope.wipo.int/search/en/detail.jsf?docId=WO2017015765) that allows computers to write data with light instead of electric current. The material, made of a cobalt dioxolene coupled to a spirooxazine, makes possible a next-generation computer memory known as light-induced random access memory. LI-RAM would be faster than current forms of RAM, require less power, and work for both short-term data processing and long-term data storage. Frank talked with Louisa Dalton about how her LI-RAM material came to be and where she gets her ideas.
{#GRAPHIC-d85e75-autogenerated}
Why write computer data with light? {#sec2}
===================================
The philosophy behind what we do is trying to decrease energy demands of all of these technologies that people use. That's very important to me on a personal level. Everything is going to be run by little computers, and it's important that they don't increase our global electrical energy demands significantly.
The semiconductor industry is up against what's called the power wall. The industry has been trying to deal with that increase in power consumption by developing resistive memory, which uses a lot less energy to run and doesn't dump as much heat. There are a lot of different types of resistive technologies that are being looked at, and they all differ in their effectiveness. What we're trying to do is come up with a resistive magnetic memory, which is good for long-term storage but is written with light.
The reason we chose light is because it is one of the fastest ways to write and transfer energy and information. Less heat, less power demand, and much faster read-write speed are the major advantages of this type of technology.
How does your material make writing binary data with light possible? {#sec3}
====================================================================
We developed a strategy in which we have an electronically bistable metal complex that can exist between a low-spin and a high-spin state. The complex switches between these two states through an intramolecular charge-transfer process. We attached a photochromic ligand to this metal complex, which changes its structure and ligand field when irradiated by light. The change in structure and ligand field causes a change in the redox potential around the metal center, and this in turn changes the driving force for charge transfer in the metal complex.
The coupling of photochromic ligands to these electronically bistable metal complexes gives rise to the photomagnetic effects that we observe at room temperature.
How long have you been working on this material? {#sec4}
================================================
This was one of the projects I proposed when I applied for faculty positions way back in 2000. When we started the project, there was essentially no work done on taking a photochromic molecule and complexing it to a transition-metal ion.
I was at Caltech at the time as a postdoc, and a lot of the faculty there and my colleagues read my proposals, and they said, "Oh, this is never going to work because you're going to excite the system and all the energy is going to get dumped into the metal, and it's just going to relax down to the ground state." I wasn't convinced that all of the excited state energy would get dumped into the metal. And I thought, you know, it's always worth trying if you don't know enough to absolutely predict the right answer.
For the ligand, we ended up settling on spirooxazines. Not only were they still photoresponsive once complexed to transition metals, they also showed an increase in photoresponsivity, particularly with cobalt and nickel. I was delighted that it actually worked.
The material that we're now working on for the light-induced RAM is a cobalt dioxolene attached to a spirooxazine ligand.
What are you doing with the material now? {#sec5}
=========================================
Now that we can generate these photomagnetic effects at room temperature in the solid state, we can incorporate them into devices. We can start thinking about making other types of magnetic memory devices.
We're collaborating with semiconductor companies, decreasing the size of the devices, and optimizing the parameters so that the material can be incorporated into semiconductor devices using technologies that are already in place. What's specifically unique about our material is that it operates on a single molecule level, and because of that, we can, in principle, make a device that is 2 by 2 nm in size and holds one bit.
Where do you get your ideas? {#sec6}
============================
I take advantage of the fact that I've moved through a lot of different fields during the course of my career. I'm part of the organic electronics community, the magnetism community, the electron transfer community. I'm sort of on the edge of a lot of communities, and because of that, there's a lot of cross-fertilization.
In our group, we like to design single-component systems in which there's a very strong electronic coupling between two different types of processes. We're perturbing one process and it's leading to a change in a secondary process. That's really the fundamental philosophy behind what we do on all levels.
{#sec10}
*Louisa Dalton is a freelance contributor to*[*Chemical & Engineering News*](http://cen.acs.org/)*, the weekly newsmagazine of the American Chemical Society. Center Stage interviews are edited for length and clarity.*
| {
"pile_set_name": "PubMed Central"
} |
INTRODUCTION {#sec1}
============
The global spread of carbapenemase-producing organisms (CPO) has been highlighted by international health authorities as a critical public health concern ([@B1][@B2][@B3]). Containment of this spread, currently recommended in North American and European infection control guidelines, includes accurate and rapid identification of colonization with CPO ([@B4], [@B5]). Different methods for detection of CPO in rectal swabs have been used, including culture with specific chromogenic media and in-house and commercial molecular tests ([@B6], [@B7]). Culture-based methods are limited by sensitivity and specificity issues depending on the composition of the medium and the targeted carbapenemases (e.g., KPC, VIM, NDM, IMP, or OXA-48 enzymes), the requirement of 24 to 48 h for growth, and the need for confirmation with phenotypic and/or molecular methods. Moreover, as the accuracy of culture methods is highly dependent on medium formulation (type and concentration of antibiotic supplementation), the increasing diversity of carbapenemase enzymes and their variable expression affect the sensitivity of culture methods ([@B8], [@B9]). In addition, bacterial load in rectal or perirectal swabs also impacts the accuracy of culture-based tests. For these reasons, molecular tests have been developed and claim to overcome most of these limitations. These molecular assays comprise multiplex-PCR-based assays ([@B10], [@B11]), isothermal amplification (LAMP)-based assays ([@B12]), and microarray-based assays ([@B13], [@B14]). Some of these assays have been evaluated using only a collection of well-characterized CPO but not with clinical samples and/or with a multisite approach.
In a previous study, we evaluated the sensitivity and specificity of a PCR-based method in a cartridge format (Xpert MDRO assay) for detecting carbapenemase genes in rectal and perirectal swabs run on the GeneXpert platform (Cepheid, Sunnyvale, CA). Testing was performed in parallel with a reference culture method ([@B15]). This molecular assay included targets for *bla*~KPC~, *bla*~NDM~, and *bla*~VIM~ genes but not for *bla*~OXA-48~ and *bla*~IMP-1~. In the current study, an updated version of the Xpert Carba-R (Cepheid, Sunnyvale, CA) was evaluated, which included all five gene targets. The evaluation was performed in a multisite prospective study for European marketing authorization under the European directive on *in vitro* diagnostic medical devices ([@B16]).
MATERIALS AND METHODS {#sec2}
=====================
Study design, prospective, and contrived specimens. {#sec2-1}
---------------------------------------------------
The study was conducted between July 2013 and February 2014 at two institutions in the United States (David Geffen School of Medicine, University of California, Los Angeles, CA and Westchester Medical Center, Valhalla, NY) and two institutions in Europe (Central Manchester University Hospitals NHS Foundation Trust, Manchester, United Kingdom and Hospital Ramón y Cajal, Madrid, Spain). Specimens included prospectively recovered rectal swabs (*n* = 383) from patients who provided informed consent to participate in the study or whose routine care included screening for CPO following local infection control policies. The protocol was approved by each institution\'s Ethics Committee. A double swab set (Venturi Transystem; Copan, CA) and transport medium (liquid Stuart transport swab; Copan) were used to collect and transport rectal swab specimens from eligible subjects. One swab was used for reference culture, susceptibility testing, and sequencing, and another swab was used for the Xpert Carba-R. To diminish the potential bias of sampling differences, the two swabs were gently rolled against one another before starting the Cepheid Xpert Carba-R assay and culture procedure.
Due to the low prevalence of organisms containing some of the carbapenemase genes, the prospectively collected clinical rectal swab specimens were supplemented with an additional 250 contrived specimens. To prepare the contrived specimens, unique, well-characterized carbapenemase-producing (*n* = 108) and non-carbapenemase-producing isolates (*n* = 142) were spiked into the negative rectal swab matrix ([Table 1](#T1){ref-type="table"}). Fifty percent of the specimens were prepared at concentrations near the analytical limit of detection (LOD) of the Xpert Carba-R assay (ranging from 1.1 × 10^2^ to 1.2 × 10^3^ CFU/swab depending on the carbapenemase gene), and the remaining specimens covered clinically relevant analyte concentrations (1×, 3×, and 10× LOD) ([@B17]). Isolates were seeded onto mock rectal swabs that were prepared by dipping swabs into an individual stool matrix that tested negative for carbapenemase genes. Nonseeded swabs that were dipped into individual stool matrix were prepared as controls.
######
Xpert Carba-R assay results by target for clinical and contrived specimens
Xpert Carba-R assay result Clinical specimens (*n* = 383) Contrived specimens (*n* = 250) All specimens (*n* = 633)
------------------------------------------- -------------------------------- --------------------------------- ---------------------------
Positive (single and/or combined targets) 42 107 149
IMP-1 0 25 25
VIM 2 24 26
NDM 2 23 25
KPC 13 19 32
OXA-48 20 15 35
VIM + OXA-48 4 0 4
NDM + KPC 1 0 1
IMP-1 + NDM 0 1 1
Negative 341 143 484
Results from the Xpert Carba-R assay for the clinical and contrived specimens were compared to those of the reference culture and sequencing results.
Reference culture, susceptibility testing, and sequencing. {#sec2-2}
----------------------------------------------------------
One of the prospectively collected rectal specimen swabs was placed in 10 ml of MacConkey broth (General Laboratory Products, Yorkville, IL, USA) containing a 10-μg meropenem disk (BD, Franklin Lakes, NJ, USA) and was vortexed. After 20 to 24 h of incubation at 35°C, MacConkey broths were sent from the collecting sites to a central reference laboratory (Clinical Microbiology Institute, OR, USA) for further processing. Upon receipt at the central reference laboratory, an aliquot of 100-μl was removed from the MacConkey broth and inoculated onto a MacConkey agar plate after which a 10-μg meropenem disk was placed in the center of the plate. Colonies on the MacConkey agar plate that grew within 27 mm of the meropenem disk after an overnight incubation at 35°C were further tested for the presence of a carbapenemase, as described by Lolans et al. ([@B18]) with some modifications. Meropenem was used instead of ertapenem to improve the specificities for Pseudomonas spp. and Acinetobacter spp., which are intrinsically resistant to ertapenem. Colonies were also tested by a standard disk diffusion method, and those classified as carbapenem-nonsusceptible (i.e., intermediate or resistant to ertapenem, imipenem, or meropenem using the Clinical and Laboratory Standards Institute \[CLSI\] interpretive criteria \[[@B19]\]) were subjected to bidirectional DNA sequence analyses for the identification of the *bla*~KPC~, *bla*~NDM~, *bla*~OXA-48~, *bla*~IMP-1~, and *bla*~VIM~ genes. The bidirectional DNA sequence analyses were performed at an independent laboratory (ACGT Inc., Wheeling, IL, USA) using primers that were different from those used in the Xpert Carba-R assay.
All contrived swab specimens were processed at the central reference laboratory in the same manner as the clinical rectal swab specimens. They were inoculated into MacConkey broth with a meropenem disk and subsequently subcultured to MacConkey agar. Isolated colonies were tested for susceptibility.
A positive reference culture result for both clinical and contrived specimens was defined as the isolation of a carbapenem-nonsusceptible organism that contained a *bla*~KPC~, *bla*~NDM~, *bla*~OXA-48~, *bla*~IMP-1~, and/or *bla*~VIM~ gene confirmed by DNA sequence analysis. A negative reference culture result was defined as either a reference culture that did not yield any carbapenem-nonsusceptible organisms or the isolation of a carbapenem-nonsusceptible organism that did not contain any of the target carbapenemase genes by DNA sequence analysis.
To minimize bias in specimen analyses, the laboratory personnel performing the reference culture and DNA sequencing were not aware of the Xpert Carba-R assay results.
Xpert Carba-R assay. {#sec2-3}
--------------------
Xpert Carba-R assay testing was performed using the second swab from the double swab set with the GeneXpert platform. Testing was performed at the laboratories in each of the 4 participating institutions according to the manufacturer\'s recommendations and as previously described ([@B15]) within 24 to 48 h from collection. The assay has a run time of ∼47 min in the instrument. Quality control for the Xpert Carba-R assays consisted of one positive and one negative control. The former was composed of an Escherichia coli isolate containing a plasmid that included DNA fragments of all five target gene sequences, and the latter was the same E. coli isolate with the cloning vector but without the cloned fragments. The controls (product M219) were produced by Maine Molecular Quality Controls, Inc. (<http://www.mmqci.com/qc-m219.php>). The two controls were run on each day that specimens were tested. Study specimens were not tested until correct results were obtained for the negative and positive controls.
All contrived swab specimens were tested with the Xpert Carba-R assay in the same manner as the clinical rectal swab specimens.
Data analysis, discrepant results, and statistics. {#sec2-4}
--------------------------------------------------
The Xpert Carba-R assay results were compared with reference culture and DNA sequencing results. An Xpert Carba-R-positive result was considered when the Xpert Carba-R assay detected the presence of at least one carbapenemase gene. For each specimen, results for each of the five carbapenemase target genes included in the Xpert Carba-R assay were reported separately, and these were compared to results for the five carbapenemase target genes obtained from reference culture and DNA sequencing.
A discrepant result was defined as a result obtained with the Xpert Carba-R assay that did not correlate with the results of reference culture and DNA sequencing of the same specimen. Discrepant testing was performed for only those specimens that were positive for a target gene by the Xpert Carba-R assay but did not show growth of colonies by culture. DNA was extracted from the MacConkey broth (100 μl) using the Qiagen DNeasy blood and tissue kit using the protocol for Gram-negative bacteria and amplified by PCR using primers corresponding to the targets that were positive by the Xpert Carba-R assay results. If one of the five targets was identified, bidirectional DNA sequence analysis was performed. A positive result was considered a true positive for discrepant resolution by the reference method in the analysis.
Discrepant testing was not performed for specimens that were negative for all target genes by the Xpert Carba-R assay but showed growth of colonies on the MacConkey agar plate and from which one of the five target genes was identified through DNA sequence analysis.
Sensitivity, specificity, and positive (PPV) and negative (NPV) predictive values were also calculated for each target carbapenemase gene ([@B20]).
Statistical analysis was performed using 95% confidence intervals (CIs) calculated with the Clopper-Pearson/Fisher exact CI ([@B21]) using Minitab version 16 (Minitab, State College, PA). Values for the Kappa coefficient, which gives a measure of the percentage of agreement between the Xpert Carba-R assay and the reference method (after resolution of discrepant results) beyond that expected by chance, were also calculated ([@B22]).
RESULTS {#sec3}
=======
Xpert Carba-R assay results with prospective rectal and contrived swab specimens. {#sec3-1}
---------------------------------------------------------------------------------
A total of 633 samples (383 clinical rectal swabs and 250 contrived specimens) were included. Overall, 149/633 (23.5%) samples were positive by the Xpert Carba-R assay, and 484/633 (76.5%) were negative. The Xpert Carba-R assay amplified 155 positive targets (26 *bla*~IMP-1~, 30 *bla*~VIM~, 27 *bla*~NDM~, 33 *bla*~KPC~, 39 *bla*~OXA-48~). In 6 samples, multiple targets were detected (4 *bla*~VIM~-*bla*~OXA-48~, 1 *bla*~IMP-1~-*bla*~NDM~, 1 *bla*~NDM~-*bla*~KPC~). Xpert Carba-R results stratified by clinical and contrived specimen type are shown in [Table 1](#T1){ref-type="table"}. Data analyses included discrepant resolution results (see below). All positive results were obtained within 32 to 48 min.
Comparison of the Xpert Carba-R assay with the reference method. {#sec3-2}
----------------------------------------------------------------
Results from the Xpert Carba-R assay by individual target compared to the reference method (culture plus sequencing) for all specimens are shown in [Table 2](#T2){ref-type="table"}. There were a total of 633 specimens, each with results for five individual targets for a total of 3,165 results. Specimens that were negative by both the Xpert Carba-R assay and the reference method (3,004 total) were stratified as follows: 606 for IMP-1, 601 for VIM, 606 for NDM, 599 for KPC, and 592 for OXA-48.
######
Results from the Xpert Carba-R assay and the reference method (culture plus sequencing) by individual target for combined clinical and contrived specimens
Xpert Carba-R assay Reference method (culture plus sequencing)
--------------------- -------------------------------------------- ---- ---- ---- ---- ------- -------
IMP-1 26 0 0 0 0 0 26
VIM 0 29 0 0 0 1 30
NDM 0 0 26 0 0 1 27
KPC 0 0 0 29 0 4 33
OXA-48 0 0 0 0 38 1 39
Negative 1 2 0 1 2 3,004 3,010
Total 27 31 26 30 40 3,011 3,165
[Table 3](#T3){ref-type="table"} shows the overall performance of Xpert Carba-R compared with that of the reference method. The results were defined as positive for the Xpert Carba-R assay if any of the carbapenemase targets were positive, and they were defined as negative for the Xpert Carba-R assay if all of the carbapenemase targets were negative. On the combined set of clinical and contrived specimens, sensitivity, specificity, PPV, and NPV were 96.6% (95% CI, 92.2% to 98.9%), 98.6% (95% CI, 97.1% to 99.4%), 95.3%, and 99.0%, respectively. The Kappa index for this comparison was 0.933 (0.900 to 0.967).
######
Overall Xpert Carba-R performance versus that of the reference method (culture plus sequencing) for combined clinical and contrived specimens
Xpert Carba-R assay Reference method (culture plus sequencing)
--------------------- -------------------------------------------- ----- -----
Positive 142 7 149
Negative 6 478 484
Total 148 485 633
The performance of the Xpert Carba-R assay for different targets, including CI, is shown in [Table 4](#T4){ref-type="table"}. Results included in [Tables 2](#T2){ref-type="table"} to [4](#T4){ref-type="table"} were also analyzed when discrepant results were resolved (see below).
######
Summary of Xpert Carba-R performance for different carbapenemase targets versus that of the reference method (culture plus sequencing) for combined clinical and contrived specimens
Target gene Sensitivity (% \[95% CI\]) Specificity (% \[95% CI\]) PPV (%) NPV (%)
------------- ---------------------------- ---------------------------- --------- ---------
IMP-1 96.3 (81.0--99.9) 100 (99.4--100) 100 99.8
VIM 93.5 (78.6--99.2) 99.8 (99.1--100) 96.7 99.7
NDM 100 (86.8--100) 99.8 (99.1--100) 96.3 100
KPC 96.7 (82.8--99.9) 99.3 (98.3--99.8) 87.9 99.8
OXA-48 95.0 (83.1--99.4) 99.8 (99.1--100) 97.4 99.7
Discrepant results and resolution. {#sec3-3}
----------------------------------
Six specimens (4 prospective and 2 contrived specimens) were negative for all target genes by Xpert Carba-R assay but were positive by the reference method (culture and DNA sequencing) and were considered to be false negatives. Bidirectional DNA sequence analysis identified 1 *bla*~IMP-1~, 2 *bla*~VIM~, 1 *bla*~KPC~, and 2 *bla*~OXA-48~ genes ([Table 2](#T2){ref-type="table"}). In addition, 18 specimens (15 prospective and 3 contrived specimens) were positive for at least one target gene by the Xpert Carba-R assay and were negative by the reference method ([Table 5](#T5){ref-type="table"}). Eleven of the 18 specimens were considered to be true positives after discrepant analyses, which included direct PCR analysis and sequencing from the MacConkey broth. Of these 11 specimens, there were 5 out of the 6 specimens in which multiple carbapenemase targets were detected by Xpert Carba-R assay. The remaining 7 specimens were considered to be false positives, as no PCR amplification was obtained from testing the MacConkey broth ([Table 5](#T5){ref-type="table"}). Targets detected by Xpert Carba-R assay were 4 KPC and one each of NDM, VIM, and OXA-48. In all of these cases, positive results were fairly late (threshold cycles \[*C~T~*s\] ranged from 26.2 to 28.2 for KPCs and were \>31 for the remaining carbapenemases).
######
Discrepant testing results[^*a*^](#T5F1){ref-type="table-fn"}
Sample Reference method Target(s) detected by Xpert Carba-R assay Discrepant analysis
-------- ------------------------- --------------------------------------------- --------------------- ---------------- ----
A No isolate NA[^*c*^](#T5F3){ref-type="table-fn"} IMP-1 IMP-1 TP
B Acinetobacter baumannii No bands[^*d*^](#T5F4){ref-type="table-fn"} KPC KPC TP
C Klebsiella pneumoniae OXA-48 OXA-48, VIM OXA-48, VIM TP
D No isolate NA OXA-48 OXA-48 TP
E Klebsiella pneumoniae OXA-48 OXA-48, VIM OXA-48, VIM TP
F No isolate NA VIM VIM TP
G No isolate NA KPC KPC TP
H No isolate NA OXA-48, VIM OXA-48, VIM TP
I No isolate NA OXA-48 OXA-48 TP
J Enterobacter cloacae KPC KPC, NDM KPC, NDM TP
K Klebsiella pneumoniae OXA-48 OXA-48, VIM OXA-48, VIM TP
L No isolate NA KPC No target gene FP
M No isolate NA NDM No target gene FP
N No isolate NA VIM No target gene FP
O No isolate NA KPC No target gene FP
P No isolate NA KPC No target gene FP
Q No isolate NA KPC No target gene FP
R No isolate NA OXA-48 No target gene FP
True-positive and false-positive results are shown.
TP, true positive; FP, false positive.
NA, not applicable.
No bands indicate that the sample did not meet PCR product band size inclusion criteria.
DISCUSSION {#sec4}
==========
Detection of fecal carriers of CPO has become a routine clinical practice in many parts of the world and is recommended by public health organizations for the containment of the spread of these isolates ([@B3], [@B23]). Different approaches are used in different countries and institutions, with some culturing rectal swabs on chromogenic media. More recently, molecular approaches have been developed to increase detection sensitivity and decrease reporting time ([@B7], [@B24]). We evaluated, through a prospective multisite study, the Cepheid Xpert Carba-R assay, a PCR-based method in a cartridge format developed for detecting carbapenemase genes in rectal swabs and run on the GeneXpert platform. This assay was previously evaluated, but only three carbapenemase targets (*bla*~KPC~, *bla*~NDM~, and *bla*~VIM~) were included ([@B15]). In the current version, two additional targets were incorporated, *bla*~OXA-48~ and *bla*~IMP-1~, to broaden the scope of the carbapenemases detected. OXA-48 is increasingly recognized in most European countries, predominating in most of the Mediterranean countries, whereas IMP-1 is increasingly detected in Asia and Latin American countries ([@B3], [@B25]).
Overall, 633 specimens, including clinical rectal swabs and contrived samples, were studied, yielding the potential detection of 3,165 carbapenemase targets. Due to the absence or low prevalence of certain carbapenemase genes in organisms found in clinical specimens, such as IMP-1, VIM, or NDM carbapenemases, that reflects the local epidemiology of participating centers, contrived specimens were also included in the evaluation. This approach is normally used in molecular diagnostic platforms for *in vitro* and clinical evaluations.
For our evaluation, the LODs of different carbapenemase targets were previously calculated (see Materials and Methods) ([@B17]). Additionally, analytical reactivity (inclusivity) and potential cross-reactivity (exclusivity) were also calculated, yielding 100% detection of target-containing organisms and no cross-reactions with other resistance genes, respectively ([@B17]). Moreover, it is notable that, in our evaluation, we used a highly sensitive method to resolve potential discrepancies between the reference method (culture plus DNA sequencing) and the Cepheid Xpert Carba-R assay. This consisted of DNA extraction and sequencing from the selective enrichment broth (MacConkey with a 10-μg meropenem disk) inoculated with the original specimens. With this approach, overall performance was excellent with sensitivity, specificity, PPV, and NPV all higher than 95%. When performance was evaluated separately for each carbapenemase, the performance was best for IMP-1, and only sensitivities and positive predictive values for VIM and KPC targets were lower than 95%. Unlike other systems for direct detection of carbapenemase genes in rectal swabs, we did not see high false-positive rates for metallo-beta-lactamase targets ([@B26]).
Our multisite prospective evaluation was performed for the initial European marketing authorization of the Xpert Carba-R assay and was carried out under the European directive on *in vitro* diagnostic medical devices ([@B16]). A limitation of our evaluation was that the Xpert Carba-R assay included targets for widespread carbapenemases at the time of the evaluation but did not include some of the new emerging OXA-48 variants, such as OXA-181 and OXA-232 carbapenemases. OXA-181 is one of the most prevalent carbapenemases in South Africa and is increasingly recognized in Asia and Europe ([@B27][@B28][@B29]), while OXA-232 is also increasingly detected in Asia ([@B30]). Recent reports demonstrated that the Xpert Carba-R version evaluated in our study was unable to detect these two variants ([@B31][@B32][@B33]). Nevertheless, this was corrected in a novel Xpert Carba-R version recently marketed, which accurately detected these OXA-48 variants ([@B34]). The absence of detection of the OXA-181 carbapenemase was also noted with other commercial molecular assays that did not originally include this target. However, as in the Xpert Carba-R assay, subsequent modification of the assay allowed detection of OXA-181 ([@B34]).
In our study, with the exception of 6 specimens ([Table 2](#T2){ref-type="table"}) in which false-negative results were obtained, the Xpert Carba-R assay accurately detected all different targeted carbapenemase genes, including those encoding the KPC, NDM, VIM, IMP-1, and OXA-48 enzymes. The absence of detection in these 6 specimens was not associated with a specific carbapenemase and may be associated with low bacterial load in the rectal swab specimens.
An additional advantage of the Cepheid Xpert Carba-R assay is the ability to detect multiple resistance genes, an outcome not previously reported in other evaluations ([@B31]). This occured in 6 cases ([Tables 1](#T1){ref-type="table"} and [5](#T5){ref-type="table"}). In 4 of them, the reference culture method failed to detect this situation, but the genes were subsequently detected by PCR and DNA sequencing from the MacConkey broth. A potential explanation for these results may be that the organisms recovered by reference culture are present in larger amounts, which would limit the growth of the other organisms present in a lower proportion, or because both carbapenemase-bearing isolates were morphologically identical and one of them was likely missed during the selection of the colonies.
However, a positive result with the Xpert Carba-R assay but negative by culture may happen if the patient had been receiving antibiotics, or the organism may carry a modified sequence of the target gene, which was not expressed or was expressed at low levels.
In conclusion, the Cepheid Xpert Carba-R assay is an accurate technique for the detection of CPO in rectal swab specimens. This assay allows rapid identification of patients colonized with CPO, which can guide infection control programs designed to limit the spread of these organisms in health care settings.
We thank Mona Patel, employee from Cepheid, for technical support during the evaluation.
Funding for this study was provided by Cepheid.
R.C. and P.R.-G. have participated in educational programs supported by Cepheid. The other authors have no conflicts of interest.
[^1]: **Citation** Tato M, Ruiz-Garbajosa P, Traczewski M, Dodgson A, McEwan A, Humphries R, Hindler J, Veltman J, Wang H, Cantón R. 2016. Multisite evaluation of Cepheid Xpert Carba-R assay for detection of carbapenemase-producing organisms in rectal swabs. J Clin Microbiol 54:1814--1819. doi:[10.1128/JCM.00341-16](http://dx.doi.org/10.1128/JCM.00341-16).
| {
"pile_set_name": "PubMed Central"
} |
Rheumatology key messagesA quarter of RA patients show only low to moderate adherence to adalimumab.Illness and treatment beliefs are the major influences on adherence to adalimumab among RA patients.Higher perceived support from health professionals and family may improve adherence to adalimumab in RA.
Introduction
============
Medication adherence is defined as the extent to which a patient's behaviour in taking their medication corresponds to agreed recommendations by their health care provider \[[@kev105-B1]\]. More than one-third of therapies are not taken as recommended, irrespective of the seriousness of disease or condition \[[@kev105-B2]\]. There is increasing recognition of lower adherence even in symptomatic diseases such as RA with medication adherence rates reported between 55% and 96% \[[@kev105-B3]\]. Little is known about biologic drug adherence in RA, and studies are further limited as adherence rates tend to be derived from proxy measures, including medication persistence (time from prescription initiation to prescription discontinuation), drug survival or medication possession ratios from administrative claims data \[[@kev105-B7]\]. With the wider use of biologic therapy in RA, together with reported low medication possession ratios and persistence rates suggested in RA in general, there is a clear need to investigate adherence rates of biologic therapy in real-world practice.
The impact of medication non-adherence may be considerable; adherent patients have more favourable outcomes \[[@kev105-B12]\], including better disease control, higher remission rates and improved physical function \[[@kev105-B13], [@kev105-B14]\], as well as lower rates of disease progression and escalation to further aggressive treatment \[[@kev105-B15], [@kev105-B16]\]. Biologic therapies also have high lifetime costs to the health care system \[[@kev105-B17]\], and lower persistence to biologic therapy is associated with higher non-pharmacy costs \[[@kev105-B18]\].
Adherence is recognized to require sustained behavioural change, influenced by both environmental and psychological factors. In RA, influences on adherence include age \[[@kev105-B10], [@kev105-B14]\], ethnicity \[[@kev105-B19]\], socio-economic factors \[[@kev105-B5], [@kev105-B20]\], complexity of treatment \[[@kev105-B21]\] and RA disease-specific factors such as inflammatory markers (ESR) and disease activity; however, it is important to note that findings are not consistent across studies. In other disease groups, the important influences of patients' illness perception and medication beliefs on adherence behaviour \[[@kev105-B22], [@kev105-B23]\], as directed by the extended Self-Regulatory Common Sense Model (SR-CSM) of illness \[[@kev105-B24]\] and treatment \[[@kev105-B25]\] have been highlighted. According to the SR-CSM, an individual's illness perception, such as beliefs about disease consequences or perceived personal control, influence coping, including self-management strategies, in response to the perception of a health threat \[[@kev105-B24]\]. A further extension to this model is the necessity-concern framework \[[@kev105-B25]\], suggesting patients' beliefs about their medication, including the perceived need for and/or concerns about medication use, are an influence on medication adherence behaviour.
The aim of the current study was to investigate the level of adherence to a biologic therapy longitudinally using an RA-specific measure of adherence. We also sought to determine the relative contribution of demographic factors, RA disease-specific influences and psychological behavioural influences on adherence in a single prospective cohort.
Methods
=======
Setting and recruitment
-----------------------
Patients were recruited through the British Society for Rheumatology Biologics Register for RA (BSRBR-RA), a UK-wide RA prospective observational cohort study established in 2001 to monitor the long-term safety of biologic therapy use in RA \[[@kev105-B26], [@kev105-B27]\]. Patient eligibility for a biologic drug followed existing National Institute for Health and Care Excellence criteria which were: satisfying 1987 ACR classification criteria for RA \[[@kev105-B28]\], having active disease with a 28-item DAS (DAS28) score \[[@kev105-B29]\] \>5.1 and failing two or more previous synthetic DMARDs (sDMARDs), including MTX. All patients were clinically diagnosed by their treating physician. It was the physician's decision to initiate a biologic, as well as the chosen biologic therapy, and no specific exclusion criteria applied. The level of provision of information about the biologic therapy or education about its use was based on the centre's routine practice and not contingent upon the individual's participation in the study. This substudy focused on patients starting s.c. adalimumab (ADA) as their first biologic drug between May 2007 and April 2009. This was the main biologic drug under active recruitment to the study at this time. All patients gave written informed consent prior to inclusion and this study was approved by the North West Research Ethics Committee (REC:MREC 00/8/053).
Data collection
---------------
### Baseline (start of treatment)
After written informed consent was obtained, the local centre provided the year of diagnosis, 1987 ACR criteria fulfilled and the DAS28 score \[[@kev105-B29]\]. A dichotomized variable was derived for an acute phase response from age- and gender-adjusted upper limits of the normal range of ESR \[[@kev105-B30]\] or CRP \[[@kev105-B31]\]. Patients provided date of birth, gender, ethnicity, work status, smoking status and postcode for calculation of socio-economic status using the country-specific Index of Multiple Deprivation \[[@kev105-B32]\]. Patients also returned the following self-reported questionnaires: Stanford HAQ \[[@kev105-B35]\]; 36-item Short Form Health Survey (SF-36) \[[@kev105-B36]\] and the EuroQol five-dimensions questionnaire (EQ-5D) \[[@kev105-B37]\] using transformed weighted health state index scores \[[@kev105-B38]\] and dichotomizing (≥0.516). Additional measures used were the Revised Illness Perception Questionnaire (IPQ-R) \[[@kev105-B39]\], capturing illness beliefs based on the SR-CSM \[[@kev105-B24]\], where higher scores across domains are indicative of a greater sense of symptomology, the acute long-lasting and cyclical nature of the disease, understanding and ultimate consequence of the disease, personal treatment control and a higher emotional state. The higher scores on the Beliefs about Medicines Questionnaire \[[@kev105-B25]\] are indicative of a stronger feeling of medication need and concern towards medication use. Also used were the Hospital Anxiety and Depression Scale (HADS) \[[@kev105-B40]\] and an adaptation of the Daily Coping Inventory, which assesses the level of coping based on the number of preclassified strategies adopted by an individual \[[@kev105-B41]\].
### Follow-up
Patients were mailed follow-up questionnaires, including the HAQ and SF-36, at 6-monthly intervals (6, 12 and 18 months). Postal reminders were sent at 2 weeks and a further reminder with repeat questionnaire at 4 weeks for baseline or follow-up non-returners. The local centre provided the DAS28 score.
### Adherence
As the main outcome measure, adherence data were collected at 6, 12 and 18 months after the baseline measures were recorded. The 19-item Compliance Questionnaire for Rheumatology (CQR) \[[@kev105-B42]\] has been validated against other adherence measures \[[@kev105-B5], [@kev105-B6], [@kev105-B20]\], including the Medication Events Monitoring system \[[@kev105-B43]\]. Patients rate their agreement with 19 statements using a 4-point Likert scale. The adjusted total score ranges from 0 to 100 (100 indicating the highest possible adherence) and is used as a continuous scale.
Analysis
--------
After assessing for attrition and assigning missing data as missing at random, appropriate application of Multiple Imputation for Chained Equations \[[@kev105-B44]\] for 40 imputations provided the imputed datasets. Complete case (CC) results are presented with reference to imputed findings where appropriate.
Multilevel mixed effects modelling analysis was performed on 329 individuals with CC baseline data to describe the longitudinal relationship between adherence (CQR score) and potential predictors \[[@kev105-B45]\]. This allowed for within- and between-patient and follow-up variability of adherence score over time. In addition, a mixed model approach assumes data are missing at random, rather than missing completely at random. This allows all individuals to be retained and their available data utilized, whether or not complete, to address potential bias issues. Sixty-two per cent of the total variance in the CQR score was represented at patient level (unconditional model: intraclass correlation 0.62, between-patient variance 67.29, time variance 40.94).
### Univariate analysis
Random intercept models were applied to each predictor variable to determine their prognostic value, controlling for age at follow-up, gender, social deprivation and disease duration. Interaction terms between the effects of each predictor on adherence at each follow-up period were retained where significant. Each significant predictor from the intercept models (*P* \< 0.05) were further modelled by including their random slope and then tested using likelihood ratio tests. The inclusion of random slopes was not warranted for all predictors.
### Multivariate analysis
A multivariate model was determined by adding and retaining any significant univariate factors with applicable interactions (*P* \< 0.05). Models returned maximum likelihood estimators and their efficiency was assessed using the Akaike's Information Criterion. Further diagnostics were performed by inspection of normality of residuals. All analysis was performed using STATA 11.2 (StataCorp, College Station, TX, USA).
Results
=======
Response rates
--------------
Of 713 patients commencing ADA, 557 (78.1%) returned a baseline questionnaire ([Fig. 1](#kev105-F1){ref-type="fig"}). A high response rate (\>75%) was maintained throughout the follow-up period. No systematic differences were observed between questionnaire returners and non-returners, including gender, age at disease onset and disease activity. More non-returners resided in the most socially deprived quartile \[50/156 (32.1%) *vs* 102/557 (18.3%)\], although in the final analysis each quartile was well represented ([Table 1](#kev105-T1){ref-type="table"}). F[ig]{.smallcaps}. 1Questionnaire returners and response rates during 18 months of follow-up T[able]{.smallcaps} 1Descriptive information of complete case baseline data of adalimumab-treated RA patientsVariableValueAge at onset, mean ([s]{.smallcaps}.[d]{.smallcaps}.), years44.25 (13.26)Age at registration, mean ([s]{.smallcaps}.[d]{.smallcaps}.), years55.92 (12.27)Female gender, *n* (%)257/329 (78.1)Ethnicity, *n* (%) White320/329 (97.3) Black African0 Black British1/329 (0.3) Indian3/329 (0.9) Pakistani1/329 (0.3) Bangladeshi0 Other4/329 (1.2)Social deprivation quartile, *n* (%) 1 (least deprived)87/329 (26.4) 299/329 (30.0) 379/329 (24.0) 4 (most deprived)64/329 (19.4)Working status, *n* (%) Working125/329 (37.9) Not working due to illness84/329 (25.5) Retired120/329 (36.6)Ever smoker, *n* (%), yes/no190/329 (57.6)Disease activity Number of baseline sDMARDs, *n* (%) 062/329 (18.8) 1161/329 (48.9) 277/329 (23.4) 329/329 (8.8)Disease duration, mean ([s]{.smallcaps}.[d]{.smallcaps}.), years11.60 (9.22)Satisfy ACR criteria, *n* (%), yes/no275/329 (83.6)Morning stiffness, *n* (%), yes/no312/329 (94.8)Involvement of \>3 joints, *n* (%), yes/no274/329 (83.3)Involvement of hand joint, *n* (%), yes/no259/329 (78.7)Symmetry, *n* (%), yes/no272/329 (82.7)Nodules, *n* (%), yes/no120/329 (36.5)RF positive, *n* (%), yes/no217/329 (66.0)Erosions on X-ray, *n* (%), yes/no193/329 (58.7)Swollen joint count (0--28), mean ([s]{.smallcaps}.[d]{.smallcaps}.)11 (6)Tender joint count (0--28), mean ([s]{.smallcaps}.[d]{.smallcaps}.)16 (7)Inflammatory marker, *n* (%), yes/no177/329 (53.8)Disease activity score (0--9.3), mean ([s]{.smallcaps}.[d]{.smallcaps}.) Baseline6.44 (0.94) 6 months (*n* = 295)3.95 (1.52) 12 months (*n* = 278)3.77 (1.59) 18 months (*n* = 257)3.57 (1.53)DAS patient global score (0--100), mean ([s]{.smallcaps}.[d]{.smallcaps}.)73.13 (16.99)Psychological factors IPQ-R domains, mean ([s]{.smallcaps}.[d]{.smallcaps}.) Disease identity (0--14)6.44 (2.33) Timeline acute/chronic (6--30)26.03 (3.28) Consequences (6--30)23.18 (3.73) Personal control (6--30)18.76 (4.24) Treatment control (5--25)17.71 (2.50) Illness coherence (5--25)18.63 (3.57) Timeline cyclical (4--20)14.67 (2.89) Emotional (6--30)19.85 (4.38) HADS (0--21), mean ([s]{.smallcaps}.[d]{.smallcaps}.) Anxiety7.57 (4.30) Depression6.79 (3.94) BMQ (5--25), mean ([s]{.smallcaps}.[d]{.smallcaps}.) Necessity21.54 (2.68) Concern14.95 (3.42) EQ-5D utility score \>0.516, *n* (%) Baseline174/329 (52.9) 6 months231/289 (79.9) 12 months215/267 (80.5) 18 months212/256 (82.8) EQ-5D (baseline) VAS health today (0--100), mean ([s]{.smallcaps}.[d]{.smallcaps}.)45.56 (20.72) EQ-5D (baseline) health in last 12 months, *n* (%) Better33/329 (10.0) Same101/329 (30.6) Worse195/329 (59.3) Coping, mean ([s]{.smallcaps}.[d]{.smallcaps}.) Problem focused (4--16)12 (10, 13) Emotionally focused (4--16)10 (8, 12) Support (family) (2--8)7 (6, 8) Support (religion) (1--4)1 (1, 2)Functional disability HAQ (0--3), mean ([s]{.smallcaps}.[d]{.smallcaps}.) Baseline (*n* = 329)1.80 (0.62) 6 months (*n* = 273)1.43 (0.77) 12 months (*n* = 255)1.43 (0.77) 18 months (*n* = 247)1.42 (0.77) SF-36 baseline domains (0--100), mean ([s]{.smallcaps}.[d]{.smallcaps}.) Physical function26.87 (22.85) Physical role25.06 (25.31) Bodily pain27.47 (16.89) General health32.77 (19.10) Vitality25.76 (19.16) Social42.63 (24.96) Emotional53.17 (34.62) Mental health57.51 (20.48) Physical Component Summary, mean ([s]{.smallcaps}.[d]{.smallcaps}.) Baseline (*n* = 329)18.72 (9.41) 6 months (*n* = 277)27.04 (13.02) 12 months (*n* = 255)27.25 (13.32) 18 months (*n* = 245)28.13 (13.42) Mental Component Summary, mean ([s]{.smallcaps}.[d]{.smallcaps}.) Baseline (*n* = 329)44.12 (11.30) 6 months (*n* = 277)48.61 (12.53) 12 months (*n* = 255)50.11 (11.69) 18 months (*n* = 245)49.60 (11.92)[^1]
Of the 557 patients returning questionnaires, 329 (59.07%) had complete-case data at baseline and were included in the analysis. For each follow-up, those with a complete CQR score contributed to the analysis. The missing CQR scores at follow-up were a combination of incomplete items needed to generate the CQR (6, 12 and 18 month follow-up; *n* = 30, 32 and 33, respectively) or returning a completely unanswered CQR (6, 12 and 18 months follow-up; *n* = 58, 67 and 88, respectively). More than 50% of CQR scores were available in individuals switching or stopping medication, with no pattern observed in those with or without a CQR score during follow-up ([supplementary Table S1](http://rheumatology.oxfordjournals.org/lookup/suppl/doi:10.1093/rheumatology/kev105/-/DC1), available at *Rheumatology* Online). The imputed data consisted of 556 individuals (one observation omitted because it was a severe outlier) and 40 imputations over baseline and 3 follow-ups.
In the complete-case dataset at baseline there were 257/329 (78%) women with a mean age at symptom onset of 44 years ([s]{.smallcaps}.[d]{.smallcaps}. 13) ([Table 1](#kev105-T1){ref-type="table"}). The majority were of white British ethnicity and across all levels of social deprivation (between 20% and 30% in each quartile). Two hundred and fourteen (62.1%) were not working due to either illness or retirement and 190 (57.1%) had ever smoked.
The mean CQR score remained \<75 over the follow-up. After dividing CQR scores into quartiles, 56 (23.2%), 54 (23.7%) and 48 (23.3%) patients for 6, 12 and 18 months, respectively, had adherence scores in the lowest quartile (CQR score 40--64) ([Table 2](#kev105-T2){ref-type="table"}). Variability was noted in individuals' adherence score over time, such that 41% of those returning all follow-ups (*n* = 59/143) reported a CQR score of 40--64 at least once over the course of the follow-up. T[able]{.smallcaps} 2Adherence as indicated by the CQR score over the follow-upFollow-up period, monthsCQR total score, mean ([s]{.smallcaps}.[d]{.smallcaps}.)CQR quartiles, *n* (%)Least adherentMost adherent40--6465--7475--8384+6 (*n* = 241)74.99 (10.40)56 (23.24)57 (23.65)70 (29.05)58 (24.07)12 (*n* = 228)74.96 (10.75)54 (23.68)57 (25.00)57 (25.00)60 (26.32)18 (*n* = 206)74.59 (10.54)48 (23.30)60 (29.13)52 (25.24)46 (22.33)[^2]
Influence of demographic factors
--------------------------------
Older age was the only demographic factor significantly associated with an increase in the CQR score \[β coefficient 0.14 ([s]{.smallcaps}.[e]{.smallcaps}. 0.04), *P* = 0.001\] ([Table 3](#kev105-T3){ref-type="table"}). T[able]{.smallcaps} 3Univariate random intercept models reflecting the predictors' influence on CQR score in ADA-treated patientsPredictorβ~0~ intercept constant, coefficient ([s]{.smallcaps}.[e]{.smallcaps}.)β~1~, coefficient ([s]{.smallcaps}.[e]{.smallcaps}.)Random intercept variance, estimate ([s]{.smallcaps}.[e]{.smallcaps}.)Overall error (residual variance), estimate ([s]{.smallcaps}.[e]{.smallcaps}.)Demographic Age at questionnaire, years67.19 (2.64)0.14 (0.04)\*\*63.44 (6.17)40.79 (2.12) Gender, female67.19 (2.64)0.76 (1.19)63.44 (6.17)40.79 (2.12) Social deprivation Quartile 2 (least deprived)67.19 (2.64)1.40 (1.03)63.44 (6.17)40.79 (2.12) Quartile 367.19 (2.64)−0.11 (1.39)63.44 (6.17)40.79 (2.12) Quartile 4 (most deprived)67.19 (2.64)1.36 (1.48)63.44 (6.17)40.79 (2.12) Ever smoked66.99 (2.66)0.43 (1.01)63.29 (6.16)40.86 (2.22)Disease activity Number of baseline sDMARDs66.60 (2.85)0.30 (0.58)63.19 (6.16)40.88 (2.22) Disease duration67.19 (2.64)−0.14 (0.06)[\*](#kev105-TF3){ref-type="table-fn"}63.44 (6.17)40.79 (2.12) Satisfy ACR criteria67.72 (2.80)−0.78 (1.34)63.17 (6.15)40.88 (2.22) Morning stiffness64.89 (3.36)2.43 (2.29)63.03 (6.14)40.87 (2.22) Involvement in \>3 joints68.25 (2.83)−1.37 (1.32)63.08 (6.14)40.86 (2.22) Involvement in hand joint68.82 (2.77)−2.17 (1.20)62.49 (6.10)40.88 (2.22) RF positive66.96 (2.73)0.30 (1.04)63.28 (6.16)40.87 (2.22) Erosions on X-ray67.14 (2.67)0.04 (1.03)63.28 (6.16)40.87 (2.22) DAS2868.04 (2.76)−0.13 (0.14)63.08 (6.32)41.30 (2.43) Presence of inflammation67.50 (2.68)−0.68 (0.67)62.45 (6.32)41.68 (2.47)Functional disability HAQ66.83 (2.65)0.09 (0.49)63.19 (6.17)39.99 (2.24) SF-36 domains Physical function67.34 (2.77)−0.002 (0.01)63.87 (6.23)40.47 (2.23) Physical role67.10 (2.68)0.002 (0.009)63.89 (6.23)40.48 (2.23) Bodily pain66.86 (2.69)0.007 (0.01)63.83 (6.23)40.48 (2.23) General health66.89 (2.68)0.003 (0.01)63.81 (6.21)40.27 (2.20) Vitality66.87 (2.67)0.01 (0.01)63.61 (6.19)40.58 (2.21) Social67.01 (2.70)0.003 (0.01)63.68 (6.23)40.64 (2.25) Emotional66.83 (2.72)0.004 (0.009)63.47 (6.21)40.68 (2.25) Mental health + time interactions66.84 (2.90)0.005 (0.02)63.05 (6.13)40.00 (2.18) Fup1[\*](#kev105-TF3){ref-type="table-fn"}mental healthInteraction χ^2^ = 11.12, *P* = 0.01--−0.03 (0.03)---- Fup2[\*](#kev105-TF3){ref-type="table-fn"}mental health--0.05 (0.03)---- Fup3[\*](#kev105-TF3){ref-type="table-fn"}mental health--0.06 (0.03)[\*](#kev105-TF3){ref-type="table-fn"}---- Physical Component Summary66.82 (2.77)−0.0003 (0.03)63.90 (6.26)40.42 (2.25) Mental Component Summary65.60 (2.89)0.03 (0.03)63.84 (6.25)40.38 (2.25)Psychological IPQ-R Disease identity69.02 (2.93)−0.31 (0.21)62.83 (6.12)40.87 (2.22) Timeline acute/chronic52.24 (4.81)0.56 (0.15)\*\*60.36 (5.92)40.83 (2.22) Consequences61.49 (4.18)0.23 (0.13)62.60 (6.10)40.87 (2.22) Personal control66.24 (3.49)0.05 (0.12)63.26 (6.16)40.87 (2.22) Treatment control55.18 (4.45)0.67 (0.20)\*\*60.73 (5.96)40.88 (2.22) Illness coherence56.71 (3.91)0.49 (0.14)\*\*60.42 (5.93)40.86 (2.22) Timeline cyclic65.94 (3.56)0.09 (0.17)63.24 (6.16)40.87 (2.22) Emotional representation63.11 (3.61)0.19 (0.11)62.54 (6.11)40.90 (2.22) HADS Anxiety + time interaction67.82 (2.83)−0.09 (0.13)63.27 (6.15)40.38 (2.19) Fup1[\*](#kev105-TF3){ref-type="table-fn"}anxietyInteraction χ^2^ = 8.11, *P* = 0.04--0.32 (0.13)[\*](#kev105-TF3){ref-type="table-fn"}---- Fup2[\*](#kev105-TF3){ref-type="table-fn"}anxiety--−0.04 (0.14)---- Fup3[\*](#kev105-TF3){ref-type="table-fn"}anxiety--0.05 (0.14)---- Depression68.07 (2.78)−0.13 (0.13)63.04 (6.14)40.87 (2.22) BMQ Necessity + time interaction27.20 (4.77)1.88 (0.19)\*\*48.16 (4.89)39.77 (2.15) Fup1[\*](#kev105-TF3){ref-type="table-fn"}necessityInteraction *X*^2^ = 16.12, *P* = 0.001--−0.43 (0.21)[\*](#kev105-TF3){ref-type="table-fn"}---- Fup2[\*](#kev105-TF3){ref-type="table-fn"}necessity--−0.54 (0.22)[\*](#kev105-TF3){ref-type="table-fn"}---- Fup3[\*](#kev105-TF3){ref-type="table-fn"}necessity--−0.84 (0.22)\*\*---- Concern + time interaction77.20 (3.64)−0.64 (0.16)\*\*60.34 (5.91)40.26 (2.19) Fup1[\*](#kev105-TF3){ref-type="table-fn"}concernInteraction *X*^2^ = 10.66, *P* = 0.014--0.52 (0.17)\*\*---- Fup2[\*](#kev105-TF3){ref-type="table-fn"}concern--0.17 (0.17)---- Fup3[\*](#kev105-TF3){ref-type="table-fn"}concern--−0.008 (0.17)---- EQ-5D Health today67.68 (2.71)−0.01 (0.01)62.95 (6.15)41.08 (2.24) Utility group \> 0.51667.01 (2.67)0.32 (0.59)63.21 (6.18)41.01 (2.24) Coping Problem focused64.91 (3.40)0.19 (0.18)62.97 (6.14)40.89 (2.22) Emotionally focused65.86 (3.08)0.16 (0.19)63.15 (6.15)40.87 (2.22) Family/professional support59.65 (3.40)1.25 (0.37)\*\*60.64 (5.95)40.87 (2.22)[^3][^4]
Influence of disease activity and physical disability
-----------------------------------------------------
Concomitant sDMARD treatment with ADA was used in 287 (81.2%) patients. Individuals had a mean disease duration of 11.6 years ([s]{.smallcaps}.[d]{.smallcaps}. 9.2). They had a high mean baseline DAS28 score of 6.4 ([s]{.smallcaps}.[d]{.smallcaps}. 0.9) and 177 (53.8%) had an increased acute phase response. At 6 months the DAS28 score was 3.95 ([s]{.smallcaps}.[d]{.smallcaps}. 1.5), with this response maintained over the follow-up. In parallel, the HAQ and SF-36 Physical Component Summary (PCS) scores improved over time with response to treatment ([Table 1](#kev105-T1){ref-type="table"}). The DAS28, high acute phase reactants and HAQ score did not predict the CQR score over time, although longer disease duration was associated with a lower adherence score \[β~1~ coefficient −0.15 ([s]{.smallcaps}.[e]{.smallcaps}. 0.06), *P* = 0.009\] ([Table 3](#kev105-T3){ref-type="table"}).
Influence of illness cognitions and mood
----------------------------------------
From the illness perception measures, there were high mean scores in patients' awareness of the long-lasting nature of RA \[timeline 26.0 ([s]{.smallcaps}.[d]{.smallcaps}. 3.28)\], with this domain also predicting increased levels of adherence \[β coefficient 0.56 ([s]{.smallcaps}.[e]{.smallcaps}. 0.15), *P* = 0.0001\] ([Tables 1](#kev105-T1){ref-type="table"} and [3](#kev105-T3){ref-type="table"}). Patients had a high coherent understanding of their illness \[mean 18.6 ([s]{.smallcaps}.[d]{.smallcaps}. 3.57)\] and treatment control \[mean 18.8 ([s]{.smallcaps}.[d]{.smallcaps}. 4.24)\], with both baseline effects significantly increasing the CQR score \[β~1~ coefficient 0.49 ([s]{.smallcaps}.[e]{.smallcaps}. 0.14), *P* = 0.0001 and 0.67 ([s]{.smallcaps}.[e]{.smallcaps}. 0.2), *P* = 0.001, respectively\]. Individuals sought more professional/family support compared with other coping strategies {median 7 \[interquartile range (IQR) 6--8\]} ([Table 1](#kev105-T1){ref-type="table"}). This support was associated with an increase in expected adherence \[β~1~ coefficient 1.25 ([s]{.smallcaps}.[e]{.smallcaps}. 0.37), *P* = 0.001\] ([Table 3](#kev105-T3){ref-type="table"}). Medication necessity was high \[mean 21.54 ([s]{.smallcaps}.[d]{.smallcaps}. 2.68)\] and had an increasing effect on the CQR score \[β~1~ coefficient 1.88 ([s]{.smallcaps}.[e]{.smallcaps}. 0.19), *P* = 0.0001)\]. Increased medication concern was associated with a reduced CQR score \[β~1~ coefficient −0.64 ([s]{.smallcaps}.[e]{.smallcaps}. 0.16), *P* = 0.0005)\], with a significant increase in this effect between baseline and 6 months \[0.52 ([s]{.smallcaps}.[e]{.smallcaps}. 0.17), *P* \< 0.005\] that remained during further follow-up ([Table 3](#kev105-T3){ref-type="table"}).
In the imputed dataset, the importance of medication beliefs was also observed in the univariate analysis. Perceived health in the previous 12 months, utility score (EQ-5D) and SF-36 PCS and Mental Component Summary (MCS) scores were also significant univariately, although with low coefficients ([supplementary Table S2](http://rheumatology.oxfordjournals.org/lookup/suppl/doi:10.1093/rheumatology/kev105/-/DC1), available at *Rheumatology* Online). Anxiety and depression scores from the HADS were not predictive of the adherence score ([Table 3](#kev105-T3){ref-type="table"}).
Independent predictors of adherence
-----------------------------------
In a multivariate analysis of the CC and imputed (IM) datasets it was found that an increased belief in medication necessity was a significant independent predictor of adherence in both the CC \[1.68 ([s]{.smallcaps}.[e]{.smallcaps}. 0.19), *P* = 0.0001\] and IM datasets \[1.65 ([s]{.smallcaps}.[e]{.smallcaps}. 0.18), *P* = 0.0001\], with this baseline effect on the CQR score diminishing over time, as indicated by the significant negative time interaction ([Table 4](#kev105-T4){ref-type="table"}). High medication concerns were also predictive of a lower CQR score \[CC = −0.50 ([s]{.smallcaps}.[e]{.smallcaps}. 0.15), *P* = 0.001; IM = −0.49 ([s]{.smallcaps}.[e]{.smallcaps}. 0.14), *P* = 0.05\], with the effect remaining important throughout follow-up. Increased professional or family support was associated with an increased CQR score \[CC = 0.81 ([s]{.smallcaps}.[e]{.smallcaps}. 0.32), *P* = 0.01; IM = 0.87 ([s]{.smallcaps}.[e]{.smallcaps}. 0.28), *P* = 0.002\]. A stronger perceived view of their illness being chronic \[0.32 ([s]{.smallcaps}.[e]{.smallcaps}. 0.14), *P* = 0.025\] and an increased feeling of treatment control \[0.41 ([s]{.smallcaps}.[e]{.smallcaps}. 0.19), *P* = 0.032\] at baseline also predicted an increased CQR score in the CC dataset. Comparing the random intercept variance of the final model and unconditional model \[63.23 ([s]{.smallcaps}.[e]{.smallcaps}. 6.16) and 43.02 ([s]{.smallcaps}.[e]{.smallcaps}. 4.46), respectively\] showed a potential 20% of individual variance in CQR score being accounted for by medication and illness beliefs and the patient adopting coping strategies at baseline. Weakened influence of an individual's baseline perception of their chronicity of illness (timeline domain) and their perceived treatment control on adherence over time were seen in the imputed dataset. T[able]{.smallcaps} 4Final random intercept models of independent predictors of CQR over time in adalimumab-treated patientsFactors associated with CQR[^a^](#kev105-TF5){ref-type="table-fn"}Complete case modelImputed modelβ coefficient ([s]{.smallcaps}.[e]{.smallcaps}.)β coefficient ([s]{.smallcaps}.[e]{.smallcaps}.)Necessity1.68 (0.19)\*\*1.65 (0.18)\*\* Fup1\*necessity−0.44 (0.21)\*−0.46 (0.20)\* Fup2\*necessity−0.54 (0.21)\*−0.58 (0.20)\*\* Fup3\*necessity−0.86 (0.22)\*\*−0.64 (0.21)\*\*Concern−0.50 (0.15)\*\*−0.49 (0.14)\*\* Fup1\*concern0.50 (0.17)\*\*0.31 (0.16)\* Fup2\*concern0.14 (0.17)0.33 (0.16)\* Fup3\*concern−0.05 (0.17)0.04 (0.17)Support0.81 (0.32)\*0.87 (0.28)\*\*Timeline acute/chronic0.32 (0.14)\*0.24 (0.13)Treatment control0.41 (0.19)\*0.27 (0.17) Fup11.53 (5.17)5.43 (4.84) Fup29.55 (5.25)8.84 (4.87) Fup318.92 (5.48)\*\*14.67 (5.32)\*β~0~ intercept constant18.44 (7.45)23.00 (6.68)$\text{σ}_{\text{u}0}^{\text{2}}$ intercept variance ([s]{.smallcaps}.[e]{.smallcaps}.)43.02 (4.46)48.86 (4.82)$\text{σ}_{\text{E}}^{\text{2}}$ overall error (residual) ([s]{.smallcaps}.[e]{.smallcaps}.)39.18 (2.11)55.92 (3.13)[^5]
To better understand the influence of medication beliefs, we utilized the final CC model to determine predicted CQR scores over time using extreme values of necessity and concern responses. Individuals showing acceptance towards their medication (high necessity and low concern) had a predicted CQR score of 78.7 ([s]{.smallcaps}.[e]{.smallcaps}. 2.0) at 6 months compared with 54.0 ([s]{.smallcaps}.[e]{.smallcaps}. 4.1) for sceptical individuals (low necessity and high concern) ([Table 5](#kev105-T5){ref-type="table"}). T[able]{.smallcaps} 5Predictive margins of CQR score during follow-up for levels of an individuals' treatment beliefIndividualBaseline BMQ domain scorePredicted CQR score (95% CI)NecessityConcern6 months12 months18 monthsAccepting25578.74 (74.81, 82.67)82.50 (78.53, 86.48)82.89 (78.97, 86.82)Ambivalent252578.85 (75.00, 82.69)75.40 (71.47, 79.33)71.93 (67.89, 75.96)Indifferent5553.89 (46.07, 61.72)59.62 (51.67, 67.56)66.42 (58.17, 74.67)Sceptical52554.00 (45.87, 62.11)52.52 (44.23, 60.81)55.46 (47.32, 63.60)[^6]
Discussion
==========
This study is one of the first longitudinal studies assessing adherence to biologic therapy in RA patients using a self-reported measure. More than 50% of individuals had a CQR score \<75, indicating compromised adherence. The levels of adherence found were comparable to previously reported rates for oral sDMARDs using the same CQR adherence measure \[[@kev105-B3], [@kev105-B4], [@kev105-B6], [@kev105-B20]\] and also support observations of low adherence measured by ADA possession ratio or persistence \[[@kev105-B7]\]. To our knowledge, this is the first study to investigate the relative contribution of demographic factors and RA disease-specific and psychological behavioural influences on adherence in a single prospective cohort.
Increased perception of treatment control as an independent predictor of increased adherence was indicative of patients starting an injectable drug such as ADA and retaining some sense of treatment control and high expectations of the new medication \[[@kev105-B46]\]. The association of coping through seeking professional/family support is verified by findings from a meta-analysis of 122 studies published between 1948 and 2001. Patients receiving social support were 3.6 times more likely to adhere to medication than those not receiving support \[[@kev105-B47]\]. The support may reflect the intense pretreatment counselling or indicate patients moving into a more dependent phase of their illness. However, little is known about the most effective type of coping and the association with adherence.
A key observation in our RA cohort was the major influence of patient beliefs on treatment adherence, seen also in other chronic conditions including asthma, hypertension and chronic pain \[[@kev105-B23]\]. The necessity-concern framework \[[@kev105-B25]\] suggests that a patient's adherence decisions are a result of the balance between their perceived need for the medication (necessity) and their concerns regarding its use. We were able to stratify individuals into specific treatment attitudes based on necessity and concern and estimate the potential influence that combinations of beliefs have on adherence. The importance of medication necessity for new users of ADA regardless of concern level was fundamental in predicting increased adherence and requires further work in establishing the stability of both beliefs over the treatment and disease pathway.
Several studies of oral sDMARD adherence, as well as studies in other conditions, have also shown similar importance of medication beliefs \[[@kev105-B4], [@kev105-B5], [@kev105-B23]\]. Non-adherence may therefore owe more to individual patient beliefs than to the actual disease or route of drug administration. A patient's level of medication belief may of course be influenced by the perceived intensity of the drug and/or its mode of administration. However, our data suggest that the influence on adherence remains qualitatively similar across therapy types.
Our findings reflect those of other studies showing that older age is associated with higher adherence \[[@kev105-B14]\]. Others have found that increased age predicted early termination of biologic therapy \[[@kev105-B10]\], although persistence studies may measure stopping due to efficacy or adverse events rather than adherence.
Measures of disease activity and functional disability were not associated with adherence in our study, despite high levels of disease activity and high HAQ scores at baseline. Others have also noted that neither disease duration nor disease activity (using ESR and CRP level) were associated with adherence in RA \[[@kev105-B5]\]. In contrast, Owen *et al.* \[[@kev105-B48]\] found increased ESR and morning stiffness were associated with higher adherence in a univariate analysis (as measured by interview). Our patients had established disease and had a high perceived understanding of the cyclical nature of the disease, a high level of illness coherence and adopted a large number of coping strategies. This long-standing experience of RA may have made them more aware of the implications of further flares and thus to keep taking medication even when feeling better. Also, our assessment of disease activity and disability was sampled over a short period of their overall disease experience. As such, it may have little effect on whether a patient chooses to take his/her medication at a specific time.
Our findings have important implications not only on ultimately reducing the economic impact of non-adherence to biologic therapies, but also in improving routine clinical outcomes. First, it highlights the need for clinicians to be vigilant for potential non-adherence in patients taking biologic therapies. It also highlights the important role of perceived health care support an individual receives, thus the supportive and empathic aspects of clinical practice should be enhanced. Our study has identified potential modifiable patient beliefs. Clinicians therefore need to address the patient's perception of medication need and concerns early in the treatment course. Further, the diminishing effect of the necessity belief over follow-up suggests that the need for medication should be reiterated throughout the treatment course. There is, however, a clear dilemma at an individual level of how patients judge personal need relative to the concern about their medication which influences the motivation of taking the medication. More than 40% of patients in this study had a strong belief in the need for treatment but simultaneously expressed strong concerns about medication use. Non-adherence is often the response to the latter. Identifying and targeting this at-risk group may be of particular value in improving overall clinical outcomes. With the increasing use of biologic therapies, approaches to improve adherence will likely reduce the economic burden by reducing wastage of drugs and avoid further drug escalation. Early evidence has shown promising potential for an SR-CSM-related behaviour intervention to target key cognitions \[[@kev105-B49]\]. Such inventions may be feasible in a routine clinical consultation to improve adherence and thus overall outcomes.
A study such as this has potential limitations. Missing data are inevitable in large observational studies using postal questionnaires. The majority of measures used were composite scores, and unless the initial methodology included a way to handle missing responses, the total score was marked as missing. The analysis of missing data only indicated that younger individuals had more complete data. Patterns of missing data showed an absence of monotone pattern and was indicative of an arbitrary one, which was effectively approached using Multiple Imputation for Chained Equations where appropriate. In addition, the mixed model methodology also allowed all available follow-up data to be utilized without missing scores having any effect on other available scores for the same individual.
Accurately defining and measuring adherence is difficult. Some of the CQR items incorporating attitudinal constructs related to medication taking may potentially confound the relationship between the medication belief scale and adherence. At the time of study, the CQR was the only validated RA adherence questionnaire.
Our study focused on patients starting ADA, because of the time period in which patients were recruited from the BSRBR-RA. There is some evidence that medication possession ratios and persistence rates differ across biologic therapies \[[@kev105-B8], [@kev105-B10], [@kev105-B11]\]. However, those studies did not use the measures employed in this study, so observed variability may be due to adverse event profiles or efficacy differences. However, our results do accord with observations in other chronic disease, suggesting they may be generalized to other ambulatory drugs in RA. It is also possible that our analyses were limited by illness and medication beliefs not being captured beyond the baseline visit. Few longitudinal studies have investigated the stability of beliefs over time, although general beliefs in non-prescribed analgesics have been shown to be stable over time \[[@kev105-B50]\]. Thurah *et al.* \[[@kev105-B4]\] also noted in 65 new users of MTX that concern levels were stable over a 9-month period. Further larger longitudinal work addressing changes in treatment beliefs would be advantageous to inform on patients behavioural influences over time. In addition, co-morbidities were not considered, which may further impact on adherence by the increasing number of medication regimes and the choice of one regime over another for multiple conditions, thus influencing illness and medication concerns. Finally, the nature of the study also prevented recording the influence of the patient--consultant relationship and contact time, the provision and extent of medication and disease information and other patient factors such as self-efficacy, which may be additional influences on medication adherence.
In conclusion, a quarter of patients showed only low to moderate adherence (CQR score \<65) to ADA, a self-administered injectable biologic therapy. Medication beliefs (high concerns and low necessity beliefs) were associated with lower adherence. Increased professional/family support, stronger perceived illness chronicity and an increased feeling of treatment control also predicted adherence over time. These findings highlight the need to prioritize the monitoring of biologic treatment adherence in a chronic symptomatic disease group such as RA.
Supplementary Material
======================
###### Supplementary Data
The British Society for Rheumatology commissioned the Biologics Register in RA (BSRBR-RA) as a UK-wide national project to investigate the safety of biologic agents in routine medical practice. D.P.M.S. and K.L.H. are principal investigators for the BSRBR-RA. I.N.B. is a National Institute for Health Research (NIHR) Senior Investigator and is funded by Arthritis Research UK, the NIHR Manchester Biomedical Research Unit and the NIHR Manchester Wellcome Trust Clinical Research Facility. J.McB. is funded by the Arthritis Research UK Primary Care Centre, Keele University (grant number 20 202). The authors acknowledge the enthusiastic collaboration of all consultant rheumatologists and their specialist nurses in the UK in providing the data. The authors would like to gratefully acknowledge the support of the NIHR through the Comprehensive Local Research Networks at participating centres. In addition, we acknowledge the support of Dr Ian Griffiths (past), Professor David Isenberg (past) and Professor Alex MacGregor (current), chairs of the BSRBR Management Committee, Professor Gabriel Panayi, Professor David G. I. Scott, Dr Andrew Bamji, Professor Deborah Bax and Professor David L Scott, presidents of the BSR during the period of data collection, for their active role in enabling the register to undertake its tasks; and to Samantha Peters (past) and Laura Guest (current), chief executive officer of the BSR, Mervyn Hogg, Nia Taylor, Alan Roach and members of the BSRBR Scientific Steering Committee. The authors also acknowledge the seminal role of the BSR Clinical Affairs Committee for establishing national biologic guidelines and recommendations for such a register. Finally, the authors would like to acknowledge the substantial contribution of Andy Tracey, Katie McGrother, Dr Xuejuan Fan and Dr Mark Lunt in database design and manipulation and Professor Alan Silman in his prior role as a principal investigator of the BSRBR-RA. I.N.B. acknowledges the support of the Manchester Academic Health Science Centre and the Manchester Biomedical Research Centre.
*Funding*: This study was funded by Arthritis Research UK Centre of Epidemiology Core Grant 17 552.
*Disclosure statement*: I.N.B. holds grants from UCB, Sanofi/Genzyme, Roche and GlaxoSmithKline and honoraria/speakers bureaus from UCB, GlaxoSmithKline, Roche, Pfizer and Medimmune outside of the submitted work. L.C. has received honoraria from Janssen and AbbVie for educational events and has received an unrestricted research award as a co-applicant from Pfizer. None of these awards are associated with the submitted manuscript. K.L.H. has received honoraria from Pfizer and AbbVie. All other authors have declared no conflicts of interest.
Supplementary data
==================
[Supplementary data](http://rheumatology.oxfordjournals.org/lookup/suppl/doi:10.1093/rheumatology/kev105/-/DC1) are available at *Rheumatology* Online.
[^1]: BMQ: Beliefs about Medicines Questionnaire; EQ-5D: EuroQol five-dimension questionnaire; HADs: Hospital Anxiety and Depression Scale; IPQ-R: Revised Illness Perception Questionnaire; SF-36: 36-item Short Form Health Survey; VAS: visual analogue scale.
[^2]: CQR: Compliance Questionnaire for Rheumatology.
[^3]: \**P* \< 0.05; \*\**P* \< 0.005.
[^4]: ADA: adalimumab; BMQ: Beliefs about Medicines Questionnaire; CQR: Compliance Questionnaire for Rheumatology; DAS28: 28-joint DAS; EQ-5D: EuroQol five-dimensions questionnaire; Fup1: follow up at 6 months; Fup2: follow up at 12 months; Fup3: follow up at 18 months; HADs: Hospital Anxiety and Depression Scale; IPQ-R: Revised Illness Perception Questionnaire; SF-36: 36-item Short Form Health Survey.
[^5]: ^a^Adjusted for age, gender, disease duration and social deprivation. \**P* \< 0.05; \*\**P* \< 0.005. CQR: Compliance Questionnaire for Rheumatology; Fup1: follow up at 6 months; Fup2: follow up at 12 months; Fup3: follow up at 18 months.
[^6]: BMQ: Beliefs about Medicines Questionnaire; CQR: Compliance Questionnaire for Rheumatology.
| {
"pile_set_name": "PubMed Central"
} |
Editorial Note on the Review Process
====================================
[F1000 Faculty Reviews](http://f1000research.com/browse/f1000-faculty-reviews) are commissioned from members of the prestigious [F1000 Faculty](http://f1000.com/prime/thefaculty) and are edited as a service to readers. In order to make these reviews as comprehensive and accessible as possible, the referees provide input before publication and only the final, revised version is published. The referees who approved the final version are listed with their names and affiliations but without their reports on earlier versions (any comments will already have been addressed in the published version).
The referees who approved this article are:
1. Fadi Charchar, School of Applied and Biomedical Sciences, Faculty of Science and Technology, Federation University, Ballarat, Victoria, Australia [^2]
2. Steven E. Artandi, Department of Medicine, Stanford University School of Medicine, Stanford, California, USA; Cancer Biology Program, Stanford University School of Medicine, Stanford, California, USA; Department of Biochemistry, Stanford University School of Medicine, Stanford, California, USA [^3]
3. Tracy Bryan, Children's Medical Research Institute, University of Sydney, Sydney, New South Wales, Australia [^4]
4. Dirk Hockemeyer, Department of Molecular and Cell Biology, University of California, Berkeley, Berkeley, California, USA [^5]
Introduction
============
Telomeres are the nucleoprotein complex at chromosome ends with essential roles in maintaining chromosomal integrity. They shorten with each cell division because of incomplete replication of the 3′ ends of DNA and thus are markers of cellular aging. Over the last decade, there has been remarkable growth in the breadth and depth of understanding the multiple roles of telomere biology in human disease. At one end of the spectrum, very rare pathogenic germline genetic variants in telomere biology genes cause exceedingly short telomeres, resulting in dyskeratosis congenita (DC) and its related telomere biology disorders. The other end of the spectrum consists of large-scale population-based studies seeking to determine associations between telomere length human disease, environmental exposures, or common genetic variants as well as the interactions between these factors.
The complexity of these interactions requires an integrated understanding of telomere basic science, clinical medicine, and epidemiology ( [Figure 1](#f1){ref-type="fig"}). Each of these topics is worthy of an in-depth critical review beyond the scope of this article. Instead, I will highlight some key findings and methodologic considerations and discuss where additional research is needed to aid in understanding the contribution of telomere biology to both rare and common human diseases.
{#f1}
Dyskeratosis congenita -- a direct connection between germline telomere biology and human disease
=================================================================================================
DC was first described in a 1906 case report of males with the mucocutaneous triad of abnormal skin pigmentation, nail dystrophy, and oral leukoplakia ( [Figure 2](#f2){ref-type="fig"}) ^[@ref-1]^. Additional similar cases were reported, including the first female case in 1963 ^[@ref-2]--\ [@ref-5]^. Patients with DC also have very high rates of bone marrow failure; stenosis of the esophagus, urethra, or lacrimal ducts (or a combination of these); head and neck squamous cell carcinoma (HNSCC); myelodysplastic syndrome (MDS); acute myeloid leukemia (AML); pulmonary fibrosis; liver disease; avascular necrosis of the hips; and other medical problems ( [Table 1](#T1){ref-type="table"}) ^[@ref-6]^. DC is inherited in X-linked recessive, autosomal dominant, or autosomal recessive patterns.
{#f2}
###### Clinical features of telomere biology disorders.
-------------------------------------------------------------------------------------------------------------------------------------
Disorder Key clinical features ^[a](#tfn1){ref-type="other"}^
------------------------------- -----------------------------------------------------------------------------------------------------
Dyskeratosis congenita (DC) Mucocutaneous triad of nail dysplasia, abnormal skin pigmentation (hyper/hypopigmented, lacy,\
reticular pigmentation), and oral leukoplakia. BMF, PF, PAVM, liver disease, avascular necrosis\
of hips or shoulders (or both), urethral stenosis, lacrimal duct stenosis, esophageal stenosis,\
HNSCC, MDS, AML, and/or developmental delay. Traditional diagnosis of DC: classic triad *or* one\
of the triad, BMF, and two other findings ^[@ref-7]^.
Revesz syndrome Features of DC plus bilateral exudative retinopathy. Intracranial calcifications and developmental\
delay also reported.
Hoyeraal Hreidarsson\ Features of DC plus cerebellar hypoplasia. Immunodeficiency has been reported as presenting\
syndrome problem.
Coats plus Bilateral retinopathy, intracranial calcifications, leukodystrophy, anemia, osteopenia, and poor\
bone healing
DC-like BMF, AA, MDS, or PF occurring in presence of at least one other DC-associated feature or family\
history suspicious of DC
Aplastic anemia Progressive multi-lineage cytopenias, non-immune mediated. May occur in the absence of DC-\
associated features.
Myelodysplastic syndrome Cytopenias with cellular dysplasia or clonal chromosomal translocations or both. May occur in the\
absence of DC-associated features.
Acute myeloid leukemia May progress from MDS or aplastic anemia. May occur in the absence of DC-associated features.
Pulmonary fibrosis May occur in the absence of DC-associated features.
Liver fibrosis Non-alcoholic, non-infectious liver disease. May occur in the absence of DC-associated features.
Familial melanoma Multiple family members with melanoma, often early age at onset
Familial lymphoproliferative\ Multiple-affected family members with chronic lymphocytic leukemia, or non-Hodgkin lymphoma
disease
Li-Fraumeni-like syndrome Cancer family history notable for angiosarcoma and other cancers
-------------------------------------------------------------------------------------------------------------------------------------
^a^Key references are noted in [Table 2](#T2){ref-type="table"}. AA, aplastic anemia; AML, acute myeloid leukemia; BMF, bone marrow failure; DC, dyskeratosis congenital; HNSCC, head and neck squamous cell carcinoma; MDS, myelodysplastic syndrome; PAVM, pulmonary arteriovenous malformation; PF, pulmonary fibrosis.
The first DC genetic locus was mapped to Xq28 in 1996 and specifically to mutations in dyskerin (encoded by *DKC1*) in 1999 ^[@ref-8]--\ [@ref-10]^. The seminal work by Mitchell and Collins was the first to show a connection between telomere biology and human disease through aberrant dyskerin function and the resultant very short telomeres now well known in patients with DC ^[@ref-11]^. Currently, *DKC1* mutations account for about 25% of classic DC cases. A combination of candidate gene sequencing, genetic linkage studies, and whole exome sequencing occurring over the last 15 years has since identified at least 14 telomere biology genes associated with DC or DC-like phenotypes: telomerase holoenzyme complex ( *DKC1*, *TERC*, *TERT*, *NOP10*, and *NHP2*), shelterin telomere protection complex ( *ACD*, *TINF2*, and *POT1*), telomere capping proteins ( *CTC1* and *STN1*), and other proteins that directly or indirectly interact with these key cellular processes ( *RTEL1*, *NAF1*, *WRAP53*, and *PARN*) ( [Figure 3](#f3){ref-type="fig"} and [Table 2](#T2){ref-type="table"}) (reviewed in [@ref-6]).
{ref-type="table"}. The asterisk indicates proteins encoded by genes with single-nucleotide polymorphisms associated with cancer. The pound sign indicates proteins encoded by genes with single-nucleotide polymorphisms associated with telomere length.](f1000research-7-15302-g0002){#f3}
###### Germline genetics of telomere biology disorders.
--------------------------------------------------------------------------------------------------------------------------------------------
Gene Protein name(s) Disorder(s) Mode of\ Year gene first associated\
inheritance with any disease, relevant\
reference(s)
----------- ------------------------------------- -------------------- ------------- -------------------------------------------------------
*DKC1* DKC1, dyskerin DC, HH XLR 1998 ^[@ref-23]^
*TERC* hTr, telomerase RNA component\ DC, AA, PF AD 2001 ^[@ref-16],\ [@ref-18],\ [@ref-24]^
(encodes an RNA)
*TERT* TERT, telomerase DC, AA, MDS, AML,\ AD 2005 ^[@ref-16],\ [@ref-17],\ [@ref-25]--\ [@ref-27]^
PF, LD, FM
HH AR
*NOP10* NOP10, NOLA3, nucleolar protein\ DC AR 2007 ^[@ref-28]^
family A, member 3
*NHP2* NHP2, NOLA2 nucleolar protein\ DC AR 2008 ^[@ref-29]^
family A, member 2
*TINF2* TIN2, TERF1 (TRF1)-interacting\ DC, HH, RS AD 2008 ^[@ref-15]^
nuclear factor 2
*WRAP53* TCAB1, telomere cajal body\ DC, HH AR 2011 ^[@ref-30],\ [@ref-31]^
associated protein 1
*CTC1* CTC1, conserved telomere\ CP, DC AR 2012 ^[@ref-32]--\ [@ref-34]^
maintenance component 1
*RTEL1* RTEL1, regulator of telomere\ PF, AA AD 2013 ^[@ref-35]--\ [@ref-40]^
elongation helicase 1
DC, HH AR
*TERF2IP* RAP1, TERF2-interacting protein FM AD 2015 ^[@ref-41]^
*PARN* PARN, poly(A)-specific\ PF AD 2015 ^[@ref-42]--\ [@ref-45]^
ribonuclease
DC, HH AR
*ACD* TPP1, telomere protection protein 1 AA, FM, FLPD AD 2014 ^[@ref-41],\ [@ref-46],\ [@ref-47]^
HH AR 2016 ^[@ref-48]^
*STN1* STN1, CST-complex subunit CP AR 2016 ^[@ref-49]^
*POT1* POT1, protection of telomeres 1 FM, FLPD, LFL AD 2014 ^[@ref-47],\ [@ref-50]--\ [@ref-52]^
CP AR 2016 ^[@ref-53]^
*NAF1* NAF1, nuclear assembly factor 1\ PF AD 2016 ^[@ref-54]^
ribonucleoprotein
--------------------------------------------------------------------------------------------------------------------------------------------
AA, aplastic anemia; AD, autosomal dominant; AML, acute myeloid leukemia; AR, autosomal recessive; CP, Coats plus; DC, dyskeratosis congenita; FLPD, familial lymphoproliferative disease; FM, familial melanoma; HH, Hoyeraal-Hreidarsson syndrome; LD, liver disease; LFL, Li-Fraumeni-like syndrome; MDS, myelodysplastic syndrome; PF, pulmonary fibrosis; RS, Revesz syndrome; XLR, X-linked recessive.
The germline mutations in DC-associated telomere biology genes result in very short telomere lengths for age ( [Figure 4](#f4){ref-type="fig"}). This knowledge made it possible to develop a diagnostic test for DC, flow cytometry with fluorescent *in situ* hybridization (flow FISH), in leukocyte subsets ^[@ref-12]^. Lymphocyte telomeres less than the first percentile for age are more than 95% sensitive and specific for differentiating patients with DC from their unaffected relatives or patients with other inherited bone marrow failure syndromes ^[@ref-13],\ [@ref-14]^. In addition to aiding diagnosis, using telomeres less than the first percentile for age has greatly added in discovering the genetic causes of DC ^[@ref-15]^.
{#f4}
Telomere biology disorders -- many names connected by pathophysiology
=====================================================================
The discovery of the multiple genetic causes and modes of inheritance has led to a growing appreciation that there is a wide range of clinical phenotypes associated with mutations in telomere biology genes. This started with the identification of *TERT* and *TERC* mutations in patients with apparently isolated aplastic anemia or pulmonary fibrosis ^[@ref-16]--\ [@ref-18]^. As defined above, classic DC is a complex multi-system disorder, but variable penetrance and expressivity of the clinical manifestations have identified a growing number of patients with one or a few features of DC as well as germline mutations in telomere biology genes and short telomeres ( [Table 1](#T1){ref-type="table"} and [Table 2](#T2){ref-type="table"}). This spectrum of illnesses has been termed telomeropathy, short telomere syndromes, or telomere biology disorders (TBDs) ^[@ref-6],\ [@ref-19]--\ [@ref-22]^. The last of these, TBD, was proposed and is favored because it is more descriptive and reflective of the underlying biology that unites these disorders ^[@ref-6],\ [@ref-22]^.
The most complex TBDs are those disorders presenting very early in childhood, namely Hoyeraal Hreidarsson syndrome (HH), Revesz syndrome (RS), and Coats plus ^[@ref-6],\ [@ref-55]--\ [@ref-60]^. In addition to having features of DC, patients with HH have cerebellar hypoplasia and immunodeficiency, whereas those with RS also have bilateral exudative retinopathy. Coats plus, a disorder characterized by retinal and gastrointestinal vascular abnormalities, poor bone healing, leukodystrophy, and cerebellar calcifications, joined the TBD spectrum when its primary cause was identified as autosomal recessive *CTC1* mutations ^[@ref-32],\ [@ref-33],\ [@ref-59]^.
The other end of the clinical spectrum includes patients with middle or later age at onset of pulmonary fibrosis, liver disease, or bone marrow failure and heterozygous germline mutations in *NAF1*, *TERT*, *TERC*, *PARN*, or *RTEL1* ^[@ref-17],\ [@ref-18],\ [@ref-42],\ [@ref-61]^. Additionally, it is important to note that most patients do not have all of the DC-associated medical complications. The mucocutaneous triad is diagnostic but varies with the age of onset and is usually progressive over time. Many patients, even members of the same family, may present with just one feature but develop more over time because of variable penetrance and expressivity of germline telomere biology defects.
There is also a growing role of mutations in the shelterin complex and cancer-prone families without DC-related clinical manifestations. Heterozygous rare, pathogenic variants in *POT1* resulting in longer telomeres have been reported in familial melanoma, familial chronic lymphocytic leukemia (CLL), and a Li-Fraumeni-like syndrome family ^[@ref-50]--\ [@ref-52]^. *POT1 s*omatic mutations in CLL have also been associated with CLL outcomes ^[@ref-62]--\ [@ref-65]^. Familial melanoma has also been associated with germline mutations in *ACD* (TPP1) and *TERF2IP* (RAP1) ^[@ref-41]^. These studies suggest an interesting dichotomy in clinical phenotypes resulting from long versus short telomeres.
Telomeres and cancer
====================
Telomeres are closely connected to cancer biology because of the role they play in chromosomal stability. There is a detailed body of work in this realm and only a few of the key features will be highlighted herein as they have been reviewed in detail elsewhere ^[@ref-66]--\ [@ref-73]^.
The primary hypotheses connecting telomeres and cancer are based on the fact that telomeres shorten with each cell division ^[@ref-74]^. In general, cellular senescence or apoptosis is triggered when telomeres reach a critically short state. It likely takes just one critically short telomere on one chromosome arm to trigger these events as suggested in a *TERC* mouse model by Hemann *et al*. ^[@ref-75]^. A cellular survival advantage is gained through bypassing apoptosis or senescence through the upregulation of telomerase, inactivation of TP53 or RB or both, initiation of alternative lengthening of telomeres (ALT), and other key biological pathways ^[@ref-76]--\ [@ref-80]^. The continued division of cells originally destined for death is hypothesized to lead to continued accumulation of mutations, and sticky chromosome ends due to abnormal telomeres can contribute to chromosomal aneuploidy. Unchecked cellular growth can occur if these genetic aberrations result in a growth advantage.
Activating somatic mutations in the *TERT* promoter have been described in melanoma, bladder, thyroid, and some central nervous system cancers ^[@ref-81]--\ [@ref-83]^. These somatic mutations in the *TERT* promoter result in increased telomerase expression and suggest that this activation could convey a growth advantage as cancer cells continue to divide despite the presence of aberrant telomeres.
Patients with DC/TBD have significantly increased risks of MDS, AML, HNSCC, and other malignancies ^[@ref-84]--\ [@ref-86]^. The 2017 update of cancer in the National Cancer Institute Inherited Bone Marrow Failure Syndrome cohort reported that cancer in patients with DC occurs at an approximately four fold higher incidence and a younger age than the general population ^[@ref-87]^. MDS and AML occurred at 578- and 24-fold greater incidence, respectively, than the general population. There was also an excess of solid cancers in patients with DC with observed/expected ratios of 74 for any HNSCC and 216 for tongue HNSCC ^[@ref-87]^. The mechanisms by which cancer develops in patients with DC/TBDs is unknown and represents an important research opportunity. Cells of patients with TBD already have a "first hit" in a key component of telomere biology. Studies of the next steps in carcinogenesis in patient-derived cells could lead to important insights into the carcinogenic process.
The advent of telomere molecular epidemiology
=============================================
Telomere molecular epidemiology has emerged with the development of high-throughput telomere length measurement methods, genome-wide genotyping platforms, and keen interest in the role of telomere biology in human disease ^[@ref-88],\ [@ref-89]^. These large, often population-based studies seek to determine (1) whether telomere length is associated with disease, (2) whether common genetic variants (that is, single-nucleotide polymorphisms, or SNPs) are associated with telomere length, (3) the degree to which SNPs contribute to telomere biology, and (4) interactions between telomere length, SNPs, and disease or phenotypes ( [Figure 1](#f1){ref-type="fig"} and [Table 3](#T3){ref-type="table"}). Although a great deal of excitement has been generated by these studies, it is important to point out that differences in telomere length between cases and controls in large population-based studies may be statistically significant but not clinically relevant. "Short telomeres" in a large case-control or cohort study are still within the clinically "normal" range and not nearly as short as telomeres of patients with TBDs ( [Figure 4](#f4){ref-type="fig"}).
###### Features of robust telomere length association studies.
-----------------------------------------------------------------------------------------------------------------------
• Strong *a priori* hypothesis of why telomere biology might be important in disease of interest\
• Comprehensive clinical phenotyping\
• Accurately measured exposure of interest\
• Large sample size with power calculations reported\
• Collection of samples prior to disease onset\
• Detailed information on how samples were collected, processed, and stored\
• Telomere length measurement methods described in detail, especially if any adaptations to published methods\
• Accurate and reproducible telomere length measurement\
• Strong statistical justification of association findings
-----------------------------------------------------------------------------------------------------------------------
Robust and accurate telomere length measurement is at the crux of all telomere length association studies. Blood or buccal cell DNA telomere length has been evaluated in a wide array of association studies, including cancer, cardiovascular disease, mental health, inflammatory diseases, environmental exposures, and many other settings. There are numerous methods to determine telomere length in tissues, single cells, and DNA preparations, each appropriate for different applications and reviewed extensively ^[@ref-90],\ [@ref-91]^. Quantitative polymerase chain reaction (qPCR) is amenable to large studies because it uses very small amounts of DNA and can be scaled up rapidly ^[@ref-92],\ [@ref-93]^. However, qPCR telomere assays generate a relative telomere length and are very sensitive to DNA extraction methods and storage ^[@ref-94]^. These challenges have led to significant challenges in reproducing data in case-control or cohort studies of qPCR relative telomere length and phenotypes ^[@ref-95],\ [@ref-96]^. The telomere restriction fragment method uses restriction enzymes to cut the subtelomeric ends of chromosomes in a DNA preparation and is most widely used in basic science laboratories, although a few groups use it in population-based studies ^[@ref-90],\ [@ref-91]^.
Blood or buccal cell telomere length association studies
--------------------------------------------------------
This section highlights just a few key topics within the growing literature of telomere length association studies. For example, individuals of African ancestry have longer telomeres than those of European ancestry and thus ancestry should be accounted for in analyses ^[@ref-97],\ [@ref-98]^. Since self-reported ancestry can be quite variable, genomic approaches may be helpful in classifying cases and controls in order to appropriately adjust for ancestry.
There is also a growing understanding of associations between environmental exposures and telomere length. Smokers have shorter telomeres than non-smokers and thus it is important to adjust for smoking in association analyses ^[@ref-99]^. Associations between prenatal exposures to smoking and air pollution as well as exposure to certain occupational chemicals have also been explored but with varying results ^[@ref-100]--\ [@ref-105]^. In each of these studies, it is essential to precisely quantify the exposure of interest in addition to using a robust and reproducible telomere length measurement.
Early cancer-telomere length association studies suggested shorter telomeres as a cancer risk factor ^[@ref-106],\ [@ref-107]^ but studies of other cancers were not consistent ^[@ref-108]--\ [@ref-112]^. Meta-analyses found that most studies with blood or buccal cell DNA collected prior to cancer diagnosis were null but that case-control studies were more likely to find associations ^[@ref-113],\ [@ref-114]^. Similarly, a direct comparison of prospectively and retrospectively collected DNA samples from patients with breast or colorectal cancer reported that telomere shortening occurred primarily after cancer diagnosis ^[@ref-115]^. Many of these inconsistencies have been attributed to reverse causation bias due to the presence of cancer, underlying inflammation, or prior therapy at the time of sample collection ^[@ref-113],\ [@ref-116]^. Currently, the most consistent studies are those of longer telomeres in pre-diagnostic samples of patients with lung cancer and melanoma ^[@ref-117]--\ [@ref-119]^. Interestingly, shorter leukocyte telomeres were associated with overall cancer mortality but not with cancer in a large prospective study of 64,637 individuals who developed 2,420 cancers ^[@ref-120]^.
There is a growing body of telomere length association studies and different aspects of mental health, including measures of perceived stress in caregivers, exposure to early life adversity, and in patients with schizophrenia, bipolar disorder, and depression ^[@ref-121]--\ [@ref-126]^. The biological mechanisms underlying these findings are unknown but current hypotheses include stress responses inducing oxidative stress, resulting in DNA damage and telomere shortening. Notably, abnormalities in brain development are present in patients with HH (cerebellar hypoplasia), RS (intracranial calcifications), and Coats plus (leukodystrophy and intracranial calcifications) ^[@ref-6],\ [@ref-55]^. The only study to date of neuropsychiatric complications in DC found higher-than-expected occurrence of developmental delay and psychiatric disorders ^[@ref-127]^. Studies of DC/TBD patients by psychiatrists and neurobiologists constitute an unstudied area highly likely to generate important insights into telomere biology and brain development.
There is also a great deal of interest in using telomere length as a measure of biological age and even in modulating telomere length through lifestyle interventions. Numerous studies suggest associations between lifestyle, exercise, and telomere lengths ^[@ref-128]--\ [@ref-133]^. However, a recent review suggests that telomere length in and of itself is not sufficient as a specific aging biomarker ^[@ref-134]^.
Current data consistently report shorter leukocyte telomeres in individuals with atherosclerotic cardiovascular disease than in unaffected controls ^[@ref-135]--\ [@ref-137]^. The biology underlying this association is thought to be related to chronic inflammation and oxidative stress coupled with aging of the vasculature. This prompted Aviv *et al*. to propose a model whereby age-dependent telomere shortening varies on the basis of the replicative needs of the specific tissue ^[@ref-138],\ [@ref-139]^. They hypothesize that skeletal muscle (a minimally replicative cell type) may represent telomere length closer to the time of birth and that the gap between skeletal muscle and leukocyte (a rapidly dividing cell type) telomere length attrition could serve to aid understanding of the associations between telomere length and human disease with each patient, in effect, serving as their own control. The first such study testing this hypothesis showed that increased attrition of telomeres in leukocytes was associated with atherosclerotic cardiovascular disease ^[@ref-140]^.
Single-nucleotide polymorphisms, telomere biology genes, and disease
--------------------------------------------------------------------
The advent of genome-wide association studies (GWAS) opened the door to understanding associations between common genetic variants (that is, greater than 1% minor allele frequency, SNPs) and human disease or phenotypes ( [Figure 5](#f5){ref-type="fig"}) (reviewed in [@ref-141]). GWAS genotype hundreds of thousands of SNPs in thousands of cases and controls and use methods to fine-tune risk estimates through large-scale replication studies and polygenic risk score computation ^[@ref-142]^.
{#f5}
Numerous GWAS of cancer etiology have identified variants in telomere biology genes as being associated with cancer risk or outcomes. SNPs in the *TERT-CLPTM1L* locus on chromosome 5p15.33 are associated with multiple cancer types, including lung, pancreatic, breast, bladder, ovarian, prostate, and testicular germ cell cancers as well as glioma, melanoma, and non-melanoma skin cancers ^[@ref-143]--\ [@ref-146]^. There are specific regions of this locus associated with different cancers, but these variants do not specifically encode deleterious coding alleles in *TERT*. They do, however, appear to be connected to telomere length through long-range regulation of this locus ^[@ref-147]^.
SNPs in *RTEL1* are associated with glioma in large GWAS of this rare brain cancer. The glioma-associated *RTEL1* SNPs are intronic, but functional studies have not yet been completed to understand their potential functions ^[@ref-148]--\ [@ref-152]^. These findings are intriguing because patients with DC or HH due to autosomal recessive *RTEL1* mutations often have abnormal brain development in the form of cerebellar hypoplasia ^[@ref-35]--\ [@ref-37],\ [@ref-153]^. Although the specific genetic loci are different, it is intriguing to speculate that there could be an important biological connection between these findings.
In addition to GWAS of cancer or other illnesses, several GWAS have been conducted to identify novel loci associated with telomere length. SNPs in known telomere biology genes, including *TERT*, *OBFC1* (encodes STN1), *CTC1*, *TERC*, *NAF1*, and *RTEL1*, as well as genes not previously known to be associated with telomere biology have been discovered ^[@ref-154]--\ [@ref-157]^. These studies illustrate the complexities of telomere length regulation by showing that even common genetic variants, and especially combinations of common genetic variants, are associated with telomere length in the general population.
The existence of telomere length GWAS in various populations set the stage for even larger studies using Mendelian randomization methods in which telomere length--associated SNPs serve as surrogates for telomere length ^[@ref-158]^. One such study used nine telomere length--associated SNPs to create a telomere length surrogate score and found longer telomere length scores associated with lung adenocarcinoma but not the other cancers ^[@ref-159]^. Renal cell carcinoma, one of the cancers with reproducible telomere length association data, was also studied using nine telomere length surrogate SNPs and it was found that genetically longer telomeres were associated with renal cell carcinoma ^[@ref-160]^.
In 2017, a Mendelian randomization study of 16 telomere length--associated SNPs from 103 GWAS with summary data on 35 cancers and 48 non-neoplastic diseases found that genetically longer telomeres associated with elevated risk of many cancers, including glioma, ovarian cancer, lung cancer, neuroblastoma, bladder, skin, testicular germ cell cancer and kidney cancer, and endometrial cancer ^[@ref-161]^. That study also found an association between genetically shorter telomeres and the risk of interstitial lung disease, celiac disease, abdominal aortic aneurysm, and coronary heart disease but not of other inflammatory or psychiatric diseases ^[@ref-161]^.
Although several studies suggest that telomere length is associated with depression, one study using Mendelian randomization and three SNPs---one each in *TERT*, *TERC*, and *OBFC1*---as surrogates for telomere length in 67,000 individuals did not find an association between depression and genetically shorter telomeres ^[@ref-162]^. These investigators used the same three SNPs to investigate genetically predicted telomere length and risk of ischemic heart disease ^[@ref-163]^. They found small but statistically significant associations in a dataset of 60,837 ischemic heart disease cases compared with controls.
The studies briefly described above have generated a great deal of enthusiasm but are not without limitations. In many instances, qPCR was used to measure the telomere lengths in GWAS and this assay can be variable between studies. The sensitivity of the assay telomere length measurement and relatively small contributions of SNPs to telomere length should be considered in interpreting large-scale telomere length Mendelian randomization studies.
The way forward
===============
The connections between telomere biology and human disease are complex and myriad and require a multi-disciplinary approach to truly understand the clinically relevant data, important basic science questions, and implications of epidemiologic analyses ( [Figure 1](#f1){ref-type="fig"}). As protectors of chromosome ends, telomeres are clearly integral to all aspects of cell biology. They are markers of biological aging and are regulated by a wide range of proteins.
Both very rare and very common germline genetic variants in telomere biology genes are associated with human disease, although the specific clinical phenotypes comprise a wide-ranging disease spectrum. Inheritance of telomere length inheritance and epigenetic regulation are also important aspects of telomere biology and should be incorporated into collaborative studies of rare and common telomere phenotypes. Additionally, optimization of telomere length measurement methods and improved understanding of environmental factors contributing to telomere biology will be essential in order to thoroughly understand these complexities. It is of the utmost importance for clinicians, epidemiologists, and basic scientists, all of whom study telomeres for a wide variety of different yet important reasons, to work together to build upon the expertise they each possess and incorporate that into improved understanding of telomere biology in human disease. This multi-disciplinary approach will enable the discovery of therapeutics and disease prevention modalities effective for patients with TBDs and for the general population.
Abbreviations
=============
AML, acute myeloid leukemia; CLL, chronic lymphocytic leukemia; DC, dyskeratosis congenita; GWAS, genome-wide association study; HH, Hoyeraal Hreidarsson syndrome; HNSCC, head and neck squamous cell carcinoma; MDS, myelodysplastic syndrome; qPCR, quantitative polymerase chain reaction; RS, Revesz syndrome; SNP, single-nucleotide polymorphism; TBD, telomere biology disorder.
The author acknowledges the invaluable contributions of all patients and their families who have taught us so much about the complexities of telomere biology disorders in all their forms.
[^1]: No competing interests were disclosed.
[^2]: No competing interests were disclosed.
[^3]: No competing interests were disclosed.
[^4]: No competing interests were disclosed.
[^5]: No competing interests were disclosed.
| {
"pile_set_name": "PubMed Central"
} |
When Aldo Leopold saw "a fierce green fire dying" in the eyes of a gray wolf he'd just shot, he recognized that his actions taken in the hope of creating a "hunters' paradise" of deer was ill conceived \[[@pbio.3000090.ref001]\]. Most of the world now recognizes that apex predators have great intrinsic value as well as providing vitally important ecosystem services. In many cases, these services outweigh some of the inconveniences to humans associated with large carnivore populations \[[@pbio.3000090.ref002]\]. At an accelerating rate during recent decades in Alaska, however, brown bears ([Fig 1](#pbio.3000090.g001){ref-type="fig"}), black bears, and gray wolves have been targeted for significant reductions in abundance in the expectation this will result in more wild ungulates (moose \[*Alces alces*\], caribou \[*Rangifer tarandus*\], and deer \[*Odocoileus hemonius sitkensis*\]) available for hunter harvest. A management priority favoring wild ungulates over large carnivores acquired the force of law with the passage in 1994 of Alaska's Intensive Management Law. This law effectively mandates management to reduce large carnivores and increase human harvests of wild ungulates. The Alaska Intensive Management efforts are occurring without rigorously collected data on the impacts of these management practices on large carnivores \[[@pbio.3000090.ref003],[@pbio.3000090.ref004],[@pbio.3000090.ref005]\] and ecosystems \[e.g., 2\].
{#pbio.3000090.g001}
Historically, gray wolves and brown bears were nearly extirpated in the conterminous United States because of persecution and habitat loss \[[@pbio.3000090.ref006]\]. Similar reductions occurred throughout the world (e.g., \[[@pbio.3000090.ref007]\] for Europe\]. In Alaska, large areas of intact habitat for large carnivores persist, although in some areas, habitats and populations are depleted by human activities such as roads, logging, mining, and other development.
The Alaska Intensive Management law sets a management priority for high levels of harvests of wild ungulates in areas where these ungulates are "important for human consumption" (Alaska Statutes 16.02.255). The law specifies that Intensive Management must occur when the Alaska Board of Game makes a finding that the harvestable numbers of ungulates are insufficient to meet human demand for game meat. Under this law, before the Board can change hunting regulations to reduce human take of ungulate species, Intensive Management must occur. Although habitat management such as controlled burns to create early-succession moose habitat is an identified Intensive Management technique under the law, the most significant Intensive Management efforts have been implemented by liberalizing hunting regulations for large carnivores \[3 for brown bears\]. The lack of significant Intensive Management efforts from habitat improvement is precluded by scale, cost, and in the case of fire, threats to human structures \[[@pbio.3000090.ref008]\]. In some places, termed Predator Control Areas, especially aggressive efforts include agency shooting of bears (both species) from helicopters, snaring of bears, and shooting female brown bears accompanied by cubs \[[@pbio.3000090.ref003]\]. By 2017, the last remaining Predator Control Area for bears was eliminated. In wolf Predator Control Areas, allowed and utilized techniques include shooting of wolves by agency staff from helicopters, land and aerial hunting by the public, and carbon monoxide poisoning of pups in dens. Alaska conveniently defines "predator control" efforts as occurring only in these Predator Control Areas; this allows the state to claim that "predator control" is ongoing in only a small portion of Alaska. However, the degree that Intensive Management is accomplished by liberalized general hunting regulations for large carnivores is far more geographically extensive than the Predator Control Areas. For brown bears, regulation liberalizations include techniques such as shooting bears in dens, baiting bears, long (sometimes year-round) open hunting seasons, elimination of resident tag fees, and liberalized individual harvest quotas of 2 per year \[[@pbio.3000090.ref003]\]. In a state of about 1,509,600 km^2^, 91% has been identified by the Alaska Board of Game as being important for human consumption of ungulates and therefore eligible for Intensive Management actions for one or more of the three wild ungulate species (compiled from 5 Alaska Administrative Code 92.108). Of this, the largest portion (60.1% of Alaska) is for moose ([Fig 2](#pbio.3000090.g002){ref-type="fig"}).
{#pbio.3000090.g002}
Intensive Management regulations for all three species of large carnivores have been liberalized in increasing proportions of the state over the last 30 years. Reported kills of brown bears by hunters more than doubled during the last 30 years in a liberalized brown bear hunting area comprising 76% of the state \[[@pbio.3000090.ref003]\]. Major liberalizations of hunting regulations for black bears and wolves also occurred, including expanded bag limits and extensions of seasons into times of the year when hides have little value.
Even Alaska's 11 National Preserves managed by the United States National Park Service are not refugia from predator reduction regulations adopted by the state of Alaska. This is because a provision in the 1980 Alaska National Interest Lands Conservation Act allows sport hunting in Alaskan National Preserves, and the state determines who qualifies as a sport hunter (normally all Alaska residents and often nonresidents). This means that Alaska's predator hunting regulations generally apply on National Preserves even though this is inconsistent with National Park Service policy guidelines, which state:
"The Service does not engage in activities to reduce the numbers of native species for the purpose of increasing the numbers of harvested species (i.e., predator control) ..." (\[[@pbio.3000090.ref009]\] Section 4.4.3).
Since about 2010, the National Park Service in Alaska has resisted adopting some of the most extreme of the state's predator-reduction hunting and trapping regulations in National Preserves. The National Park Service currently has the legal authority to do this (Federal Register, 80 FR 205, 23 October 2015). The federal administration that took office in 2017, however, is currently proposing to reverse this 2015 rule, thereby constraining the ability of the National Park Service to resist adopting Alaska's liberal hunting and trapping regulations on National Preserves (80 Federal Register 64325, RIN 1024-AE38). Similarly, nationwide, the current administration is also attempting to require that the U.S. Fish and Wildlife Service attempt to "align" hunting and trapping regulations on National Wildlife Refuges with regulations of the encompassing state. This constrains the ability of national wildlife refuge managers to manage hunting in ways consistent with national interests. In Alaska, this proposal makes it difficult for national wildlife refuge managers to resist adopting the state's predator control regulations such as baiting for brown bears on the Kenai National Wildlife Refuge \[[@pbio.3000090.ref003]\]. If successful in forcing managers of National Wildlife Refuges to adopt state hunting and trapping regulations, national priorities for wildlife management within refuges could be compromised throughout the United States.
Recent ecological studies have demonstrated the fundamental importance of apex predators in stabilizing ecosystems \[[@pbio.3000090.ref002],[@pbio.3000090.ref010],[@pbio.3000090.ref011]\]. The removal or significant reduction of large carnivores can trigger a chain of events that can create a downward spiral toward ecosystem simplification \[[@pbio.3000090.ref010]\]. Because research strongly suggests that apex predators regulate ecosystem structure and function, it would make sense for wildlife managers to carefully evaluate the ecological role of apex predators prior to implementing programs designed to reduce large carnivores. Hunting, especially selective hunting for large trophy animals, also can have adverse behavioral and genetic consequences for hunted populations \[[@pbio.3000090.ref012],[@pbio.3000090.ref013]\]. A recent study concluded that Alaska's wolf predator control management, adjacent to the Yukon--Charley Rivers National Preserve, had adverse impacts on wolf populations within the national preserve where there was no wolf control \[[@pbio.3000090.ref014]\].
Science-based management of large carnivores in most of Alaska will require the political will and wisdom to repeal Alaska's Intensive Management law. Alternatively or additionally, it will require professional wildlife managers to resist adoption of predator reduction regulations that are not conducted as experiments and/or do not include adequate monitoring programs of both carnivores and ungulates; this was a key recommendation in the 1997 report of National Research Council \[[@pbio.3000090.ref005]\]. Furthermore, in Alaska and other states, the U.S. Department of the Interior needs to meet its legal mandate to manage for natural and healthy ecosystems in ways that are in the national interest. In Alaska, this will require not aligning hunting and trapping regulations on National Park Preserves and National Wildlife Refuges with state regulations that are designed to reduce naturally occurring densities of large carnivores. The state of Alaska also should be candid with the public about the absence of science supporting the efficacy of predator control programs to achieve established objectives with regard to ungulate harvests instead of making unsupported claims of "success" for wolf reduction efforts in publicly distributed booklets about Intensive Management (e.g., \[[@pbio.3000090.ref015]\]). For bears, there are not even any claimed successes for increased harvests of adult moose or caribou resulting from increased bear harvests \[[@pbio.3000090.ref003]\]. Appointments by the Alaska Governor to the Alaska Board of Game, which sets Alaska hunting regulations, should include members who recognize the importance and value of large carnivores both to ecosystem function \[[@pbio.3000090.ref002]\] as well as to the state's economy and wildlife viewing enthusiasts \[[@pbio.3000090.ref016]\]. Mechanisms and funding must be in place to ensure science-based management that includes adequate monitoring and research of predator--prey relationships and trends \[[@pbio.3000090.ref003],[@pbio.3000090.ref005],[@pbio.3000090.ref017]\]. Information campaigns and other grass roots efforts by concerned citizens and nongovernmental organizations are likely needed to remedy current unsound management practices for large carnivores in Alaska.
Alaska is unique in the world as a place where brown bear, black bear, and gray wolf populations are intentionally targeted for population reductions in efforts to increase human harvests of wild ungulate prey species and this priority is mandated by state statute. Similar management priorities do not occur in Europe \[[@pbio.3000090.ref007],[@pbio.3000090.ref018]\]. The situation in Europe, however, is complicated by the private ownership of wildlife in many countries, which creates economic incentives to control predators from landowners who lease hunting rights. Predators are controlled in some other areas of the world to reduce losses of domestic livestock. In parts of Canada, there are concerns that reductions in wolf abundance may be necessary to bolster woodland caribou populations depleted by habitat losses \[[@pbio.3000090.ref019]\].
In most of the world, there has been a paradigm shift moving away from predator control to supposedly benefit ungulates \[[@pbio.3000090.ref020]\]. Enlightened scientific management at the scale of ecosystems is needed to put Alaska back on the path to avoiding the errors in predator management experienced a century ago when Leopold first noticed the fierce green fire fading in the eyes of a dying wolf.
**Provenance:** Not commissioned; externally peer reviewed
[^1]: The authors have declared that no competing interests exist.
[^2]: Current address: Retired, Missoula, Montana, United States of America
[^3]: Current address: Retired, Anchorage, Alaska, United States of America
[^4]: Current address: Retired, Anchorage, Alaska, United States of America
| {
"pile_set_name": "PubMed Central"
} |
Introduction
============
Early embryo growth arrest is a unique form of spontaneous abortion, which is characterized by absent fetal development or fetal death in the first-trimester of pregnancy ([@b1-mmr-16-01-0591],[@b2-mmr-16-01-0591]). With the environmental pollution and increasing life stress in populations, a previous epidemiological study demonstrated that early embryo growth arrest occurs in \>4% of pregnant women in China and the incidence is increasing rapidly ([@b3-mmr-16-01-0591]). Although a number of factors, such as genetic abnormality, immune system deficits and viral infection have been indicated in the pathogenesis of early growth arrest, the underlying cellular and molecular mechanisms remain poorly understood. Furthermore, although \~50% of incidence of early embryo growth arrest can be ascribed to chromosome abnormality, the etiology of the other clinical cases remains ambiguous ([@b3-mmr-16-01-0591],[@b4-mmr-16-01-0591]).
The placenta functions as the exchange organ between the fetal and maternal bodies, and it serves a major role in providing nutrients and carrying over metabolic substrates via trans-placental exchange. Fetal development and pregnancy maintenance are dependent on normal placental growth ([@b5-mmr-16-01-0591]). In mammals, epigenetic modification is commonly overlaid during embryogenesis and early development of the fetus. DNA methylation is the most important epigenetic regulation process in genomic imprinting, transposon silencing, X-inactivation and gene repression ([@b6-mmr-16-01-0591]--[@b8-mmr-16-01-0591]). DNA methylation is regulated by DNA methyltransferases (DNMTs), including DNMT1, DNMT3A, DNMT3B and DNMT3L. The dynamics of DNA methylation is vital for embryonic and placental development in physiological and pathological conditions ([@b9-mmr-16-01-0591]--[@b12-mmr-16-01-0591]).
The three main DNMTs (DNMT1, DNMT3A and DNMT3B) are responsible for two special methylation processes that are required for tissue-specific methylation patterns. DNMT1 mediates maintenance DNA methylation, while DNMT3A and DNMT3B are responsible for *de novo* methylation ([@b7-mmr-16-01-0591],[@b13-mmr-16-01-0591]). *DNMT3L* is highly homologous to *DNMT3A* and *DNMT3B* in sequence, but has no catalytic activity *in vitro*. Previous studies have revealed that mutation or deletion of *DNMT1*, *DNMT3A* and *DNMT3B* triggered cell death in human embryonic stem cells ([@b12-mmr-16-01-0591],[@b14-mmr-16-01-0591]). At present however, little is known about the patterns of DNA methylation or DNMTs expression in human placenta during early pregnancy. Some previous studies have reported low expression of 5-methyl-cytosine and relative hypo-methylation in the developmental placenta of different pregnancy stages ([@b15-mmr-16-01-0591]). Previous results based on gene knockout experiments indicated that all the four DNMTs are involved in human trans-acting imprinting defects. For instance, *DNMT3B* gene mutations of specific residues in the C-terminal catalytic domain are known to cause immunodeficiency-centromeric instability-facial anomalies syndrome ([@b16-mmr-16-01-0591],[@b17-mmr-16-01-0591]). Aberrant expression of DNMTs may be responsible for the high rate of abortion and abnormal embryo growth. In addition, *DNMT3a* or *DNMT3b* knockout mice exhibited embryonic development arrest because of the failure to initiate *de novo* methylation following implantation ([@b18-mmr-16-01-0591],[@b19-mmr-16-01-0591]). In addition, wide demethylation and developmental arrest was reported in *DNMT1* knockout mice embryos at the early stage of gestation ([@b20-mmr-16-01-0591]).
To test the involvement of abnormal DNA methyltransferases in early embryo growth arrest, the authors measured the level of DNMTs expression in chorionic villi obtained from pregnant women diagnosed with early embryo growth arrest and control patients. A significant downregulation of the DNMT3A protein was observed in these embryo growth arrest cases; however, their *DNMT3a* mRNA expression levels were comparable to the controls. The discordance of the mRNA and protein expression of DNMT3A indicated that the translation process of DNMT3A was specifically interfered during early embryo growth arrest.
Materials and methods
=====================
### Clinical data and tissue samples
Chorionic villous samples were obtained from 80 pregnant women (gestational age ranging from 6 to 9 weeks) who underwent termination of pregnancy at the Department of Obstetrics in the Qingdao Municipal Hospital between January 2013 and June 2014. A total of 40 of the 80 cases were diagnosed with embryo growth arrest by B-mode ultrasound; the other 40 cases were having a normal pregnancy, but patients had voluntarily requested for the termination of pregnancy. Patient age was between 22 and 34 years old, and other serious medical conditions were ruled out, such as chromosome abnormities, endocrine diseases, infection, immunological diseases and other serious maternal complications. Each sample (\~20 g) was immediately frozen in liquid nitrogen for 30 min, and subsequently stored at −80°C for reverse transcription-quantitative polymerase chain reaction (RT-qPCR) and western blot analysis. The remainder of each sample was fixed with 10% buffered formalin and was paraffin-embedded for immunohistochemistry. Written informed consent for the publication of any associated data and accompanying images was obtained from all patients involved. The current study was approved by the Ethical Review Committee of Qingdao University (Qingdao, Shandong, China).
### RNA extraction, cDNA synthesis and RT-qPCR
Total RNA was extracted from the chorionic villi with a PureLink^™^ RNA Mini kit (Thermo Fisher Scientific, Inc., Waltham, MA, USA), according to the manufacturer\'s instructions. RNA quantity and quality was determined by a NanoDrop 2000 spectrophotometer (Thermo Fisher Scientific, Inc., Wilmington, DE, USA). cDNA was then synthesized from 1 µg total RNA with SuperScript^™^ III Reverse transcriptase (Thermo Fisher Scientific, Inc.) according to the manufacturer\'s instructions. The mRNA expressions of DNMTs and β-actin were measured by the Master Cycler ep realplex PCR system (Eppendorf, Hamburg, Germany) using a QuantiFast SYBR-Green PCR kit (Qiagen GmbH, Hilden, Germany) according to the manufacturer\'s protocols. The PCR cycling parameters were set as follows: 95°C for 5 min, followed by 40 cycles of PCR reaction at 95°C for 5 sec, and finally 60°C for 30 sec. β-actin was used as an internal control. All reactions were run in triplicate. The threshold cycle is defined as the fractional cycle number at which the fluorescence passes the fixed threshold. The data were obtained by normalizing *DNMT1*, *DNMT3a*, *DNMT3b* and *DNMT3L* genes threshold values with corresponding β-*actin* threshold value, and then analyzed with 2^−ΔΔCq^ method ([@b21-mmr-16-01-0591]). The primer sequences are presented in [Table I](#tI-mmr-16-01-0591){ref-type="table"}.
### Immunohistochemical staining and analysis
The biopsied tissues were initially fixed in 10% buffered formalin, embedded in paraffin and cut into 4 µm sections. Tissue sections were then de-paraffinized and dehydrated in graded ethanol and dimethylbenzene, respectively. Antigen retrieval was performed by boiling sections in EDTA buffer (pH 8.0) for 2.5 min in a pressure cooker. The endogenous peroxidase activity was blocked using 0.3% hydrogen peroxide. After rinsing in phosphate-buffered saline (PBS), the sections were incubated for 1 h with polyclonal rabbit anti-DNMT1 antibody (1:200, NB100-264, Novus Biologicals, LLC, Littleton, CO, USA), polyclonal rabbit anti-DNMT3A antibody (1:400, NB100-265, Novus Biologicals, LLC), polyclonal rabbit anti-DNMT3B antibody (1:800, NB100-266, Novus Biologicals, LLC) or polyclonal rabbit anti-DNMT3L antibody (1:100, ab115522, Abcam, Cambridge, UK). After rinse in PBS for several times, the sections were incubated with a biotinylated goat anti-rabbit secondary antibody (UltraSensitive^™^ SP kit, Fuzhou Maixin Biotech Co., Ltd., Fuzhou, China) at 37°C for 30 min. Sections were then stained using 3, 3-diaminobenzidine chromogen solution (Zhongshan Golden Bridge Biotechnology Co., Ltd., Beijing, China) and counterstained with hematoxylin (Fuzhou Maixin Biotech Co., Ltd). All slides were observed separately by two pathologists.
### Western blotting detection of DNMTs
The frozen tissue was re-suspended in passive lysis buffer (Promega Corporation, Madison, WI, USA) including freshly added protease inhibitors and phosphatase inhibitors cocktails (Sigma-Aldrich, Darmstadt, Germany). Protein concentrations in the supernatant fractions were determined using a bicinchoninic acid assay kit (Thermo Fisher Scientific, Inc.). Protein lysates were separated with 8 or 12% SDS-acrylamide gels, and transferred onto nitrocellulose membranes. After blocking for 60 min in a 5% skim milk solution, membranes were incubated overnight at 4°C with monoclonal rabbit anti-DNMT1 antibody at 1:1,000 (D59A4, Cell Signaling Technology, Inc., Danvers, MA, USA), monoclonal rabbit anti-DNMT3A at 1:1,000 (D23G1, Cell Signaling Technology, Inc.) and polyclonal rabbit anti-DNMT3B at 1:1,000 (NB100-266, Novus Biologicals, LLC), polyclonal rabbit anti-DNMT3L at 1:1,000 (ab115522, Abcam) or monoclonal mouse anti-β-actin at 1:5,000 (A5441, Sigma-Aldrich). Signals from horseradish-peroxidase-conjugated secondary antibodies were visualized by enhanced chemilluminescence solution (GE Healthcare Life Sciences, Chalfont, UK) and processed with an UVP visualizer (UVP, Inc., Upland, CA, USA). Quantification was performed using VisionWorks LS version 7.0 from the UVP visualizer itself. Experiments were repeated in triplicate.
### Statistical analysis
The statistical analysis was performed using SPSS version 17.0 software (SPSS Inc., Chicago, IL, USA). Student\'s t test was used to compare the quantitative data between the two groups and a one-way analysis of variance (ANOVA) followed by Bonferroni post hoc test was used for multiple comparisons. The results are presented as the mean ± standard error of the mean and a two-sided P\<0.05 was considered to indicate a statistically significant difference.
Results
=======
### DNMT mRNA expression in chorionic villi of early embryo growth arrest patients
First, the authors measured the level of mRNA expression of DNMT1, DNMT3A, DNMT3B and DNMT3L in chorionic villi obtained from the placental tissue of both groups. DNMT3B presented the highest level of expression compared to other three DNMTs in both groups (One-way ANOVA, n=40 for each group, P\<0.0001, [Fig. 1](#f1-mmr-16-01-0591){ref-type="fig"}), indicating that DNMT3B is the major type of DNA methyltransferase expressed in chorionic villi. In addition, it was identified that the mRNA expression of four DNMTs (DNMT1, DNMT3A, DNMT3B and DNMT3L) in chorionic villi of early embryo growth arrest patients were comparable to that in the controls, respectively (unpaired Student\'s t-test, n=40 for each group, P\>0.05, [Fig. 1](#f1-mmr-16-01-0591){ref-type="fig"}).
### DNMT protein expression in chorionic villi of early embryo growth arrest patients by IHC
Next, IHC studies were conducted to determine whether the protein expression of DNMTs was altered in patients with early embryo growth arrest. IHC staining demonstrated that all four types of DNMTs were expressed in trophoblast cells and chorionic villi. DNMT3A and DNMT3B positive staining was localized in the cell nucleus ([Fig. 2A](#f2-mmr-16-01-0591){ref-type="fig"}) while DNMT1and DNMT3L were in the cytoplasm (data not shown). Immunostaining intensity was calculated by Image Pro plus 6.0 analysis software (Media Cybernetics Inc., Rockville, MD, USA) and represented with semi-quantitative integrated optical density (IOD). The total IOD value of each sample was calculated by summing up the staining intensity from six randomly selected visions per section and across six representative sections. The current results revealed that the expression of DNMT3A protein in chorionic villi of early embryo growth arrest patients was significantly lower than that in chorionic villi of the controls (unpaired Student\'s t-test, n=40 for each group, P\<0.01, [Fig. 2B](#f2-mmr-16-01-0591){ref-type="fig"}). The protein levels of DNMT1, DNMT3B and DNMT3L were similar between two groups (unpaired Student\'s t test, n=40 per group, P\>0.05, [Fig. 2B](#f2-mmr-16-01-0591){ref-type="fig"}).
### DNMTs protein expression in chorionic villi of early embryo growth arrest patients by western blotting
To confirm the above results, western blot analysis was conducted to compare the DNMT expression in chorionic villi from patients with early embryo growth arrest with healthy controls. Data were analyzed with the UVP gel imaging system. Consistently, it was identified that, in comparison to the controls, DNMT3A protein expression was specifically decreased in chorionic villi obtained from embryo growth arrest patients (unpaired Student\'s t -test, n=40 per group, P\<0.05, [Fig. 3B](#f3-mmr-16-01-0591){ref-type="fig"}). No significant difference was presented between groups as for DNMT1, DNMT3B and DNMT3L expression (unpaired Student\'s t-test, n=40 per group, P\>0.05, [Fig. 3](#f3-mmr-16-01-0591){ref-type="fig"}).
Discussion
==========
In recent years, serious pregnancy complications, such as early embryo growth arrest, recurrent spontaneous miscarriage, intrauterine growth restriction and preeclampsia, have globally been on the increase despite improved medical environment ([@b1-mmr-16-01-0591],[@b22-mmr-16-01-0591]). These complications are the leading causes of mortality and they put huge emotional and financial burdens on the caregivers and patients themselves. Early embryo growth arrest is a stop in intrauterine embryonic development in the early period of pregnancy ([@b1-mmr-16-01-0591]). Distinct from spontaneous abortion, the majority of the pregnant women with early embryo growth arrest present no obvious symptoms, such as bleeding or abdominal pain ([@b1-mmr-16-01-0591],[@b2-mmr-16-01-0591]). Early diagnosis of embryo growth arrest primarily depends on B-mode and Doppler ultrasound examination ([@b23-mmr-16-01-0591]). Although previous etiology studies have focused on chromosome abnormality ([@b1-mmr-16-01-0591],[@b24-mmr-16-01-0591]--[@b26-mmr-16-01-0591]), sperm deformity, immune factors, environment factors and hormonal regulation, the molecular mechanisms of embryo growth arrest remain uncertain.
DNA methylation is crucial during the entire embryonic development. Specifically, it is reported to be important in various cellular processes such as cell differentiation, gene transcription and genomic imprinting ([@b10-mmr-16-01-0591],[@b11-mmr-16-01-0591]). DNA methylation is established in epiblast cells within embryonic days 4.5 and 6.5 ([@b12-mmr-16-01-0591]). Previous studies indicated that genomic imprinting, regulated by DNA methylation, may serve a critical role in placenta growth ([@b27-mmr-16-01-0591],[@b28-mmr-16-01-0591]). Methylation levels of human placenta increase in a gestational stage-dependent manner ([@b26-mmr-16-01-0591]). In addition, recent studies demonstrated the association between changes of placental DNA methylation patterns and pregnancy disorders ([@b29-mmr-16-01-0591]). Furthermore, dynamic regulation of DNA methylation by DNMTs exists in placenta ([@b11-mmr-16-01-0591],[@b12-mmr-16-01-0591]). Therefore, understanding the role and underlying mechanisms of DNA methylation in placental development may offer a new strategy to prevent serious pregnancy complications.
In the present study, the authors investigated the expression patterns of four DNMTs (DNMT1, DNMT3A, DNMT3B and DNMT3L) in chorionic villi of pregnant women and demonstrated that all four DNMTs were expressed in placental villi and trophoblast cells. The expression levels of DNMT3B mRNA and protein were the highest both in patients with early embryo growth arrest and in individuals with normal pregnancy, suggesting that DNMT3B may be one of the prominent DNA methyltransferases involved in placenta development during early pregnancy. More importantly, specific downregulation of the DNMT3A protein was observed with both immunostaining and western blot analysis in the chorionic villi of pregnant women with embryo growth arrest, while none of other three DNMTs (DNMT1, DNMT3B or DNMT3L) was changed at both mRNA and protein levels. To the best of the authors\' knowledge, the maintenance DNA methylation is a crucial and indispensable element in embryo implantation and development process ([@b10-mmr-16-01-0591],[@b11-mmr-16-01-0591]). Previous studies suggested that both DNMT3A and DNMT3B are essential for *de novo* methylation in the embryonic stem cells and early embryos ([@b30-mmr-16-01-0591]). In a mouse model, the DNMT3B protein is more enriched in the embryonic stem cells than DNMT3A ([@b9-mmr-16-01-0591],[@b19-mmr-16-01-0591]). However, the DNMT3A protein begins to be expressed again in the middle embryonic development (E10.5-E14.5), while DNMT3B is not ([@b9-mmr-16-01-0591]). These studies support the theory that epigenetic alterations in early embryos may be carried forward to subsequent developmental stages ([@b31-mmr-16-01-0591]). Moreover, *DNMT3A* transcription is presumably the vital point for embryonic development process. Therefore, the finding that an abnormal decrease of chorionic villous DNMT3A protein in the placenta of early embryo growth arrest patients may be important for the pathogenesis of early embryo growth arrest. Decreased protein expression accompanied by normal mRNA expression of DNMT3A also indicated that DNMT3A protein translation or post-translation processing was specifically interfered in early embryo growth arrest.
The placenta serves an important role in managing intrauterine embryonic growth and development through nutrients and waste transfer ([@b32-mmr-16-01-0591]). Previous studies have indicated that placental epigenetic profiles may affect fetal growth through the interaction between intra-uterine and extra-uterine environments ([@b33-mmr-16-01-0591]--[@b35-mmr-16-01-0591]). Administration of DNA methyltransferase inhibitors to pregnant rats resulted in smaller placentas with severe histological damage at different gestational ages ([@b36-mmr-16-01-0591]). Another previous study reported that DNMT3B mRNA expression decreases following the increase in DNMT3A and DNMT1 expression in early pregnancy placenta ([@b37-mmr-16-01-0591]). Other studies indicated that the DNMT1 protein level was significantly decreased and the global DNA methylation level was significantly downregulated in chorionic villi of women with early pregnancy loss ([@b27-mmr-16-01-0591],[@b29-mmr-16-01-0591]). Although different subtypes of DNMTs were reported to be downregulated in the present study compared to the previous reports, the downregulation trends of DNMTs expression was the same. Our findings indicated that the downregulation of DNMT3A expression in chorionic villi is specific in early embryo growth arrest. Insufficient maintenance methylation due to the lower level of DNMT3A protein expression may be associated with abnormal embryonic development in human early pregnancy loss.
In summary, DNA methylation has a critical role in placenta development, and changes in methylation pattern can lead to adverse birth outcome. Placental epigenome may serve as a modulator of disease pathogenesis. The current findings confirmed that downregulation of DNMT3A protein expression and the ensuing disturbance of the maintenance DNA methylation may serve an important role in the pathogenesis of early embryo growth arrest. Further studies are required to disclose the specific target genes regulated by the changes of DNA methylation in placental pathogenesis. In addition, studies are required to elucidate the distinct nature of DNA methylation in the maternal-fetal interaction during pregnancy.
The current work was supported by the Municipal Science and Technology Foundation of Qingdao (grant no. 2012-WSZD009).
DNMT
: DNA methyltransferase
IUGR
: intrauterine growth restriction
ICF syndrome
: immunodeficiency-centromeric instability-facial anomalies syndrome
{#f1-mmr-16-01-0591}
{#f2-mmr-16-01-0591}
{#f3-mmr-16-01-0591}
######
Sequences of specific gene primers used for reverse transcription-quantitative polymerase chain reaction.
Gene Forward sequence (5′-3′) Reverse sequence (5′-3′)
--------- -------------------------- --------------------------
β-actin CACTCTTCCAGCCTTCCTTC GTACAGGTCTTTGCGGATGT
DNMT1 GACCCGTCTCTTGAAGGTGG CTTCTCCTGCATCAGCCCAA
DNMT3A GGTCACGCAAAACAGAACCC CCTTGGTGAAACCCTTTGCG
DNMT3B CTCTTCCTCAGCTGTGTGGG CTGTCGGCACTGTGGTTTTG
DNMT3L GATTCTTTGCCCCCATAGCCT TTCAAGGTTCCAGTGGTCCG
| {
"pile_set_name": "PubMed Central"
} |
Introduction {#Sec1}
============
The neonatal cardiac tissue dilplay high tissue proliferation and differentiation, but these processes subside within the first week after birth^[@CR1],[@CR2]^. Indeed, neonatal cardiomyocytes proliferate well up to the first postnatal week^[@CR3]^. After this period, even though there is DNA synthesis, it is mostly associated with multinucleation. In young animals, cardiomyocytes turnover rate is around 1% per year, and this decreases with age^[@CR4],[@CR5]^. DNA synthesis is one of the steps in this process and can lead to multinucleation in cardiac myocytes. There are different methodological approaches to assay them^[@CR6],[@CR7]^.
Proliferation and differentiation are both energy-demanding processes, and, as a result, changes in energy metabolism are expected as the neonatal cardiac properties rapidly evolve within this short postnatal period. Indeed, other tissues undergoing rapid expansion, such as fetal liver primitive hematopoietic stem cells, require an efficient energy source to fuel cellular expansion^[@CR8]^. Furthermore, mesenchymal stem cell differentiation also involves increased oxidative metabolic capacity that is specific for the type of differentiation. Osteocytes and adipocytes require enhanced respiratory capacity, whereas commitment to chondrogenesis leads to a loss of respiratory capacity. Blocking mitochondrial plasticity in these models prevents differentiation, demonstrating that changes in oxidative metabolism are required for this process^[@CR9]^.
In the heart, postnatal development is believed to involve a shift from glycolytic fermentation to oxidative phosphorylation (reviewed in^[@CR10]^). In perfused rabbit hearts, glycolytic rates were found to decrease between the 1^st^ and 7^th^ postnatal days, with a concomitant increase in fatty acid oxidation^[@CR11]^. These results are consistent with a shift from fermentative to oxidative metabolism; however, lactate oxidation rates remain unchanged at both time points, which is not consistent with this view^[@CR11]^. Further support for the concept that heart development following birth involves a shift to more oxidative metabolism comes from a PGC-1α/β mouse model^[@CR12]^, which lacks the transcriptional co-activators responsible for promoting mitochondrial biogenesis and presents postnatal heart maturation defects, dying shortly after birth. However, PCG-1 family proteins display other functions that may contribute toward the abnormalities, including the regulation of fatty acid oxidation. More recently, Puente *et al*. elegantly demonstrated that postnatal heart maturation involves enhanced oxidative tissue damage and that prevention of this damage by antioxidants or hypoxic conditions maintain the proliferative capacity of neonatal cardiomyocytes^[@CR13]^. Based on these results and the knowledge that mitochondrial mass increases during postnatal cardiac development^[@CR14]^, Puente *et al*. proposed that, after birth, atmospheric oxygen promotes an increase in mitochondrial oxidative activity resulting in enhanced production of reactive oxygen species. Interestingly, however, although markers of mitochondrial mass have been quantified^[@CR13],[@CR14]^, no direct functional studies comparing oxygen consumption rates in hearts with high proliferation capacity (P1) and low proliferation capacity (P7) have been conducted. In this study, we measured respiratory activity in neonatal rat hearts and found, quite surprisingly, that oxygen consumption activity does not correlate well with expression patterns of oxidative genes, and decreases during maturation. Furthermore, we demonstrate that maintaining high oxygen consumption capacity is necessary to sustain the proliferative capability of neonatal cardiomyocytes.
Results {#Sec2}
=======
Higher glycolysis and oxygen consumption in the P1 heart {#Sec3}
--------------------------------------------------------
There is evidence that the postnatal heart becomes dependent on oxidative catabolic pathways as proliferative capacity declines, in a process possibly induced by atmospheric oxygen^[@CR13]^. Indeed, we found that RNASeq expression analysis of genes involved in oxidative catabolic pathways, such as β-oxidation, the tricarboxylic acid cycle, and oxidative phosphorylation, showed a gradual overall upregulation during the postnatal period (Fig. [1A](#Fig1){ref-type="fig"}). A concomitant exchange in the expression profile of glycolytic enzymes occurred within the first days after birth. RNASeq results were confirmed by qRT-PCR experiments showing that mitochondrial genes were downregulated in P1 relative to P7, while glycolytic Aldolase c gene (Aldoc) was upregulated (Fig. [1B](#Fig1){ref-type="fig"}).Figure 1Gene expression analysis. (**A**) Heat map showing gene expression assayed by RNASeq. The heat maps show gene clusters involved in glycolysis, β-oxidation, oxidative phosphorylation and the tricarboxylic acid cycle in rat hearts from 2, 6, 8, and 12 days after birth (n = 6 for both). (**B**) Quantitative RT-PCR analysis of genes involved in cell metabolism. Bars indicate the relative expression (2^−ΔΔct^) for P1 versus P7 hearts (n = 6 for both) of Aldolase c (a glycolysis enzyme) and mitochondrial Lrrk2, Ndufv3, and Ucp2; t-test, p \< 0.05.
Despite the widespread and robust results in gene expression patterns observed, most metabolic regulation occurs via allosteric and post-translational mechanisms^[@CR15]--[@CR17]^, which may not reflect gene expression values, so we functionally evaluated oxidative metabolism in P1 and P7 hearts. Tissue samples collected from P1 and P7 pups, and oxygen consumption rates, corrected for tissue protein content, were compared using high resolution respirometry^[@CR18]^ using either glucose or fatty acids as substrates (Fig. [2A](#Fig2){ref-type="fig"}). Interestingly, we found that despite the upregulation of genes involved in oxidative metabolism, basal oxygen consumption rates decreased significantly between P1 and P7, with both glucose and fatty acids as substrates, although more strikingly with the former. Not only did glucose-supported oxygen consumption decrease, but the production of lactate (Fig. [2B](#Fig2){ref-type="fig"}) also dropped, indicating that both complete oxidation of glucose to CO~2~ and water through oxidative phosphorylation and fermentation of glucose to lactate dampened during postnatal maturation. Confirming that glycolytic activity decreased between P1 and P7, we found that the activity of lactate dehydrogenase (Fig. [2C](#Fig2){ref-type="fig"}) and hexokinase (Fig. [2D](#Fig2){ref-type="fig"}) decreased during this period, despite the stimulated gene expression depicted in Fig. [1A](#Fig1){ref-type="fig"}. To further characterize metabolic changes in neonatal cardiac cells, we measured mitochondrial content by quantifying markers of mitochondrial biogenesis and mass in heart tissue samples. Citrate synthase activity (Fig. [3A](#Fig3){ref-type="fig"}), VDAC and COX IV levels (Fig. [3B and C](#Fig3){ref-type="fig"}, respectively) were equal between P1 and P7 heart tissue. Similarly, mitochondrial DNA normalized by genomic DNA is equal in P1 and P7 ventricles (Fig. [3D and E](#Fig3){ref-type="fig"}) and cultured cells (Supp. Figure [1](#MOESM1){ref-type="media"}). Overall, these results highlight the limitations of using gene expression patterns to evaluate functional oxidative status in the developing heart. They further demonstrate that the postnatal period involves dampening of anaerobic glycolysis and oxidative catabolic activity, which do not involve changes in mitochondrial content.Figure 2High oxygen consumption and glycolytic capacity in P1 heart tissue. (**A**) Bars indicate oxygen consumption (mean ± SEM) determined by use of an Oroboros O2k high-resolution respirometer in P1 (n = 5) and P7 (n = 6) hearts with glucose (5 mM) or oleic acid (100 µM) as substrates; two-way ANOVA, \*p \< 0.05 vs. P1, ^\#^p \< 0.05 vs. glucose in P1. (**B**) Lactate levels (mean ± SEM) measured in the oxygen consumption conditioned buffer for P1 and P7 hearts (n = 6 for both); t-test, \*p \< 0.05 vs. P1. (**C**) Lactate dehydrogenase activity (mean ± SEM) in P1 and P7 hearts (n = 5 for both); t-test, \*p \< 0.05 vs. P1. (**D**) Hexokinase activity (mean ± SEM) in P1 and P7 hearts (n = 6 for both); t-test, \*p \< 0.05 vs. P1. Figure 3Postnatal metabolic maturation does not alter mitochondrial mass. (**A**) Citrate synthase activity (mean ± SEM) in P1 and P7 hearts (n = 6 for both); t-test, NS. (**B**) VDAC1 protein expression (mean ± SEM) in P1 and P7 hearts (n = 6 for both) measured by Western Blotting; t-test, NS. (**C**) COX IV protein expression (mean ± SEM) in P1 and P7 hearts (n = 6 for both) measured by Western Blotting; t-test, NS. (**D**) Mitochondrial Cox I and (**E**) Mterf1DNA quantification relative to genomic DNA in P1 and P7 heart tissue (n = 6 for both); t-test, NS.
Enhanced maximal and basal oxygen consumption rates in P1 cardiomyocytes {#Sec4}
------------------------------------------------------------------------
Tissue oxygen consumption allows for an evaluation of basal oxidative activity, but cannot readily determine maximal oxidative capacity or how much of this activity reflects ATP production. To assess these parameters and further confirm our results in tissue samples, we prepared P1 and P7 cardiomyocytes cultures and compared their oxygen consumption rates using extracellular flux analysis^[@CR19],[@CR20]^ (Fig. [4](#Fig4){ref-type="fig"}). Figure [4A](#Fig4){ref-type="fig"} shows a typical flux analysis experiment confirming that basal respiration is higher in P1 cells (results are quantified in Fig. [4B](#Fig4){ref-type="fig"}). The addition of the ATP synthase inhibitor oligomycin allows for the determination of the respiratory rates linked to ATP production (oligomycin-inhibited) and the proton leak (oligomycin-insensitive). In both (Fig. [4C and D](#Fig4){ref-type="fig"}, respectively) the values decreased in P7 relative to P1 cardiomyocytes. A subsequent addition of the uncoupler CCCP promotes maximal respiratory rates, which were higher in P1 cells (Fig. [4E](#Fig4){ref-type="fig"}). The spare respiratory capacity (calculated as maximal respiration minus basal), often associated with protection against damaging conditions^[@CR21]^, was equal in P1 and P7 cells (Fig. [4F](#Fig4){ref-type="fig"}). Finally, the addition of the electron transport inhibitors rotenone and antimycin allowed for the determination of non-mitochondrial oxygen consumption rates, which were slightly lower in P7 cells (Fig. [4G](#Fig4){ref-type="fig"}) but contributed little to overall oxygen consumption. We subtracted these rates from all other measurements. Altogether, these results confirm the findings with intact cardiac tissue, indicating that basal respiratory activity decreased between the first and seventh postnatal days, and indicate that this decrease is related to a decrease in ATP-linked oxygen consumption. Also, it shows that the maximal respiratory capacity of P1 cardiomyocytes is higher than P7. Given the surprising results of our oxygen consumption measurements, we decided to visualize the area of functional mitochondria in live cells using the membrane potential-sensitive fluorescent dye tetramethylrhodamine methyl ester (TMRM), which accumulates in mitochondria sustaining active membrane potentials. We observed that P1 cardiomyocytes displayed larger areas of active mitochondria (Fig. [4H](#Fig4){ref-type="fig"}), once again confirming that a functional decrease in mitochondrial oxidative activity occurs between P1 and P7.Figure 4P1 cardiomyocytes have a significant mitochondrial oxidative metabolism. (**A**) Time scan measurements of real-time oxygen consumption (OCR) using a Seahorse Flux Analyzer in cardiomyocyte cultures from P1 and P7 (n = 3 for both). Panels B-H show quantitative comparisons derived from experiments (mean ± SEM), such as those in panel (A,B) Basal OCR; t-test, \*p \< 0.05 vs. P1. (**C**) ATP-linked OCR; t-test, \*p \< 0.05 vs. P1. (**D**) Proton leak; t-test, \*p \< 0.05 vs. P1. (**E**) Maximal OCR; t-test, \*p \< 0.05 vs. P1. (**F**) Reserve capacity; t-test, NS. (**G**) Non-mitochondrial; t-test, \*p \< 0.05 vs. P1. (**H**) TMRM-stained the cell-integrated pixel area (mean ± SEM) in cardiomyocyte cultures from P1 and P7 pups (n = 3 for both); t-test, \*p \< 0.05 vs. P1.
Robust oxidative metabolism is necessary for cardiomyocyte proliferation {#Sec5}
------------------------------------------------------------------------
The neonatal mammalian heart can regenerate following injury, mainly through cardiomyocyte proliferation. We previously demonstrated that robust regeneration occurred after cardiac apex resection in P1 neonatal rats, with cardiomyocyte neoformation and preserved cardiac function, while P7 pups failed to display this regenerative capacity^[@CR2]^. We then determined oxidative metabolism rates in cardiac tissue samples from resected animals. Interestingly, we found that P1 resected cardiac ventricle tissue consumed oxygen at faster rates in the presence of glucose even five days after the procedure compared to P7 animals (Fig. [5A](#Fig5){ref-type="fig"}). Oxygen consumption rates were equal between sham and resected tissues, indicating that the resection process itself was not responsible for the higher respiratory rates, but rather that these rates reflected the developmental stage of the newborn animals. Lactate levels were equal in all samples (Fig. [5B](#Fig5){ref-type="fig"}), indicating that fermentation did not change in this period, and O~2~/CO~2~ ratios were also unchanged, indicating maintenance of the proportion of oxidized lipid and carbohydrate (Fig. [5C](#Fig5){ref-type="fig"}).Figure 5Cardiac cell proliferation does not alter tissue metabolic profiles. (**A**) Bars indicate oxygen consumption (mean ± SEM) with glucose (5 mM) measured using Oroboros O2k high-resolution respirometry in P1, P7, and sham resected neonatal hearts 5 days (+5) after the procedure (S1 + 5d, n = 8 and R1 + 5d, n = 3; S7 + 5d, n = 7 and R7 + 5d, n = 4); two-way ANOV*A*, \*p \< 0.05 vs. 1d + 5. (**B**) Lactate production (mean ± SEM) measured in the oxygen consumption conditioned buffer from P1 and P7 sham and resected neonatal hearts five days after the procedure (S1 + 5d, n = 8 and R1 + 5d, n = 3; S7 + 5d, n = 7 and R7 + 5d, n = 4); two-way ANOVA, NS. (**C**) O~2~/CO~2~ rate (mean ± SEM) calculated from O~2~ and CO~2~ concentration in the oxygen consumption conditioned buffer from P1 and P7 sham and resected neonatal hearts 5 days after the procedure (S1 + 5d, n = 9 and R1 + 5d, n = 3; S7 + 5d, n = 7 and R7 + 5d, n = 4); two-way ANOVA, NS.
These results suggest that high levels of oxidative metabolism support the proliferative activity required for cardiac regeneration that occurs only within a narrow postnatal period. To further explore these findings, we measured cell proliferation *in vitro* in P1 and P7 cardiomyocytes cultures using high-content screening assays for serum-induced proliferation. To accurately differentiate cardiomyocytes from fibroblasts in culture, we used monoclonal antibodies against sarcomeric tropomyosin and vimentin, respectively, in P1 and P7 cultures. Data are presented as proliferation rates *in vitro* after 24 hours of culture. The cardiomyocyte proliferation rate was significantly greater in P1 compared to P7 cultures (Fig. [6](#Fig6){ref-type="fig"}), confirming their enhanced proliferation in response to serum. Fibroblast proliferation rates were similar in P1 and P7 cultures (Fig. [6A](#Fig6){ref-type="fig"}). Corroborating this, CDK1 protein and cell cycle gene markers were higher in P1 heart tissues, indicating maintenance of cell cycle activity (Fig. [6B](#Fig6){ref-type="fig"} and Supp. Figure [2](#MOESM1){ref-type="media"}, respectively). CDK1 gene expression differences tended in the same direction (Supp. Figure [2](#MOESM1){ref-type="media"}; p = 0.0592).Figure 6P1 cardiomyocytes maintain high proliferative rates in culture. (**A**) Bars indicate P1 (n = 7) and P7 (n = 4) cardiomyocyte and fibroblast proliferative rates (mean ± SEM) after 24 hours in culture; two-way ANOVA, \*p \< 0.05 vs. P1. (**B**) CDK1 protein expression (mean ± SD); t-test, \*p \< 0.05 vs. P1); t-test, \*p \< 0.05 vs. P1, in P1 and P7 hearts (n = 6).
Next, we sought to test the contribution of oxidative metabolism to the proliferative capability of P1 cells by promoting a partial, nonlethal, chemical inhibition of respiration and following its effect on the proliferation of P1 cells. The mitochondrial complex I inhibitor rotenone, which promotes the accumulation of NADH, and thus inhibition of both the tricarboxylic acid cycle and β-oxidation, was used at a dose of 5 nM, which did not affect cell viability (Fig. [7A](#Fig7){ref-type="fig"}). This dose, however, reduced mitochondrial oxidative metabolism with a marked effect on ATP-linked oxygen consumption (Fig. [7B,C](#Fig7){ref-type="fig"}). Interestingly, the reduction in mitochondrial oxygen consumption was sufficient to reduce cell cycle activity, measured by differential condensed chromosomes and Ki67 expression (Fig. [7D and E](#Fig7){ref-type="fig"}, respectively), and to inhibit the proliferative capacity of P1 cardiomyocytes (Fig. [7F](#Fig7){ref-type="fig"}). Conversely, this seems not to affect fibroblast proliferation (Fig. [7F](#Fig7){ref-type="fig"}). Overall, these results show that high mitochondrial respiratory activity in P1 cells is required to sustain cardiomyocyte proliferation.Figure 7Decreased oxygen consumption reduces P1 cardiomyocyte proliferative rates. (**A**) Cell viability (mean ± SEM) measured in P1 cardiomyocyte culture under control conditions and treated for 48 hours with 5 nM rotenone (n = 3 for both); t-test, NS. (**B**) Basal OCR (mean ± SEM) using a Seahorse Flux Analyzer in P1 cardiomyocyte cultures, control condition and treated with 5 nM rotenone for 48 hours (n = 3 for both); t-test, \*p \< 0.05 vs. control. (**C**) ATP-linked OCR (mean ± SEM) in P1 cardiomyocyte cultures, control condition and treated with 5 nM of rotenone for 48 hours (n = 3 for both); t-test, \*p \< 0.05 vs. control. (**D**) Cell cycle analysis (mean ± SEM) in P1 cardiomyocyte cultures under control conditions and treated for 48 hours with 5 nM rotenone (n = 3 for both); two-way ANOVA, \*p \< 0.05 vs. control. (**E**) Percentage of Ki67^+^ cardiomyocytes (mean ± SEM) in P1 cardiomyocyte cultures (n = 4 for both); t-test, \*p \< 0.05). (**F**) Bars indicate P1 cardiomyocyte and fibroblast proliferative rates (mean ± SEM) of after 24 hours in culture under control conditions (n = 7) or when treated with 5 nM rotenone (n = 6) for 48 hours; two-way ANOVA, \*p \< 0.05 vs. control.
Discussion {#Sec6}
==========
The main finding of the present work is that the early postnatal high cardiac metabolism and accompanying cell proliferation correlates with oxidative energy metabolism. The energy metabolism demand subsides as proliferation ceases and oxidative energy metabolism prevails as the main energetic source for the cardiac tissue after that.
Recent evidence indicates that rodent hearts display a regenerative capacity restricted to a period of few days after birth, after which cardiomyocytes permanently differentiate. This transition is complex, and data suggest a role for atmospheric oxygen tension as a trigger^[@CR13]^. Also, one may hypothesize that the process involves a shift from more fermentative to more oxidative metabolism. Indeed, gene expression patterns, reported here (Fig. [1](#Fig1){ref-type="fig"}) and previously^[@CR22]^, and quantitative mass spectrometry analysis^[@CR13]^ show a striking increase in the expression of all energy metabolism pathways that require oxygen.
However, the gene expression pattern may be limited as a proxy of metabolic activity, since much of metabolic regulation involves post-translational modifications and allosteric mechanisms^[@CR15]--[@CR17]^. Indeed, our functional assessment of respiratory activity in neonatal pup hearts, using both tissue slices (Fig. [2A](#Fig2){ref-type="fig"}) and cultured cells (Fig. [4](#Fig4){ref-type="fig"}) shows that oxidative metabolism decreases between the first and seventh days after birth. This decrease involves an overall dampening in metabolic fluxes because it occurs in the presence of both glucose and fatty acids (Fig. [2A](#Fig2){ref-type="fig"}). A parallel decrease in lactate production is also observed (Fig. [2B](#Fig2){ref-type="fig"}), indicating that anaerobic glycolysis is also blunted during perinatal heart development, suggesting that overall energetic demands decrease in the days after birth. These effects are not accompanied by changes in mitochondrial mass (Fig. [3](#Fig3){ref-type="fig"}). Indeed, cardiomyocyte extracellular flux analysis experiments indicate that the decrease in oxidative metabolism between P1 and P7 was mostly related to lower rates of ATP-linked mitochondrial respiration (Fig. [4](#Fig4){ref-type="fig"}). Overall, these results indicate that gene expression analysis is a poor indicator for functional metabolic status and suggest that metabolic activity is the preferred indicator to be evaluated whenever possible. It also may apply to differentiating cells, in which significant changes in protein turnover may occur.
Our results indicate that a decrease in oxidative metabolism over the first days after birth accompanies postnatal cardiac maturation. Metabolic changes that occur during differentiation have been evaluated only sporadically in other tissues, and have varying results. Neuronal differentiation also involves a decrease in oxidative metabolism, but, differently from our findings with cardiomyocytes, with a concomitant increase in fermentation^[@CR23]^. Mesenchymal stem cell differentiation to adipocytes and osteocytes involves increases in oxygen consumption rates, but differentiation into chondrocytes involves a large decrease in respiratory activity^[@CR9]^. Overall, these studies show that there is no specific rule regarding oxidative metabolism and reproductive capacity, but rather that oxidative metabolic rates follow energetic necessities, which are variable during differentiation.
The mechanisms behind this shift from high oxidative metabolic rates to lower oxidative rates in neonatal cardiomyocytes are complex and remain to be determined. Recently, Puente *et al*.^[@CR13]^ demonstrated that hypoxia depresses postnatal cardiomyocyte cell cycle arrest and mitochondrially-targeted antioxidants prevent it. Because the significant accumulation of markers of oxidative damage also occurs in this process, the authors concluded that redox signaling processes triggered by atmospheric air mediate cell cycle arrest. This finding raises the interesting possibility that redox signaling may also be responsible for the decrease in oxygen consumption that occurs between P1 and P7.
Irrespective of the mechanism whereby oxidative metabolism is dampened days after birth, we established that the maintenance of high electron transfer rates is pivotal for cardiomyocyte proliferation (Fig. [7](#Fig7){ref-type="fig"}), and therefore regeneration, consistent with the high energy expenditure required for cardiomyocyte proliferation. Primitive fetal liver hematopoietic stem cells use oxidative phosphorylation to generate ATP more efficiently for extensive cell expansion^[@CR8]^. Recently, hypoxia and glucose deprivation was shown to reduce ATP levels in murine induced stem cell cardiomyocytes and neonatal cardiomyocytes, with an increase in oxidant production and cell death specially in the former^[@CR24]^. Interestingly, our results show that the relationship between oxygen consumption and the ability to proliferate *in vitro* is restricted to cardiomyocytes.
Overall, our results using functional techniques demonstrate that a marked decrease in oxygen-dependent metabolic rates occurs during postnatal cardiac development and that the high respiratory rates present in neonatal cells are necessary for cardiomyocyte proliferation. Based on these results, it is tempting to speculate that interventions designed to extend the narrow "regenerative window" of newborn hearts may include a coadjutant stimulus to promote mitochondrial biogenesis or decrease mitophagy^[@CR25]^ and may also be relevant for strategies aimed at expanding cardiac cells *in vitro* for transplantation or stimulation of endogenous cardiac proliferation therapies.
Methods {#Sec7}
=======
Animals {#Sec8}
-------
We used in this study Wistar rats in postnatal day 1 (P1) and postnatal day 7 (P7). They were maintained with the mother rat in a 12:12 hour light--dark cycle and temperature-controlled environment (22 °C) with free access to food and water.
The Institutional Review Board of the University of São Paulo Medical School, Brazil aprooved the experimental procedures (\#285/12), which followed the US National Institutes of Health and institutional guidelines for care and use of laboratory animals.
Cardiomyocyte isolation and culture {#Sec9}
-----------------------------------
We euthanized the P1 and P7 neonatal rats by decapitation, and the hearts were excised and rinsed in cold ADS buffer. We pooled a total of 8--10 neonatal hearts in each isolation. We removed the atria and minced and digested the ventricles in 3 mL of pre-warmed digestion buffer containing collagenase II (Worthington Chemicals) and pancreatin (Sigma-Aldrich) in a shaking incubator set at 37 °C for 15 minutes. After this, we transferred the digested material to 15 mL tubes for inactivation with media containing two mL Dulbecco's Modified Eagle's Medium and M199 (4:1) containing 10% Donor Horse Serum, 5% Newborn Calf Serum, and 1% penicillin/streptomycin (Invitrogen, Life Technologies). We centriguged the supernatant collected from each digest at 220 × g for 5 minutes, and the cell pellet was resuspended in fresh medium and maintained at 37 °C. We repeated six times the incubation and centrifugation steps and resuspended the final cell pellet from all digests in a 100-µm cell strainer, followed by two mL of fresh media. Adherent cells were attached for 45 minutes in culture dishes at 37 °C in a humidified atmosphere of 95% air and 5% CO~2~. The non-adherent fraction (myocytes) was centrifuged at 220 × g for 5 minutes and counted in a Neubauer chamber. We plated the cells on laminin-coated plates. After 48 hours, cardiomyocytes formed a monolayer of viable and spontaneously beating cells, and they were trypsinized and frozen for later use.
Tissue oxygen consumption assays {#Sec10}
--------------------------------
Mitochondrial oxygen consumption was monitored using a computer-interfaced Clark-type electrode, as previously described^[@CR18]^, with continuous stirring at 37 °C. For cardiac ventricle tissue, oxygen consumption assays were performed using Oroboros O2k high-resolution respirometry (Oroboros Instruments, Bioblast). We added each cardiac ventricle slice (1--2 mm) into the equipment chamber with two mL of Krebs buffer without substrates, which allowed the maintenance of the physical structure of the thin neonatal cardiac tissue. After recording basal respiration, we added glucose (5 mM) or oleic acid (100 µM) and registerd the oxygen consumption rate for a further 5 minutes. We normalized the results by basal respiration and tissue protein content. We stored conditioned buffer and heart tissue for future analysis.
Seahorse oxygen consumption rate (OCR) measurements {#Sec11}
---------------------------------------------------
We measured oxygen consumption rates (OCR) in P1 and P7 primary cardiomyocyte cultures using an XF24 Extracellular Analyzer (extracellular flux, 24-well plate, Seahorse Bioscience), as previously described^[@CR19],[@CR20]^ and also exposed P1 cardiomyocytes to 5 nM rotenone (Sigma-Aldrich). The cells were seeded as three or four replicates in a laminin-coated XF24 24-well culture microplate at 40,000 cells/well (0.21 cm^2^ growth area) in 500 μL of growth medium and incubated for 48 hours at 37 °C in a humidified atmosphere of 95% air and 5% CO~2~. Before the assay, the medium was removed and replaced by 500 μL of bicarbonate-free assay medium (114 mM NaCl, 4.7 mM KCl, 1.2 mM KH~2~PO~4~, 1.16 mM MgSO~4~, 2.5 mM CaCl~2~, pH 7.2, and 2.8 mM glucose). The cells were pre-incubated for 1 hour at 37 °C in air. We determined respiration-driven ATP synthesis and proton leak-driven respiration by the addition of oligomycin (4 μg/mL). After three measurement cycles, five μM of the uncoupler carbonyl cyanide-p-trifluoromethoxy-phenylhydrazone (CCCP) was added to determine maximal respiratory capacity. After a further three measurement cycles, one μM rotenone was added to block complex I in addition to 1 μM antimycin A to inhibit complex III, thereby ablating mitochondrial oxygen consumption. We normalized OCR by the amount of cellular protein content in each well.
Lactate production and O~2~/CO~2~ ratios {#Sec12}
----------------------------------------
Conditioned buffers from tissue oxygen consumption assays were used to measure lactate concentrations, diluted oxygen, and dioxide carbon concentrations in duplicate, using *ABL 800 Flex* equipment, according to the manufacturer's instructions. We used protein content to normalize the measures.
Enzymatic activities {#Sec13}
--------------------
Enzymatic activities were determined in cardiac ventricle tissues using kinetic assays. For hexokinase activity, we homogenized the heart samples in buffer containing 50 mM of Trietanolamin, 50 mM Tris-HCl, one mM EDTA, two mM of MgCl~2~, and 30 mM β-mercaptoethanol (pH 7.5). The suspension was centrifuged and the resulting supernatant was added in duplicate to a 96-well plate with 75 mM Tris-HCl, 0.8 mM EDTA, 7.5 mM MgCl~2~, 1.5 mM KCl, and 4 mM β-mercaptoethanol (pH 7.5), containing 0.4 mM NADP^+^, ATP (1.8 mM), creatine phosphokinase (1.8 U), phosphocreatine (0.4 mM), 1% TRITON-X 100, and glucose-6-phosphate dehydrogenase. The time scan started after the addition of glucose, and the activity was estimated using a spectrophotometer at 340 nm. For lactate dehydrogenase activity, we homogeneized the heart samples in buffer containing 77 mM of Tris-HCl, one mM EDTA, two mM MgCl~2~, and four mM β-mercaptoethanol (pH 7.5). We centrifuged the suspension, and the added the resulting supernatant in duplicate to a 96-well plate with 120 mM Tris-HCl, 3.4 mM NADH, and 1% TRITON X-100 (pH 7.5). We began the the time scan by adding 20 mM of pyruvate and estimated the activity following the absorbance at 340 nm. For citrate synthase activity, we homogenized the heart samples in buffer containing 50 mM Tris and one mM EDTA (pH 7.4) and centrifuged the suspension and added the resulting supernatant in duplicate to a 96-well plate with 100 mM Tris, 0.2 mM DTNB, 30 mM acetyl-coenzyme A and 1% TRITON X-100 (pH 8.1). We began the time scan by adding 0.5 mM oxaloacetic acid, and estimated the activity following the absorbance at 412 nm.
Cell viability {#Sec14}
--------------
P1 cardiomyocyte culture viability was determined using the MTT assay. Cells were plated in duplicate at 20,000 cells/well in 96-well plates and maintained in serum-free medium for the measurement. We added 20 µL of MTT solution (Thiazolyl Blue Tetrazolium Bromide, Sigma-Aldrich) at 5 mg/mL in PBS to each well after 45 hours of treatment with 5 nM rotenone. The plate was shaken and incubated again at 37 °C for 3 hours. We removed the culture medium and added 200 µL of DMSO and performed the reading at 560 nm. Absorbance measured was directly related to cell viability.
Apical resection surgery {#Sec15}
------------------------
P1 and P7 neonatal rats underwent heart apex resection as previously described^[@CR2]^. Neonates were anesthetized by hypothermia for 10 minutes and kept in the inhaler with isoflurane during the procedure (Isoforine - Cristália). After a small skin incision, we performed a ventrolateral thoracotomy at the third intercostal space by dissection of the intercostal muscles. The heart was exposed, and the apex was resected using iridectomy scissors. We closed the thoracic wall incision using 7.0 non-absorbable silk sutures. For recovery, the neonates were placed on a warmed plate (29 °C) under a heat lamp with oxygen flow until fully conscious. In Sham surgery, we did not perform the resection. We euthaneized the P1, P7, and sham rats by decapitation five days after apical resection and collected the hearts.
High content screening cardiomyocyte proliferation rates {#Sec16}
--------------------------------------------------------
We seeded P1 and P7 primary cardiomyocytes in duplicate on laminin-coated 96-well plates at 10,000 cells/well in Dulbecco's Modified Eagle's Medium and M199 (4:1) containing 10% Donor Horse Serum, 5% Newborn Calf Serum, and 1% penicillin/streptomycin (Invitrogen, Life Technologies), and incubated at 37 °C in a humidified atmosphere of 95% air and 5% CO~2~ for 24 and 48 hours. We also exposed P1 cardiomyocytes to 5 nM rotenone (Sigma-Aldrich). After each time point, for imaging, we incubated the cells for 30 min at 37 °C in the presence of Hoechst 33342 (1 µg) and the membrane potential-sensitive fluorescent dye tetramethylrhodamine methyl ester (TMRM, 15 nM, Invitrogen, Life Technologies). After the live acquisition, cells were fixed with 4% paraformaldehyde, permeabilized with 0.1% Triton X-100 for 15 minutes at 4 °C, and blocked with 5% serum albumin bovine for 60 minutes at room temperature. For immunostaining, anti-tropomyosin (sarcomeric) antibody (T9283 -- 1:200, Sigma-Aldrich), anti-vimentin antibody (ab92547 -- 1:300, Abcam) and anti-Ki67 antibody (ab16667 - 1:300, Abcam) were added and incubated overnight at 4°C, followed by washing. Secondary antibodies (Alexa 488 and 555--1:500, Molecular Probes, Invitrogen, Life Technologies) were added and incubated for 1 hour at room temperature, followed by washing. We labeled cell nuclei with DAPI nuclear stain (1:100). Images were acquired with an IN Cell Analyzer 2200 high-content imaging system (GE Healthcare, USA) configured with the standard size CMOS camera. Images of live cells were acquired using 40x (0.6NA) objective (9 images per well-brightfield, Hoechst 33342, Cy3); images of fixed cells were acquired using 20x (0.45 NA) objective (20 images per well-brightfield, DAPI, FITC, Cy3). Analysis protocols were developed with IN Cell Investigator^TM^ software (CM and fibroblasts differential counting GE Healthcare, USA) as well as using a python script (for %KI67^+^ cardiomyocytes enumeration). The area of functional mitochondria was quantified using the cell-integrated area of Cy3 channel to investigate TMRM-stained area (in pixels). Cardiomyocytes and fibroblasts were identified, differentiated, and quantified using the cell-integrated intensity of the FITC channel for Tropomyosin positive cells and Cy3 channel for Vimentin positive cells. We estimated the proliferation rate as the ratio between the number of cells after 48 hours in culture normalized to the number of cells after 24 hours in culture. The percentage of Ki67 + cardiomyocytes was obtained by counting and summing all cardiomyocytes (defined as tropomyosin^+^ cells, FITC channel) with or without KI67 staining in the nucleus for every 20 images per well. The average of %KI67 + cardiomyocytes per well is reported. We analyzed cell cycle phases by differential total DAPI signal per nucleus and categorized into G1, S and G2/M cell cycle phases. We used a function in the Spotifre Decision Site data visualization software (Tibco) provided along with IN Cell Investigator software.
RNASeq analysis {#Sec17}
---------------
We prepared the libraries, and next-generation sequencing - RNASeq (also known as WTSS -- Whole-transcriptome Shotgun Sequencing) in the "Large-Scale Sequencing Laboratory" facility located at the University of São Paulo Medical School, FMUSP, São Paulo-SP, Brazil. Total RNA from samples was purified (RNeasy Mini Kit, Qiagen) and analyzed with the Agilent 2100 Bioanalyzer (Agilent Technologies). Samples with RIN higher than 9.0 and A260/280 higher than 1.8 were considered adequate for library preparation with TruSeq Stranded Total RNA with "Ribo-Zero" Gold kit (Illumina). Libraries were individually quantified through real-time PCR (qPCR) and sequenced with HiSeq. 2500 equipment (Illumina) according to the manufacturer's instructions with pair-end reads (2 × 100 cycles). Data were analyzed using bioinformatics tools. We estimated differential gene expression with the DESEq v1.12.1 package using the R-Bioconductor. Annotations of genes were made using Biomart (R-Bioconductor) software.
Western blots {#Sec18}
-------------
We denatured tissue lysate proteins (40 µg) by heating at 100 °C for 5 minutes; separeted the proteins using SDS-PAGE and transferred to PVDF membranes (Amersham Hybond^TM^--P, GE Healthcare). We blocked the blotted membranes with 5% bovine serum albumin overnight at 4 °C. We detected individual proteins by blotting with specific primary antibodies for 1 hour at room temperature using anti-VDAC (ab15895, 1:1000, Abcam), anti-COX IV (\#4840, 1:1000, Cell Signaling), anti-CDK1 (ab32384, 1:1000, Abcam), anti-LC3 (\#2775, 1:1000, Cell Signaling) and anti-beta-actin (ab8227, 1:2000, Abcam) followed by secondary probing with HRP-conjugated (1:1000 dilution). Immunodetection was determined using the enhanced chemiluminescence (ECL) method. We used ImageJ software (NIH software) for densitometry normalized it with beta-actin expression.
Gene expression measurements by RT-PCR and mitochondrial DNA quantification {#Sec19}
---------------------------------------------------------------------------
We isolated total RNA using PureLinK^TM^ RNA Mini Kit (Life Technologies, USA) following the manufacturer's instructions and performed cDNA synthesis with a SuperScript^TM^ III First-Strand Synthesis System for RT-PCR kit (Invitrogen, Life Technologies). Reverse transcription polymerase chain reaction (RT-PCR) was performed in duplicate with a QuantStudio^TM^ 12 K Flex Real-Time PCR System (MicroAmp® Optical 384-well Reaction Plate), using Taq polymerase SYBR® Green PCR Master Mix (Applied Biosystem, Foster City, CA) under the following conditions: initial denaturation at 95 °C for 10 minutes; 40 cycles of denaturation at 95 °C for 15 seconds, annealing and extension at 60 °C for 1 minute; final extension at 72 °C for 5 minutes. The fluorescent signal was collected at the end of every cycle. Data are presented as 2^−∆∆Ct^, with ∆∆Ct = ∆Ct~experimental~ − ∆Ct~control~ and standard deviation was calculated using the following formula: $\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$s^{\prime} =\sqrt{{s}_{{experimental}}^{2}+{s}_{{control}}^{2}}$$\end{document}$ when s is the standard deviation of the corresponding delta-ct's on dataset. For mitochondrial DNA relative quantification qPCR, Mterf1 and mitochondrial Cox1 genes was assayed in based on protocol previously described^[@CR26]^. Primers sequences and details are presented on Supplemental Table [1](#MOESM1){ref-type="media"}.
Statistical analysis {#Sec20}
--------------------
Results are presented as means ± standard error of the mean. The unpaired Student *t* test and two-way ANOVA with post hoc Bonferroni's test were used to compare groups as appropriate. RT-PCR and mitochondrial DNA normalized by genomic DNA t-tests were analyzed using the ∆Ct~experimental~ dataset. All statistical analyses were performed using GraphPad Prism 5.0 (GraphPad Softwares Inc., CA, USA). A p ≤ 0.05 value was considered statistically significant.
Electronic supplementary material
=================================
{#Sec21}
Supplementary Information
**Electronic supplementary material**
**Supplementary information** accompanies this paper at 10.1038/s41598-017-15656-3.
**Publisher\'s note:** Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
This work was supported by grants to J.E.K. from Fundação Zerbini; São Paulo Research Foundation \[FAPESP 13/17368-0\]; Conselho Nacional de Desenvolvimento Científico e Tecnológico \[404733-2012-3\]; CAPES PROBITEC \[23038-4156-2012-80\] and Ministério da Saúde do Brasil \[MS-DECIT-PROADI_SUS -- H. Samaritano\] and to A.J.K. from Centro de Pesquisa, Inovação e Difusão de Processos Redox em Biomedicina \[FAPESP 13/07937-8\] and Núcleo de Apoio à Pesquisa de Processos Redox em Biomedicina \[Universidade de São Paulo\]. A.E.T.S.C, V.B., M.F.F and A.A.K. are recipients of FAPESP fellowships 2012/10109-7, 2011/19678-1, 2013/04871-6 /2015/25776-7 and 2013/26440-7, respectively. We wish to thank Juliana Sanajotti Nakamuta for helpful discussions. Camille C. Caldeira-da-Silva, Elida Adalgisa Neri, Anelise Casagrande da Silva, Mariliza Velho Rodrigues, Sueli Oba, Joelcimar Martins da Silva and Mario Costa Cruz for technical assistance, as well as Leno Luis Dima for animal manipulation and the Core Facility for Scientific Research -- University of São Paulo (CEFAP-USP/ CONFOCAL) and William Scaff for the High Content Screening analysis and Antônio Carlos Lerário for bioinformatics data analysis of RNASeq data.
A.E.T.S.C., V.B., M.F.F., A.J.K. and J.E.K. proposed and designed the experiments. A.E.T.S.C., M.F.F. and A.A.K. performed the experiments. A.E.T.S.C., V.B., M.F.F., A.A.K., A.J.K. and J.E.K. analyzed the data and discussed the interpretation of results. A.E.T.S.C., V.B. and M.F.F. prepared the figures. A.E.T.S.C., M.F.F., A.J.K. and J.E.K. wrote the main manuscript text. All authors reviewed the manuscript.
Competing Interests {#FPar1}
===================
The authors declare that they have no competing interests.
| {
"pile_set_name": "PubMed Central"
} |
All nucleotide accession files are available from the HIV Sequence Database (accession numbers: AY423971, AY423984, AY424079, AY424138, AY424163, DQ388514-DQ388517, DQ411851-DQ411854, DQ422948, DQ435682, DQ435684, FJ443190, FJ443274, FJ443316, FJ443382, FJ443533, FJ443575, FJ443639, FJ443670, FJ443711, FJ443808, FJ443841, FJ443999, FJ444017, FJ444059, FJ444103, FJ444124, FJ444215, FJ444284, FJ444315, FJ444325, FJ444366, FJ444384, FJ444395, FJ444421, FJ444437, FJ444529, FJ444561, FJ444586, FJ444600, FJ444612, FJ496194, FJ496204, HM215253, HM215254, HM215262, HM215263, HM215278, HM215286, HM215289, HM215300, HM215302, HM215305, HM215307, HM215308, HM215311, HM215316-20, HM215324, HM215326, HM215328, HM215335-36, HM215338, HM215341, HM215344, HM215352, HM215354, HM215360, HQ595742-43, HQ595747-51, HQ595753, HQ595756-7, HQ595759, HQ595763, HQ595766, HQ615941-61, JN681219-JN681223, JN681225-34, JN681236, JN681238-49, JN681252-58, JN967790-94, JN967797-8, JN977604, JQ061131, JQ352785, JQ352789, JQ352790, JQ352794, JQ352799, JQ352801, JX131327, KC154012-20, KC154022-28, KC156123, KC156125, KC156129-30, KC247557, KF114881-2, KF114884-95, KJ598853, KJ700457-8).
Introduction {#sec001}
============
The development of effective biomedical intervention strategies to prevent HIV-1 infection remains a global priority. To support these efforts, two large immunization trials in high incidence, clade C epidemic regions in southern Africa are imminent. The first, a Phase 3 efficacy trial using a vaccine similar to the one used in the RV144 trial modified to include clade C antigens will be tested to determine if the protection observed in the RV144 vaccine trial in Thailand can be replicated in this high incidence setting (<http://vaccineenterprise.org/content/P5Partnership>). The second is a Phase 2b trial to evaluate if passive administration of the VRC01 monoclonal antibody, that targets the viral CD4 binding site (CD4bs), reduces HIV-1 acquisition \[[@ppat.1005742.ref001]\]. Both interventions rely on the induction of HIV-specific antibodies against the HIV-1 envelope glycoprotein. HIV-1 is extraordinarily diverse, and evaluation of potential coverage by these intervention strategies would therefore need to take envelope diversity into account. As there is a severe HIV-1 transmission bottleneck that may affect viral phenotype \[[@ppat.1005742.ref002]--[@ppat.1005742.ref007]\], studies that aim to elucidate the target for active and passive immunization should ideally be done on viruses that are collected soon after transmission.
Although correlates of protection from HIV-1 infection are not fully understood, both neutralizing and non-neutralizing antibodies are known to play a crucial role. The importance of neutralizing antibodies is demonstrated in non-human primate models where passive administration of broadly neutralizing antibodies (bnAbs) conferred complete protection from simian-human immunodeficiency virus (SHIV) challenge \[[@ppat.1005742.ref008],[@ppat.1005742.ref009]\]. While bnAbs with extraordinary coverage have been tested for safety in HIV-1 infected and uninfected humans \[[@ppat.1005742.ref001]\], the VRC01 trial will be the first to evaluate the effectiveness of bnAbs for prevention. While a vaccine should ideally elicit bnAb responses, no such vaccine has been developed to date. However results from the RV144 efficacy vaccine trial in Thailand found that non-neutralizing antibodies may offer an alternative route to protection \[[@ppat.1005742.ref010]\]. The HIV-1 envelope is comprised of several variable domains, and antibody responses directed to V1/V2 variable loops \[[@ppat.1005742.ref010],[@ppat.1005742.ref011]\], as well as to V2 and V3 linear peptides \[[@ppat.1005742.ref012]\], were inversely correlated with infection risk in RV144. Genetic signatures in breakthrough infections that were associated with reduced acquisition risk were also identified \[[@ppat.1005742.ref011],[@ppat.1005742.ref013],[@ppat.1005742.ref014]\].
The HIV-1 global pandemic is highly diverse, comprising of many different clades (also known as genetic subtypes) and recombinant circulating forms (chimeras between more than one clade). Although neutralizing antibody responses elicited by one clade are generally effective at neutralizing viruses across clades, there is evidence of enhanced potency when clades are matched \[[@ppat.1005742.ref015]--[@ppat.1005742.ref019]\]. Even within a clade, diversity has been shown to have an effect illustrated by the fact that the clade-matched neutralization advantage was more pronounced in regions with lower viral diversity such as Thailand, compared to epidemics in southern Africa with higher viral diversity \[[@ppat.1005742.ref015]\]. The effects of viral diversification on neutralization targets have also been seen by tracking the epidemic over time. Two cohorts in Europe, both predominantly infected with clade B, showed that HIV-1 has evolved at a population level to become more resistant to serum and bnAb neutralization over a twenty-year period \[[@ppat.1005742.ref020]--[@ppat.1005742.ref022]\]. This high diversity and increasing neutralization resistant phenotype is a potential problem for both active and passive immunization.
Immunization strategies need to block the virus(es) that establishes infection, and it is thus important to understand the properties of these transmitted founder viruses. There have been conflicting observations in cross-sectional studies as to whether viruses collected soon after transmission are more sensitive or more resistant to neutralization compared to viruses from chronic infection. In one study acute clade B viruses tended to be more sensitive to neutralization by VRC01, as well as pooled IgG from HIV-1 infected subjects compared to chronic viruses \[[@ppat.1005742.ref005]\]. A separate clade C study generally detected no significant difference in neutralization susceptibility to VRC01 or sera \[[@ppat.1005742.ref006]\]. In contrast to the clade B study, a large multi-clade study showed that early viruses were more neutralization resistant than late viruses when tested against clade-mismatched plasma, with no difference observed for clade-matched plasma \[[@ppat.1005742.ref015]\]. Small sample size in some studies together with differences in study design and reagents all potentially confounded these studies, motivating for a large standard panel of viruses collected soon after transmission to more accurately evaluate targets of passive immunization studies and vaccines.
Here, using a large panel of 200 HIV-1 clade C Env-pseudotyped viruses generated from acute/early infection, we investigated the genotypic and neutralization properties of these viruses that may impact non-neutralizing and neutralizing antibody recognition. This collection is a good representation of viruses from southern Africa, the region of the world most affected by the HIV-1 epidemic and a major region for passive and active immunization efficacy trials. This panel, representing the largest collection of clade C Env-pseudotyped viruses from acute/early infection, is a valuable resource to the field, providing a reagent set that will enable establishing the cross-reactive potential of newly isolated monoclonal antibodies and the characterization of vaccine responses in the critically important HIV-1 clade C epidemic in southern Africa.
Methods {#sec002}
=======
Ethics statement {#sec003}
----------------
The CAVIMC-CAVD HIV-1 Clade C Virus Neutralization Phenotype Study was reviewed and approved by the research ethics committee of the Faculty of Health Sciences of the University of Cape Town (168/2007; 513/2012). All participants provided written informed consent for study participation.
Study participants {#sec004}
------------------
Samples used to generate functional *env* clones originated from Botswana (BW, n = 6), Zambia (ZM, n = 13), Malawi (MW, n = 23), Tanzania (TZ, n = 28), and South Africa (ZA, n = 130). South African samples originated from eight provinces: Western Cape (Cape Town, ZAwc n = 12), Eastern Cape (ZAec n = 2), North-West (ZAnw n = 7), Kwazulu-Natal (ZAkzn n = 68), Mpumalanga (ZAmp n = 6), Northern Cape (ZAnc n = 3), Limpopo (ZAlp n = 5) and Gauteng (Soweto and Johannesburg, ZAgp n = 27) ([Table 1](#ppat.1005742.t001){ref-type="table"} and [S1 Table](#ppat.1005742.s009){ref-type="supplementary-material"}). Of the 200, 199 were assumed to be sexually transmitted, and one was transmitted by breast feeding. Samples were catalogued as originating from an individual soon after transmission if they had a documented HIV-1 negative test within the previous 100 days, or if they were HIV-1 PCR positive and antibody negative at the time the sample was collected.
10.1371/journal.ppat.1005742.t001
###### Characteristics of the acute/early clade C panel virus donors by infection stage and country/region of origin.
{#ppat.1005742.t001g}
*Infection Stage*
------------------------------------ ------- --------- ------------------- -------- ---------
Botswana BW 6 1 5 0
Malawi MW 23 13 7 3
Tanzania TZ 28 2 1 25
South Africa ZA
Western Cape (Cape Town) ZAwc 12 1 0 11
Eastern Cape ZAec 2 0 2 0
North-West ZAnw 7 1 0 6
Kwazulu-Natal ZAkzn 68 19 8 41
Mpumalanga ZAmp 6 5 1 0
Northern Cape ZAnc 3 3 0 0
Limpopo ZAlp 5 5 0 0
Gauteng (Johannesburg and Soweto) ZAgp 27 14 4 9
Zambia ZM 13 3 1 9
**Total** **200** **67** **29** **104**
^a^Pre-seroconversion/acute viruses (PCR positive but HIV-1 seronegative, Ab-) (Fiebig Stage I /II) (56).
^b^Indeterminate viruses (early seroconversion with an indeterminate western blot, Ab+/-) (Fiebig Stage III /IV).
^c^Post-seroconversion/early viruses (previous HIV negative diagnosis \<100 days, Ab+) (Fiebig Stage V/VI).
Single genome amplification (SGA) and sequencing {#sec005}
------------------------------------------------
cDNA synthesis, followed by single genome amplification of *env* (\~2.5kb), was performed according to a method described previously \[[@ppat.1005742.ref004],[@ppat.1005742.ref023]\]. Both *env* strands were directly sequenced using an ABI PRISM 3100 Genetic Analyser using BigDye terminator reagents (Applied Biosystems, Warrington, UK), and sequence reads were assembled, edited and consensus sequences generated using Sequencher version 5.2.3 (Gene Codes Corporation, Ann Arbor, MI USA). Amplicons used for sequencing were generated using the limiting dilution approach, and for 90% (n = 179) of samples, at least five SGA derived sequences were generated per sample allowing for accurate transmitted founder consensus inference \[[@ppat.1005742.ref024]--[@ppat.1005742.ref026]\] ([S1 Table](#ppat.1005742.s009){ref-type="supplementary-material"}).
Pseudovirus production and infectivity assay {#sec006}
--------------------------------------------
For pseudovirus production, full-length *rev-env* cassettes were cloned into one of two mammalian expression vectors, pcDNA 3.1 Directional/V5-His-TOPO (Invitrogen, Carlsbad, CA) or pTarget (Promega, Madison, WI). The amplicon sequence selected for cloning was the one closest to a participant's consensus that was generated from at least five sequences. The resulting clones were sequenced to ensure an exact match to the original amplicon sequence and where cloned inserts differed from the parental sequence, mutagenesis was performed to ensure a match with the parental sequence. Env-pseudotyped viruses were generated by co-transfecting envelope clones with a clade B backbone pSG3ΔEnv (NIH AIDS Research and Reference Reagent Program) in HEK293T cells as previously described \[[@ppat.1005742.ref027],[@ppat.1005742.ref028]\]. Pseudovirus functionality was determined by measuring luciferase expression after infecting TZM-bl cells (NIH AIDS Research and Reference Reagent Program, ARRRP). Relative luminescence units (RLUs) of ≥100,000 were considered ideal and 30,000 RLUs were accepted in cases where readings were 2.5 x the background; \<2.5 times the background were considered negative.
Co-receptor usage {#sec007}
-----------------
Co-receptor usage was inferred by Geno2Pheno \[[@ppat.1005742.ref029],[@ppat.1005742.ref030]\] and webPSSM \[[@ppat.1005742.ref031],[@ppat.1005742.ref032]\] using a false positive rate (FPR) cut-off of 5 for Geno2pheno \[[@ppat.1005742.ref033]\]. Viruses predicted to use CXCR4 were tested for co-receptor usage in the Trofile assay \[[@ppat.1005742.ref034]\]. Viruses used in this entry assay were produced in an identical manner as described above except the HEK293T cells were co-transfected with *env* clones and the backbone, pNL4-3.lucRΔenv (NIH AIDS Research and Reference Reagent Program). Positive tropism controls RP1.12, an X4 virus \[[@ppat.1005742.ref035]\], and QH0515.1, an R5 virus \[[@ppat.1005742.ref036]\] were used. Viruses were normalized using the in-house p24 assay and standardized quantities of Env-pseudotyped viruses were tested for their ability to infect U87 cell lines expressing CD4 and either CCR5 (U87_R5) or CXCR4 (U87_X4) \[[@ppat.1005742.ref037]\].
Sequence analysis and phylogenetic inference {#sec008}
--------------------------------------------
Sequences were aligned using GeneCutter (<http://hiv.lanl.gov/content/sequence/GENE_CUTTER/cutter.html>) \[[@ppat.1005742.ref038]\] which enables the maintenance of alignments in codon space by employing HMMER \[[@ppat.1005742.ref039]\] and is trained specifically for HIV-1 on a full-length genome alignment. Sequences were examined for evidence of inter-clade recombination using RIP \[[@ppat.1005742.ref040]\], Rega HIV subtyping tool (400bp sliding window with 20bp steps size) \[[@ppat.1005742.ref041]\] and with jpHMM \[[@ppat.1005742.ref042]\]. Only sequences that were clade C throughout the entire *env* reading frame were included in the panel. For phylogenetic analysis and similarity comparisons, regions with more than 5% gaps were excluded. Candidate vaccine strains, Ce1086 (FJ444395 from Malawi in 2004); TV1 (AF391230 from South Africa in 1998); 96ZM651 (AF286224 from Zambia in 1996); and other clade C and clade C-related sequences from India and China (CRF07_BC, CFR08_BC and BC unique recombinants which were predominantly clade C across *env*, n = 59) were included as references. Phylogenies were computed using FastTree \[[@ppat.1005742.ref043]\] on nucleotide sequences using the GTR substitution model. Gp120 amino acid distance calculations for comparing vaccine coverage and relatedness were determined using the HIVb substitution model in DIVEin \[[@ppat.1005742.ref044]\]. The sequences from the placebo arm of the RV144 study (n = 66) were the same as those analyzed previously \[[@ppat.1005742.ref013]\]. For investigation of the outward evolution of protein sequences, additional un-rooted phylogenies were computed in PhyML, version 3.0 \[[@ppat.1005742.ref045]\] with the HIVb substitution model \[[@ppat.1005742.ref046]\]. Outward evolution or divergence measured in branch length was calculated from protein phylogenies inferred in PhyML with trees rooted using the minimum sum of variance method \[[@ppat.1005742.ref047]\]. Trees were visualized in FigTree v1.4.2 (<http://tree.bio.ed.ac.uk/software/figtree/>) and cluster support stated as \>80% of a 100 resampled replicates. Env characteristics such as loop length, glycan density and net charge were determined for both conventional variable loops as well as for hypervariable regions within variable loops using a Los Alamos HIV-1 sequence database web tool which excises and characterizes either complete variable loops or hypervariable regions (<http://www.hiv.lanl.gov/content/sequence/VAR_REG_CHAR>). Weblogos of signature sites and linear peptide portions V2 (HXB2 161--179) and V3 (HXB2 300--322), were used to show amino acid frequency by position and were performed using a web tool at the Los Alamos HIV-1 database AnalyzeAlign (<http://www.hiv.lanl.gov/content/sequence/ANALYZEALIGN/analyze_align.html>). Signature analysis to distinguish between viruses from pre-seroconversion, indeterminate and post-seroconversion was performed using a phylogenetically corrected signature analysis strategy \[[@ppat.1005742.ref048]\].
Serum panel {#sec009}
-----------
Fifty-four serum samples were collected from antiretroviral (ARV) drug-naïve, chronically HIV-1-infected individuals originating from 3 SAAVI (South African AIDS Vaccine Initiative) clinical trial sites in South Africa: Durban, Kwazulu-Natal (n = 16, CAPRISA); Cape Town, Western Cape (n = 20; Desmond Tutu HIV Foundation, DTHF); and Soweto, Gauteng (n = 18, Perinatal HIV Research Unit, PHRU) ([S2 Table](#ppat.1005742.s010){ref-type="supplementary-material"}). Serum samples were pre-screened for potency and breadth against three clade C pseudoviruses (CAP8.6F, CAP255.16 and Du156.12), a clade C consensus (ConC), a clade B pseudovirus (6535), a clade B consensus (ConB) and a single clade A pseudovirus (Q23.17). From this, a panel of 30 sera was selected, 10 per site, representing differing breadth (selecting some low, medium and high) for assaying against the 200 Acute/Early clade C panel ([S3 Table](#ppat.1005742.s011){ref-type="supplementary-material"}). Serum samples were collected between March 2011 and October 2013 and all were heterologous, except for two samples from Durban. All analyses were adjusted to correct for the inclusion of these autologous measurements.
Neutralization assays {#sec010}
---------------------
Neutralization of Env-pseudotyped viruses was measured using the validated TZM-bl assay. Briefly, reductions in Tat-regulated firefly luciferase (Luc) reporter gene expression were measured in 96-well culture plates after a single round of virus infection in TZM-bl cells as described previously \[[@ppat.1005742.ref027]\]. Assay stocks of Env-pseudotyped viruses were produced by transfection in HEK293T cells, as described above, and titrated in TZM-bl cells. Serum samples (inactivated at 56°C, for 1 h) and bnAbs were tested at 1:20 dilution and then at 3-fold dilutions up to seven times in duplicate wells for each dilution starting at either 10 μg/ml (PG9, PGT128) or 50 μg/ml (4E10) or 25 μg/ml (CAP256-VRC26.25). VRC01 was initially tested at10 ug/ml, and was later evaluated at 50 ug/ml. Neutralization was determined as the serum dilution or bnAb concentrations at which a 50% reduction in RLU was detected compared to virus control well RLUs, reported as the 50% inhibitory dilution (ID~50~) for serum samples and the 50% inhibitory concentrations (IC~50~) for bnAbs. In some cases we also report IC~80~ values for bnAbs (80% reduction in RLU). Background luminescence read from cells-only control wells was subtracted. Positive controls consisted of a HIV-1 IgG pool purified from five HIV-1 clade C-positive plasma samples (HIVIG-C) and a pool of purified IgGs from HIV-1 clade B positive plasma samples (HIVIG-B, NIH AIDS Research and Reference Reagent Program). bnAbs PG9 and PGT128 were provided by D. Burton (Scripps Research Institute). bnAbs VRC01 and CAP256-VRC26.25 were provided by J. Mascola (NIH Vaccine Research Center). bnAb 4E10 was commercially obtained from PolymunScientific (Klosterneuburg, Austria). All assays were conducted in laboratories adhering to Good Clinical Laboratory Practice (GCLP).
Statistical analysis {#sec011}
--------------------
Statistical comparisons were performed in GraphPad Prism 5.0 (GraphPad Prism version 5.00 for Windows, GraphPad Software, San Diego California USA, [www.graphpad.com](http://www.graphpad.com)) or in RStudio, version 0.98.501 (R Core Team. 2013. R: a language and environment for statistical computing, <http://www.r-project.org>). Statistical significance was considered where p-values were ≤0.05. Differences in viral characteristics (neutralization susceptibility measured as ID~50~) and divergence of vaccine strains from the acute/early clade C and RV144 breakthrough viruses) were assessed using nonparametric Mann-Whitney tests for distributions between two un-paired groups. One-sided tests were performed where existing associations exist. Correlations of variable loop properties with IC~50~ values, phylogenetic branch length over time and ID~50~ values with branch length, were tested using Kendall's rank correlation test as implemented in the R package Kendall v2.2, statistics provided for one-sided tests unless stated otherwise.
Differences in neutralization sensitivity between viruses over calendar time from three different periods (1998--2005, 2006--2007 and 2008--2010) were evaluated using a non-parametric Jonckheere-Terpstra test for trend among ordered groups (time interval groups in this case, with time periods pre-selected based on numbers). We tested an increase in neutralization resistance hypothesis.
Following the strategy used in Seaman et al., \[[@ppat.1005742.ref016]\] tier categorization was performed by grouping k-means clustering whereby the 200 C-clade viruses were assigned to one of four subgroups (tier 1A, & B, tier 2 and tier 3) ranging from highly sensitive to resistant phenotypes. Subsequently, rank ordering of viruses according to their average log~10~ ID~50~ titers to further resolve tier 1B and tier 2 was performed \[[@ppat.1005742.ref016],[@ppat.1005742.ref049]\]. We also tested the association between the ID~50~ titers and the early stage of infection, as well as the presence or absence of the glycan at position 332 using a Gaussian fixed-effect generalized model (GLM) with the titers as dependent variable and stage, the presence of the 332 glycan and year as independent variables.
Nucleotide sequence accession numbers {#sec012}
-------------------------------------
Listed in [S1 Table](#ppat.1005742.s009){ref-type="supplementary-material"}.
Results {#sec013}
=======
Southern African HIV-1 acute/early clade C envelope panel {#sec014}
---------------------------------------------------------
To generate a panel of 200 functional clade C envelope (*env* gene) clones from viruses collected soon after transmission, plasma samples were obtained from individuals estimated to be infected for less than a 100 days: 67 individuals were HIV-1 PCR positive but HIV-1 seronegative (pre-seroconversion, or Ab-) \[[@ppat.1005742.ref050]\]; 29 were in early seroconversion with an indeterminate western blot (indeterminate, or Ab+/-); and the remaining 104 were HIV-1 seropositive with a negative diagnosis within the previous 100 days (post-seroconversion, or Ab+) ([S1 Table](#ppat.1005742.s009){ref-type="supplementary-material"}). Samples originated from five southern African countries (South Africa, Botswana, Zambia, Tanzania and Malawi) ([Table 1](#ppat.1005742.t001){ref-type="table"}).
Phylogenetic analysis showed that the viruses generally did not cluster according to country of origin, suggesting a generalized regional epidemic with intermixing of strains ([Fig 1](#ppat.1005742.g001){ref-type="fig"}), although African and Asian C clade sequences form separate clades ([S1 Fig](#ppat.1005742.s001){ref-type="supplementary-material"}). One bootstrap-supported cluster of four sequences, and two clusters of two sequences each, were identified from Tanzania; and seven clusters of two sequences each were identified from South Africa. The majority of South African sequences (68/130) originated from one particular region (KwaZulu Natal). We found no strong bootstrap-supported evidence of local founder effects within South Africa, although some clades consisting of sequences only of South African origin were evident, with one smaller clade of 10/68 sequences originated in Kwazulu-Natal observed (highlighted in [Fig 1](#ppat.1005742.g001){ref-type="fig"}). [S1 Fig](#ppat.1005742.s001){ref-type="supplementary-material"} provides a more detailed breakdown of all geographic origins within South Africa, and includes the vaccine strains that will be used in the upcoming trial, which are located in three phylogenetically distinct parts of the tree ([S1 Fig](#ppat.1005742.s001){ref-type="supplementary-material"}).
{ref-type="table"}).\
Branches are colored according to country/region. Bootstrap values \> 80% of 100 resampled replicates are illustrated as filled circles on nodes. South African samples from Soweto/Johannesburg (Gauteng province), Cape Town (Western Cape Province), and KwaZulu-Natal are highlighted in light green, red, and blue respectively, sequences from other locations in South Africa are shown in grey. Clades that were from the same geographic region are highlighted. Only one of these regional grouping (\>2 sequence clusters) had strong bootstrap support (4 sequences from Tanzania, highlighted in orange).](ppat.1005742.g001){#ppat.1005742.g001}
Transmitted viruses almost exclusively utilize the co-receptor CCR5 \[[@ppat.1005742.ref003],[@ppat.1005742.ref005],[@ppat.1005742.ref006]\]. In order to determine co-receptor phenotype in the clade C panel we inferred viral co-receptor usage based on the V3 loop amino acid sequence using two genotyping methods \[[@ppat.1005742.ref029]--[@ppat.1005742.ref033]\]. Of the 200 viruses, 192 were predicted to use CCR5 while eight were predicted as CXCR4-using by both methods ([S4 Table](#ppat.1005742.s012){ref-type="supplementary-material"}). Experimental analysis using U87 cells expressing either the CXCR4 or CCR5 co-receptor \[[@ppat.1005742.ref034]\], however found all eight to be exclusively CCR5-using with no evidence for dual tropism. Thus, all 200 clade C viruses were considered to be CCR5-tropic.
Majority of clade C panel viruses have a moderately resistant tier 2 phenotype {#sec015}
------------------------------------------------------------------------------
HIV-1 tier classification is a useful way to define a virus' neutralization susceptibility profile, with tier 1 viruses being highly sensitive and tier 3 highly resistant to neutralization \[[@ppat.1005742.ref016],[@ppat.1005742.ref049]\]. Neutralization assays were performed using 30 clade C HIV-1+ serum samples from chronic infection, which were prescreened to reflect sera with a range of neutralization potency and breadth ([S2](#ppat.1005742.s010){ref-type="supplementary-material"} and [S3](#ppat.1005742.s011){ref-type="supplementary-material"} Tables). Following a procedure published by Seaman et al (2010) \[[@ppat.1005742.ref016]\], Envs were grouped using k-means clustering, and 4 clusters each representing a group of viruses with similar patterns of sensitivity were identified ([Fig 2](#ppat.1005742.g002){ref-type="fig"}). This was used for an initial tier classification, with tier 1A, 1B and tier 3 forming highly significant and robust clusters, and tier 2 capturing everything between. This classification was recapitulated when neutralization sensitivity of each of the 200 viruses was measured as the geometric mean ID~50~ titer (GMT) determined for all sera against which each virus was tested ([Fig 3A](#ppat.1005742.g003){ref-type="fig"}). The higher the GMT relative to other viruses in the panel, the more neutralization sensitive the virus. The rank of the titers closely reflected the tiers assigned by k-means clustering, although there was a small amount of intermixing between tier 2 and 3. The way Envs are grouped in a k means is generally highly dependent on the sera used for the evaluation, as is evident by the large number of Envs that sometimes are grouped with a more sensitive, sometimes with the more resistant cluster, depending on resampling. Consistent with previous approaches \[[@ppat.1005742.ref016]\], the ambiguous calls that were not consistently highly sensitive in the bootstrap (tier 1), or consistently highly resistant (tier 3), were classified as tier 2. [Fig 3A](#ppat.1005742.g003){ref-type="fig"} illustrates the continuum of geometric means.
![K-means (n = 4) neutralization clustering of clade C Env pseudoviruses for tier classification (n = 200).\
Neutralization sensitivity was assessed by assaying pseudoviruses against 30 sera from chronically infected individuals. Individual viruses are indicated to the right of the heatmap with individual serum listed at the bottom. Viruses are ranked according to respective sensitivities and categorized into tiers as determined by k-means clustering. Neutralization sensitivity (log~10~ ID~50~ titer) is denoted by color key provided in the inset. Viral isolates clustering, together with \>95% probability, was performed to assess the stability of the Env pseudovirus clusters based on a random-with replacement resampling of the sera (rows), or to assess the stability of the serological clusters based on resampling of the viruses (columns), and were boxed on the heatmap indicating categorization into their respective tiers. Bootstrap support for clusters was determined by resampling the data sets from the individual serum and re-evaluating the k-means clusters 10,000 times giving an indication of how many times each virus was a member of the originally assigned tier classification in the resampled datasets. Env pseudoviruses that clustered with the assigned tier are shown in the column on the left as blue for tier 1A (very sensitive), yellow for tier 1B, magenta for tier 2 (intermediate), and black for tier 3 (very resistant). This follows the tier classification of Seaman et al (2010) \[[@ppat.1005742.ref016]\]. Similarly sera potential is illustrated at the top with the most uniformly potent sera indicated as blue, potent sera as yellow, modestly potent as magenta, and least potent as black. Color intensity indicates the probability of falling within a cluster and the degree of blending the frequency of falling into each of the respective tiers, with strong unblended colors, associated with 95% bootstrap support, boxed. The poor resolution of the k- means bootstrapping emphasizes that the sensitivity of Envs is a continuum, and will be dependent on the specific sera or antibodies used for the evaluation.](ppat.1005742.g002){#ppat.1005742.g002}
![Geometric mean titer and tier 1A and 1B classification (n = 200).\
(A) Viruses are rank ordered according to neutralization sensitivity to 30 clade C chronic infection serum samples, from the least sensitive to the most sensitive along the x-axis by average log~10~ GMTs. Two viruses were classified as highly sensitive tier 1A and an additional 17 as above-average sensitive tier 1B. A previously determined cut off ID~50~ = 200, was used to distinguish between tier 1A and tier 1B is indicated on the graph, with tier 1B classified viruses above this cut off colored in red and those below in orange \[[@ppat.1005742.ref016]\]. Twelve pseudoviruses classified by both Seaman et al., (2010) and this study were found to be discrepant ([S1 Table](#ppat.1005742.s009){ref-type="supplementary-material"}): Du156.12 consistently falls near the boundary of the tier 2 and tier 1B, was classified as tier 1B here however was previously classified as tier 2; and ZM197M and SM109F were classified as tier 2 here, however were previously classified as 1B. (B) Maximum likelihood phylogenetic analysis of southern African clade C acute/early envelope nucleotide sequences (n = 200) with branches colored according to tier. Bootstrap values \> 80% of 100 resampled replicates are illustrated as filled circles on nodes.](ppat.1005742.g003){#ppat.1005742.g003}
The majority of viruses exhibited a moderately resistant tier 2 phenotype (75%, n = 150), whereas 8.5% (n = 17) were classified as possessing a more sensitive tier 1B phenotype, and 15.5% (n = 31) classified as possessing a more resistant tier 3 phenotype ([Fig 3A](#ppat.1005742.g003){ref-type="fig"}). Phylogenetic analysis showed that the most sensitive or most resistant viruses generally did not cluster together ([Fig 3B](#ppat.1005742.g003){ref-type="fig"}). To further interrogate if there were signatures in the viral sequences associated with tier phenotype, we considered every amino acid in each position using either a simple Fisher's exact test testing for the association between the presence and absence of the amino acid and the phenotype of interest, and a phylogenetically corrected Fisher's test (described fully in Gnanakaran et al, 2011) \[[@ppat.1005742.ref051]\]. We identified no phylogenetic signatures associated with tier phenotype based on a false discovery rate (FDR) threshold of q \< 0.2, however, using a simple Fisher's test, a single association with tier 3 was identified, K683R. Lys (K) is the most common amino acid in this position. Arg (R) was enriched among Tier 3 viruses, found 16/30 times (53%), while among Tier 1 or 2 viruses it was present only 28/170 times (16%), (p-value of 0.00004, with a q = 0.04 for the full set of signature comparisons). This amino acid is located in the membrane-proximal external region (MPER) embedded in the 4E10 bnAb epitope, and the enrichment of R in Tier 3 viruses suggests that this position and region may be implicated in neutralization resistance.
The neutralization distribution of this clade C panel roughly approximated the tier distribution of a multi-clade panel of Env-pseudotyped viruses previously described (19%, 66% and 15% for tiers 1B, 2 and 3 respectively) \[[@ppat.1005742.ref016]\]. Unique to our study was the identification of 1% (n = 2) of viruses with a highly sensitive tier 1A phenotype. This highly neutralization sensitive phenotype is not normally associated with circulating viruses. Both tier 1A envelope sequences matched a single genome derived amplicon, suggesting that these highly neutralization sensitive viruses do occur *in vivo*, although it remains possible that they were the result of an artificially introduced error during amplification, as both had a single mutation away from the derived transmitted founder sequence: the CH0505.w4.3 clone has a W680G mutation in the MPER \[[@ppat.1005742.ref052]\] and the SO032_A2.8--1 clone had an E406G mutation in V4. We therefore do not know if these viruses were transmitted or if they evolved post-infection, or may have been the result of PCR error and we thus excluded them in all analysis pertaining to neutralization sensitivity.
Neutralization sensitivity to polyclonal sera changed over the first 100 days of infection {#sec016}
------------------------------------------------------------------------------------------
There are conflicting data on whether viruses sampled in early infection are more or less sensitive to serum neutralization compared to viruses sampled later in infection. We investigated whether, over the first 100 days of infection, we could detect any difference in neutralization sensitivity. To select viruses that best reflected the virus *in vivo* that had recently traversed the transmission bottleneck, we included only envelopes generated using the single genome amplification (SGA) approach \[[@ppat.1005742.ref023]\]; and excluded viruses that were classified as multi-variant transmission or for which there was insufficient information to classify their multiplicity of infection ([S1 Table](#ppat.1005742.s009){ref-type="supplementary-material"}). This subset of 139 Env-pseudotyped viruses was classified into three groups according to time of infection where sequential gain of HIV-1 antibody responses was used as a marker of time from infection \[[@ppat.1005742.ref050]\]: pre-seroconversion group (n = 58) (no detectable antibodies Ab-; estimated infection \<15 days); indeterminate group (n = 26) (evolving antibody responses Ab+/-, estimated infection \<37 days); and early post-seroconversion group (n = 55) (near/full seroconversion Ab+; estimated duration of infection \< 100 days).
We found that pre-seroconversion viruses were more resistant to neutralization by the 30 clade C sera compared to post-seroconversion viruses, with pre-seroconversion viruses having a significantly lower GMT (p = 0.001, one-sided Wilcoxon rank sum test) ([Fig 4A](#ppat.1005742.g004){ref-type="fig"}). There was no significant difference in GMT between the indeterminate group and the post-seroconversion group. To confirm that the result was real, given the possibility of error in measurements, we built an error model based on the many duplicated neutralization data points available \[[@ppat.1005742.ref053]\]. We randomly resampled from the Gaussian distribution based on the observed error to add noise to our individual points, and generated 10,000 replicate datasets using the uncertainty model. For both comparisons, acute (Ab-) versus early (Ab+) or indeterminate (Ab +/-), more than 97% of replicates show similar trends as [Fig 4A](#ppat.1005742.g004){ref-type="fig"}, i.e. the acute group is less sensitive than both indeterminate and early groups. Furthermore, using 10,000 replicates of artificial ID~50~ data generated as above, we also found that the acute group (Ab-) had higher normalized sum of ranks from the Wilcoxon rank sum test statistic than the indeterminate group (Ab+/-) in 99.8% replicates, and than the early group (Ab+) in 100% replicates (p \< 10^−4^). This suggests that the results shown in [Fig 4A](#ppat.1005742.g004){ref-type="fig"} are robust under realistic models of experimental uncertainty.
{#ppat.1005742.g004}
We also found that the pre-seroconversion viruses were neutralized less frequently by 30 clade C sera compared to the post seroconversion viruses (median breadth of 36% versus 50%; p = 0.0474; two sided Mann-Whitney test) ([Fig 4B](#ppat.1005742.g004){ref-type="fig"}). This was largely driven by the resistant viruses as when only neutralization sensitive viruses were analyzed, there was no significant difference in sensitivity between pre-seroconversion, indeterminate and post-seroconversion viruses (median IC~50~ 85.90 μg/ml, 89.74 μg/ml and 95.06 μg/ml) ([S8 Fig](#ppat.1005742.s008){ref-type="supplementary-material"}). This behavior was confirmed with purified IgGs from clade C infections (HIVIG-C), however no difference was observed when assayed against pooled purified IgGs collected from clade B infections (HIVIG-B) ([S2 Fig](#ppat.1005742.s002){ref-type="supplementary-material"}), suggesting a possible clade-associated specificity may be driving this differential resistance pattern. Taken together, these data suggest that viruses present prior to seroconversion are less likely to be neutralized by clade-matched polyclonal HIV-1 sera than those viruses present post-seroconversion.
Several studies have shown that increases in length and glycosylation density across the V1, V2 and V4 regions of gp120 are associated with neutralization resistance \[[@ppat.1005742.ref015],[@ppat.1005742.ref020],[@ppat.1005742.ref054]--[@ppat.1005742.ref056]\]. We investigated whether these features could account for observed differences in neutralization sensitivity seen in the 200 clade C viruses. Similar to previous studies, when we analyzed all viruses together, we found that longer V1, V2 and V4 length, and higher glycan density were strongly associated with neutralization resistance ([S3 Fig](#ppat.1005742.s003){ref-type="supplementary-material"}). However, we found no significant difference in V1, V2 and V4 loop length or glycan density between the pre-seroconversion, indeterminate and post-seroconversion groups of viruses ([Fig 4C and 4D](#ppat.1005742.g004){ref-type="fig"}), indicating that observed differences in neutralization phenotype between these groups were not attributable to differences in variable loop length or glycan density.
The glycan at position 332 and time post infection are independent predictors of neutralization resistance {#sec017}
----------------------------------------------------------------------------------------------------------
We then investigated whether differences in neutralization phenotype could be attributable to differences in particular antibody specificities. Viruses were assayed with bnAbs targeting: the V2-glycan site (CAP256-VRC26.25 and PG9), the gp41 MPER epitope (4E10), the CD4 binding site (VRC01), and the V3/C3 glycan supersite (PGT128). Here IC~50~ was used, where low concentrations indicate increased viral sensitivity. For antibodies targeting the CD4bs, V2-glycan and MPER specificities we found no difference in neutralization susceptibility between pre-seroconversion and post-seroconversion viruses, suggesting that these epitope targets did not differ between the groups. However, pre-seroconversion viruses were significantly more resistant to neutralization by PGT128 compared to the post-seroconversion viruses (median IC~50~ of 4.42 μg/ml compared to 0.06 μg/ml respectively, p = 0.0174, Mann-Whitney two sided test) ([Fig 5A](#ppat.1005742.g005){ref-type="fig"}), with 50% of pre-seroconversion and 27% of post-seroconversion viruses resistant to this bnAb.
![Role of the glycan at position 332 (N332+) in influencing neutralization susceptibility of clade C viruses.\
Each point represents the geometric mean of viruses tested for neutralization sensitivity using either bnAbs (A) or a panel of 30 South African sera (B and C). (A) Comparison of neutralization sensitivity of clade C viruses to bnAbs VRC01, PG9, PGT128, CAP256-VRC26.25 and 4E10. The highest concentration tested for each bnAb and the percentage of viruses neutralized (breadth) are indicated. Env pseudotyped clade C viruses (n = 139), from single infections for which SGA-derived sequences were available, were partitioned into three infection stage groups per sequential gain of HIV-1 specific antibody responses as a marker of time from infection; pre-seroconversion (Ab-), indeterminate (Ab-/+) and post-seroconversion (Ab+). Throughout, pre-seroconversion viruses are indicated in blue, with indeterminate in grey and post-seroconversion in red. Significant differences between medians indicated with the relevant p-value for a two-sided Mann-Whitney test. (B) Comparison of glycan frequency changes between pre-seroconversion (Ab-) and post-seroconversion (Ab+) infection stages for six sites comprising the glycan patch \[[@ppat.1005742.ref057]\]. N-linked glycans were predicted according to the Nx(T/S) sequon, where x is not a proline. (C) Comparison of neutralization sensitivity between viruses categorized into those containing N332+ (n = 94), those with N334+ (n = 22) and those without a glycan at either position 332 or 334 (n = 21). Significance tested and shown with p-values provided for a Mann-Whitney, two-sided test. (D) A comparison of susceptibility using only viruses containing the N332 glycan, per infection stage. Dot plots show medians in red horizontal lines and interquartile ranges in black lines. Significance tested by non-parametric two-sided Mann-Whitney tests. Uncorrected p-values are provided.](ppat.1005742.g005){#ppat.1005742.g005}
PGT128 specificity is largely dependent on the presence of a glycan at position 332 \[[@ppat.1005742.ref058],[@ppat.1005742.ref059]\], and similar to what we previously reported \[[@ppat.1005742.ref059]\], there was an underrepresentation of this glycan in pre-seroconversion clade C viruses ([Fig 5B](#ppat.1005742.g005){ref-type="fig"}). To determine the contribution of the 332 glycan in influencing neutralization resistance to polyclonal sera, we divided the 139 viruses into three groups: those with the glycan at position 332 (N332+), those with the glycan at position 334 (N334+) (the glycan at position 332 shifts to position 334 in some cases) and those with no glycan at either position 332 or 334 (N332-/N334-). Similar to previous studies, we found that viruses with the 332 glycan were indeed significantly more sensitive to serum neutralization compared to viruses lacking this glycan (median ID~50~ of 48.98 μg/ml, 35.73 μg/ml and 41.30 μg/ml for N332+, N334+ and N332-/N334- viruses respectively, p = 0.0085; Mann-Whitney two sided test) ([Fig 5C](#ppat.1005742.g005){ref-type="fig"}). However, when we considered only viruses containing the 332 glycan, the pre-seroconversion viruses still had significantly lower sensitivity to serum neutralization (median ID~50~ of 34.36 μg/ml, 62.52 μg/ml, and 56.36 μg/ml, for Ab-, for Ab+/-, Ab+ respectively, p = 0.0219), indicating that the lack of the glycan at position 332 is not the sole determinant of increased resistance of pre-seroconversion viruses ([Fig 5D](#ppat.1005742.g005){ref-type="fig"}). We then fitted a Gaussian fixed-effect generalized model (GLM) to test for an association between GMT with infection stage and the 332 glycan. Partitioning viruses into pre-seroconversion (Ab-) and indeterminate with post-seroconversion (Ab+/- and Ab+), the model showed that both the presence/absence of the 332 glycan and the stage of infection reliably predicted GMT, without any interaction between infection stage and glycosylation state ([S4 Fig](#ppat.1005742.s004){ref-type="supplementary-material"}). The log~10~ GMT was on average 1.36 fold higher when the 332 glycan was present (p = 0.0094), and 0.72 lower when the sample was pre-seroconversion (p = 0.0034).
PGT128-like antibodies have been shown to recognize other glycans centered around the 332 glycan, the so-called high-mannose patch, which include glycans, N136/7, N156, N295, N301 and N334 \[[@ppat.1005742.ref057]\]. N156 and N301 are conserved in clade C viruses and are therefore unlikely to contribute to differences in neutralization phenotype. N295 has been reported to substitute for N332 \[[@ppat.1005742.ref060]\], but was found in higher frequencies in pre-seroconversion viruses and thus unlikely to be playing a role here. This together with N136/7, which occurs at similar frequencies in both pre-seroconversion and post-seroconversion groups, this could not account for increased neutralization resistance in the pre-seroconversion viruses ([Fig 5B](#ppat.1005742.g005){ref-type="fig"}).
We sought to determine whether other genotypes were statistically associated with pre-seroconversion viruses. We identified a site in V5, G464, which was present in all post-seroconversion viruses in our panel, but in only 80% of pre-seroconversion/indeterminate viruses (p = 0.00024, q = 0.096; Fisher's exact test). This site is located in the CD4 binding domain, however the relevance of this finding is not yet clear.
Increased neutralization resistance as the clade C epidemic matures {#sec018}
-------------------------------------------------------------------
As this large collection of clade C viruses from soon after transmission was collected over 13 years, we sought to determine whether the southern African epidemic has increased in resistance to neutralization over time. For this we tested serum collected from clade C infected individuals between 2011 and 2013 against the clade C panel. We first investigated whether viruses diverged over this period, by determining Gp160 phylogenetic tree branch length (measured as distance from root) from 1998 to 2010. We found that early viruses (warmer colors) displayed shorter branches (closer to the root) compared to later viruses (cooler colors) ([Fig 6A](#ppat.1005742.g006){ref-type="fig"}). Even when measuring divergence conservatively by determining protein distances on alignments excluding all hypervariable regions, we found a significant positive correlation of branch length over time (one sided Kendall's τ = 0.157, p = 0.0009) ([Fig 6B](#ppat.1005742.g006){ref-type="fig"}) indicating that the clade C epidemic diversified appreciably over the 13 year time period. We also observed a significant negative correlation of branch length with serum neutralization sensitivity, as measured by geometric mean ID~50~ titer (two sided Kendall's τ = -0.141, p = 0.0015) ([Fig 6C](#ppat.1005742.g006){ref-type="fig"}), but could find no direct correlation of polyclonal serum neutralization sensitivity over time (two sided Kendall's τ = -0.017, p = 0.7338) ([S5A Fig](#ppat.1005742.s005){ref-type="supplementary-material"}). We also found no significant association between variable loop length and glycan density over time ([S5B and S5C Fig](#ppat.1005742.s005){ref-type="supplementary-material"}), suggesting that this trait has remained relatively constant. Taken together these data provides indirect evidence of diminishing sensitivity to within-clade sera over time.
{#ppat.1005742.g006}
We then evaluated changes in virus susceptibility to known bnAbs. Similar to polyclonal serum, we identified a significant negative association of branch length with bnAb neutralization sensitivity (measured in geometric mean IC~50~ titer for all bnAbs in aggregate) (Kendall's τ = 0.1157, p = 0.0075) ([S6A Fig](#ppat.1005742.s006){ref-type="supplementary-material"}). Viruses were grouped into three time periods according to date of collection to generate three groups of relevant sample size (1998--2005, 2006--2007 and 2008--2010). Although the first time period spanned 8 years, the majority of samples (\>73%) were collected over a three-year period from 2003--2005. When we evaluated each of the five bnAbs individually, we found increased resistance as measured over three time periods to PG9, VRC01 and 4E10 (p = 0.013, p = 0.030 and p = 0.004 respectively, Jonckheere-Terpstra test), but not to PGT128 or CAP256-VRC26.25 ([Fig 6D](#ppat.1005742.g006){ref-type="fig"}). Thus, we demonstrated that viruses are becoming more neutralization resistant to certain bnAbs as the epidemic progresses.
VRC01 monoclonal antibody coverage in a clade C epidemic {#sec019}
--------------------------------------------------------
The role of VRC01 in preventing HIV acquisition is being evaluated in a Phase 2b study in Africa (ClinicalTrials.gov Identifier: NCT02568215), where trial participants will receive an intravenous infusion of VRC01 at a dose of 10 mg/kg or 30 mg/kg every 8 weeks. As VRC01 serum concentrations will decay over time between transfusions, we were interested in determining the effectiveness of this antibody at different concentrations including, 50 ug/ml, 10 ug/ml and 1 ug/ml. At these concentrations, 84%, 80% and 56% of the panel were neutralized respectively (IC~50~) ([Fig 7](#ppat.1005742.g007){ref-type="fig"}). Assayed with the same concentration range, but measuring the concentration that resulted in 80% inhibition (IC~80~), VRC01 neutralized 78%, 68% and 30% of viruses, respectively. The median IC~50~ titer against sensitive viruses at \<10 ug/ml was 0.42 μg/ml.
{#ppat.1005742.g007}
We have shown that viruses are becoming increasingly resistant to VRC01 as the epidemic matures, where overall potency at \<10 ug/ml was shown to decrease from a median IC~50~ titer of 0.48 μg/ml to 0.69 μg/ml to 0.84 μg/ml over three time periods measured (1998--2005, 2006--2007 and 2008--2010) ([Fig 6D](#ppat.1005742.g006){ref-type="fig"}). This was observed despite VRC01 breadth (percentage viruses neutralized at IC~50~ \< 10 μg/ml) staying relatively constant over this time period (77%, 78% and 79% respectively).
Protein divergence of clade C acute/early panel viruses from candidate vaccine strains {#sec020}
--------------------------------------------------------------------------------------
The RV144 trial in Thailand showed 31.2% vaccine efficacy with immune correlates of risk identified as antibodies predominantly against V1V2 region of gp120 \[[@ppat.1005742.ref010]\]. Building on the success of RV144, a Phase 3 trial is planned for South Africa using vaccines similar to those used in RV144, however the canarypox vector has been modified to express the clade C (96ZM651) gp120 and the protein boost now comprises two clade C gp120 Env proteins (TV1 and Ce1086). To get insight into how such a vaccine will perform in the South African setting, we genotypically compared the clade C viruses to candidate C vaccines, and Thailand CRF01_AE breakthrough viruses to the AE gp120 immunogen used in the RV144 trial.
Phylogenetic comparison of the three clade C candidate vaccine strains showed that they fall within the 200 clade C panel sequences from southern Africa and cluster separately from clade C viruses sampled in India and China and recombinant CRF07_BC and CRF08_BC viruses from China ([S1 Fig](#ppat.1005742.s001){ref-type="supplementary-material"}), suggesting they are broadly representative of the southern African epidemic. The mean gp120 amino acid sequence distance between the panel viruses and clade C candidate vaccine protein boost sequences ranged from 22.19% (95% CI of mean, 21.90%--22.48%) for Ce1086 to 24.15% (95% CI of mean, 23.85%--24.44%) for TV1. This was significantly greater distances than that observed in RV144, where the mean gp120 amino acid sequence distance between the CRF01_AE viruses from breakthrough infections in the placebo arm to the clade-matched vaccine immunogen, CM244 was 15.24% (95% CI 14.69%--15.79%) (Ce1086 and TV1 individually both to CM244 p\<0.0001) ([Fig 8A](#ppat.1005742.g008){ref-type="fig"}). Taken together the mean gp120 amino acid sequence distance of Ce1086 and TV1 to panel viruses was 23.14% (95% CI of mean 22.91%--23.37%), which was 8% more distant when compared to distances between CRF01_AE viruses and the RV144 vaccine immunogens.
![Relationship of clade C acute/early panel *env* genes/envelope proteins (n = 200) to clade C candidate vaccine strains compared to the relationship of CRF01_AE breakthrough viruses from RV144 (n = 66 placebo arm) to the RV144 vaccine.\
**(A)** Gp120 amino acid distances (excluding signal peptide) for clade C viruses to clade-matched vaccine prime- and boost-strains (96ZM651, TV1 and Ce1086); and CRF01_AE viruses from breakthrough infections in the placebo arm of RV144 to clade-matched RV144 vaccine strains (92TH023 and CM244). Box plots for RV144 distances in light grey and clade C distances in dark grey, with means in red. Significance shown in p-values provided for a two-sided Mann-Whitney test. **(B)** Amino acid distances across V2 (HXB2 161--179) and V3 (HXB2 300--322) linear B cell epitopes for clade C viruses as well as for RV144 placebo CRF01_AE viruses to respective clade-matched vaccine protein boost strains, Ce1086 and TV1 and CM244. Significance shown in p-values provided for a two-sided Mann-Whitney test. **(C)** Logo conservation plots across V2 and V3 linear B cell epitope peptides which includes vaccine signature sites as identified in RV144 \[[@ppat.1005742.ref011],[@ppat.1005742.ref013]\]. These illustrate clade C acute/early panel sequence amino acid frequency and corresponding vaccine strain residue conservation. Residues shared between clade C and either vaccine strain Ce1086 or TV1 are shown in black with the remaining residues in grey. Positions of RV144 identified signatures associated with reduced infection risk K169, I181X, I307 and F317X are shaded in blue. **(D)** Comparison of RV144 signature site frequencies in acute/early clade C panel (southern Africa) compared to Thai CRF01_AE sequences from the placebo arm (Thai). In RV144, a match to the vaccine at position 169 (with a K) and at 307 (with I) was associated with protection; while a mismatch to the vaccine at positions 181 (I181X) and 317 (F317X) was associated with protection \[[@ppat.1005742.ref011],[@ppat.1005742.ref013]\].](ppat.1005742.g008){#ppat.1005742.g008}
Divergence from the candidate vaccine V2 and V3 linear peptides {#sec021}
---------------------------------------------------------------
Binding antibodies to 19-mer V2 (HXB2 161--179) and 21-mer V3 (HXB2 300--322) linear peptides (so-called "hotspot" regions) were correlated with reduced risk of infection in RV144 \[[@ppat.1005742.ref012]\]. We were interested in estimating the potential coverage of vaccines by determining the protein distance across V2 and V3 linear peptides between the C-clade vaccine boost proteins (TV1 and Ce1086) and the C-clade panel viruses.
We found high mean V2 amino acid distances of 40.51% (95% CI 39.13%--41.89%), with lower mean V3 amino acid distance of 17.10% (95% CI 16.20%--18.00%) ([Fig 8B](#ppat.1005742.g008){ref-type="fig"}). Considering amino acids that were conserved in at least 90% of viruses across these hotspot regions, only 53% (10/19) of the V2 residues were conserved compared to 62% (13/21) in the V3 region ([Fig 8C](#ppat.1005742.g008){ref-type="fig"}). We found that mean distances between the V2 hotspot region and viruses from the RV144 placebo group was significantly lower than in the clade C epidemic: 19.63% (95% CI 17.38%--21.88%) compared to 40.51% (95% CI 39.13%-- 41.89%) respectively (p \< 0.0001), but similar distance were observed in V3, 19.22% (95% CI 16.78%-- 21.66%) compared to 17.10% (95% CI 16.20%--18.00%), Thailand and South Africa respectively) ([Fig 8B](#ppat.1005742.g008){ref-type="fig"}). We cannot be sure however what the significance of this is, whether the absence of divergence in hotspots important in the RV144 trial will be critical for clade C. We have shown that the clade C epidemic is diverging over time and we were interested to see what effect this had on C-clade vaccines distances to clade C viruses across V2 and V3 hotspot regions. We found no significant difference between distances to 70 viruses collected after 2008 and the 130 remaining viruses collected during the preceding decade ([S7 Fig](#ppat.1005742.s007){ref-type="supplementary-material"}), indicating that V2 and V3 hotspot regions are not diverging significantly over the period measured.
Frequency of sites correlated with reduced infection risk in RV144 as found in clade C panel viruses {#sec022}
----------------------------------------------------------------------------------------------------
In the RV144 genetic sieve analysis, which compared sequences in the vaccine and placebo arm of breakthrough infections to the vaccine strains, a number of signature sites associated with reduced infection risk were identified \[[@ppat.1005742.ref011],[@ppat.1005742.ref013],[@ppat.1005742.ref014]\]. Two of these genetic signatures (K169; I307) present in the RV144 vaccine strain were associated with increased vaccine efficacy against breakthrough viruses with matching amino acid residues at these positions, and have supportive experimental data to show that mutations found in vaccine breakthrough infections were associated with decreased antibody binding \[[@ppat.1005742.ref011],[@ppat.1005742.ref013]\]. The protective signature K169 was present in both TV1 and Ce1086, whereas I307 was entirely absent. We determined the frequency of these protective signatures in our panel of clade C viruses, and compared them to the frequency in 66 viruses from the placebo arm of RV144 ([Fig 8D](#ppat.1005742.g008){ref-type="fig"}). Both genetic signatures were found at a lower frequency in clade C viruses compared to RV144: the K169 site was found in 69% (95% CI 62%-75%) of clade C viruses compared to 86% (95% CI 78%--95%) of CRF01_AE viruses, and the I307 was found in 60% (95% CI 53%-67%) of clade C and 67% (95% CI 55%-78%) of CRF01_AE viruses respectively ([Fig 8D](#ppat.1005742.g008){ref-type="fig"}). The variation in these sites in clade C is illustrated by a sequence logo ([Fig 8C](#ppat.1005742.g008){ref-type="fig"}).
Two sites have been identified where a mismatch to the vaccine was associated with reduced risk of infection (I181X; F317X). The protective signature I181X was present in Ce1086 but absent in TV1, whereas neither Ce1086 nor TV1 contained the protective signature F317X. I181X was found at a similar frequency in clade C viruses compared to RV144 CRF01_AE viruses (33%, 95% CI 26%--40% compared to 27%, 95% CI 16%--38% respectively) whereas F317X was found at a lower frequency (5%, 95% CI 2%--8% compared to 20%, 95% CI 10%--30% respectively).
Discussion {#sec023}
==========
The extraordinary diversity of HIV-1 is a barrier to achieving protection in active and passive immunization studies. Several large prophylactic trials in clade C epidemic regions of Africa that rely on antibody-mediated protection are imminent, including passive immunization with the broadly neutralizing antibody (bnAb) VRC01, and a vaccine efficacy study of the RV144 regimen tailored for clade C. Using a large panel of 200 clade C pseudotyped viruses from acute/early infection, we observed that pre-seroconversion viruses were more resistant to antibody neutralization compared to post-seroconversion viruses. Additionally, we provide evidence of antigenic drift in certain bnAb targets including VRC01, which we estimated will only block \~80% of clade C viruses at its most efficacious dose. Furthermore, the higher divergence of clade C viruses from candidate clade C vaccines, compared to CRF01_AE viruses to the clade matched vaccine used in the RV144 vaccine, may make protection harder to achieve in clade C epidemic regions. Our study therefore provides a comprehensive analysis of viral traits that affect antibody recognition and interrogates how planned clinical trials may perform in the clade C epidemic of southern Africa.
Studies in non-human primate models have shown that vaccines have a limited window within which they can block infection, making the properties of pre-seroconversion viruses highly relevant \[[@ppat.1005742.ref061]\]. The increased resistance of pre-seroconversion viruses seen in our study was only detected when viruses were assayed with clade C (and not clade B) polyclonal sera, suggesting clade-specific epitopes are either shielded or absent in these viruses. We propose that an under-representation of the N332 glycan (critical for the PGT128-like bnAbs), shown here and previously \[[@ppat.1005742.ref059]\], is partially responsible for the more resistant phenotype. This is supported by the fact that antibody responses dependent on the N332 glycan are common in clade C sera \[[@ppat.1005742.ref062]\]. The lack of the N332 glycan however was not the sole determinant of this resistant phenotype, as we still observed this effect even when only including viruses with the N332 glycan. Interestingly, an analysis using a large multi-clade panel of 219 viruses, tested against 170 polyclonal sera, also found viruses collected early in infection to be less sensitive to neutralization by polyclonal HIV-1 sera \[[@ppat.1005742.ref015]\], suggesting that this phenomenon is more generalizable. However, the signal was only detected for inter-clade comparisons in Hraber et al., \[[@ppat.1005742.ref015]\], which may have been due in part to a smaller sample size for within clade comparisons, and/or differences in the classification of viruses according to infection stage.
The reasons why pre-seroconversion viruses have a more resistant phenotype are unclear. It is possible that these traits provide the virus with a competitive advantage, either at the level of transmission, or in the early establishment of infection in the new host. Another possibility is that following transmission, and up to seroconversion replication in the absence of neutralizing antibodies, similar to cultivation in vitro \[[@ppat.1005742.ref063]\], lead to adaptations that result in reduced shielding and thus the evolution of a more sensitive neutralization phenotype. Other studies examining viral genotypic properties shortly after transmission have also observed rapid adaptation with initial shortening and subsequent elongation of envelope variable loops as HIV-1 antibody responses evolve \[[@ppat.1005742.ref064]\]. If this viral property is advantageous to establishing clinical infection, it may enhance transmission risk in a scenario where transmission predominantly occurs from donors who are in the acute/early stage of infection, which is thought to occur in roughly 50% of cases \[[@ppat.1005742.ref065],[@ppat.1005742.ref066]\]. It will nevertheless be important to further elucidate the factors responsible for the increased resistance of pre-seroconversion viruses, to ensure that vaccines are designed to elicit antibodies that target the most vulnerable bnAb sites.
Persistent cycles of immune pressure have resulted in an accumulation of HIV-1 escape mutations at a population level. Recently, two studies, in predominantly clade B European cohorts, have shown that HIV-1 is also becoming increasingly neutralization resistant over time \[[@ppat.1005742.ref020]--[@ppat.1005742.ref022]\]. In our study we show for the first time that this is also occurring in the clade C heterosexual epidemic in southern Africa, where viruses are becoming increasingly resistant to some bnAbs, including those targeting the CD4bs (VRC01), V2-glycan (PG9-like but not CAP256-VRC26-like epitopes) and MPER (4E10). While Bunnick et al. \[[@ppat.1005742.ref020]\] found that the viruses were increasing in loop length and glycan density as the epidemic progresses, similar to Bouvin-Pley et al, \[[@ppat.1005742.ref021]\], we did not observe this genotypic change in our cohort. However, our study, as well as many others, have found these properties to be associated with neutralization resistance \[[@ppat.1005742.ref002],[@ppat.1005742.ref015],[@ppat.1005742.ref055],[@ppat.1005742.ref067]--[@ppat.1005742.ref069]\]. It is possible we did not detect significant changes in these properties due to our shorter sampling period of 13 years, as compared to a period of 20 years for other studies. These observations nevertheless, suggest that vaccine strains may need to be updated to be more representative of circulating diversity.
It was concerning to note that approximately 20% of clade C viruses in this panel were resistant to VRC01 and that they have become substantially less sensitive to this bnAb as the epidemic has matured, with an observed nearly doubling of concentration needed to reach 50% inhibition over thirteen years ([Fig 6D](#ppat.1005742.g006){ref-type="fig"}). Thus, if antibodies are to be an effective prevention modality, they will need to have increased neutralization breadth, so as to reduce the impact of natural resistance. Furthermore, it may be necessary to use a mixture of antibodies that target multiple sites on the virus to both increase coverage and curb viral escape routes. This strategy is a focus of many research programs and there are several newer generation antibodies, and antibody combinations, currently in clinical development \[[@ppat.1005742.ref009],[@ppat.1005742.ref070],[@ppat.1005742.ref071]\]. Utilizing this same virus panel with 15 bnAbs that target four regions (CD4 binding site, V1V2-glycan region, V3-glycan region and MPER), a comprehensive evaluation was performed to identify the best-in-class single antibodies, together with optimal combination of antibodies for HIV prevention and treatment \[[@ppat.1005742.ref072]\].
The clade C version of the RV144 vaccine regimen is currently being tested for safety and immunogenicity, with Phase 3 trials due to start in South Africa in 2016. We analyzed envelope sequences at various sites or regions relevant to the correlates of risk in the RV144 vaccine trial, in order to consider how this vaccine may perform in the South African clade C epidemic relative to the vaccine used in the partially efficacious RV144 vaccine trial in Thailand. We found that the clade C gp120 protein boosts (Ce1086 and TV1) were generally representative of currently circulating viruses, however protein pairwise distance scores between the viruses and candidate vaccine boosts were approximately 8% greater than between RV144 vaccine immunogens and breakthrough viruses. Together with the low frequency of the RV144 signature sites, this may make vaccine protection harder to achieve in southern Africa compared to Thailand, a suggestion previously proposed by Hraber et al. based on neutralization properties \[[@ppat.1005742.ref015]\]. However, in RV144 the major immune correlate of risk was V2 binding antibody responses, and while these responses were clade sensitive, they were found to be cross-reactive with clade C V2 peptides, suggesting that high diversity in V2 may have a limited impact \[[@ppat.1005742.ref011],[@ppat.1005742.ref012]\]. Together these findings have important implications for future vaccine development although a major caveat associated with this conclusion is that the proposed correlates of risk identified in RV144 may be different in a clade C setting.
This clade C panel of acute/early Env pseudotyped viruses is the largest collection of functional clade C envelope clones available, and as such, provides a unique resource to the field. The samples are geographically representative of the southern African epidemic, and all samples were from acute/early infections. Together this makes them valuable reagents to support HIV-1 vaccine and passive immunization clinical trials, and important reagents for characterization of the breadth and potency of newly isolated bnAbs. However, one limitation of this work was that, due to the difficulty in obtaining very early samples, this study took many years to accumulate sufficient sample numbers, and only 63% of the panel was relatively current (obtained between 2006--2010).
In conclusion, elucidation of viral traits associated with resistance to antibody responses, and how these change over time, will be important to inform vaccine design efforts. This panel will be used to select representative viruses for evaluation of nAb responses in vaccine trials conducted in southern Africa. Furthermore, it has been used extensively to inform passive immunization studies and to prioritize bNAbs for clinical testing in clade C populations. This study has highlighted aspects that are highly relevant to vaccine design and provides insight into how efficacy vaccine trials, currently underway or imminent, will fare in this region of the world.
Supporting Information {#sec024}
======================
###### Maximum likelihood phylogenetic analysis of southern African clade C acute/early envelope nucleotide sequences (n = 200).
Branches are colored according to country/regional origin. Fifty-nine clade C and clade C related sequences included in the tree from India and China are illustrated in grey. Circulating recombinant forms CRF07_BC, CFR08_BC (<http://www.hiv.lanl.gov/content/sequence/HIV/CRFs/CRFs.html>) as well as unique inter-clade B and C recombinant forms (URFs) that were clade C across Env are included. Southern African candidate vaccine strains, TV1, Ce1086 and 96ZM651 are indicated. Country abbreviations used are: Botswana (BW), Malawi (MW), Tanzania (TZ) Zambia (ZM) and South Africa (ZA). South African sequences were further partitioned by region and annotated by symbols: Eastern Cape (ZAec), Western Cape (ZAwc), Northern Cape (ZAnc), Mpumalanga (ZAmp), Kwazulu-Natal (ZAkzn), Limpopo (ZAlp), Gauteng (ZAgp) and North-West (ZAnw) ([Table 1](#ppat.1005742.t001){ref-type="table"}). Bootstrap values \> 80% of 100 resampled replicates are illustrated as filled circles on nodes.
(TIF)
######
Click here for additional data file.
###### Variation in neutralization sensitivity of clade C viruses overall and per infection stage to IgG pools from clade B infected individuals (HIVIG-B) and from clade C infected individuals (HIVIG-C).
Pseudotyped clade C viruses were partitioned into three infection-stage groups per sequential gain of HIV-1 specific antibody responses in the donors as a marker of time from infection; pre-seroconversion (Ab-) in blue, indeterminate (Ab-/+) in grey and post-seroconversion (Ab+) in red filled circles. Significant differences in sensitivity for Ab- and Ab+ viruses are shown as p-values for a two-sided Mann-Whitney test. Graphs show median IC~50~ values illustrated by horizontal lines in red. Higher IC~50~ titers reflect increased neutralization resistance.
(TIF)
######
Click here for additional data file.
###### Strong associations exist between clade C neutralization sensitivity and hypervariable loop properties such as loop length, glycan density and net charge.
Associations tested between virus sensitivity and hypervariable loop properties; length (A); glycan density (B); and net charge (C). Significance of associations was tested using Kendall's rank correlation tau. Three exceptionally sensitive viruses, two tier 1A (SO032_A2.8--1, CH0505.w4.3) and one tier 1B, (6644.v2.c33), were excluded from this analysis.
(TIF)
######
Click here for additional data file.
###### Modelling the contribution of N332 and infection stage on neutralization sensitivity of viruses.
Viruses from pre-seroconversion are indicated in blue and those from post-seroconversion in red. Open and filled circles depict viruses without and with N332 respectively. Medians are shown by horizontal lines in red and interquartile ranges in black. Significant differences between medians (in red) indicated with the relevant p-value for a two-sided Mann-Whitney test.
(TIF)
######
Click here for additional data file.
###### Direct correlation of polyclonal sera neutralization sensitivity and Env characteristics over the course of the clade C epidemic.
No discernible significant association of polyclonal sera neutralization (A) or Env characteristics, variable loop length (B) and glycan density (C) over time. Correlations were performed using Kendall\'s rank correlation tau test.
(TIF)
######
Click here for additional data file.
###### Diversification of the southern African clade C epidemic, over a period spanning 1998--2010 is associated with overall increased resistance to bnAb neutralization.
\(A\) Protein ML phylogeny derived distances, estimated as branch lengths from root using the minimum sum of variance method, correlated to bnAb susceptibility measured as the geometric mean IC~50~ titer against all five bnAbs, VRC01, PG9, PGT128, CAP256 and 4E10 in aggregate. Placeholder constants of 50 were used for IC~50~ \>50 μg/μl, \>10 μg/μl \>25 μg/μl. (B) Sensitivity to bnAb correlated to calendar year. Correlations tested with Kendall's Tau.
(TIF)
######
Click here for additional data file.
###### Gp120 Amino Acid divergence of acute/early clade C viruses to vaccine inserts over the course of the clade C epidemic.
No significant difference in Gp120 amino acid distances of panel viruses to vaccine inserts Ce1086 and TV1, when comparing viruses collected from earlier in the epidemic (before 2008, n = 130) to those collected later in the epidemic (after 2008, n = 70).
(TIF)
######
Click here for additional data file.
###### Comparison of neutralization susceptibility only for neutralization-sensitive clade C panel viruses.
Viruses were partitioned into three infection stage groups according to their sequential gain of HIV-1 specific antibody responses as a marker of time from infection; pre-seroconversion (Ab-) (n = 58) indicated in blue, indeterminate (Ab-/+) (n = 26) indicated in grey and post-seroconversion (Ab+) (n = 55) indicated in red. Serum neutralization potency as measured by GMT. Three very sensitive viruses, two tier1A (SO032_A2.8--1, CH0505.w4.3) and one tier 1B, (6644.v2.c33), were excluded from the analysis. Mann-Whitney two-sided tests employed with median values shown in red and interquartile ranges in black.
(TIF)
######
Click here for additional data file.
###### Description of 200 southern African acute/early clade C panel viruses.
(DOCX)
######
Click here for additional data file.
###### Pre-screening of 54 of chronic serum samples against a panel of seven pseudoviruses, including 4 clade C's (CAP8.6F, CAP255.16, Du156.12 and a clade C consensus); 2 clade B's (6535 and a clade B consensus); and 1 clade A (Q23.17).
(DOCX)
######
Click here for additional data file.
###### Demographic characteristics of the subset of thirty sera and plasma samples selected based on evidence for cross-reactivity, sufficient sample availability, and geographical representation.
(DOCX)
######
Click here for additional data file.
###### Co-receptor usage predictions and *in vitro* Trofile assay confirmation.
(DOCX)
######
Click here for additional data file.
For providing plasma samples, we thank the following networks/institutions: the SANBS (Southern African National Blood Services); CHAVI (Centre for HIV/AIDS Vaccine Immunology); HVTN (HIV Vaccine Trials Network); CAPRISA (Centre for the AIDS Programme of Research in South Africa); the Zambia-Emory HIV Research Project; HISIS (HIV super infection study); MHRP (US Military HIV Research Program); Malawi and HIV-1 in Pregnancy Program; and participants and study team of the Tshedimoso study in Botswana (Botswana/Harvard Partnership). For providing serum samples, we thank the following institutions and individuals: Desmond Tutu HIV foundation (DTHF), Heidi Freislich; CAPRISA and The Perinatal HIV Research Unit (PHRU), Fatima Laher.
[^1]: The authors have declared that no competing interests exist
[^2]: Conceived and designed the experiments: BK MSS DCM CW. Performed the experiments: CR RT AR JG BRC GPB JCM TED MSS. Analyzed the data: CR BK MSS EEG KW PH CW DJS. Contributed reagents/materials/analysis tools: BK MSS DJS HG KMG HT PH GEG JK MJM MV KM LB MH LMo JM MLR SAK QAK JHK BHH FG RS LMa DCM CW PLM NT. Wrote the paper: BK MSS DCM CW LMo CR.
| {
"pile_set_name": "PubMed Central"
} |
1. Introduction {#sec1}
===============
Macrocyclic species based on transition metal compounds and multidentate ligands is an interesting field in chemistry and has been the subject of extensive research due to their potential applications in building block macrocyclic-based chemistry \[[@B1], [@B2]\] and environmental chemistry \[[@B3]\] and biomedical \[[@B4]\]. The chemistry of Schiff-base is an important field in coordination chemistry \[[@B5]\]. This is due to their ability to react with a range of metal ions forming stable complexes which have applications in different fields \[[@B6], [@B7]\]. One interesting application in the field of coordination chemistry has been to investigate the use of Schiff-base ligands to develop phenoxo-bridged binuclear complexes with homometallic and/or heterometallic centres. Complexes based on Schiff-base ligands play important roles in biomedical \[[@B8], [@B9]\], biomimetic, and catalytic systems \[[@B10], [@B11]\] and in supporting liquid crystalline phases \[[@B12]\]. A number of Schiff-base complexes have been used as oxygen carriers to mimic complicated biological systems \[[@B7], [@B11]\]. Furthermore, metal complexes of chromium, manganese, nickel, copper, zinc, and ruthenium with a wide variety of Schiff-bases are active oxidants for stoichiometric conversion and have been used as catalysts for carbonylation, hydrogenation, hydroformylation, and epoxidation reactions \[[@B13], [@B14]\]. Recently, we reported the formation of polymeric chain assemblies of some phenoxo-bridged binuclear transition metal complexes with multidentate Schiff-base ligand, namely, sodium(E)-6,60-((1E,10E)-(ethane-1,2-diylbis(azan-1-yl-1-ylidene) bis(methan-1-yl-ylidene) bis-(4-methyl-2((E)(pyridine-2-ylmethylimino)methyl)phenolate) H~2~L. As part of our group to explore the use of multidentate Schiff-base ligands for the building blocks of macrocyclic compounds, we describe here the formation of macrocyclic species of some phenoxo-bridged tetranuclear transition metal ions with the new macrocyclic Schiff-base. The ligand was derived via a template approach in which ethylenediamine fragments facilitated the linkage of the two units (10,21-bis-iminomethyl-3,6,14,17-tricyclo\[17.3.1.1^8,12^\] tetracosa-1(23),2,6,8,10,12(24),13,17,19,21,-decaene-23,24-disodium).
2. Experimental {#sec2}
===============
2.1. Materials and Methods {#sec2.1}
--------------------------
All reagents were commercially available and used without further purification. Solvents were distilled from appropriate drying agents immediately prior to use.
2.2. Physical Measurements {#sec2.2}
--------------------------
Melting points were obtained on a Buchi SMP-20 capillary melting point apparatus and are uncorrected. IR spectra were recorded as KBr discs using a Shimadzu 8400 FTIR spectrophotometer in the range 4000--400 cm^−1^. Electronic spectra of the prepared compounds were measured in the region 250--900 nm for 10^-3 ^M solutions in DMF at 25°C using a Shimadzu 160 spectrophotometer. ^1^H- and ^13^C-NMR spectra were acquired in DMSO-d~6~ solution using a Brucker AMX400 MHz spectrometer with tetramethylsilane (TMS) as an internal standard. Mass spectra obtained by positive fast atom bombardment (FAB) were recorded on a VG autospec micromass spectrometer. Elemental analyses (C, H, and N) were carried out on a Heraeus instrument (Vario EL). Metals were determined using a Shimadzu (A.A) 680 G atomic absorption spectrophotometer. Chloride was determined using potentiometer titration method on a (686-Titro processor-665Dosimat-Metrohm Swiss). Conductivity measurements were made with DMSO solutions using a PW 9526 digital conductivity meter, and room temperature magnetic moments were measured with a magnetic susceptibility balance (Johnson Matthey Catalytic System Division).
3. Synthesis {#sec3}
============
3.1. Preparation of the Precursor Sodium (2,4,6-Triformyl Phenolate) (STFP) {#sec3.1}
---------------------------------------------------------------------------
To a solution of p-hydroxybenzaldehyde (12.2 g, 10 mmol), hexamethylenetetramine (28.2 g, 20 mmol) in glacial acetic acid (50 mL), paraformaldehyde (30 g, 100 mmol) was added. The mixture was allowed to stir continuously until the deep orange viscous solution was obtained, and then heated up to 90°C for two hours. The solution was allowed to cool to room temperature, and then concentrated H~2~SO~4~ (10 mL) was carefully added. The resulting solution was refluxed for 30 min, and on treatment with distilled water (400 mL), a light orange precipitate was formed, which was stored overnight at 4°C. The orange product was isolated by filtration and washed in small amount of cold methanol to obtain *2,4,6-triformyl phenol (tfp)* \[[@B15]\]. The yielded product *(TFP)* (59%, 10.5 g) was mixed with equimolar amounts of NaOH (2.35 g, 5.8 mmol) in ethanol (25 mL). The mixture was allowed to stir for 30 min to give the sodium salt product (STFP) as a red-orange powder \[[@B16]\]. Yield: 65%, 7.66 g; m.p. 150°C. IR data (cm^−1^): 2995 (C--H) arom, 2924 and 2791 (C--H) aldehydic, 1683 (C=O), and 1240 (C--O).
3.2. Preparation of Na~4~L {#sec3.2}
--------------------------
A solution of sodium *2,4,6-triformyl phenol*ate (STFP) (0.5 g, 2.4 mmol) in methanol (15 mL) was added slowly with stirring to a mixture of ethylenediamine (0.22 g, 3.6 mmol) dissolved in methanol (15 mL), and then 2--4 drops of glacial acetic acid were added to the reaction mixture. The mixture was allowed to reflux with stirring. After 4 h, the reaction was cooled to room temperature and then allowed to slowly evaporate to give a pale yellow viscous residue which was stirred for 30 min with a hot mixture of DMF/methanol 3 : 1 (15 mL). Solvent containing the required ligand was transferred to a vessel by decantation, and then solvent was removed under reduced pressure and residue was kept under vacuum for drying for 24 h. Yield: 0.77 g, 33%, m.p. = 178°C. IR data (cm^−1^): 3008 *ν*(C--H) arom, 2723 *ν*(C--H) iminic, 1632 and 1622 *ν*(C=N), 1350 *ν*(phenoxide). The ^1^H-NMR spectrum of the ligand in DMSO-d~6~ showed peaks at *δ* ~H~ (400 MHz, DMSO-d~6~): 8.5 (8H, s, N=C--H); 8.3 (4H, s, N=C--H); 7.5 (8H, s, Ar--H); 4.5 (16H, m, N--CH~2~); 3.1 (8H, m, N--C--H~2~) and at *δ* ~c~ (100.63 MHz, DMSO-d~6~): 57.49 (CH~2~N=C); 59.73 (CH~2~N=C); 117.91 (Ar); 119.13 (Ar); 121.7 (Ar); 123.4 (Ar); 143.1 (Ar); 160.08 (C--O); 162.23 (C=N) imine; 164.01 (C=N) imine. The positive (FAB) mass spectrum of Na~4~L showed a peak at *m*/*z* 967.103 (10%) corresponding to (M+Na)^+^ and the following fragments; 854.56 (32%) \[M-(Na~2~O+CH~2~CH~2~)\]^+^, 788.90 (26%) \[M-{(Na~2~O+CH~2~CH~2~)^+^(NaCH=NO)}\]^+^, 728.14 (32%) \[M-(Na~2~O+CH~2~CH~2~)+(NaCH=NO)+(NH~2~CH~2~CH~2~NH~2~)\]^+^, 533.21 (100%) \[M-(Na~2~O+CH~2~CH~2~)+(NaCH=NO)+(NH~2~CH~2~CH~2~NH~2~)+(C~10~H~10~NH--C=NNa)\]^+^, 480.68 (29%) \[M-(Na~2~O+CH~2~CH~2~)+(NaCH=NO)+(NH~2~CH~2~CH~2~NH~2~)+(C~10~H~10~NH--C=NNa)+(NaCH~2~NH~2~)\]^+^.
3.3. General Synthesis of Complexes {#sec3.3}
-----------------------------------
A solution of the Schiff-base ligand (1 mmol) dissolved in a 3 : 1 mixture of DMF/MeOH (25 mL) was allowed to stir for 15 min. A methanolic solution (15 mL) of the metal(II) salt (4.1 mmol) was then added dropwise, (metal (II) salts are hydrated chloride; MCl2·XH2O (M = Mn^II^, X = 4; Co^II^, Ni^II^ and Cu^II^; X = 6, 6 and 2, respectively. Zinc chloride was no hydrated)). The reaction mixture was heated under N~2~ for 2 h on a water bath, resulting in the formation of a solid mass which was washed several times with hot methanol and then dried at room temperature. Elemental analysis data, colours and yields for the complexes are given in [Table 1](#tab1){ref-type="table"}.
^1^H-NMR spectrum of \[Zn^II^ ~4~(L)\]Cl~4~ in DMSO-d~6~ showed peaks at *δ* ~H~ (400 MHz, DMSO-d~6~): 3.8 (8H, m, N--C--H~2~), 5.2 (16 H, m, N--C--H~2~), 7.3 (8H, d, 9.8 Hz, Ar--H), 8.1 (4H, br, N=C--H), and 8.7 (8H, br, N=C--H).
3.4. Determination of Biological Activity {#sec3.4}
-----------------------------------------
Bioactivities were investigated using agar-well diffusion method \[[@B17]\]. The wells were dug in the media with the help of a sterile metallic borer with centres at least 24 mm. Recommended concentration (100 **μ**L) of the test sample 1 mg/mL in DMSO was introduced in the respective wells. The plates were incubated immediately at 37°C for 20 hours. Activity was determined by measuring the diameter of zones showing complete inhibition (mm). To examine the role of DMSO in the biological screening, separate studies were conducted with the solutions alone of DMSO, which showed no activity against any bacterial strains. All these complexes were found to be potentially active against these bacterial strains, except for the strain of *Pseudomonas aeruginosa*.
4. Results and Discussion {#sec4}
=========================
4.1. Chemistry {#sec4.1}
--------------
A template approach was implemented to obtain the Schiff-base Na~4~L in a reasonable yield ([Scheme 1](#sch1){ref-type="fig"}). Using Na^+^ ion was found to be essential to form the ligand since otherwise only a polymeric mixture, partially soluble in hot DMF, was recovered via direct approach. The ligand was prepared from the reaction of sodium *2,4,6-triformyl phenol*ate (STFP) with ethylenediamine in mole ratios 2 : 3, respectively. The Schiff-base is soluble with stirring in DMF and DMSO but not in other common organic solvents. The ligand was characterised by elemental analysis ([Table 1](#tab1){ref-type="table"}), IR ([Table 2](#tab2){ref-type="table"}) and UV-Vis ([Table 3](#tab3){ref-type="table"}) spectroscopy, and ^1^H- and ^13^C-NMR spectroscopy. The IR spectrum of the free Schiff-base shows characteristic bands at 1632, 1622, 1350, and 1031 cm^−1^ due to the *ν*(C=N), *ν*(phenoxide), and *ν*(C--O) functional groups, respectively. The UV-Vis spectrum of Na~4~L exhibits an intense absorption peak at 295 nm, assigned to *π* → *π*\*. The peak at 322 nm assigned to *n* → *π*\* transition.
The bridged phenoxy tetranuclear complexes with Mn^II^, Co^II^, Ni^II^, Cu^II^ and Zn^II^ were synthesised by heating 1 mmole of the ligand with 4.1 mmole of the metal chloride in a mixture of DMF/MeOH. Complexes of general formula \[M^II^ ~4~(L)\]Cl~4~ ((M = Mn^II^, Co^II^, Ni^II^, Cu^II^, and Zn^II^) were obtained, [Scheme 1](#sch1){ref-type="fig"}). The complexes are air-stable solids, soluble in hot DMSO and DMF but not in other common organic solvents. The coordination geometries of the complexes were deduced from their spectra. The analytical data ([Table 1](#tab1){ref-type="table"}) agree well with the suggested formulae. Conductivity measurements of Mn^II^, Co^II^, Ni^II^, Cu^II^, and Zn^II^ complexes in DMF lie in the 291.08--297.14 cm^2^ Ω^−1^ mol^−1^ range, indicating their 1 : 4 electrolytic behaviour ([Table 1](#tab1){ref-type="table"}) \[[@B18]\].
4.2. FTIR and NMR Spectra {#sec4.2}
-------------------------
The most important infrared bands for the complexes together with their assignments are listed in [Table 2](#tab2){ref-type="table"}. The IR spectra of the complexes exhibited ligand bands with the appropriate shifts due to complex formation. The *ν*(C=N) imine stretching band at 1632 cm^−1^ in the free Schiff-base is shifted to lower frequency and is observed at around 1589 cm^−1^ for the complexes. The bands are assigned to a *ν*(C=N) stretch of reduced bond order. This can be attributed to delocalisation of metal electron density (*t* ~2*g*~) to the *π*-system of the ligand \[[@B19], [@B20]\], indicating coordination of nitrogen of the C=N moieties to the metal atoms \[[@B21]\]. In addition, the IR spectra of the complexes display peaks around 1620 cm^−1^, which may be attributed to the *ν*(C=N) imine stretching of the uncoordinated moieties. Further, bands in the region of 1518--1550 cm^−1^ in all the complexes suggest phenoxide bridging with the metal atoms \[[@B22], [@B23]\]. At lower frequency, the complexes exhibited bands around 619--688 and 516--584 cm^−1^, which could be assigned to *ν*(M--N) and *ν*(M--O) vibration modes, respectively \[[@B19], [@B24]\]. Due to the larger dipole moment change for M--O compared to M--N, the *ν*(M--O) usually appears at higher frequency than the *ν*(M--N) band \[[@B25]\]. The electronic spectra and magnetic moment data of the complexes are summarised in [Table 3](#tab3){ref-type="table"}.
The ^1^H-NMR spectrum in DMSO-d~6~ of the free Schiff-base shows peaks at 8.5 and 8.3 ppm assigned to --CH=N-- (imine) protons, indicating that the azomethine protons are nonequivalent. In addition, the spectrum revealed two peaks around 4.5 and 3.7 ppm assigned to the CH~2~N fragment. The appearance of two chemical shifts may be because of the formation of two types of azomethine, (i) the one that is involved in the formation of the submacrocyclic component, and (ii) the one that facilitated the linkage between the two submacrocyclic parts. The ^13^C-NMR displays two peaks at ca. 59 ppm and two signals at ca. 162 ppm, indicating that the CH~2~N groups and the azomethine moieties are in a different environment. The NMR data is in accordance with the IR result in which two different peaks for C=N group were observed. The peak at 7.4 ppm is assigned to protons of aromatic ring. The ^1^H-NMR spectrum of \[Zn^II^ ~4~(L)\]Cl~4~ showed that the peaks of the azomethine protons are nonequivalent. Peaks observed around 8.7 ppm are related to the coordinated azomethine which are shifted slightly downfield, compared with those observed for the free ligand. A peak recorded at 8.1 ppm is attributed to the free azomethine groups (uncoordinated). The doublet at 7.3 ppm is assigned to protons of aromatic rings. The appearance of these protons as a doublet is due to mutual coupling and/or a fluctuation behaviour generated by (--CH~2~CH~2~--) moieties \[[@B15]\]. In general, the spectrum showed broader peaks compared with that for the free ligand. This may point out that a fluctuation behaviour occurred in DMSO solution.
4.3. Mass Spectra {#sec4.3}
-----------------
The mass spectrum of the ligand was consistent with the proposed structural formula ([Section 3](#sec3){ref-type="sec"}). The positive ion FAB mass spectrum for \[Cu^II^ ~4~(L)\]Cl~4~ showed several peaks corresponding to successive fragmentation of the molecule. The mass spectrum of Cu(II) complex does not display a peak may refer to molecular ion peak. The first peak observed at *m*/*z* 1179 represents the molecular ion peak of the complex losing 2Cl moieties. Three distinct peaks were observed in the mass spectrum at *m*/*z* 1116, 932, and 902 which can be assigned to the fragments \[M-{(2Cl+Cu+H)}\]^+^, \[M-(4Cl+Cu+(CH~2~CH~2~)~3~+N~2~)\]^+^, and \[M-(4Cl+Cu+(CH~2~CH~2~)~3~+N~2~+O~2~)+H\]^+^, respectively. The FAB(+) mass spectrum for \[Co^II^ ~4~(L)\]Cl~4~ showed several peaks corresponding to successive fragmentation of the molecule. However, the spectrum failed to show a peak that refers to molecular ion peak. The first peak observed at *m*/*z* 1106 represents the molecular ion peak of the complex losing (2Cl+CH~2~CH~2~CN) fragment. Four distinct peaks were observed in the mass spectrum at *m*/*z* 897, 751, 518, and 328, can be assigned to the fragments \[M-{(2Cl+CH~2~CH~2~CN)+(2Cl+Co+CH~2~CH~2~+(CN)~2~}\]^+^, \[M-{(2Cl+CH~2~CH~2~CN)+(2Cl+Co+CH~2~CH~2~+(CN)~2~+(CH~2~CH~2~)~2~+Co+O~2~)}\]^+^, \[M-{(2Cl+CH~2~CH~2~CN)+(2Cl+Co+CH~2~CH~2~+(CN)~2~+(CH~2~CH~2~)~2~+Co+O~2~)+(C~12~H~16~N~4~O)}\]^+^, and \[M-{(2Cl+CH~2~CH~2~CN)+(2Cl+Co+CH~2~CH~2~+(CN)~2~+(CH~2~CH~2~)~2~+Co+O~2~)+(C~12~H~16~N~4~O)+(CHNH)~2~+CH~2~CH~2~+2Co)}\]^+^, respectively. The FAB (+) mass spectrum for \[Ni^II^ ~4~(L)\]Cl~4~ showed several peaks corresponding to successive fragmentations of the molecule. However, no peak related to molecular ion peak was detected in the spectrum. The first peak observed at *m*/*z* 1106 represents the molecular ion peak of the complex losing (2Cl+CH~2~CHCN)^+^ fragment. Five distinct peaks were observed in the mass spectrum at *m*/*z* 1031, 666, 638, 610, and 500, which can be assigned to the fragments \[M-{(2Cl+CHCH~2~CN)+(NiO)}\]^+^, \[M-{(2Cl+CHCH~2~CN)+(NiO) + (C~14~H~13~N~4~Ni)}\]^+^, \[M-{(2Cl+CHCH~2~CN)+(NiO)+(C~14~H~13~N~4~Ni)+(CH~2~CH~2~)}\]^+^, \[M-{(2Cl+CHCH~2~CN)+(NiO)+(C~14~H~13~N~4~Ni)+(CH~2~CH~2~)+(CH~2~CH~2~)}\]^+^, and \[M-{(2Cl+CHCH~2~CN)+(NiO)+(C~14~H~13~N~4~Ni)+(CH~2~CH~2~)+(C~6~H~7~NO)}\]^+^, respectively.
4.4. Electronic Spectra and Magnetic Moment Measurements {#sec4.4}
--------------------------------------------------------
The electronic spectra of the complexes with the ligand exhibited various extents of hypsochromic shift of the bands related to the intraligand *π*→*π*\* transition. The electronic spectrum of the tetranuclear-Mn(II) Schiff-base complex showed additional peaks at 318 and 423 nm assigned to the charge transfer (CT) and d--d transitions, respectively, in a distorted tetrahedral geometry \[[@B26], [@B27]\]. The observed magnetic moment for the Mn(II) complex 5.1 B.M is typical for tetrahedral geometry \[[@B19]\]. The electronic spectrum of the Co(II) complex is consistent with tetrahedral assignment \[[@B26], [@B28]\]. The spectrum of the Co(II) complex exhibited band characteristic of tetrahedral Co(II) complexes \[[@B26]--[@B29]\]. The magnetic moment was consistent with the tetrahedral environment around Co(II). The observed bands for the Ni(II) complex and its diamagnetic behaviour agrees well with the proposed square planar geometry \[[@B26], [@B30]\]. The electronic spectrum of the Cu(II) complex displays a broad band assigned to ^2^B~1~g → ^2^Eg transition, corresponding to square planar geometry \[[@B29], [@B30]\]. A magnetic moment of 1.51 B.M. is typical for four-coordinate copper complexes \[[@B31]\]. The spectrum of the Zn(II) complex exhibited bands assigned to ligand field *π*→*π*\* and L → M charge transfer \[[@B26], [@B32]\]. The metal normally prefers tetrahedral geometry. The magnetic moment values for the tetranuclear macrocyclic complexes at RT are lower than the predicted values, indicating the presence of some antiferromagnetic interactions. This may occur from metal-metal interactions through the phenolic oxygen atoms and/or extensive electron delocalisation, which may be related to the formation of layer structures \[[@B15], [@B19], [@B33]\].
5. Biological Activity {#sec5}
======================
The free Schiff-base macrocyclic ligand and its metal complexes were screened against *Staphylococcus aureus, Escherichia coli*, and*Pseudomonas aeruginosa*to assess their potential as an antimicrobial agent by disc diffusion method. The measured zone of inhibition against the growth of various microorganisms is listed in [Table 4](#tab4){ref-type="table"}. It is found that the metal complexes have higher antimicrobial activity against Gram negative species only compared with the free ligand. Hence complexation increases the antimicrobial activity. Such increased activity of the metal complexes can also be explained on the basis of chelation theory \[[@B33]\]. According to this, the chelation reduces the polarity of the metal atom mainly because of the partial sharing of its positive charge with donor group and possible *π*-electron delocalisation over the whole ring. This increases the lipophilic character of the metal chelate system which favours its permeation through lipid layer of the cell membranes.
6. Conclusion {#sec6}
=============
In this paper, the synthesis and coordination chemistry of some macrocyclic-based tetranuclear metal complexes derived from the Schiff-base Na~4~L are investigated. A template approach was used to prepare the ligand in a reasonable yield. The complexes were prepared by mixing at reflux 1 mmole of the Schiff base with 4 mmole of the appropriate metal chloride. Tetranuclear complexes of the general formulae \[M~4~(L)\]Cl~4~ (where M = Mn, Co^II^, Ni^II^, Cu^II^, and Zn^II^) was obtained. Physicochemical analysis showed four cationic coordinate metal complexes were formed.
{#sch1}
######
Colours, yields, elemental analyses, and molar conductance values.
Compound Colour Yield (%) m.p. Found (calcd.) (%) Λ~*M*~ (cm^2^Ω^−1^mol^−1^)
------------------------ ------------- ----------- ------ -------------------- ---------------------------- -------------- -------------- -------------- --------
Na~4~L Pale yellow 47 178 60.4 (61.6) 4.52 (4.9) 17.7 (18.4) --- ---
\[Mn^II^ ~4~(L)\]Cl~4~ Brown 44 285 46.6 (47.4) 3.4 (3.7) 13.6 (13.8) 17.6 (18.1) 10.7 (11.7) 294.01
\[Co^II^ ~4~(L)\]Cl~4~ Red-brown 50 320 46.2 (46.9) 3.4 (3.6) 13.4 (13.7) 18.1 (19.2) 10.9 (11.5) 295.58
\[Ni^II^ ~4~(L)\]Cl~4~ Green 46 305 46.36 (46.9) 3.33 (3.6) 13.39 (13.7) 18.14 (19.1) 10.88 (11.5) 292.71
\[Cu^II^ ~4~(L)\]Cl~4~ Green 42 270 45.9 (46.2) 3.3 (3.6) 13.2 (13.5) 19.8 (20.4) 10.9 (10.3) 291.08
\[Zn^II^ ~4~(L)\]Cl~4~ Yellow 42 265 44.86 (45.8) 3.4 (3.5) 13.2 (13.4) 19.9 (20.8) 10.1 (9.3) 297.14
######
FTIR frequencies in (cm^−1^) of the compounds.
Compound *ν*(C=N)~iminic~ *ν*(Phenoxide) *ν*(M--N) *ν*(M--O)
------------------------ ------------------ ---------------- ----------- -----------
Na~4~L 1632, 1622 1350 --- ---
\[Mn^II^ ~4~(L)\]Cl~4~ 1575, 1618 1528 619 516
\[Co^II^ ~4~(L)\]Cl~4~ 1579, 1620 1518 663 584
\[Ni^II^ ~4~(L)\]Cl~4~ 1581, 1620 1525 688 565
\[Cu^II^ ~4~(L)\]Cl~4~ 1577, 1617 1550 663 554
\[Zn^II^ ~4~(L)\]Cl~4~ 1583, 1622 1540 632 555
######
Magnetic moment and UV-Vis spectral data in DMF solutions.
Compound *μ* ~eff~ (BM) Band position (*λ* nm) Extinction coefficient *ε* ~max~ (dm^3^ mol^−1^cm^−1^) Assignments
------------------------ ---------------- ----------------------------- --------------------------------------------------------- -------------
Na~4~L 295 920 *π* → *π*\*
322 850 CT
\[Mn^II^ ~4~(L)\]Cl~4~ 5.11 267 640 *π* → *π*\*
318 530 CT
423 360 ^6^A~1~g→^4^T~1~g
\[Co^II^ ~4~(L)\]Cl~4~ 3.40 278 860 *π* → *π*\*
343 432 CT
461 123 ^4^T~1~g^(F)^→^4^T~1~g^(P)^
\[Ni^II^ ~4~(L)\]Cl~4~ 0.02 271 758 *π* → *π*\*
316 410 CT
664 87 ^1^A~1~g→^1^A~2~g
\[Cu^II^ ~4~(L)\]Cl~4~ 1.51 283 323 *π* → *π*\*
303 212 CT
462 103 ^2^B~1~g→^2^Eg
\[Zn^II^ ~4~(L)\]Cl~4~ Diamagnetic 291 574 *π* → *π*\*
311 1235 CT
######
Biological activity for Schiff-base macrocyclic ligand and its complexes.
-------------------------------------------------------------------------------------------------------------------------------
Compounds *Staphylococcus aureus* (+) *Escherichia coli* (−)\ *Pseudomonas aeruginosa* (−)
-------------------------- ----------------------------- ------------------------- ------------------------------ ----- --- ---
Free ligand − \+ − \+ − −
\[Mn^II^ ~4~(L)\]Cl~4~ \ ++ +++ \+ ++ − −
\[Co^II^ ~4~(L)\]Cl~4~
\[Ni^II^ ~4~(L)\]Cl~4~ ++ +++ \+ ++ − −
\[Cu^II^ ~4~(L)\]Cl~4~ ++ +++ \+ ++ − −
\[Zn^II^ ~4~(L)\]Cl~4~ − \+ ++ +++ − −
-------------------------------------------------------------------------------------------------------------------------------
(−): No inhibition/inactive, (+): (3--5) mm/active, (++): (6--8) mm/more active, (+++): (9--14) mm/highly active.
[^1]: Academic Editors: A. I. Matesanz and A. Souldozi
| {
"pile_set_name": "PubMed Central"
} |
Understanding human population genetic structure remains important for gaining insights into human history and demography, as well as for investigating genetic diseases in relation to geography and ancestry[@b1][@b2][@b3]. Historically, human population divergence is assessed using approaches from a variety of disciplines including archaeology, palaeontology, linguistics, climatology and genetics. Early studies of genetic divergence were conducted by investigating the genetic variability of mitochondrial DNA and Y chromosomes[@b4][@b5]. Currently, genome-wide SNP sites are used to measure population differentiation (e.g., the HapMap genotype data)[@b6][@b7][@b8], and to search for outlier regions that are potentially associated with geographically restricted genetic diseases[@b9]. Different classes of genetic markers vary widely in many important characteristics, such as their mode of inheritance (paternal, maternal, or biparental), mutation rate, and degree of selective neutrality. As a result population genetic divergence can vary depending on the class of genetic marker investigated. Copy-number-variable (CNV) loci are an important cause of genetic variation in human genomes, and give rise to differences of 4.8--9.5% in the overall length of human genomes[@b10][@b11]. However population genetic divergence at the genome-wide CNV loci has not been investigated in detail[@b12][@b13], nor has genome-wide divergence at the CNV loci been compared with that at SNP sites.
Genetic variation at CNV loci in *Homo sapiens* and other species has been extensively reviewed from a number perspectives[@b12][@b14]. Topics covered include the mechanisms for generating copy number variation, natural selection on duplication and deletion variants, the impacts of demographical changes on CNV loci, associations with SNP loci, and the role of CNV loci in causing diseases[@b10][@b11][@b12][@b15][@b16][@b17]. At the population level, the evolutionary dynamics of CNV loci can be studied within the framework of population genetics[@b12][@b14]. Emerson *et al*.[@b18] used an infinite-site model to investigate purifying selection on copy number variation in specific gene regions in *Drosophila melanogaster*. Sjödin and Jakobsson[@b12] suggested the use of a K-allele model[@b19] or a stepwise mutational model[@b20] to describe the mutation process at CNV loci. The neutrality of CNV loci has also been analyzed[@b21][@b22]. We recently developed a three-allele model to test neutrality at CNV loci, and demonstrated selective neutrality at 856 CNV loci scored in 1184 healthy individuals from the HapMap genotype data set[@b23]. The evolution of these CNV loci can be essentially explained by a mutation-drift process[@b23]. Here, we proceed with the same dataset to investigate population genetic divergence at the genome-wide CNV loci.
In comparison with variation at SNP sites, variants at CNV loci have several distinct features. First, CNV variants often differ in length by 1kbp or more[@b24][@b25], whereas SNP variants differ by a single base pair. Thus although CNV loci (\~4.8\~9.5% of human genomes) are much less abundant than SNP sites in human genomes, they represent an important type of chromosomal structural variation[@b11]. Second, more complex processes are involved in generating copy number variants, including non-allelic homologous recombination (NAHR)[@b26], non-homologous end joining (NHEJ), and insertion of transposable elements (TEs)[@b27][@b28]. These differ dramatically from the mechanisms generating point mutation (transitions and transversions) at SNP sites. Third, the average mutation rate at CNV loci is expected to be much higher than the point mutation rates at SNP sites[@b29], resulting in a much younger average age of alleles for CNV than for SNP loci in natural populations[@b30]. Given these differences in the properties of CNV and SNP markers, we anticipate that they will vary in their degree of population genetic divergence.
To test this hypothesis, we employ genotype data at CNV loci from the HapMap Phase III populations. This has two advantages. The first is that genetic divergence among these populations has been fully investigated at genome-wide SNP sites[@b31], providing the opportunity for direct comparison with results for the genome-wide CNV loci. Analysis of CNV loci has so far only been conducted with partial HapMap Phase III populations[@b31] or at a particular gene site[@b32]. Our result should differ from existing analysis because they will include more populations with a wider range of ancestry. Increasing the number of individuals will affect both the genetic divergence and the number of common CNV loci. The second advantage for using the HapMap dataset is that exact discrete copy numbers are available for each diploid genotype at each CNV locus[@b31]. Although techniques for detecting CNV loci have recently been improved, discrete copy-number genotypes at each CNV locus, which are also essential for accurate case-control association testing with CNV loci[@b33], are rarely archived in publically accessible data. Furthermore, the sample sizes in previous studies at CNV loci are often too small, and hence are inappropriate for population genetic structure analysis[@b18][@b34][@b35]. The large sample sizes in HapMap Phase III populations means that the probabilities of making either false-positive or negative CNV calls are negligible[@b23].
In this study we analyze genetic divergence at the genome-wide CNV loci and compare it with that at the genome-wide SNP sites in exactly the same populations. To further address the population genetic properties of CNV loci and reinforce our explanations of evolution at CNV loci, we test LDs at both gametic and zygotic levels among all pairs of CNV loci. We compare the patterns of gametic and zygotic LDs at CNV loci with those previously reported at SNP sites[@b36][@b37]. Recent theoretical studies indicate that zygotic LD is more informative than gametic LD for inferring the effects of different evolutionary forces (mating system, gene flow, selection, and genetic drift)[@b38][@b39]. In the absence of functional epistatic selective effects among loci, gametic LD (lower order) is always greater than the maximum zygotic LD in value. Other processes, including mating system, gene flow and genetic drift, do not change this pattern although they can generate LD (statistical associations between loci)[@b38][@b39]. The difference between the values of gametic LD and maximum zygotic LD can be used to infer whether epistasis exists between loci. Such differences tested previously at the genome-wide SNP sites with the HapMap Phase III populations[@b37], have shown the existence of epistases among many SNP sites. Here, we also investigate this property at the genome-wide CNV loci by presuming that individual CNV loci are directly/indirectly or equally involved in fitness changes. Information from LD analyses among CNV loci helps us to view the difference in population genetic divergence between SNP and CNV loci from a different perspective. Overall our objective is to infer the roles of mutation and migration in producing human population genetic divergence at the genome-wide CNV loci by comparing the single and multilocus population genetic structure of SNP and CNV loci.
Results
=======
Population genetic divergence
-----------------------------
Maximum likelihood estimates (MLEs) of allele frequencies are summarized in [Table S1](#S1){ref-type="supplementary-material"}. Although all CNV loci are polymorphic in the pooled population, they exhibit various levels of polymorphisms among populations ([Table 1](#t1){ref-type="table"}). More than 80% of CNV loci are polymorphic in African populations (ASW, LWK, MKK, and YRI), but less than 60% in non-African populations except MEX (62.38%). Three Asian populations (CHB, CHD, and JPT) have about 45% polymorphic CNV loci.
African populations have 1.84--1.90 alleles per CNV locus while Asian populations have about 1.50 alleles per CNV locus. The rest of the 11 populations have intermediate numbers of alleles per locus (*N*~*a*~ = 1.6--1.66). Similarly, African populations have high gene diversity over all CNV loci (*H*~*e*~ = \~0.13) and small standard deviations (\~0.15); while Asian populations have low gene diversity (\~0.11) but large standard deviations (\~0.16) over all CNV loci. The rest of the 11 populations have intermediate gene diversity and standard deviations ([Table 1](#t1){ref-type="table"}).
Genetic differentiation measured by *G*~*st*~ is 0.0498 ± 0.0491 among all CNV loci, and most individual *G*~*st*~ values are around 0.05, with a few CNV loci having relatively large *G*~*st*~ values ([Fig. 1](#f1){ref-type="fig"}). Substantial variations exist among chromosomes, especially for the small *G*~*st*~ values that are outside the 95% CIs ([Fig. 2](#f2){ref-type="fig"}). The proportions of CNV loci exhibiting a significantly low level of population genetic divergence are 72.72% on Chr 1, 51.35% on Chr 4,76.6% on Chr 5, 84.48% on Chr 6, 76.67% on Chr 7, 56.26% on Chr 9, 62.8% on Chr 11, 52.63% on Chr 17, 60.87% on Chr 19, and 90.91% on Chr 22. The rest of the chromosomes have less than 50% of CNV loci with a significantly low level of population differentiation. None of the chromosomes has any CNV locus exhibiting a significantly high level of population differentiation ([Fig. 2](#f2){ref-type="fig"}).
The average pairwise multilocus *G*~*st*~ ranges from 0.0038 ± 0.00001 (CHB-CHD) to 0.0421 ± 0.0001 (JPT-LWK), with the mean of 0.0255 ± 0.0114 over all pairs ([Table 2](#t2){ref-type="table"}). The average pairwise multilocus *G*~*st*~ in African populations ranges from 0.0059 ± 0.00001 (LWK-YRI) to 0.0128 ± 0.00002 (MKK-YRI), with the mean of 0.0081 ± 0.0025 over population pairs. The average pairwise multilocus *G*~*st*~ in non-African populations ranges from 0.0038 ± 0.00001(CHB-CHD) to 0.0352 ± 0.0001(TSI-JPT), with the mean of 0.0212 ± 0.0109 over population pairs. The average pairwise multilocus *G*~*st*~ among African and non-African populations ranges from 0.0206 ± 0.00004 (MKK-TSI) to 0.0421 ± 0.0001 (JPT-LWK), with the mean of 0.0324 ± 0.0064. over population pairs.
Compared with the pairwise multilocus *F*~*st*~ previously reported at the genome-wide SNP sites[@b7], the pairwise multilocus *G*~*st*~ at the genome-wide CNV loci is generally more than three times lower (average ratio of *F*~*st(SNP)*~/*G*~*st(CNV)*~ = 3.3081 ± 1.1837; [Table 2](#t2){ref-type="table"}). The ratios of *F*~*st(SNP)*~/*G*~*st(CNV)*~ range from 1.3481 ± 0.0171 (LWK-YRI) to 2.1023 ± 0.0087 (MKK-LWK) in African populations, with the mean of 1.6849 ± 0.3294 over population pairs; from 0.2649 ± 0.0265 (CHB-CHD) to 3.6545 ± 0.0253 (CEU-CHD) in non-African populations, with the mean of 2.5048 ± 0.9240 over population pairs; and from 3.3.5497 ± 0.0200 (ASW-GIH) to 4.8624 ± 0.0258 (CHD-MKK) among African and non-African populations, with the mean of 4.2584 ± 0.3548 over population pairs ([Table 2](#t2){ref-type="table"}).
Inter-chromosomal variations in pairwise *G*~*st*~ values are substantial among different population pairs ([Figure S1a](#S1){ref-type="supplementary-material"}), indicating the presence of differential divergences among chromosomes during the formation of populations. The pairs among African and non-African populations have large variations among chromosomes, especially on Chrs 9, 10, 16, 20, and 22 ([Figure S1a](#S1){ref-type="supplementary-material"}), while the pairs among African populations or among non-African populations exhibit relatively stable divergences among chromosomes (e.g., CHB-JPT and CEU-CHB; [Figure S1b](#S1){ref-type="supplementary-material"}).
Pairwise Nei's genetic distances at multiple CNV loci range from 0.001 ± 0.000004 (CHB-CHD) to 0.0241 ± 0.0001 (CHD-YRI), with a mean of 0.0124 ± 0.0067 over all pairs ([Table S2](#S1){ref-type="supplementary-material"}). The average genetic distance is 0.0029 ± 0.0010 among African populations, 0.0085 ± 0.0049 among non-African populations, and 0.0174 ± 0.0040 among African and non-African populations. Cluster analysis with the unweighted pair group method with arithmetic mean (UPGMA) shows that the three subgroups (African, Asian, and the rest of the populations) are clearly distinguished ([Fig. 3](#f3){ref-type="fig"}). Bootstrapping resample trees (1000) using PHYLIP[@b40] indicate that African and non-African populations can be separated with a probability of 100% (data not shown here).
Consider an average mutation rate of the order 10^−5^ at a CNV locus[@b29], the equal effective population sizes among the 11 populations, and 25 years per generation. From the average distance and its approximate variance *V(t*), the population isolation time is generally about *t* = 0.0124 × 5 × 10^4^ × 25 ± 0.0067 × 5 × 10^4^ × 25 = 15500 ± 8375 years among populations, *t* = 3625 ± 1250 years among African populations, about *t* = 10625 ± 6125 years among non-African populations, and about *t* = 21750 ± 5000 years among African and non-African populations.
Gametic and zygotic LDs
-----------------------
Statistical tests indicate that very few pairs of CNV loci, 0.027\~0.073%, exhibit significant gametic LDs in the 11 populations ([Table 3](#t3){ref-type="table"}; [Table S3](#S1){ref-type="supplementary-material"} for details). Most pairs of CNV loci have insignificant gametic LDs in each population. Among the significant gametic LDs, African populations generally have a lower proprtion of CNV locus pairs with significant gametic LDs than do most non-African populations ([Table 3](#t3){ref-type="table"}). The average significant r-squares are higher for CNV loci from the same chromosome (\~0.76) than from different chromosomes (\~0.16). Among the significant gametic LDs on the same chromosomes, more pairs come from partially overlapped CNV loci in each population ([Table 3](#t3){ref-type="table"}).
Patterns of gametic LDs are different among populations. African populations have more significant gametic LDs from different chromosomes than from the same chromosomes, while non-African populations except CEU and MEX have more significant gametic LDs from the same chromosomes than from different chromosomes. No common pairs of CNV loci have significant gametic LDs on different chromosomes among 11 populations, but twelve common pairs from overlapped CNV loci (except one on Chr 7) exist, with 1 on Chrs 1, 7,9,11, and 12, 3 on Chr 5, and 4 on Chr 6 ([Table S4](#S1){ref-type="supplementary-material"}).
Tests of zygotic LDs also indicate that a very few CNV loci have significant zygotic LDs, 0 \~ 0.0359% ([Table 4](#t4){ref-type="table"}), which is generally less than the proportion of significant gametic LDs ([Table 3](#t3){ref-type="table"}). Most CNV loci with significant zygotic LDs are partially overlapped on the same chromosomes ([Table S5](#S1){ref-type="supplementary-material"}). African populations have fewer significant zygotic LDs than do most non-African populations in significant *D*~*ij*~ (*i, j* = 0, 1, 2) except *D* ~*3j*~ (j = 0, 1, 2, 3). There are twenty-two common pairs of CNV loci (mostly overlapped) of significant zygotic LDs in 11 populations, with 1 pair on Chr 1, 3 on Chr 5, 7 on Chr 6, 3 on Chr 7, 1 on Chr 9, 2 on Chr 10, 2 on Chr 11, 1 on Chr 12, 1 on Chr 13, and 1 on Chr 20 ([Table 4](#t4){ref-type="table"}). These locus pairs also have significant gametic LDs, while some CNV loci with significant zygotic LDs have no significant gametic LDs in 11 populations ([Table S4](#S1){ref-type="supplementary-material"}).
For all CNV loci the maximum zygotic LD is smaller than the gametic LD in value, indicating that no epistatic effects exist between CNV loci. Both gametic and zygotic LD analyses indicate that these CNV loci are essentially in linkage equilibrium except for a few overlapped loci in each population.
Joint migration and mutation rates
----------------------------------
From the pairwise multilocus *G*~*st*(CNV~) ([Table 2](#t2){ref-type="table"}) and the pairwise multilocus *F*~*st*(SNP)~[@b7][@b31], the ratios of the joint migration and nutation rates at CNV loci (*m*~*c*~ + 3*μ*~*c*~/2) to those at SNP sites (*m*~*s*~ + 2*μ*~*s*~) are estimated according to [equations (9)](#eq32){ref-type="disp-formula"} and (12) ([Table 5](#t5){ref-type="table"}). The ratios range from 0.2624 ± 0.0263 (CHB-CHD) to 5.7238 ± 0.0375 (CHD-MKK), with the mean of 3.6600 ± 1.4188 over all pairs. The ratios change from 1.4126 ± 0.0144 (ASW-LWK) to 2.1402 ± 0.0088 (MKK-YRI) in African populations, with the mean of 1.6988 ± 0.3411 over population pairs; from 0.2624 ± 0.0263 (CHB-CHD) to 3.9942 ± 0.0311 (CEU-CHD) in non-African populations, with the mean of 2.6796 ± 1.0224 over population pairs; and from 3.8132 ± 0.0266 (ASW-GIH) to 5.7238 ± 0.0375 (CHD-MKK) among African and non-African populations, with the mean of 4.8157 ± 0.4929 over population pairs.
Using the average pairwise *F*~*st*(SNP)~ = 0.0956 ± 0.0567[@b7][@b31] and the average pairwise *G*~*st*(CNV)~ = 0.0255 ± 0.0114 across all population pairs, we obtain (*m*~*c*~ + 3*μ*~*c*~/2)/(*m*~*s*~ + 2*μ*~*s*~) = 4.0396 ± 3.2341, where a large standard deviation arises from the variation among populations. The above estimates indicate that the joint migration and mutation rates are generally much greater at the genome-wide CNV loci than at the genome-wide SNP sites.
The ratio of the mutation rate to the migration rate at CNV loci can be approximately quantified. According to [equations (13)](#eq37){ref-type="disp-formula"} and [(14)](#eq38){ref-type="disp-formula"}, estimates of *μ*~*c*~/*m* are summarised in [Table 5](#t5){ref-type="table"}, which range from 0.0352 ± 0.0177 (TSI-CEU) to 3.1492 ± 0.0250 (CHB-MEX), with a mean of 1.8153 ± 0.9016 over population pairs (except for a negative value for the CHB-CHD pair). The mutation rate is generally smaller than the migration rate among African populations (0.2392 ± 0.0115\~0.7601 ± 0.0055; [Table 5](#t5){ref-type="table"}), but is greater than the migration rate among non-African populations (1.2036 ± 0.5881) or among African and non-African populations (2.4655 ± 0.4384).
Estimate of *μ*~*c*~/*m* is 2.0264 ± 2.1561 from the rate (*m*~*c*~ + 3*μ*~*c*~/2)/(*m*~*s*~ + 2*μ*~*s*~) = 4.0396 ± 3.2341 in the 11 populations, and 2.0352 ± 2.0909 from (*m*~*c*~ + 3*μ*~*c*~/2)/(*m*~*s*~ + 2*μ*~*s*~) = 4.0529 ± 3.1364 in four populations (CEU, YRI, CHB, and JPT)[@b7][@b8] (average pairwise *F*~*st*(SNP)~ = 0.1265 ± 0.0675; average pairwise *G*~*st*(CNV)~ = 0.0354 ± 0.0158 in the present study). These estimates indicate that the mutation rate at CNV loci is generally about twice as large as the migration rate.
Discussion
==========
Our results indicate a closer population genetic relationship at CNV loci than at SNP sites among 11 HapMap Phase III populations. Previous reports indicate a similar pattern at specific loci among African, European and East Asian populations (HapMap Phase II data)[@b41], or among HapMap Phase II populations (*F*~*st*~ = \~0.11 at the genome-wide SNP sites)[@b42]. A general similarity in relative population genetic structure at CNV loci and SNP sites is also reported with more populations (29) and fewer CNV loci (396) and individuals (405 in total), but the difference is not quantified[@b13]. LD analyses indicate that these CNV loci are essentially in linkage equilibrium except for a few overlapped loci. Epistasis does not exist for any pair of CNV loci, presuming that these CNV loci are not selectively neutral or equally additive in influencing fitness. This result is different from those at the genome-wide SNP sites where epistasis occurs among many intron SNPs[@b37]. The results provide additional support for a recent report indicating that the 856 CNV loci are selectively neutral in each population[@b23]. The evolutionary processes for the low level of population divergences are different from those at the nonsynonumous SNP sites with *F*~*st*~ \< 5% where negative selection is thought to be involved[@b31].
Note that our analyses are based on the three-allele system for describing the evolution at a CNV locus because the maximum number of allele copies is four in a diploid genotype. These 856 CNV loci are shown to exhibit neutrality among 1184 healthy individuals[@b23]. A system of more than three alleles is needed when more than four allele copies occur in a genotype at any CNV locus. This could likely occur when fewer individuals are surveyed or when unhealthy individuals are included because the number of common CNV loci could become fewer with smaller sample sizes. Under this situation, a neutrality test at CNV loci is needed for small sample sizes, and the extent of population genetic divergence could be different from the results reported here. This needs further verification.
Nei's genetic distance at the genome-wide CNV loci is generally comparable to those between human populations at the common protein or blood group loci[@b43]. However, African populations have even smaller genetic divergence at CNV loci. In the process of mutation-drift at the 856 CNV loci[@b23], population differentiation is expected to occur more recently owing to the high mutation rates at CNV loci. The time estimates since divergence are much shorter than those for general population genetic divergence in humans estimated from common protein loci (\~120 Kyrs between human populations[@b43]), or than the postulated time (\>100 Kyrs) for modern humans to leave Africa and colonize the rest of the world. Because the assumption for = 2 *μt*[@b43][@b44] is violated due to the unequal effective population sizes among populations[@b45][@b46], the varying mutation rates among loci, and the finite number of alleles at a CNV locus (not the infinite-allele model)[@b23], the preceding estimates might provide a reference for the minimum divergence times.
Patterns of genetic divergence at CNV loci may reflect the historical divergence in forming modern human origins. The common pattern at both CNV and SNP loci is that the smallest genetic divergence is present among African populations, followed by among non-African populations, and then among African and non-African populations. Polymorphisms at CNV loci decrease from African to non-African populations. More alleles per CNV locus in African populations suggest a longer-term accumulation of mutants. These patterns are consistent with the Out of African model rather than with the multiregional model for modern human origins[@b47][@b48]. Genetic drift effects reduce genetic diversity in non-African populations. Further inferences on the evolutionary processes occurring among non-African populations would require additional information besides the comparison of polymorphisms at CNV loci. Nevertheless, the genetic relationships among non-African populations show a clear separation of Asian populations from non-Asian populations. Evidence at genome-wide CNV loci supports the hypothesis that CHB and CHD have a very close genetic relationship. This is slightly different from the genetic relationships revealed by the patterns of zygotic and gametic LDs at the genome-wide SNP sites where JPT and CHD have a very close genetic relationship[@b37]. Genetic drift effects could explain the relative small differentiation in polymorphism at CNV loci in Asian and European populations. Both CHB and CHD have relatively smaller genetic drift effects than JPT[@b45], and hence have higher polymorphisms (1.50 vs1.48 alleles per CNV locus). CEU probably has relatively smaller genetic drift effects than do CHB and JPT[@b45], and hence has more alleles per CNV locus (1.66 alleles per CNV locus). A relatively high level of polymorphisms in MEX among non-African populations probably arise from an admixture of individuals with multiple distinct ancestries, which is consistent with previous explanations[@b37][@b49].
Because both mutation and migration reduce population genetic divergence[@b50], the combined patterns of genetic divergence at CNV and SNP loci provide us with an opportunity to address their relative roles. Previous reports[@b51] indicate that the mutation rates are about 1.7 × 10^−6^ to 1.0 × 10^−4^, about 100\~10000 times of the point mutation rate at SNP sites (1.8--2.5 × 10^−8^). Fu *et al*.[@b29] indicates that the mutation rate for most CNV loci is about order of 10^−5^ per CNV locus per generation. On average, a mutation rate of the order 10^−5^ at the 856 CNV loci could be inferred from the estimate of the population-scaled mutation rate *θ* (=4 *Nμ*) = 0.1415 ± 0.0144[@b23], given *N* \~ 3000[@b45]. Patterns of the *μ*~*c*~/*m*estimates suggest a dominant role that the mutation process plays in shaping population genetic divergence at CNV loci, especially in the non-African populations ([Table 5](#t5){ref-type="table"}). The low *μ*~*c*~/*m* in African populations could likely arise from their closer genetic relationships where the inter-population gene exchanges are historically more frequent or from natural evolutionary convergence where their genetic compositions become similar since ancestral populations. However, statistical tests indicate that the mutation-drift process can explain the variation at CNV loci in African populations, implying that the latter process could be the main reason for low genetic divergence[@b23].
In comparison with the previous results (*G*~*st*~ \~ 0.11) at a few CNV loci[@b10] (67 CNV loci and *n* = 270 in total) or at the locus of a specific gene CCL4L[@b32] in four HapMap populations (YRI, CEU, and CHB + JPT), our investigation shows much lower population genetic divergence at the 856 CNV loci among these four populations (mean *G*~*st*~ = 0.0345 ± 0.0158; [Table 2](#t2){ref-type="table"}). This result indicates that the CNV loci shared among 1184 healthy individuals exhibit smaller population genetic divergence. Also, compared with the pairwise *F*~*st*~ across chromosomes at the genome-wide SNP sites ([Fig. 2](#f2){ref-type="fig"} in Baye[@b8]), a similarity in pattern at the genome-wide CNV loci exists ([Figure S1](#S1){ref-type="supplementary-material"}). The difference is the presence of low population genetic divergence at CNV loci.
A caveat in the above inferences is that it is based on the assumption of equilibrium among the processes of mutation, drift, and migration at CNV and SNP loci in human populations. Like conventional population genetics analyses in different organisms, such an equilibrium might not be attained in reality, and a dynamic model of evolution is more realistic for further investigation. However, concerning the estimates of , the qualitative conclusion about the major effects of mutation on population genetic divergence cannot be rejected at the genome-wide CNV loci[@b29], especially in non-African populations.
Although small LDs are difficult to detect owing to the statistical power, very few CNV loci exhibit significant gametic and zygotic LDs from either the same or different chromosomes. This is different from the patterns at the genome-wide SNP sites (Hu and Hu[@b37] for zygotic LDs with the recombination rate \<10%, Reich *et al*.[@b36] for gametic LDs with the recombination rate \<16%, and Koch *et al*.[@b52] for gametic LDs with the recombination rate \>25%). The CNV loci on the same chromosomes (except a few overlapped loci) are distributed over a wide range of distances, with an average recombination rate of 3.3% (0\~35%). The significant correlations among CNV loci do not exist across populations[@b53]. The generally concordant pattern of no significant gametic and zygotic LDs provides no evidence for the presence of functionally epistatic CNV loci[@b26][@b27], different from the results at genome-wide SNP sites[@b37].
Patterns of LDs also suggest that the effects of mutation on reducing LDs are stronger than the effects of migration that increases LDs. The gametic LDs at CNV loci gradually decay with time in African populations, and the same is the case for the zygotic LDs at CNV loci[@b53], except for the overlapped CNV loci (but not for one pair of CNV loci on Chr 7 with a physical distance of 2658 bp that requires a longer time to decay). The gametic LDs at CNV loci initially formed by the founder effects in non-African populations also decay with time due to the mutation and recombination effects. The same is the case for the zygotic LDs[@b38]. If recombination is the dominant process in eroding LDs, a certain proportion of CNV loci could maintain significant LD within very short distances except for overlapped loci. Such an expected pattern is not observed ([Tables S4](#S1){ref-type="supplementary-material"} and [S5](#S1){ref-type="supplementary-material"}). High mutation rates causing low LDs between CNV and SNP loci are also discussed[@b54]. Thus, the mutation effects could be greater than the recombination effects in eroding both gametic and zygotic LDs although recombination and mutation effects are both involved in reducing LDs[@b55].
Finally, our investigation suggests differential evolutionary processes at CNV and SNP loci along chromosomes. Although mosaic patterns occur in genome architecture in terms of different measures of genetic diversity or from different perspectives[@b53], the DNA segments with CNV loci themselves display individual blocks each with a small level of population genetic divergence. These blocks are different from the gametic or zygotic LD blocks at SNP sites since recombination within CNV loci should rarely occur. The LD blocks between CNV loci cannot be maintained due to the effects of the high mutation rates.
Methods
=======
Genotype data at CNV loci
-------------------------
Genotype data in 11 HapMap Phase III populations, released by The International HapMap 3 Consortium, was downloaded from <ftp://ftp.ncbi.nlm.nih.gov/hapmap/cnv_data/hm3_cnv_submission.txt.> The data differs from most accessible data sets in that it provides the discrete copy numbers per CNV locus. The copy numbers at a CNV locus are derived through a two-step process according to Altshuler *et al*.[@b31] The first step is to detect copy number variation on each chromosome by analyzing the probe-level intensity data from both the Affymetrix and Illumina arrays. QuantiSNP[@b56] and Birdseye[@b57] algorithms are used to identify CNV loci separately. Common CNV loci are further identified, and refined to ensure qualified copy number variant calls. The second step is to determine the discrete copy numbers for each CNV locus from the probe-level intensity data. CNVtools[@b33] and a two-dimensional model (Gaussian mixture)[@b31], are used to infer the copy numbers from the maximum posterior likelihood function. A meta-approach combining the two algorithms and other criteria are used to further refine the discrete copy number classes to ensure reliable copy number estimates per diploid genomes. This second step for estimating the copy number per CNV locus is not conducted in most archived CNV data sets although later techniques for CNV locus detection are now more advanced.
Diploid genotypes were recorded in integers (0, 1, 2, 3, and 4): 0 for the genotype without any allele copy in both gametes, 1 for the genotype with one allele copy in one gamete but without any copy in the other gamete, 2 for the genotype with one allele copy in each gamete, 3 for the genotype with one allele copy in one gamete and two allele copies in the other gamete, and 4 for the genotype with two allele copies in each gamete. From the individual IDs in the HapMap project, eleven populations were extracted from the pooled data (hm3_cnv_submission.txt): ASW (African ancestry in Southwest USA), CEU (Utah residents with Northern and Western European ancestry from the CEPH collection), CHB (Han Chinese in Beijing, China), CHD (Chinese in Metropolitan Denver, Colorado), GIH (Gujarati Indians in Houston, Texas), JPT (Japanese in Tokyo, Japan), LWK (Luhya in Webuye, Kenya), MEX (Mexican ancestry in Los Angeles, California), MKK (Maasai in Kinyawa, Kenya), TSI (Toscans in Italy), and YRI (Yoruba in Ibadan, Nigeria). Sample size for each population is shown in [Table 1](#t1){ref-type="table"}. The number of CNV loci per Chr ranges from 11 on Chr 22 to 68 on Chr 2, with 856 common CNV loci in total. Mean size of CNV loci per Chr is \~0.02 Mb, ranging from 26 to 456897 bp. The physical distance between adjacent CNV loci per Chr is \~3.3 Mb on average, ranging from 0 (partially overlapped loci) to 34804235 bp. There are 29 CNV loci that are partially overlapped on chromosomes.
Allele frequency
----------------
Because the maximum number of allele copies is four at a CNV locus in the diploid genotype dataset of HapMap Phase III populations, a three-allele system is used to describe the genotype composition. Note that a system of more than three alleles is needed if the number of allele copies is more than 4 in a diploid genotype[@b23][@b58]. Let *A*~0~, *A*~1~, and *A*~2~ be the alleles with 0-, 1-, and 2-copies at a CNV locus, respectively. Allele *A*~1~ may be the most abundant variant in a population (the segment on the reference genome), while alleles *A*~0~ and *A*~2~ are likely less abundant at a CNV locus. Owing to lack of information needed to separate distinct genotypes with the same copy numbers in diploids, allele frequencies under Hardy-Weinberg equilibrium (HWE) were estimated using the expectation-maximization (EM)[@b23][@b29][@b59][@b60]. Polymorphism was measured in terms of the number of observed alleles per CNV locus (*N*~*a*~), the percentage of polymorphic loci, *P*(99%), and the genetic diversity in a population ( where *p*~*u*~ is the *u*th allele frequency) which is equal to the expected heterozygosity (*H*~e~) under HWE.
Genetic divergence
------------------
Population genetic differentiation was measured by *G*~*st*~[@b44]: *G*~*st*~ = 1 − *H*~*s*~/*H*~*t*~ where *H*~s~ is the mean of the expected heterozygosity (*H*~*e*~) per locus over all populations and *H*~*t*~ is the expected heterozygosity per locus in the pooled population. The 95% confidence intervals (CIs) for *G*~*st*~ was derived using the bootstrapping approach. To relate the population genetic differentiation to the time since the populations diverge from a single ancestral population, genetic distance was measured[@b46]. This distance develops under a specific evolutionary processes. Nei's genetic distance[@b44] was used to measure population genetic divergence: *D* = −ln(*I*) where in which *p*~*lu*1~ and *p*~*lu*2~ are the frequencies of alleles *u*1 and *u*2 at the *l*th locus from populations 1 and 2, respectively. Under the neutral process (mutation and genetic drift), Nei's genetic distance is linearly related to the time since divergence (*t*), i.e. [@b44][@b46], and its approximate variance *V(t*) = *V(D*)/4*μ*^2^, given a mutation rate *μ*. Standard deviations for *G*~*st*~ and Nei's genetic distance were calculated using the jackknife method[@b46].
LD tests
--------
To assess the properties of CNV loci relevant for interpreting population genetic divergence, both the gametic and zygotic LDs were tested in each population. Assuming that CNV loci are involved in fitness, a comparison of gametic LD with the maximum zygotic LD in value can be used to determine whether epistasis occurs or not among loci[@b37][@b38][@b39]. If the maximum zygotic LD (high order LD) is greater than the gametic LD (low order) in value, epistasis exists between loci, which otherwise does not occur (additive or neutral effects). This relationship has been applied to analyzing genome-wide SNP sites[@b37], providing the evidence of epistasis among many intron SNP sites in each of the 11 populations. For a pair of CNV loci each with three alleles, there are 9 types of two-non-allele gametes. Let *d*~*ij*~ (*i, j* = 0, 1, 2) be the gametic LD between allele *i* at the first locus and allele *j* at the second locus, and *p*~*ij*~ be the gametic frequency in the population. MLE of the frequency of a genotype pair, (*s, t* = 0, 1, 2, 3, 4), can be obtained using the direct counting method. An EM method is used to estimate the gametic frequency through an iterative calculation, which is described below:
where *δ*~*ij*~, a Kronecker delta variable, is equal to 1 when *i* = *j*, and 0 when *i*≠*j*. Note that the E- and M-steps are combined into one formula in [equation (1)](#eq8){ref-type="disp-formula"}. Thus, given the initial gametic frequency *p*~*ij*~ (*i, j* = 0, 1, 2), the gametic frequency at the next step *p*′~*uv*~ can be calculated using [equation (1)](#eq8){ref-type="disp-formula"}. Then, replace *p*~*ij*~ in [equation (1)](#eq8){ref-type="disp-formula"} with *p*′~*uv*~ and recalculate *p*′~*uv*~ at the next step. This iterative calculation is repeated until the convergence of gametic frequencies is attained.
The gametic LD, *d*~*ij*~, is then estimated as where and are the MLEs of the frequencies of allele *i* at the first locus and allele *j* at the second locus, respectively. A chi-square statistic with 1 degree of freedom (df) is used to test H~0~: *d*~*ij*~ = 0 [@b46], i.e.
R-square, , is used to measure gametic LD, which ranges from 0 to 1. [Appendix S1](#S1){ref-type="supplementary-material"} gives the power calculation for the gametic LD test. The power tends to a concave upward curve as the allele frequency increases because the variance under H~0~ or under H~1~ (*d*~*ij*~ ≠ 0) has a maximum value at the intermediate allele frequencies. A large variance increases the uncertainty and hence reduces the power, given a sample size (*n*), a significance level (*α*), and gametic LD. The power also increases as the sample size or the gametic LD increases.
Let *D*~*ij*~ be the zygotic LD between genotypes *i* at the first locus and *j* at the second locus (*i, j* = 0, 1, 2, 3, 4) in the population. The MLE of zygotic LD, , from the sample of size *n* can be obtained by where is the MLE of the joint frequency of genotypes *i* at the first locus and *j* at the second locus, and (or ) is the frequency of genotype *i* (or *j*). To test H~0~: *D*~*ij*~ = 0, a chi-square statistic with 1 df is set as
The normalized r-square is set as , which ranges from 0 to 1[@b37][@b39][@b61]. [Appendix S2](#S1){ref-type="supplementary-material"} derives the power calculation for the zygotic LD test. Similarly, the power increases as the sample size or the zygotic LD increases. The power may be relatively lower for testing zygotic LD than for testing gametic LD due to the doubling of sample size in gametic LD tests.
The significance tests of gametic and zygotic LDs were conducted at the genome-wide level in each population, and hence a Bonferroni adjusted p-value was set as 0.05/the number of all pairs of CNV loci across 22 chromosomes, ranging from 1.88 × 10^−7^\~6.91 × 10^−7^ owing to different numbers of polymorphic loci in the 11 populations. To minimize the impacts of minor allele frequency (MAF) on amplifying gametic LD test or on increasing false-positive errors, those alleles with their frequencies being out of the range \[0.05, 0.95\] in the samples were excluded in testing gametic LD. For the same reason, those genotypes with genotypic frequencies beyond the range \[0.05, 0.95\] in the samples were excluded in testing zygotic LD. Sample sizes ranging from 77 to 171 can provide appropriate statistical power for genotypic frequencies within the range \[0.05, 0.95\] ([Appendix S2](#S1){ref-type="supplementary-material"}). Since the constraints and hold, only four gametic LDs and sixteen zygotic LDs were tested for each pair of CNV loci. Note that CNV loci were not filtered out by frequency except in this LD analysis.
Joint mutation and migration rates
----------------------------------
Consider a neutral CNV locus with three alleles. Let *μ*~*c*~ be the mutation rate of one allele to any of the other two alleles at a CNV locus. The probability density distribution (pdf) for the allele frequency under an equilibrium among genetic drift, mutation, and migration effects can be approximated by synthesizing Kimura's[@b19] and Wright's[@b50] work, i.e.
where *N* is the effective population size, *m*~*c*~ is the migration rate per generation for an allele at a CNV locus, *Q* is the migrant allele frequency, and *θ*~c~ (aka "population diversity") is the population-scaled mutation rate (=4*Nμ*~c~). *F*~*st*~ per locus is derived as
The practical population differentiation with *F*~*st*~[@b62] is measured by *G*~*st*~[@b44] for a three-allele locus.
Similarly, the pdf of allele frequency at a bi-allelic SNP locus under an equilibrium among genetic drift, mutation, and migration effects can be approximated by synthesizing Kimura's[@b19] and Wright's[@b50] work,
where *m*~*s*~ is the migration rate per generation, *Q* is the migrant allele frequency, and *θ*~s~ is equal to 4*Nμ*~s~ in which *μ*~s~ is the mutation rate at an SNP locus. *F*~*st*~ per locus is derived as
The relative extent of genetic divergence at the genome-wide SNP sites versus at the genome-wide CNV loci is measured by the ratio of *F*~*st(SNP*)~/*G*~*st(CNV*)~, and its standard deviation can be estimated from the variance approximation:
where and are the means of *F*~*st*(SNP)~ and *G*~*st*(CNV)~, respectively, and *cov(F*~*st*(SNP)~, *G*~*st*(CNV)~) is the covariance between *F*~*st(SNP)*~ and *G*~*st(CNV)*~. The above expression is derived by the delta method[@b63]. Estimate of the variance of the ratio can be approximated by assuming that the covariance, cov(*F*~*st*(SNP)~, *G*~*st*(CNV)~) is negligible at the genome-wide scale. Correlations between CNV and SNP loci are weak, which could arise from the effects of transposition events, recurrent mutation/reversions, or the preference of CNV loci at the low density of SNP sites on chromosomes[@b12][@b54].
From [equations (5](#eq26){ref-type="disp-formula"}) and ([7](#eq28){ref-type="disp-formula"}), the ratio of the joint migration and nutation rates at CNV loci to those at SNP sites is estimated as
Similarly, the variance of this ratio can be estimated using the delta method[@b51]. Let *X* = *F*~*st(SNP*)~(1 − *G*~*st(CNV*)~) and *Y* = *G*~*st(CNV*)~(1 − *F*~*st(SNP*)~). Again, assume that cov(*F*~*st(SNP)*~, *G*~*st(CNV)*~) is neglected at the genome-wide scale. The variance *V(X*) is given by
*V(Y*) can be obtained by replacing *F*~*st(SNP)*~ and 1 − *G*~*st(CNV)*~ in [equation (10)](#eq33){ref-type="disp-formula"} with *G*~*st(CNV)*~ and 1 − *F*~*st(SNP)*~, respectively. Similarly, *V(XY*) can be obtained by replacing 1 − *G*~*st(CNV)*~ in [equation (10)](#eq33){ref-type="disp-formula"} with *G*~*st(CNV*)~. The covariance cov(*X, Y*) is given by
The variance of the ratio *V(X*/*Y*) can be estimated from the following expression,
The variance can be appropriately estimated by *V(X*/*Y*) in [equation (12)](#eq35){ref-type="disp-formula"}, especially when the sample sizes are large.
It is appropriate to assume that the migration rate is the same, on average, at the neutral CNV and SNP loci (*m*~*c*~ = *m*~*s*~ = *m*) although local variation might occur among loci (e.g., due to the genetic hitchhiking effects). Also, compared with the migration rate, the point mutation rate at the SNP sites can be neglected. Thus, the ratio of the mutation rate to the migration rate at CNV loci can be estimated:
The standard deviation of the *μ*~*c*~/*m* estimate can be obtained according to [equation (12)](#eq35){ref-type="disp-formula"}, i.e.
Additional Information
======================
**How to cite this article**: Hu, X.-S. *et al*. High mutation rates explain low population genetic divergence at copy-number-variable loci in *Homo sapiens. Sci. Rep.* **7**, 43178; doi: 10.1038/srep43178 (2017).
**Publisher\'s note:** Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Supplementary Material {#S1}
======================
###### Supplementary Dataset 1
###### Supplementary Dataset 2
###### Supplementary Dataset 3
###### Supplementary Tables, Appendices and Figure S1
We sincerely appreciate Inke König and three anonymous reviewers for very helpful comments on this article. The work is supported by the startup funding from South China Agricultural University to XSH (4400-K16013), and by the Forest Sciences and Technology Innovation Project in Guangdong Province to XYC (2015KJCX009).
The authors declare no competing financial interests.
**Author Contributions** X.S.H. conceived and designed the study. X.S.H. analyzed data and wrote the manuscript. Y.H. analyzed the data. L.T.D. provided logistic assistance. F.C.Y., R.A.E. and X.Y.C. revised the manuscript. All authors approved the manuscript.
{#f1}
{#f2}
{#f3}
###### Sample sizes and polymorphisms at the genome-wide CNV loci in 11 human populations.
Populations Sample sizes *P*(99%)[\*](#t1-fn1){ref-type="fn"} *N*~a~ Mean *H*~e~ ± *S*~d~
------------- -------------- -------------------------------------- -------- ----------------------
ASW 83 85.16 1.90 0.1290 ± 0.1457
CEU 165 57.71 1.62 0.1123 ± 0.1622
CHB 84 46.14 1.50 0.1072 ± 0.1651
CHD 85 46.61 1.50 0.1047 ± 0.1636
GIH 88 53.85 1.58 0.1121 ± 0.1629
JPT 86 44.51 1.48 0.1079 ± 0.1669
LWK 90 80.49 1.85 0.1268 ± 0.1483
MEX 77 62.38 1.66 0.1130 ± 0.1599
MKK 171 83.53 1.88 0.1248 ± 0.1522
TSI 88 56.78 1.60 0.1123 ± 0.1626
YRI 167 80.14 1.84 0.1322 ± 0.1496
^\*^*P*(99%): the percentage of polymorphic loci where the frequency of the most common allele was ≤0.99.
*N*~a~: the number of observed alleles per CNV locus; *H*~*e*~: the expected heterozygosity under Hardy-Weinberg equilibrium.
###### Comparison of the pairwise *G*~*st*(CNV)~ at the genome-wide CNV loci with the pairwise *F*~*st*(SNP)~ at the genome-wide SNP sites[@b7].
ASW CEU CHB CHD GIH JPT LWK MEX MKK TSI YRI
----- ----------------- ----------------- ----------------- ------------------ ------------------ ------------------ ------------------ ------------------ ------------------ ------------------ ------------------
ASW 0.0275 (0.0001) 0.0357 (0.0001) 0.0353 (0.0001) 0.0267 (0.0001) 0.0336 (0.0001) 0.0071 (0.00001) 0.0258 (0.0001) 0.0087 (0.00001) 0.0255 (0.00005) 0.0061 (0.00001)
CEU 3.7081 (0.0230) 0.0307 (0.0001) 0.0307 (0.0001) 0.0140 (0.00004) 0.0331 (0.0001) 0.0348 (0.0001) 0.0116 (0.00003) 0.0248 (0.00005) 0.0038 (0.00001) 0.0375 (0.0001)
CHB 3.9697 (0.0210) 3.5964 (0.0251) 0.0038 (0.00001) 0.0289 (0.0001) 0.0063 (0.00002) 0.0413 (0.0001) 0.0246 (0.0001) 0.0295 (0.0001) 0.0332 (0.0001) 0.0401 (0.0001)
CHD 4.0432 (0.0213) 3.6545 (0.0253) 0.2649 (0.0265) 0.0280 (0.0001) 0.0073 (0.00003) 0.0415 (0.0001) 0.0243 (0.0001) 0.0295 (0.0001) 0.0326 (0.0001) 0.0403 (0.0001)
GIH 3.5497 (0.0200) 2.4946 (0.0227) 2.6315 (0.0221) 2.7458 (0.0229) 0.0304 (0.0001) 0.0333 (0.0001) 0.0147 (0.00004) 0.0221 (0.00005) 0.0139 (0.00004) 0.0328 (0.0001)
JPT 3.9680 (0.0209) 3.3981 (0.0234) 1.1119 (0.0163) 1.0957 (0.0144) 2.5390 (0.0181) 0.0421 (0.0001) 0.0260 (0.0001) 0.0306 (0.0001) 0.0352 (0.0001) 0.0406 (0.0001)
LWK 1.4017 (0.0142) 4.1900 (0.0242) 4.2389 (0.0187) 4.2400 (0.0210) 3.9617 (0.0195) 4.1856 (0.0207) 0.0331 (0.0001) 0.0082 (0.00002) 0.0332 (0.0001) 0.0059 (0.00001)
MEX 3.6383 (0.0206) 2.6817 (0.0118) 2.0810 (0.0217) 2.9173 (0.0221) 2.3857 (0.0152) 2.6898 (0.0205) 4.0130 (0.0196) 0.0214 (0.0001) 0.0112 (0.00003) 0.0330 (0.0001)
MKK 1.6680 (0.0576) 4.1655 (0.0217) 4.8353 (0.0256) 4.8624 (0.0258) 4.2751 (0.0243) 4.7094 (0.0248) 2.0688 (0.0129) 4.4724 (0.0255) 0.0206 (0.00004) 0.0128 (0.00002)
TSI 3.8807 (0.0210) 1.0415 (0.0262) 3.3390 (0.0231) 3.4390 (0.0236) 2.4423 (0.0159) 3.1983 (0.0219) 4.2654 (0.0225) 2.8533 (0.0195) 4.7502 (0.0307) 0.0334 (0.0001)
YRI 1.5206 (0.0662) 4.1929 (0.0203) 4.6181 (0.0199) 4.6174 (0.0199) 4.3656 (0.0230) 4.5913 (0.0242) 1.3481 (0.0171) 4.3467 (0.0230) 2.1023 (0.0087) 4.5808 (0.0201)
The above diagonal values are the mean multilocus *G*~*st*(CNV)~ estimates, and the below diagonal values are the ratios of *F*~*st*(SNP)~/*G*~*st*(CNV)~. Standard deviations are shown in parentheses.
###### Means and standard deviations of significant gametic LDs (r-squares) in 11 human populations[\*](#t3-fn1){ref-type="fn"}.
ASW CEU CHB CHD GIH JPT LWK MEX MKK TSI YRI
--------- ------------- ------------- ------------- ------------- ------------- ------------- ------------- ------------- ------------- ------------- -------------
*d*~00~ 0.031% 0.053% 0.045% 0.054% 0.034% 0.041% 0.027% 0.044% 0.034% 0.031% 0.045%
0.84 ± 0.32 0.78 ± 0.32 0.79 ± 0.33 0.72 ± 0.35 0.75 ± 0.34 0.76 ± 0.34 0.80 ± 0.34 0.79 ± 0.33 0.68 ± 0.43 0.86 ± 0.24 0.64 ± 0.44
21(16:5) 28(17:11) 26(16:10) 30(16:14) 27(16:11) 25(15:10) 21(15:6) 28 (17:11) 27(16:11) 25(17:8) 27(15:12)
0.21 ± 0.04 0.09 ± 0.02 0.18 ± 0.03 0.17 ± 0.01 0.19 ± 0.02 0.18 ± 0.02 0.18 ± 0.04 0.21 ± 0.04 0.10 ± 0.02 0.18 ± 0.03 0.10 ± 0.02
59 37 9 13 9 5 42 34 60 11 77
*d*~01~ 0.035% 0.058% 0.050% 0.062% 0.037% 0.047% 0.032% 0.048% 0.038% 0.033% 0.050%
0.79 ± 0.32 0.78 ± 0.31 0.80 ± 0.32 0.74 ± 0.34 0.80 ± 0.30 0.75 ± 0.33 0.80 ± 0.33 0.80 ± 0.39 0.65 ± 0.42 0.85 ± 0.23 0.63 ± 0.43
23(18:5) 30(19:11) 28(18:10) 32(18:14) 27(18:9) 28(17:11) 23(17:6) 29(19:10) 28(16:12) 26(18:8) 29(17:12)
0.21 ± 0.04 0.10 ± 0.01 0.20 ± 0.03 0.17 ± 0.01 0.19 ± 0.03 0.18 ± 0.02 0.19 ± 0.04 0.21 ± 0.04 0.10 ± 0.02 0.17 ± 0.03 0.10 ± 0.02
68 41 11 17 12 6 52 40 69 13 89
*d*~10~ 0.034% 0.064% 0.057% 0.062% 0.044% 0.051% 0.029% 0.048% 0.040% 0.041% 0.051%
0.82 ± 0.32 0.75 ± 0.34 0.75 ± 0.34 0.71 ± 0.35 0.76 ± 0.33 0.74 ± 0.34 0.77 ± 0.36 0.78 ± 0.32 0.68 ± 0.42 0.83 ± 0.26 0.64 ± 0.44
23(17:6) 31(18:13) 27(16:11) 31(16:15) 28(17:11) 27(16:11) 22(15:7) 29(18:11) 27(16:11) 27(18:9) 27(15:12)
0.22 ± 0.09 0.13 ± 0.17 0.27 ± 0.23 0.21 ± 0.16 0.23 ± 0.15 0.26 ± 0.26 0.18 ± 0.04 0.23 ± 0.11 0.10 ± 0.02 0.21 ± 0.14 0.11 ± 0.07
68 47 17 18 19 10 47 40 76 21 92
*d*~11~ 0.041% 0.071% 0.064% 0.073% 0.048% 0.059% 0.034% 0.056% 0.045% 0.047% 0.058%
0.79 ± 0.32 0.74 ± 0.34 0.74 ± 0.35 0.72 ± 0.34 0.80 ± 0.30 0.74 ± 0.33 0.78 ± 0.34 0.82 ± 0.30 0.64 ± 0.44 0.82 ± 0.27 0.65 ± 0.43
28(19:9) 34(20:14) 31(19:13) 35(19:16) 29(19:10) 31(18:13) 25(17:8) 30(20:10) 29(16:13) 30(19:11) 33(18:15)
0.22 ± 0.08 0.13 ± 0.34 0.26 ± 0.21 0.20 ± 0.12 0.22 ± 0.30 0.24 ± 0.24 0.19 ± 0.04 0.22 ± 0.09 0.10 ± 0.01 0.20 ± 0.13 0.11 ± 0.07
80 53 19 23 22 12 55 49 85 25 104
^\*^The percentages in the same row as *d*~*ij*~ (*i, j = 0, 1*) in the table are the proportions of significant gametic LDs among all pairs of LD tests. The data in the second row under each *d*~*ij*~ is the gametic LD among CNV loci from the same chromosomes. The data in the third row under each *d*~*ij*~ is the observed numbers of pairs with significant LDs from the same chromosomes (non-overlapped locus pairs: overlapped locus pairs). The data in the fourth row under each *d*~*ij*~ is the significant gametic LDs among CNV loci from different chromosomes. The data in the fifth row under each *d*~*ij*~ is the observed numbers of pairs with significant LDs among CNV loci from different chromosomes.
###### Percentages of the pairs of CNV loci with significant zygotic LDs in 11 human populations[\*](#t4-fn1){ref-type="fn"}.
LD ASW CEU CHB CHD GIH JPT LWK MEX MKK TSI YRI
--------- -------- -------- -------- -------- -------- -------- -------- -------- -------- -------- --------
*D*~00~ 0.0045 0.0082 0.0141 0.0189 0.0123 0.0193 0.0046 0.0063 0.0039 0.0093 0.0067
*D*~01~ 0.0041 0.0066 0.0077 0.0101 0.0047 0.0083 0.0025 0.0042 0.0031 0.0059 0.0034
*D*~02~ 0.0019 0.0033 0.0026 0.0025 0.0019 0.0028 0.0013 0.0021 0.0008 0.0025 0.0009
*D*~03~ 0.0008 0.0016 0.0026 0.0025 0.0019 0.0028 0.0008 0.0014 0.0008 0.0017 0.0009
*D*~10~ 0.0041 0.0115 0.0116 0.0126 0.0075 0.0124 0.0038 0.0063 0.0043 0.0085 0.0051
*D*~11~ 0.0143 0.0271 0.0308 0.0327 0.0226 0.0332 0.0143 0.0253 0.0129 0.0221 0.0111
*D*~12~ 0.0113 0.0222 0.0218 0.0214 0.0151 0.0276 0.0114 0.0211 0.0125 0.0161 0.0098
*D*~13~ 0.0023 0.0025 0.0026 0.0025 0.0019 0.0041 0.0008 0.0014 0.0027 0.0017 0.0013
*D*~20~ 0.0034 0.0057 0.0051 0.0050 0.0029 0.0069 0.0008 0.0028 0.0008 0.0042 0.0013
*D*~21~ 0.0143 0.0246 0.0244 0.0264 0.0160 0.0304 0.0105 0.0232 0.0129 0.0178 0.0111
*D*~22~ 0.0139 0.0279 0.0283 0.0252 0.0160 0.0359 0.0127 0.0246 0.0145 0.0187 0.0128
*D*~23~ 0.0023 0.0016 0 0 0 0.0014 0.0004 0 0.0020 0 0.0013
*D*~30~ 0.0004 0.0025 0 0.0013 0 0 0 0.0007 0 0 0
*D*~31~ 0.0026 0.0049 0 0.0038 0.0019 0.0069 0.0017 0.0035 0.0024 0.0025 0.0026
*D*~32~ 0.0041 0.0049 0.0051 0.0038 0.0047 0.0069 0.0021 0.0049 0.0016 0.0034 0.0034
*D*~33~ 0.0015 0 0 0 0 0.0014 0.0008 0.0007 0.0004 0.0008 0.0013
^\*^*D*~*ij*~ is the zygotic LD between genotype *i* at the first locus and *j* at the second locus (*i, j* = 0, 1, 2, 3).
###### Ratios of the joint mutation and migration rates at CNV loci to those at SNP sites (above diagonal), and the ratios of the mutation rate to the migration rate at CNV loci (below diagonal).
ASW CEU CHB CHD GIH JPT LWK MEX MKK TSI YRI
----- ----------------- ----------------- --------------------------------- ----------------- ----------------- ----------------- ----------------- ----------------- ----------------- ----------------- -----------------
ASW 4.008 (0.0303) 4.4667 (0.0288) 4.5564 (0.0293) 3.8132 (0.0266) 4.8110 (0.0312) 1.4126 (0.0144) 3.9084 (0.0278) 1.6765 (0.0587) 4.1896 (0.0250) 1.5129 (0.0664)
CEU 2.0054 (0.0202) 3.9223 (0.0309) 3.9942 (0.0311) 2.5468 (0.0239) 3.6992 (0.0284) 4.7303 (0.0335) 2.7259 (0.0115) 4.5349 (0.0262) 1.0528 (0.0266) 4.7910 (0.0286)
CHB 2.3111 (0.0192) 1.9482 (0.0206) 0.2624 (0.0263) 2.7677 (0.0256) 1.1119 (0.0164) 4.9274 (0.0269) 2.9478 (0.0260) 5.7238 (0.0375) 3.6286 (0.0282) 5.4445 (0.0289)
CHD 2.3709 (0.0195) 1.9962 (0.0207) ---[\*](#t5-fn1){ref-type="fn"} 2.8878 (0.0266) 1.0967 (0.0145) 4.9298 (0.0299) 3.0640 (0.0266) 5.5163 (0.0368) 3.7503 (0.0289) 5.4487 (0.0288)
GIH 1.8755 (0.0177) 1.0312 (0.0159) 1.1785 (0.0171) 1.2586 (0.0178) 2.6720 (0.0208) 4.4186 (0.0269) 2.4310 (0.0159) 4.6233 (0.0290) 2.4969 (0.0168) 4.9364 (0.0321)
JPT 2.5407 (0.0208) 1.7994 (0.0189) 0.0746 (0.0109) 0.0644 (0.0097) 1.1147 (0.0139) 4.8732 (0.0294) 2.8197 (0.0244) 5.3380 (0.0352) 3.4744 (0.0264) 5.4210 (0.0350)
LWK 0.2751 (0.0096) 2.4869 (0.0223) 2.6183 (0.0180) 2.6199 (0.0199) 2.2790 (0.0179) 2.5822 (0.0196) 4.4772 (0.0272) 2.0917 (0.0135) 4.7997 (0.0314) 1.3588 (0.0173)
MEX 1.9390 (0.0185) 1.1506 (0.0077) 1.2985 (0.0173) 1.3760 (0.0177) 0.9540 (0.0106) 1.2131 (0.0162) 2.3181 (0.0181) 4.8450 (0.0363) 2.9185 (0.0204) 4.9055 (0.0319)
MKK 0.4510 (0.0391) 2.3566 (0.0175) 3.1492 (0.0250) 3.0109 (0.0246) 2.4155 (0.0194) 2.8920 (0.0235) 0.7278 (0.0090) 2.5633 (0.0242) 5.1655 (0.0365) 2.1402 (0.0088)
TSI 2.1264 (0.0167) 0.0352 (0.0177) 1.7524 (0.0188) 1.8335 (0.0193) 0.9980 (0.0112) 1.6496 (0.0176) 2.5331 (0.0210) 1.2790 (0.0136) 2.7770 (0.0244) 5.2357 (0.0291)
YRI 0.3419 (0.0443) 2.5273 (0.0190) 2.9630 (0.0193) 2.9658 (0.0192) 2.6243 (0.0214) 2.9473 (0.0234) 0.2392 (0.0115) 2.6037 (0.0213) 0.7601 (0.0059) 2.8238 (0.0194)
^\*^negative value.
Standard deviations are shown in parentheses.
| {
"pile_set_name": "PubMed Central"
} |
Introduction
============
Electrochemotherapy is a local treatment of cancer which combines the use of a medical device with pharmaceutical agents to achieve local tumor control in solid cancers. The procedure consists of applying short high-intensity pulsed electric fields to cells, in response to which, the plasma membrane's permeability to various molecules transiently increases. This facilitates cellular uptake of cytotoxic agents, thus increasing their cytotoxicity \[[@B1]-[@B4]\]. The treatment is based on the phenomenon termed electroporation, which occurs, when externally delivered electric field induces a sufficiently large transmembrane voltage. Electroporation is, in addition to its use in electrochemotherapy, used also as non-viral gene delivery method to cells *in vitro* and *in vivo* -- gene electrotransfer \[[@B5],[@B6]\]. Furthermore, electroporation as sole modality can be used as tumor ablation in the form of irreversible electroporation, also referred to as non-thermal irreversible electroporation \[[@B7]-[@B14]\].
Electrochemotherapy uses electroporation to allow increased uptake of chemotherapeutic drugs into tumor cells \[[@B2],[@B15]\]. Since the initial research *in vitro*, which was performed more than 20 years ago, electrochemotherapy has gone through extensive basic research on the mechanisms of its action, the required electric field, pulse parameters \[[@B16]\] and sequences for successful treatment, screening of appropriate drugs for use with electric pulses \[[@B17],[@B18]\], *in vivo* studies on different animal models \[[@B19]\], as well as clinical trials in cutaneous and subcutaneous tumors \[[@B19],[@B20]\] and deep-seated tumors \[[@B21]-[@B23]\]. The aim of this review is to give the reader a concise overview over the development of this biomedical engineering-based cancer treatment modality. Its foundation and discussion is critical to guide the development of electrochemotherapy, as adequate preclinical evidence for any kind of medical treatment requires significant time and financial resources. This present paper provides an overview of the current status of electrochemotherapy and sets the groundwork for discussion on further development and standardization of the technique in relation to other local and/or ablation treatment modalities.
This review is divided into several sections comprising the following topics:
• Overview of preclinical research and clinical data on use of electrochemotherapy in treatment of cutaneous and subcutaneous tumors. Implementation into standard of care, focusing on cutaneous and subcutaneous metastases.
• Other uses of electrochemotherapy, such as treatment of visceral and deep-seated tumors.
• The last section outlines the challenges for future studies which need to be addressed, taking into account important aspects of cancer treatment, such as improvement in patient's quality of life and in time to progression or overall survival.
Physical/theoretical background of electrochemotherapy
------------------------------------------------------
Electrochemotherapy relies on using hydrophilic drugs in combination with application of high-voltage electric pulses applied to tumors. The two drugs that have been used most often in electrochemotherapy are bleomycin (BLM), and cisplatin (*cis*-diamminedichloroplatinum (II) -- CDDP). Both of these drugs are membrane low- or non-permeable, and their intracellular target is DNA \[[@B24]\]. The applied electric pulses are most commonly delivered in trains of eight 100-μs-long pulses. By using these electric pulses, the efficacy of the chemotherapeutic drugs is potentiated significantly. The potentiation is most expressed for BLM \[[@B2]\] and CDDP \[[@B25]\]. Potentiation of other drugs has also been studied \[[@B17],[@B18]\] but proved to be much lower or insignificant (Table [1](#T1){ref-type="table"}).
######
**Summary of drugs tested for*in vitro*and*in vivo*potentiation in combination with electroporation pulses**
**Drug tested** ***In vitro*potentiation** ***In vivo*potentiation**
--------------------------------------------- ----------------------------------------------------------- ---------------------------
Bleomycin Yes; 100--5000-fold \[[@B1],[@B17],[@B18],[@B26]-[@B28]\] Yes
Cisplatin Yes; 1.8--12.2-fold \[[@B17],[@B18],[@B28],[@B29]\] Yes \[[@B28]-[@B30]\]
Calcium Yes; more hundred-fold \[[@B31],[@B32]\] Yes \[[@B31]\]
Netropsin Yes; 200-fold \[[@B1]\] --
Carboplatin Yes; 1.6--13-fold \[[@B17],[@B18],[@B27],[@B29]\] --
2-*N*-methyl-9-hydroxy-ellipticinium (NMHE) Yes; 4-fold \[[@B1]\] --
Vincristine Yes; 1.3--3.4-fold \[[@B18]\] --
ActiNomycin D Yes; 2--3-fold \[[@B1]\] --
Cytarabine Yes; 2-fold \[[@B18]\] --
Oxaliplatin Yes \[[@B29]\] --
Platinum (II) complex 3P-SK Yes \[[@B29]\] Yes \[[@B29]\]
Platinum (II) complex PtAMP Yes \[[@B29]\] --
Mitomycin C Yes but low; 1.3--1.4-fold \[[@B18],[@B33]\] --
Vinblastine Yes but low; 1.1--1.3-fold \[[@B18]\] --
5-fluorouracil No or low; 1.25-fold \[[@B18],[@B26]\] --
Paclitaxel No or low; 1.3-fold \[[@B17],[@B18]\] --
Doxorubicin No or low; 0.67--2-fold \[[@B17],[@B18]\] --
Nimustine hydrochloride (ACNU) No \[[@B27]\] --
Methotrexate No \[[@B1]\] --
9-OH-ellipticine No \[[@B1]\] --
Didemnin B No \[[@B1]\] --
Melphalan No \[[@B1]\] --
Mithramycin No \[[@B1]\] --
Taxotere No \[[@B1]\] --
DauNorubicin No \[[@B17]\] --
Adriamycin No \[[@B32]\] No \[[@B28]\]
Etoposide No or ND \[[@B17],[@B18],[@B27]\] --
Ancitabine ND \[[@B18]\] --
Gemcitabine ND \[[@B18]\] --
ND drug was tested but the potentiation could Not be determined due to methodological limitations.
-- Not tested.
Upon the application of external electric pulses to cells (either in a culture medium, or in tissue), the electric field causes relocation of charges on the cell membrane. This causes the buildup of transmembrane voltage -- termed induced transmembrane voltage, which is superimposed to the cells' normal resting transmembrane voltage (or resting membrane potential, as it is termed in physiology literature). The increase in transmembrane voltage follows Schwann's equation \[[@B34],[@B35]\], and in areas, where transmembrane voltage exceeds a certain voltage, a sufficiently strong electric field through the membrane is established leading to pore formation which allows the passage of water, charged molecules, as well as larger molecules. Although these pores are too small and short-lived to be observed using conventional or electron microscopy, indirect evidence supporting their existence comes from simulations of lipid bilayers using molecular dynamics simulations \[[@B36]-[@B38]\].
Cellular membranes can remain permeable, for large and charged molecules, for minutes after the external electric field delivery ceases \[[@B4],[@B39],[@B40]\]. The mechanism of membrane "resealing" or repair requires active cellular mechanisms and therefore also energy \[[@B41]\]. A possible explanation of the comparatively long-term permeability of cell membranes is chemical alteration of membrane lipids \[[@B42]-[@B44]\]. In order to successfully, and reproducibly achieve cell membrane electroporation, which is the prerequisite for successful electrochemotherapy, appropriate pulse generators are needed. Initially, capacitor discharge pulse generators were used, but later square wave generators became the most prominent as they provide better reproducibility of pulses and control of electroporation \[[@B45]\].
Preclinical research and electrochemotherapy mechanisms
-------------------------------------------------------
Introduction of electrochemotherapy into clinical trials and nowadays into clinical practice is based on extensive preclinical data, on its effectiveness on different tumors and on solid evidence of its mechanisms of action. Several mechanisms of action have already been identified: increased membrane permeability and intracellular drug accumulation; vascular effects; and involvement of immune response.
The first and the predominant one is increased cellular uptake of BLM and CDDP, by exposure of cells or tumors to electric field. Increased cytotoxicity of BLM or CDDP was demonstrated *in vitro,* with several fold potentiation \[[@B2],[@B18],[@B25]\] (Table [1](#T1){ref-type="table"}), as already mentioned above. The *in vitro* data were confirmed and elaborated *in vivo* on different animal tumor models \[[@B4],[@B46]-[@B49]\]. Sufficient drug accumulation in cells is one of the most prominent underlying mechanisms responsible for effective electrochemotherapy \[[@B2],[@B30],[@B50]\]. Since the intracellular drug accumulation due to membrane permeabilization is a consequence of exposure of the cells to sufficiently high local electric field, therefore adequate electric field distribution in the tumors needs to be established \[[@B51],[@B52]\]. All these preclinical data on the pharmacological and physical parameters needed for effective electrochemotherapy have translated into clinical use of electrochemotherapy. Currently, BLM is predominantly used in electrochemotherapy, based on higher potentiation of its cytotoxicity. However, the use of CDDP still remains to be fully explored. CDDP namely has the advantage of being effective already on its own, which however was not further explored after the initial clinical study \[[@B53]\]. Namely, widespread clinical use of CDDP in standard of care could be augmented by sensitizing specific tumors to CDDP by delivering electric pulses \[[@B54]\]. From the clinical perspective it is also helpful that there is another drug of choice, when BLM is contraindicated, or the allowed cumulative dose of BLM is reached \[[@B55]\].
Electrochemotherapy also has two distinct vascular effects. Since the exposure of tumors to electric fields predisposes stromal cells to drug uptake, electrochemotherapy also has an effect on endothelial cells of tumor vessels. This action leads to endothelial cell death (apoptosis) and consequently to abrogation of tumor blood flow. This first effect was named vascular disrupting effect of electrochemotherapy \[[@B56]\]. The second effect is the vasoconstricting effect, demonstrated in tumors, and confirmed on normal and tumor vessels by intravital microscopy \[[@B57]\]. This effect, termed vascular lock, induces prolonged entrapment of the drug within the tumors, providing better action of BLM or CDDP \[[@B58]\]. However, it also prevents inflow of the drug into tumors, if given after the delivery of electric pulses.
Evidence for electric pulses and electrochemotherapy actions on tumors exists in histological, physiological and numerical models \[[@B56],[@B58]\]. Furthermore, it was also observed and demonstrated in clinical cases, where electrochemotherapy was used for the treatment of bleeding tumors \[[@B59],[@B60]\]. However, relative contributions of vascular disrupting and vasoconstricting action of electrochemotherapy in overall electrochemotherapy effectiveness still remain to be determined. Some clinicians emphasize its great importance, predominantly in well vascularized tumors. However, vascular disrupting effect is not observed on bigger blood vessels, such as major hepatic arteries and veins, allowing treatment of tumors in vicinity of them, being the advantage over radiofrequency ablation which is a thermal ablative technique ineffective along bigger blood vessels due to the heat sink effect. Safety and efficacy of electrochemotherapy was recently demonstrated in tumors that were close to bigger vessels and in treatment of liver metastasis located between the major hepatic vessels \[[@B22]\].
Last but not least, involvement of immune response of the organism after electrochemotherapy was demonstrated to be important. Evidently, due to the heterogeneity of tumor cells in the tumors, in relation to their orientation, their size, and uneven drug distribution in the tumors, not all cells in the tumors can be effectively eradicated by electrochemotherapy \[[@B61]\]. This is due to the fact that not all the cells can be electroporated and/or the drug is not available to all the cells in the tumor. Both reasons stem from the inhomogeneity of the tumor \[[@B62]\]. Therefore, similarly as with other physical methods, e.g. radiotherapy, the remaining fraction of cells, when sufficiently low, can be eradicated by the effective immune response of the organism. In relation to this, evidence that immune competence of the organism is needed for the complete eradication of the tumors after electrochemotherapy was provided; namely in immunodeficient organisms the curability rate of the tumors was significantly lower than in immunocompetent ones \[[@B61]\]. This is thought to be related to antigen shedding after electrochemotherapy from the destroyed cells, which would activate immune cells within the tumors. Some indications also point to the recruitment of antigen presenting cells and boosting of CD11c and CD11b cells after electrochemotherapy of melanoma \[[@B63],[@B64]\].
It needs to be stressed however, that the effect of electrochemotherapy is local; limited to the treated nodules, as the neighboring non-treated nodules do not respond. To gain systemic component some attempts were made to boost the immune response by cytokines (IL-2, IL-12, IL-15, GM-CSF, TNFα) providing some preclinical and clinical data on the possibility of combining electrochemotherapy and immunotherapies, to provide systemic antitumor activity \[[@B49],[@B65]-[@B69]\].
Based on the mechanisms of action of electrochemotherapy, antitumor activity should be observed on all types of tumors tested, regardless of their histological origin. In fact electrochemotherapy is effective on all tumor types, but the effectiveness seems to vary somewhat, depending on the tumor type \[[@B50]\]. Also, recent review of clinical data provided evidence of different level of effectiveness of electrochemotherapy in different tumor types \[[@B20]\]. Although the underlying mechanisms are still to be fully determined, several mechanisms were already proposed. The first could be that there is intrinsic variability in tumor cell sensitivity to the drugs \[[@B50],[@B68],[@B70]\], the second that successful membrane permeability to some degree depends on tumor type although being sufficiently high for cell permeabilization \[[@B20],[@B50],[@B71]\], the third, that drug distribution and thus its availability in different tumors varies depending on the vascularization of the tumors \[[@B58]\], and the fourth, that the immunogenicity of the tumors is also involved \[[@B68]\].
Recently an attempt was made to address bioavailability of drug particularly in larger tumors by adding intratumoral delivery of drug in addition to systemic, i.e. intravenous drug delivery. Based on these results it seems that the variability in response due to pharmacological variances in drug distribution particularly in larger tumors could be circumvented by intratumoral drug administration \[[@B72]-[@B74]\]. Furthermore, the pharmacological *in vivo* data in patients need to be obtained in order to verify the drug dosage in the tumors that is required for effective electrochemotherapy. The drug dosage needed for treatment of smaller tumors may be lower than for larger ones (more than 2 cm in diameter), as well as for well vascularized tumors, like hepatocellular carcinoma, vs. metastases of colorectal carcinoma in liver that are less vascularized. Such studies would also clarify the therapeutic window, namely, is there really only 20 minutes time for the therapy \[[@B75]\] or the time could be longer; the current recommendations are based on observations of only a single patient. Gathering such pharmacokinetic and pharmacodynamics data would thus be of great clinical significance.
Clinical data on electrochemotherapy
------------------------------------
### Formation of standard operating procedures
The initial clinical studies that were reporting on electrochemotherapy with BLM or CDDP, even though using somewhat different protocols, provided sufficient evidence for safe and effective use of electrochemotherapy, both in human and veterinary oncology \[[@B70],[@B76],[@B77]\]. Another milestone was achieved with preparation of the standard operating procedures (SOP) for electrochemotherapy using the Cliniporator device that were prepared in 2006 during the ESOPE project (European Standard Operating Procedures of Electrochemotherapy) \[[@B78]\]. The SOP were prepared based on the experience of the leading European cancer centers on electrochemotherapy \[[@B55]\]. The aim of this document was to define guidelines for safe and effective use of electrochemotherapy for treatment of cutaneous and subcutaneous tumors, when using the newly developed electric pulse generator Cliniporator. In particular, the SOP was aimed at oncologists that have no or little experience performing electrochemotherapy, and standardizing the therapeutic procedure. The treatment procedures were described in details, to the point of all necessary materials, and electrode selection for treatment, in specific clinical situation.
Briefly, in the SOP the decision tree was designed and was to be followed helping clinicians in decisions how to treat the patient, according to the number, size and thickness/depth of nodules to be treated. The limit for recommendation of intratumoral drug administration was set at the point of 5--7 nodules and tumor nodules up to 2 cm in diameter. However, on multiple nodules and tumor nodules larger than 0.8 cm, an intravenous drug administration was recommended. Furthermore, the choice of electrodes was suggested based on the size and the thickness/depth of the tumor nodule. The superficial and small (up to 1 cm) nodules were recommended to be treated by plate or needle row electrodes, whereas larger nodules were to be treated by hexagonal needle electrodes \[[@B55],[@B78]\] (Figure [1](#F1){ref-type="fig"}). The procedure can be performed both in local or general anesthesia, which was left to the choice of the treating surgeon, but few and smaller nodules were recommended to be treated in local, while others in general anesthesia. The SOP also suggested that electrochemotherapy could be repeated in consecutive sessions.
{#F1}
The preparation of the SOP was an important step which enabled the spread of the technology throughout Europe, providing the pulse generator certified for use in patients, standardized electrodes, and simple guidelines for medical doctors on how to use electrochemotherapy in treatment of superficial tumors \[[@B55]\]. Electrochemotherapy is being introduced into standard clinical practice in Europe and SOP undoubtedly facilitated its increasing usage in clinics \[[@B79]\]. The number of ongoing clinical trials (currently there are 6 ongoing studies in clinicaltrials.gov, and 5 in clinicaltrialsregister.eu, however, 3 entries are present in both databases) and the number of published studies is increasing (Figure [2](#F2){ref-type="fig"}), confirming the relevance of electrochemotherapy in cancer patient's care.
{#F2}
Since the technology (electrochemotherapy) has spread throughout Europe, vast experience on its clinical applications is being gathered. It has expanded to the point that electrochemotherapy is now being used in different clinical situations, which were not predicted in the originally published SOP. Therefore, new SOP for cutaneous and subcutaneous tumors and eventually for deep-seated tumors are needed, in order to assure the effectiveness of the treatment and even more importantly the safety for the patients.
Implementation of electrochemotherapy into standard of care: focus on cutaneous and subcutaneous tumors
-------------------------------------------------------------------------------------------------------
New treatments provide benefit to the patients but at the same time they can put them at some new risk. Electrochemotherapy, in addition to already established skin directed therapies, like surgery, radiotherapy, topical and intralesional therapies, and other ablative techniques, offers itself due to its relative ease being performed on an outpatient basis as well as due to its safety and low toxicity. The feasibility and safety of electrochemotherapy was evaluated in the first few clinical studies on electrochemotherapy \[[@B2],[@B80],[@B81]\]. Due to good safety and toxicity profile, and only minor side effects, it was concluded that electrochemotherapy on superficial tumors is a safe treatment \[[@B19],[@B78],[@B82],[@B83]\]. No serious effects related to electrochemotherapy have ever been reported. However, some minor immediate effect are observed, like minor irritation and uncomfortable sensation or pain associated with contraction of muscles in the vicinity of the electrodes that immediately subside after delivery of each electric pulse if performed in local anesthesia \[[@B78],[@B84]-[@B86]\]. In addition, some late effects can also occur, like slight erythema, edema and sometimes necrosis of the tumor. All these effects are local, transient, minimal and well tolerated by patients; therefore electrochemotherapy can be performed in an outpatient setting. However, in some latest studies on locally advanced and metastatic soft tissue sarcomas \[[@B87]\] and large cutaneous recurrences of breast cancer \[[@B88]\], they observed longer-lasting periods of pain after electrochemotherapy treatment.
Beside good safety profile, electrochemotherapy on superficial tumors proved to be effective local tumor treatment. The effectiveness of single-session electrochemotherapy was recently extensively addressed in systematic review including meta-analysis \[[@B20]\]. After more than two decades of clinical application of electrochemotherapy, there was an indispensable need to consolidate current experience on electrochemotherapy from the effectiveness point of view and to establish the actual overall effectiveness of electrochemotherapy derived from data reported in clinical studies to-date. The responses of 1894 tumors from 44 eligible clinical studies published until October 2011 were included in the evaluation. The data demonstrated that the effectiveness of single-session electrochemotherapy on cutaneous and subcutaneous tumors is 59.4% for complete response (CR) and 84.1% for objective response (OR). Furthermore, outcomes of studies published before and after publication of the SOP were compared. Based on results obtained, we could presume that SOP improved outcome of electrochemotherapy (see Table [2](#T2){ref-type="table"}). Partial and objective rates significantly increased, whereas non-response rate significantly decreased. However, CR rate remains almost the same. We probably cannot ascribe this improvement in electrochemotherapy outcome solely to the SOP, as other factors like experience gained by clinicians, as well as their training, and exchange of the experience, definitely contributed their share. Recently, an increasing number of new clinical studies on electrochemotherapy have been published, after the publication of the meta-analysis review. Therefore, we updated results of systematic review by including outcomes of 472 tumors from 16 additional studies published between October 2011 and August 2013 (Table [2](#T2){ref-type="table"}). These results confirm positive influence of SOP on effectiveness of electrochemotherapy on cutaneous and subcutaneous tumors.
######
Response rate of the tumors treated by electrochemotherapy, pooled from individual studies
**Publication period** **No. of studies** **No. of patients** **No. of tumors** **OR (%)** **CR (%)** **PR (%)** **NR (%)**
-------------------------- -------------------- --------------------- ------------------- -------------------- -------------- ------------------- --------------------
**Before SOP** 19 175 592 458 (77.4%)^a,\ b^ 362 (61.1%) 96 (16.3%)^c,\ d^ 134 (22.6%)^e,\ f^
**ESOPE** 1 41 171 145 (84.8%) 126 (73.7%) 19 (11.1%) 26 (15.2%)
**After SOP (Oct 2011)** 25 294 1192 1047 (87.8%)^a^ 712 (59.7%) 335 (28.1%)^c^ 145 (12.2%)^e^
**After SOP (Aug 2013)** 41 519 1664 1478 (88.8%)^b^ 1031 (62.0%) 447 (26.9%)^d^ 186 (11.2%)^f^
The studies are grouped in those before, and after the ESOPE study, when the standard operating procedures were published, in 2006. OR = objective response (complete response + partial response); CR = complete response; PR = partial response; NR = no response (no change + progressive disease); ^a,\ b,\ c,\ d,\ e,\ f^ = statistically significant difference (*p* \< 0.001, Chi-square test).
In addition to overall effectiveness of electrochemotherapy, the systematic review \[[@B20]\] addressed also differences in effectiveness of electrochemotherapy of cutaneous and subcutaneous tumors in clinical setting due to heterogeneous treatment conditions (i.e. chemotherapeutic drug, route of drug administration, and tumor type). Equal effectiveness of electrochemotherapy was demonstrated for BLM and CDDP administered intratumorally. However, significantly higher effectiveness for intratumoral than for intravenous administration of BLM was established. The results of electrochemotherapy after the intratumoral or intravenous drug administration basically cannot be compared. Due to the obvious difference in the drug concentrations in the tumors; after intratumoral administration higher drug concentration is delivered to the tumor than after intravenous administration. Currently most of the studies report clinical data after intravenous administration, so there may be imbalance between the two groups that are being compared. Furthermore, the differences in the response rate may be also due to the difference in the tumor size treated; tumors that were treated by intravenous route may have been larger, since the SOP was designed to utilize intravenous route for such situations. Recently it has been suggested that treatment of larger tumors with combined intravenous and intratumoral drug administration can be used with good success \[[@B73]\]. Future studies should be designed to evaluate the intratumoral route for larger tumors. Previous clinical \[[@B82]\] and preclinical \[[@B49]\] studies demonstrated the potential success of such approach.
It has been demonstrated that effectiveness of electrochemotherapy varies with tumor type \[[@B20]\]. Use of often repeated statement about equal effectiveness of electrochemotherapy, regardless of tumor type based on available clinical results appears to be unjustified. Among all types of tumors, the highest effectiveness of electrochemotherapy was achieved on basal cell carcinoma and the lowest on squamous cell carcinoma tumors, presumably because squamous cell carcinoma tumors were usually larger in size than basal cell carcinoma tumors, and also reflecting the fact that basal cell carcinomas tend to recur and metastasize at a lower rate than all other skin tumors.
Currently, there are several registries that pool the data on patients treated with electrochemotherapy; the INSPECT and at least two Italian registries \[[@B89]\]. Some of them gather data on specific tumor type, melanoma, head and neck tumors, and the other (INSPECT) on any kind of tumor histology. These data, if appropriately collected, will help development of electrochemotherapy, and indicate whether drug dosage, route of administration, or electric pulse treatment needs to be adjusted for higher effectiveness of electrochemotherapy. Currently all the clinical data report on tumor response, however in future studies the data on patients response are needed in order to indicate the effect of electrochemotherapy on tumor progression free interval or overall survival. Only then another meta-analysis can unambiguously demonstrate the differences in the response rate of different tumor types, based on well synchronized data collection, and sufficiently long follow-up of patients \[[@B90]\].
Although skin tumors are primarily treated by surgery and radiotherapy, cutaneous metastases are not rare. Furthermore, improved treatment of metastatic cancer results in longer life expectancy, which also results in increased incidence of skin metastases. Skin metastases can represent important problems for patients and physicians -- bleeding metastasis, local infections, pain, and the consequences of radical treatment by surgery may lead to disfigurement and greatly impaired quality of life. The treatment of skin metastases is thus becoming increasingly important. Not just improve in patients' quality of life but also long-lasting local disease control of cutaneous metastases treated with electrochemotherapy was confirmed in some clinical studies \[[@B73],[@B91],[@B92]\].
The current prevalent use for electrochemotherapy in skin tumors is focused on palliation in patients with numerous skin metastases, or patients who have undergone many previous treatments. This use is the most documented and investigated, and is also most supported, e.g. by the NICE guidelines, as we detail in the following section. Electrochemotherapy has also been used in less aggressive non-melanoma types of skin cancer for treatment of primary tumors, such as basal cell carcinoma. In such cases, treatment needs to be very efficient to be considered for use in comparison with surgery or radiotherapy. The main use of electrochemotherapy in such cases would be in areas where there is too little tissue to be able to perform radical resections or where surgery outcome could lead to greatly diminished patients' quality of life, or where a good cosmetic outcome is desired, such as, the head and neck area \[[@B93]\], the cheeks \[[@B94],[@B95]\], the nose \[[@B82],[@B96]\] or near the eyes \[[@B97]\].
In order to get wide acceptance of any kind of treatment, it has to be included in (national) guidelines for a specific treatment, authorized by national or international body, and then it will also enter into the national health system and become eligible for reimbursement. The acceptance is when the effectiveness of the new therapy is evidence based, compared to the other ablative techniques that are currently used for cutaneous tumors/metastases. Therefore only comparative studies will unambiguously demonstrate its effectiveness. Another issue is the acceptance by the patients, morbidity and hospitalization, where electrochemotherapy may have its advantages over other ablative techniques.
Currently, electrochemotherapy clinical application is mostly limited to Europe and is routinely applied in everyday clinical practice in 130 clinical centers (information obtained from IGEA SpA upon request in January 2014). Accounting for differences among EU countries, electrochemotherapy iscurrently reimbursed by the national health insurances in Switzerland, Austria, Germany, Denmark, Spain, UK, Italy, Portugal, and Slovenia. Reimbursement activities are ongoing in Poland and France (data obtained from IGEA Srl in January 2014). In most of these countries it has been acknowledged in their treatment guidelines, as well as recommended in European for the treatment of melanoma \[[@B98]\]. However, except for melanoma, electrochemotherapy still awaits recognition in guidelines for treatment of other cancers. The exception is NICE guidelines, where besides melanoma, electrochemotherapy was recognized as treatment option also for the treatment of primary basal cell carcinoma and squamous cell carcinoma \[[@B99]\], and in German guidelines provided by Dermatologic society \[[@B100]\]. Further advantage of electrochemotherapy is that it can be implemented in developing countries, where big facilities lake radiotherapy departments are not available, and the cost of systemic therapy is too high. In such environment, electrochemotherapy could be the option to help patients with cancer. A further step forward would be also use of non-chemotherapeutic drugs that do not require specific preparation or handling in oncology centers. The first attempt has already been made, the use of calcium, which has proved its activity in preclinical studies on electrochemotherapy (Table [1](#T1){ref-type="table"}) \[[@B31]\], and the first clinical trial on calcium use in electrochemotherapy has been initiated (ClinicalTrials.govidentifier: NCT01941901). This approach would give electrochemotherapy advantage to be implemented outside the oncology centers, also in private practice of dermatology clinics.
Summarizing, electrochemotherapy is currently accepted for treatment of cutaneous and subcutaneous metastases and palliation of skin tumors, but lacks evidence for using electrochemotherapy as curative treatment of primary tumors. For this, additional studies need to be designed and coordinated, which will compare electrochemotherapy with current standard of care, like surgery and radiotherapy in the case of basal cell carcinoma.
Current developments of electrochemotherapy
-------------------------------------------
### Visceral and deep-seated tumors
Electrochemotherapy is now being developed and evaluated for treatment of visceral or deep-seated tumors \[[@B22],[@B23]\]. In order to do this, several steps need to be accomplished, including adjustment of the technology and clinical studies that will prove the safety and effectiveness of the treatment in phase I/II studies. The technology is already being modified for the treatment of tumors larger than 3 cm in diameter, since the current technology provides treatment of tumors smaller than 3 cm. The other adjustments are in the line of safety, predominantly in organs adjacent to the heart, where the delivered electric pulses can interfere with functioning of the heart. Since the heart is especially susceptible to induction of life-threatening arrhythmias if electrical pulses are delivered during the so-called vulnerable period of the ventricles which coincides with T wave in ECG, synchronization of pulses presents another engineering challenge. The vulnerable period can be avoided by synchronizing the delivery of the pulses with the QRS complex of the heartbeat \[[@B101]-[@B103]\]. This is an important consideration particularly for clinical trials that are on-going, i.e. for treatment of liver \[[@B22]\], bone \[[@B104]\] and brain metastases \[[@B105]\], as well as for the treatment of big breast chest wall recurrences \[[@B88],[@B106]\].
Several characteristics make electrochemotherapy a potential candidate for treatment of visceral and deep-seated tumors. These are the ability to treat volumes with diameters in excess of 3 cm, which is a typical limit for thermal therapies, such as radiofrequency and microwave ablation; the advantage of organ sparing effect, and possibility of treatment in the vicinity of major blood vessels. Along with these benefits come additional engineering challenges, among them the requirement of complete tumor nodule coverage by sufficiently high electric field \[[@B51],[@B52]\], that can be achieved by patient specific pre-treatment planning \[[@B107]\], the availability of suitable electrodes for delivery of electric pulses \[[@B108],[@B109]\], synchronization of pulse delivery with ECG \[[@B102],[@B103],[@B110]\], and intraoperative guidance of electrode positioning.
The overview of the implementation of electrochemotherapy in treatment of deep-seated tumors demonstrates that new electrodes are being developed and used in treatment of various tumor types and locations (Table [3](#T3){ref-type="table"}).
######
Different types of electrodes developed for electrochemotherapy of visceral and deep-seated tumors
**Type of electrodes** **Location and type of tumors treated** **Institution performing treatment** **References/Clinical trial number**
--------------------------------------------------------------------- ------------------------------------------------------------------------------- --------------------------------------------------------------------------------------- --------------------------------------
**Long needle (Figure**[3](#F3){ref-type="fig"}**)** Metastases of colorectal tumors in liver bone metastases soft tissue sarcomas Institute of Oncology Ljubljana, Slovenia Istituto Ortopedico Rizzoli, Bologna, Italy \[[@B22]\], NCT01264952
\[[@B104]\]
\[[@B87]\]
Veneto Region Oncology Research Institute of Padova, Italy
**Endoluminal (Figure**[4](#F4){ref-type="fig"}**)** Colorectal, gastric and esophageal tumors Cork Cancer Research Center, Ireland \[[@B108]\], NCT01172860
**Expandable "umbrella" type (Figure**[5](#F5){ref-type="fig"}**)** Brain tumors Copenhagen University Hospital at Herlev, Denmark \[[@B105]\], NCT01322100
Basically three types of electrodes have been developed and are being used:
• The first ones are long needle electrodes that are inserted into the tumor and surrounding tissue in order to safely cover the tumor and achieve an appropriate margin (Figure [3](#F3){ref-type="fig"}). Such electrodes are being tested for treatment of liver metastases of colorectal tumors, where safety and effectiveness have been reported \[[@B22]\]. Such technology can be implemented also for treatment of other types of tumors, such as hepatocellular carcinoma or cholangiocarcinoma. This long needle technology is being tested also for treatment of soft tissue sarcomas \[[@B87]\] and for bone metastases \[[@B104]\], based on solid preclinical evidence \[[@B111]\]. However it could be implemented also for other large (more than 3 cm) tumors located deeper under the skin, like in squamous cell head and neck tumors.
{#F3}
• The second are endoluminal electrodes that are being tested for the treatment of colorectal cancer but could be used also for the treatment of esophageal tumors (Figure [4](#F4){ref-type="fig"}). The ongoing clinical trial in treatment of recurrent colorectal tumors has included so far just a few patients, but has proven feasibility of such approach with good antitumor effectiveness \[[@B112]\].
{#F4}
• The third are the electrodes that are aimed at treatment of brain tumors \[[@B105],[@B109]\] (Figure [5](#F5){ref-type="fig"}). The electrodes have proven to be suitable for the treatment of the brain tumors, through the skull. Unfortunately due to slow recruitment the trial was stopped, but a new one is being prepared (*personal communication*). Such electrodes could potentially be used also for treatment of liver tumors, similarly as percutaneous radiofrequency ablation, with the development of appropriate monitoring system of electrode guidance into the tumors.
![**Retractable brain electrodes. A)** Rendering of the electrode in fully retracted state. **B)** Electrode in fully extended state. Recommended voltage is 1000 V for the fully extended electrodes. **C)** Diagram of the insertion into brain \[[@B105]\].](1475-925X-13-29-5){#F5}
Conclusion and remaining challenges
-----------------------------------
A new treatment or medical technology can be adopted in the clinics after it has proven to be safe and effective. This is the prerequisite, but not sufficient step. It needs to be established as a practical and safe use among many indications for treatment of cancer at various stages of the disease. Electrochemotherapy is now at the stage of palliative treatment in most indications, or in the case other treatments have failed. Therefore, controlled studies will probably not suffice to bring electrochemotherapy along the established ablative techniques (e.g. radiofrequency ablation) or surgery -- controlled, randomized clinical trials are needed, but will be difficult to execute. The other option, based on the current evidence, is that controlled individualized trials will be launched, extending the treatment indications for also less advanced stages of the disease. In the current situation it is of utmost importance that no serious adverse events are reported, since this may shed uncertainty on its development. Consistent results regarding effectiveness after use by clinicians who are not original developers is also a critical step in the growth of a new treatment.
Electrochemotherapy is in its early phase of clinical acceptance, currently the focus being on skin tumors and metastasis. Further work is however required. In treating skin tumors and metastasis electrochemotherapy needs to be compared against current standard treatments; e.g. for basal cell carcinoma it needs to be compared to radiotherapy and/or surgery. This will also require a five-year follow-up!
In general the body of data needs to be collected in multicenter prospective randomized controlled studies in which response per patient rather than per tumor nodule is observed. Additionally, focus will need to be on tumor recurrence free interval, and since electrochemotherapy is a local treatment, quality of life improvement should probably be one of most important outcomes, rather than patients' lifespan. Some of these data will also become available from clinical data registries like InspECT, IMI and GIDO.
The preparation of new/updated Standard Operating Procedures (SOP) will undoubtedly have to deal with differences in responses between differently sized tumors and different types of tumors. Original SOP and decision tree regarding the number of nodules and route of drug administration (intratumoral or intravenous) were based on the experience of the developers and not based on controlled studies determining appropriate dose or administration route. While effective and facilitating the advancement of this treatment modality, further refinement of SOP will be needed to get full acceptance from the medical community. Data on clinical pharmacokinetics and pharmacodynamics of bleomycin and cisplatin in electrochemotherapy would also be important to obtain. Among the studies still needed are dose response evaluations for both intravenous and intratumoral routes. This may be necessary for several indications particularly relating to size, location and type of tumor. In a properly designed study different responses to electrochemotherapy with respect to the tumor size could be addressed; as well as the need for combination of intravenous and intratumoral drug delivery as suggested for efficient treatment of large tumors.
The other challenge would be also to reduce intensity or extent of muscle contractions in order to eliminate the use of muscle relaxants and consequently also pain during electrochemotherapy. The first efforts in this direction have been done by suggesting application of high-frequency bipolar electroporation pulses \[[@B113]\].
The remaining challenge is also to address specificity and sensitivity of electrochemotherapy treatment and to determine its therapeutic index. It had not been systematically addressed before how does healthy tissue respond to electrochemotherapy.
In addition, electrochemotherapy can be implemented in combination with surgery as neoadjuvant treatment. Specifically, with such approach tumors located in surgically difficult to reach positions, due to the involvement of some critical physiological structures (e.g. nerves and blood vessels) or due to potential damage on organs and functions (e.g. sight, hearing, speech, eating) can be treated in order to reduce tumor burden before surgical removal. This approach has already been demonstrated to be useful and effective \[[@B95],[@B114]-[@B118]\]. The downsizing of the tumors can facilitate the surgical intervention, with long term tumor control, and organ sparing effect. The other approach is that electrochemotherapy would be used after surgical removal of the bulk of the tumor, and electrochemotherapy would be used to treat the remaining tumor mass and achieve appropriate margins. There is no report on such approach yet, however it has been successfully used in treatment of sarcoids in horses (*personal communication, no reference is available*). Either of these approaches needs further developments, which only with time will provide enough evidence to implement them into a broader clinical practice.
Electrochemotherapy could be used also as concomitant treatment with other treatment modalities. One of the well explored applications is the use of electrochemotherapy as radiosensitizing approach. BLM and CDDP are well known radiosensitizers. Therefore, electrochemotherapy could, by increased intratumoral accumulation, potentiate radiation response of tumors without normal tissue damage. This was proven in animal models in several reports, with CDDP and BLM, in single and fractionated irradiation regimen \[[@B50],[@B119],[@B120]\]. However there are only a few clinical reports following this idea \[[@B121]-[@B123]\].
The other treatment combinations would be in combination with targeted drugs, and immunostimulatory approaches, either in the form of biologic therapy or in the form of gene therapy. Some attempts have already been made, predominantly in combination of electrochemotherapy with gene electrotransfer on preclinical level \[[@B49],[@B66]-[@B69]\], however the approach needs clinical verification in clinical trials. It is presumed that such combined therapy can add a systemic component, by boosting the specific immune response to tumors that was elicited by antigen shedding from tumors after electrochemotherapy. However, there is no clinical evidence that after electrochemotherapy the non-treated nodules would respond.
In today's economy additional health economics studies will have to be performed clearly showing acceptable price for the benefit that electrochemotherapy can bring to the patients \[[@B124]\]. Electrochemotherapy certainly has potential in this respect as the drugs currently used are inexpensive and there is the potential for the development of low cost instruments and electrodes.
Treating deep-seated tumors requires even more effort, time and resources. Even though the technology of image guided transcutaneous insertion of electrodes is available \[[@B125]\], pretreatment planning development for electrochemotherapy is still in its early development \[[@B126]\]. Currently electric field distribution prediction is calculated in 3D models using finite element methods numerical modeling, taking into account different electric properties and increase in conductivity due to electroporation. Optimization of placement of electrodes is thus performed based on criteria function which for electrochemotherapy requires tumor tissue to be covered with sufficiently high electric field \[[@B127]\]. Uncertainties in tissue conductivity and exact positioning of electrodes relative to the target tissue and each other can however greatly affect tumor treatment outcome \[[@B128]\]. In addition, cell kill based on electrochemotherapy model still needs to be developed in order to be able to predict tumor treatment outcome. Any approach, which is not covered by the SOP, can benefit greatly from treatment planning. Although conclusive proof of treatment planning's benefit has not yet been established through controlled studies, there are several indications, that it is beneficial. At the very least, it can serve as tool for scaling from smaller to larger tumors, since the scaling is non-linear due to electroporation's effect on tissue electric properties. Algorithms designed for treatment planning can also be used in novel electrode designs \[[@B109]\].
Achieving acceptance of a new clinical treatment for cancer is a huge project that takes considerable time, orchestrated efforts and a lot of resources, whereas biomedical engineers tend to look at medical progress in the short term. Appropriately controlled studies that take all parameters into consideration are a critical part of bringing a new approach to broad acceptance. Focusing on specific cancer targets to gain initial acceptance will open up additional treatment options and facilitate acceptance of the therapy. Apart from x-ray all other techniques and treatments needed 20 years or so to gain its place. Needless to say the route is not always straightforward, nor is it always successful. The regulations in Europe and USA are quite different, and specific, and costly. However, overall there are no big differences between the two markets \[[@B129],[@B130]\]. There are new large markets that will affect usual strategies that companies have in bringing new technologies and treatments to the patient.
Competing interests
===================
DM holds patents on electrochemotherapy that are licensed to and is occasionally consulting IGEA S.r.l, Italy -- a producer of device and equipment for electrochemotherapy. The other authors declare no competing interests.
Authors' contributions
======================
DM conceived the review, and participated in its design and coordination and helped to draft the manuscript. BM made the systematic review of newly published clinical papers and wrote parts of the manuscript and corresponding figures and tables. BK made the electrode figures and wrote parts of the manuscript. RH helped draft the manuscript and participated with critical information to achieve a comprehensive review and evaluation. GS wrote parts of the manuscript and supervised the preclinical and clinical parts. All authors read and approved the final version of the manuscript.
Acknowledgements
================
The manuscript is a result of a networking effort of the COST TD1104 Action (<http://www.electroporation.net>). The research has been performed within the scope of LEA EBAM. In the past research on electrochemotherapy performed by the authors has been funded by EU and Slovenian Research Agency, United States National Institutes of Health, American Cancer Society and the Center for Molecular Delivery at the University of South Florida.
| {
"pile_set_name": "PubMed Central"
} |
Introduction
============
Lung cancer is one of the most commonly diagnosed tumors and continually leads to many deaths worldwide.[@b1-ott-9-6083] There were an estimated 1.8 million new cases of lung cancer in 2012, which occurred in the developing countries, accounting for 58% of the total cases. An estimated 1,987,909 new cases of lung cancer were diagnosed and 1,732,185 deaths were due to lung cancer in 2012. Moreover, lung cancer is the most common tumor in males.[@b2-ott-9-6083] The incidence of lung cancer is still obviously increasing.[@b3-ott-9-6083] The 5-year survival rate for lung cancer patients was only 16% in the past four decades, and the main obstacle in improving lung cancer prognosis was delayed diagnosis.[@b4-ott-9-6083] There are many reasons for the occurrence of lung cancer, but the major external risk factor was easily identified -- cigarette smoking. It has been reported that current smokers are approximately 20 times more likely than never smokers to develop lung cancer.[@b5-ott-9-6083],[@b6-ott-9-6083] Thus, smoking cessation is very important for public health. In addition, early detection and diagnosis play a crucial role in decreasing the mortality of lung cancer. Survival of lung cancer patients undergoing lung resection was more than 80%, which suggested that early detection and diagnosis helped immensely in the early treatment of lung cancer.[@b7-ott-9-6083],[@b8-ott-9-6083] Increasing evidence has indicated that polymorphisms of numerous susceptible genes play a significant association in the development of lung cancer. Thus, these polymorphisms could be useful as biomarkers for lung cancer detection.
Interleukin-10 (IL-10) cytokine, which is essential for T~H~2 responses, plays a central role in anti-inflammation and is also capable of blocking tumor immune surveillance to inhibit T-cell immunity.[@b9-ott-9-6083] A recent study has shown that the expression of IL-10 was associated with progression of tumors including lung cancer.[@b10-ott-9-6083] Some genetic polymorphisms of IL-10 gene, especially in the promoter region, could affect the expression of gene-encoded proteins associated with lung cancer susceptibility and prognosis.[@b11-ott-9-6083] Three SNPs at positions -1082, -592, and -819 of IL-10 gene promoter have been reported to be associated with expression of IL-10 gene.[@b12-ott-9-6083] Moreover, many individual studies have investigated the role of *IL-10 -1082A/G, -592C/A*, and -*819T/C* polymorphisms in lung cancer risk. Nevertheless, the results of these studies remained ambiguous and inconclusive owing to the reduced power of single studies. In this study, an updated meta-analysis was performed to estimate the effect of *IL-10 -1082A/G, -592C/A*, and -*819T/C* polymorphisms on lung cancer risk.
Material and methods
====================
Literature search strategy
--------------------------
The databases of PubMed, Embase, Web of Science, China National Knowledge Infrastructure, WanFang, and Database of Chinese Scientific and Technical Periodicals (VIP) were searched up to July 2016 to identify relevant studies published concerning the correlation of *IL-10 -1082A/G, -592 C/A*, and -*819T/C* polymorphisms with lung cancer risk. The following MESH terms and their synonyms were used: ("Lung Neoplasms" \[MESH\] or "lung cancer" or "lung tumor" or "lung carcinoma") and ("Interleukin-10" or "IL-10"). Language restriction was not applied in the process of retrieval. The reference lists of retrieved article were hand searched.
Inclusion criteria and exclusion criteria
-----------------------------------------
The following inclusion criteria were used: 1) studies investigating the association between *IL-10 -1082A/G, -592C/A*, and -*819T/C* polymorphisms and lung cancer risk; 2) studies with a case--control or cohort design; 3) studies with sufficient genotyping data available for calculation of odds ratios (ORs) and 95% confidence intervals (95% CIs); and 4) studies performed in humans. The exclusion criteria were as follows: 1) not a case--control or cohort study, 2) studies without available genotype data, 3) studies with duplicate data, and 4) meta-analysis and reviews.
Quality score assessment
------------------------
The Newcastle--Ottawa scale was used to assess the quality of studies included. Two investigators independently calculated the score of each study. The scores ranged from 0 to 9, and a score of ≥7 indicated a high quality study.
Data extraction
---------------
Two investigators independently searched articles and extracted the original data according to the inclusion criteria and exclusion criteria. The extracted information from each eligible study included the following: 1) the first author's name, 2) year of publication, 3) ethnicity, 4) source of controls, 5) number of cases and controls, and 6) genotype frequencies. Disagreement in data extraction was settled by consulting with the third investigator.
Statistical analysis
--------------------
The strength of association between *IL-10 -1082A/G, -592C/A*, and -*819T/C* polymorphisms and lung cancer risk was evaluated by calculating crude ORs and 95% CIs under an allele, a homozygous, a dominant, and a recessive model, respectively. Chi-square-based *Q*-test and *I*^2^ statistic test were performed to evaluate the heterogeneity assumption. A random-effects model (the DerSimonian and Laird method) was used to calculate the crude ORs of each study in case of *P*\<0.05 or *I*^2^\>50%, whereas a fixed-effects model (the Mantel--Haenszel method) was used if heterogeneity did not exist.[@b13-ott-9-6083],[@b14-ott-9-6083] Potential publication bias was evaluated using Begg's funnel plots and Egger's test. Sensitivity analysis was conducted to estimate the stability of the results by omitting each single study. The overall analysis and stratified analysis were performed using Stata 12.0 software (Stata Corp LP, College Station, TX, USA).
Results
=======
Characteristics of published studies
------------------------------------
Based on the inclusion criteria, a total of 372 publications were obtained on literature searching. After removing the articles that did not deal with the association of IL-10 polymorphism with lung cancer risk, reviews, and meta-analysis, 14 were left for further analysis. After further evaluation by reading full texts, four publications were excluded because of insufficient data on IL-10 genotype. Finally, 10 articles with 19 studies that investigated the correlation of *IL-10 -1082A/G, -592C/A*, and -*819T/C* polymorphisms with lung cancer risk were used for the further statistical analysis. Of these, eleven studies were conducted in Asians and eight studies were performed in Caucasians. In addition, eight publications were in English and two publications were in Chinese. The characteristics of all the studies included are listed in [Table 1](#t1-ott-9-6083){ref-type="table"} and shown in [Figure 1](#f1-ott-9-6083){ref-type="fig"}.
Quantitative analysis
---------------------
A total of 10 publications amounting to 19 case--control studies were included in this meta-analysis. In the overall analysis, the presence of *IL-10 -1082A/G* polymorphisms was associated with increased lung cancer risk, while *IL-10 -592C/A* polymorphisms were shown to decrease the lung cancer risk (for A-1082G, A vs G, OR =1.17, CI =1.04--1.30, *P*\<0.05; AA vs \[AG + GG\], OR =1.21, CI =1.05--1.40, *P*\<0.05; for C-592A, C vs A, OR =0.84, CI =0.77--0.91, *P*\<0.05; CC vs AA, OR =0.74, CI =0.64--087, *P*\<0.05; CC vs \[CA + AA\], OR =0.83, CI =0.74--0.94, *P*\<0.05; \[CC + CA\] vs AA, OR =0.79, CI =0.69--0.91, *P*\<0.05). When stratified by ethnicity, *IL-10 -1082A/G, C-592A*, and *T-819C* polymorphisms were associated with increased lung cancer risk in Asians (for -1082A/G, AA vs \[AG + GG\]: OR =1.20, CI =1.05--1.39, *P*\<0.05; for C-592A, C vs A, OR =1.36, CI =1.20--1.53, *P*\<0.05; CC vs AA, OR =1.85, CI =1.45--2.37, *P*\<0.05; CC vs \[CA + AA\], OR =1.36, CI =1.15--1.61, *P*\<0.05; for -*819T/C*, T vs C: OR =1.21, CI =1.06--1.38, *P*\<0.05; TT vs CC, OR =1.54, CI =1.18--2.01, *P*\<0.05; \[TT + TC\] vs CC, OR =1.51, CI =1.17--1.95, *P*\<0.05). Furthermore, when stratified based on disease type, a significant correlation was detected between *IL-10 T-819C* polymorphism and non-small-cell lung cancer (NSCLC; for -*819T/C*, T vs C: OR =1.57, CI =1.15--2.16, *P*\<0.05; TT vs CC, OR =3.15, CI =1.58--6.30, *P*\<0.05; \[TT + TC\] vs CC, OR =3.07, CI =1.58--5.93, *P*\<0.05). Interestingly, no significant association between *IL-10 -1082A/G, C-592A*, and *T-819C* polymorphisms and lung cancer risk was found in Caucasians. Furthermore, significantly decreased risk of lung cancer risk was found in the overall analysis of C-592A polymorphism, whereas significantly increased risk was detected in the Asian population ([Table 2](#t2-ott-9-6083){ref-type="table"}).
Heterogeneity analysis
----------------------
Under the homozygous genetic models, significant heterogeneity existed among the studies on *IL-10 -1082A/G* and -*819T/C* polymorphisms (for 1082A/G, *I*^2^=71.8%, *P*=0.007; for -*819T/C*, *I*^2^=88.0%, *P*=0.000). The forest plots suggested that ethnicity and disease type might be one of the sources of heterogeneity for *IL-10 -1082A/G* and -*819T/C* polymorphisms. However, heterogeneity was significantly lowered after the subgroup analyses based on ethnicity and disease type were conducted. This indicated that the ethnicity and disease type might lead to the heterogeneity. Furthermore, a meta-regression revealed that score, sample size, and genotyping method were not the sources of heterogeneity (for *IL-10 -1082A/G* and -*819T/C*, *P*\>0.05; [Figures 2](#f2-ott-9-6083){ref-type="fig"}[](#f3-ott-9-6083){ref-type="fig"}--[4](#f4-ott-9-6083){ref-type="fig"}).
Publication bias and sensitivity analysis
-----------------------------------------
On the basis of the Begg's test and Egger's test, no publication bias was found for the association between *IL-10 -1082A/G, -592C/A*, and -*819T/C* polymorphisms and lung cancer risk. Moreover, sensitivity analysis indicated that the results were stable after sequentially omitting each study ([Figures 5](#f5-ott-9-6083){ref-type="fig"}[](#f6-ott-9-6083){ref-type="fig"}--[7](#f7-ott-9-6083){ref-type="fig"}).
Discussion
==========
IL-10 was shown to have a significant effect on the tumorigenesis of different cancers and could even protect tumors by inhibiting the function of macrophages and dendritic cells, which present tumor cell antigens to the cytotoxic T lymphocytes.[@b15-ott-9-6083] Previous studies have found an association between IL-10 polymorphism and oral cancer, breast cancer, gastric cancer, colorectal cancer, esophageal cancer, hepatocellular cancer, cervical cancer, thyroid cancer, and nasopharyngeal cancer.[@b16-ott-9-6083]--[@b24-ott-9-6083] As for lung cancer, many studies have revealed that the loss of IL-10 expression could promote progression and poor clinical outcomes of lung cancer; however, the opposite effects were also found in some patients.[@b25-ott-9-6083] Interestingly, the absence of IL-10 had opposite effects on the early and late stages of NSCLC.[@b26-ott-9-6083] Hence, the expression or polymorphisms of IL-10 gene might serve as a good indicator of susceptibility and prognostic outcome for lung cancer in different stages of the disease.[@b27-ott-9-6083] These lung cancer biomarkers might be used for screening, early detection, early diagnosis, prognosis, prediction, stratification, and therapy response monitoring.[@b3-ott-9-6083] Although previous case--control studies showed that the IL-10 polymorphism might contribute to the risk of lung cancer, and the sample size of the studies was too small. Thus, this meta-analysis was performed.
Through this comprehensive meta-analysis, the association between *IL-10 -1082A/G, -592C/A*, and -*819T/C* polymorphism and the risk of lung cancer was evaluated. To our knowledge, this is the first systematic meta-analysis to date that deals with the association between *IL-10 -1082A/G, -592C/A*, and -*819T/C* polymorphisms and lung cancer risk. In the overall analysis, we found that *IL-10 -1082A/G, -592C/A*, and -*819T/C* polymorphisms were significantly associated with lung cancer risk. Interestingly, *IL-10 -1082A/G, -592C/A*, and -*819T/C* polymorphisms were significantly associated with lung cancer risk in Asians, but not in the Caucasians. Furthermore, the results indicated that individuals who carried the *IL-10 -1082G* allele had an increased risk of lung cancer, and those with the *IL-10 -819T* allele were shown to have a decreased risk of lung cancer compared to those with the -819C allele. The results of previous case--control studies by Shih et al,[@b28-ott-9-6083] Seifart et al,[@b29-ott-9-6083] Colakogullari et al,[@b30-ott-9-6083] and Peddireddy et al[@b31-ott-9-6083] for *IL-10 -1082A/G* polymorphism were consistent with this meta-analysis and indicated that the *IL-10 -1082A/G* polymorphism could increase the risk of lung cancer. The results of Hsia et al,[@b32-ott-9-6083] Shih et al,[@b28-ott-9-6083] and Zhang et al[@b33-ott-9-6083] for *IL-10 -819T/C* polymorphism showed that *IL-10 -819T/C* polymorphism was significantly associated with an increased risk of lung cancer. Nevertheless, different results were also found in the studies of Hsia et al[@b32-ott-9-6083] and Hart et al[@b34-ott-9-6083] concerning the *IL-10 -1082A/G* polymorphism. However, several of the included studies showed no significant correlation of *IL-10 A-592C* and -*819T/C* polymorphisms with lung cancer risk.[@b29-ott-9-6083],[@b30-ott-9-6083],[@b32-ott-9-6083],[@b34-ott-9-6083]--[@b36-ott-9-6083] In addition, there was adequate evidence indicating the different frequency distribution of *IL-10 -592C/A* and -*819T/C* allele between Asians and Caucasians in the meta-analysis. And this result was consistent with the data given in the HapMap database. Thus, a decreased risk of lung cancer in Caucasians and an increased risk of lung cancer in Asians was associated with the *IL-10 -592C/A* and -*819T/C* polymorphisms, although no significant association of *IL-10 -592C/A* and -*819T/C* polymorphisms with lung cancer risk in Caucasians was found. In addition, based on the results of the subgroup analysis, we found that *IL-10 -819T/C* polymorphism was associated with susceptibility to NSCLC. A previous study reported that IL-10 might be involved in the process of tumor escape from the immune response, and increased production of IL-10 has been shown to correlate with the survival of NSCLC patients.[@b28-ott-9-6083] However, only one study was included in the subgroup analysis, and this lowered the statistical power. For the *IL-10 -1082A/G* and -*592C/A* polymorphism, no significant correlation with NSCLC risk was detected. The analysis of small-cell lung cancer was not performed because the sample of small-cell lung cancer was too small and the classification of small-cell lung cancer was not clear. On the basis of the meta-analysis, to a large extent, there was no significant association between *IL-10 -1082A/G*, -*592C/A* polymorphism and NSCLC risk. To confirm this conclusion, studies with larger sample number must be conducted. The study results showed that subgroup analysis is very important for a systematic review and meta-analysis. From the studies included, there is evidence that the association between IL-10 polymorphism and lung cancer risk is still not clear. Therefore, performing a meta-analysis and systematic review was necessary to clarify this relevance. These results are shown in [Figure 3](#f3-ott-9-6083){ref-type="fig"}.
Limitations
===========
Several limitations of this meta-analysis should be acknowledged. First, the small number of eligible studies might limit the meta-analysis. There were only five or seven studies available for three SNPs of IL-10 gene in the stratification analysis in particular, which might have attenuated the statistical power of the meta-analysis. Second, the missing individual data, for example, cancer stage and cancer subtype might also have an impact on the interpretation. Third, this meta-analysis could not address the gene--environmental interactions in the correlation of IL-10 polymorphism with lung cancer risk. The conclusions only depend on the ORs, which might lead to confounding bias. Finally, the inclusion of data that was inconsistent with HWE might affect the accuracy of the results. However, the large sample size could weaken the impact of the association of IL-10 polymorphism with lung cancer risk.
In conclusion, this meta-analysis of ten case--control studies suggested that the *IL-10 -1082A/G, -592C/A*, and -*819T/C* polymorphisms were associated with the increased risk of lung cancer in Asians. Nevertheless, considering the limitations of the study, further large-scale, well-designed, and population-based studies are still needed to assess the correlation between IL-10 polymorphism and lung cancer risk.
The authors would like to thank all the participants involved in the research.
**Disclosure**
The authors report no conflicts of interest in this work.
{#f1-ott-9-6083}
{#f2-ott-9-6083}
{#f3-ott-9-6083}
{#f4-ott-9-6083}
{#f5-ott-9-6083}
{#f6-ott-9-6083}
{#f7-ott-9-6083}
######
Characteristics of eligible studies included
Author Year Controltype Nationality Ethnicity Disease Control Case *P* for HWE Score Genotyping method
-------------------------------------- ------ ------------- ---------------------------- ------------ ---------------- ----------- -------- -------- -------- ----------- -------- -------- -------- ------------- ------- -------------------
**A-1082G** **Total** **AA** **AG** **GG** **Total** **AA** **AG** **GG**
Peddireddy et al[@b31-ott-9-6083] 2016 HB India Asians NSCLC 250 130 84 36 246 156 69 21 \<0.05 7 PCR-RFLP
Hsia et al[@b32-ott-9-6083] 2014 HB People's Republic of China Asians Lung neoplasms 716 561 130 25 358 273 69 16 \<0.05 7 PCR-RFLP
Hart et al[@b34-ott-9-6083] 2011 HB Norway Caucasians NSCLC 435 104 226 105 436 120 207 109 0.41 7 TaqMan
Hao et al[@b37-ott-9-6083] 2009 HB People's Republic of China Asians Lung neoplasms 52 46 6 44 36 7 0.99 8 TaqMan
Colakogullari et al[@b30-ott-9-6083] 2007 HB Turkey Caucasians Lung neoplasms 59 33 21 5 44 11 30 3 0.53 7 PCR-SSP
Shih et al[@b28-ott-9-6083] 2005 HB People's Republic of China Asians NSCLC 205 194 11 0 154 115 39 0 0.69 8 PCR-RFLP
Seifart et al[@b29-ott-9-6083] 2005 HB Germany Caucasians Lung neoplasms 243 86 115 42 39 6 21 12 0.74 7 PCR-RFLP
**A-592C** **Total** **CC** **CA** **AA** **Total** **CC** **CA** **AA**
Zhang et al[@b33-ott-9-6083] 2015 HB People's Republic of China Asians Lung neoplasms 336 75 176 85 330 110 156 64 0.37 7 PCR-RFLP
Hsia et al[@b32-ott-9-6083] 2014 HB People's Republic of China Asians Lung neoplasms 716 71 277 368 358 40 145 173 0.08 7 PCR-RFLP
Hart et al[@b34-ott-9-6083] 2011 HB Norway Caucasians NSCLC 433 243 175 15 434 264 144 26 \<0.05 7 TaqMan
Liang et al[@b35-ott-9-6083] 2011 HB People's Republic of China Asians NSCLC 120 7 44 69 116 11 36 69 0.99 8 PCR-RFLP
Vogel et al[@b36-ott-9-6083] 2008 PB Denmark Caucasians NSCLC 744 452 250 42 403 241 149 13 0.34 8 PCR-RFLP
Colakogullari et al[@b30-ott-9-6083] 2007 HB Turkey Caucasians Lung neoplasms 59 27 25 7 44 19 23 2 0.74 7 PCR-SSP
Shih et al[@b28-ott-9-6083] 2005 HB People's Republic of China Asians NSCLC 205 13 76 116 154 18 70 66 0.91 8 PCR-RFLP
**T-819C** **Total** **TT** **TC** **CC** **Total** **TT** **TC** **CC**
Zhang et al[@b33-ott-9-6083] 2015 HB People's Republic of China Asians Lung neoplasms 336 145 144 47 330 108 135 87 0.25 7 PCR-RFLP
Hsia et al[@b32-ott-9-6083] 2014 HB People's Republic of China Asians Lung neoplasms 716 372 265 79 358 212 128 18 \<0.05 7 PCR-RFLP
Colakogullari et al[@b30-ott-9-6083] 2007 HB Turkey Caucasians Lung neoplasms 59 7 26 26 44 2 23 19 0.9 7 PCR-SSP
Shih et al[@b28-ott-9-6083] 2005 HB People's Republic of China Asians NSCLC 205 104 86 15 154 66 58 30 0.63 8 PCR-RFLP
Seifart et al[@b29-ott-9-6083] 2005 HB Germany Caucasians Lung neoplasms 242 14 88 140 40 2 14 24 0.97 7 PCR-RFLP
**Abbreviations:** HB, control hospital based; PB, control population based; NSCLC, non-small-cell lung cancer; PCR, polymerase chain reaction; RFLP, restricted fragment length polymorphism; SSP, sequence specific primer.
######
Meta-analysis results of the association between *IL-10 -1082A/G, -592C/A*, and -*819T/C* polymorphisms and lung cancer risk
IL-10 Allele model Homozygous model Dominant model Recessive model
------------- ------------------- ------------------ ------------------- ----------------- ------------------------- -------- ------------------------- --------
**A-1082G** **A vs G** **AA vs GG** **AA vs (AG** + **GG)** **(AA** + **AG) vs GG**
Caucasians 1.37 (0.98--1.35) 0.08 1.33 (0.97--1.83) 0.08 1.21 (0.94--1.55) 0.14 1.21 (0.92--1.58) 0.17
Asians 1.14 (0.96--1.36) 0.14 0.99 (0.65--1.51) 0.95 1.20 (1.05--1.39) \<0.05 1.15 (0.92--1.43) 0.21
NSCLC 1.02 (0.89--1.18) 0.75 1.01 (0.77--1.33) 0.95 1.05 (0.87--1.27) 0.58 0.98 (0.76--1.27) 0.87
Total 1.17 (1.04--1.30) \<0.05 1.23 (0.98--1.54) 0.07 1.21 (1.05--1.40) \<0.05 1.15 (0.92--1.43) 0.21
**C-592A** **C vs A** **CC vs AA** **CC vs (CA** + **AA)** **(CC** + **CA) vs AA**
Caucasians 0.96 (0.83--1.11) 0.55 0.88 (0.59--1.33) 0.55 0.96 (0.80--1.14) 0.62 0.89 (0.60--1.34) 0.58
Asians 1.36 (1.20--1.53) \<0.05 1.85 (1.45--2.37) \<0.05 1.36 (1.15--1.61) \<0.05 0.74 (0.62--0.87) \<0.05
NSCLC 0.98 (0.87--1.10) 0.70 0.96 (0.77--1.21) 0.74 0.98 (0.84--1.14) 0.75 0.97 (0.79--1.20) 0.79
Total 0.84 (0.77--0.91) \<0.05 0.74 (0.64--0.87) \<0.05 0.83 (0.74--0.94) \<0.05 0.79 (0.69--0.91) \<0.05
**T-819C** **T vs C** **TT vs CC** **TT vs (TC** + **CC)** **(TT** + **TC) vs CC**
Caucasians 0.96 (0.65--1.41) 0.82 1.36 (0.44--4.17) 0.59 1.50 (0.50--4.50) 0.47 0.85 (0.53--1.38) 0.52
Asians 1.21 (1.06--1.38) \<0.05 1.54 (1.18--2.01) \<0.05 1.15 (0.97--1.37) 0.11 1.51 (1.17--1.95) \<0.05
NSCLC 1.57 (1.15--2.16) \<0.05 3.15 (1.58--6.30) \<0.05 1.37 (0.90--2.09) 0.14 3.07 (1.58--5.93) \<0.05
Total 0.97 (0.86--1.10) 0.63 0.95 (0.76--1.20) 0.68 0.96 (0.82--1.14) 0.66 0.97 (0.79--1.19) 0.77
**Abbreviations:** NSCLC, non-small-cell lung cancer; OR, odds ratio; 95% CI, 95% confidence interval.
| {
"pile_set_name": "PubMed Central"
} |
Background
==========
In environmental health research, focus has shifted from relatively simple to more complex issues. Empirical single agent -- single effect studies have been supplemented by research on risks of complex environmental exposures in varying economic, cultural and political settings. Environmental health impact assessment has become a valuable tool for decision support. These types of assessments increasingly use so-called environmental burden of disease (eBoD) measures to express health impacts. The eBoD can be viewed as the gap -- caused by environmental factors -- between current health status and an alternative situation in which environmental exposures are reduced or eliminated. Burden of disease estimates enable comparison of divergent environmental health problems. This in turn enables policy makers to set priorities. However, scientists often have to make many assumptions when assessing the eBoD. Knowledge and data are often incomplete, and diverging perceptions exist about what the most important aspects of a problem are. Assessments are often highly interdisciplinary, complex and multifaceted, and the uncertainty about results can be significant \[[@B1]\]. This may affect decision making based on these assessments.
A 2005 comparison of 17 eBoD studies published between 1996 and 2005 (internal RIVM/MNP publication by Knol et al.) showed that there are significant differences between eBoD estimates that concern -- at first sight -- similar issues. Smith et al. \[[@B2]\], for example, estimate the fraction of the total global disease burden attributable to the environment to be 25--33%, whereas Melse and de Hollander \[[@B2],[@B3]\] estimate this to be 7.5 to 11% (for OECD countries only: 2--5%). Such differences can sometimes not be fully explained by reading the assessment reports. Methods, assumptions and input data are often insufficiently explained, which hampers interpretation and comparability of results. Fox-Rushby and Hanson \[[@B4]\] show that 9 out of 16 papers on burden of disease published between 1993 and 2000 did not declare the underlying assumptions.
Even though it is never possible to reduce uncertainty to zero in these complex assessments, there is significant room for improvement in dealing with uncertainty \[[@B1]\]. Various eBoD studies have addressed the need for uncertainty and sensitivity analyses (for example \[[@B2],[@B4]-[@B10]\]), but, as yet, these analyses are based primarily on statistical uncertainty of some parameters and input data. Other sources of uncertainty are often touched upon in the discussion sections of publications, but usually not in a systematic manner. However, many environmental health issues are not straightforward and uncertainties cannot be captured in simple confidence intervals \[[@B11],[@B12]\]. Only if both scientists and policy makers realize the potential extent of uncertainties and the way they may affect the assessment results, can these assessments lead to truly informed policy making. In order to achieve this, a typology of different dimensions of uncertainty can help to structure, assess and potentially reduce uncertainties, and moreover to improve the dialogue about uncertainties between scientists and policy makers.
The present study explores the different types of uncertainty that may play a role in eBoD studies expressed in Disability Adjusted Life Years, structured using a typology. The impact that uncertainties can have on assessment results -- and thereby on decision making -- will be illustrated using examples from the existing eBoD literature. Some suggestions are given as to how to address and communicate uncertainties to policy makers. This paper aims to create awareness among environmental health impact assessors about the potential impact and importance of uncertainties, and to provide a practical approach and structure to deal with uncertainties in eBoD assessments.
Disability Adjusted Life Years
------------------------------
An increasingly popular metric to express the environmental burden of disease is the DALY (Disability Adjusted Life Years). DALYs indicate the potential number of healthy life years lost in a population, i.e. burden of disease. Not only life years lost due to premature mortality, but also years spent with reduced quality of life due to diseases are included. For diseases, severity weights (also referred to as disability weights) are used to quantify the reduced quality of life. They are developed by expert panels and range from 0 for complete health to 1 for death. Diseases with a severity weight ranging from 0.05 to 0.1 include for example low back pain, uncomplicated diabetes, or mild angina. Examples of more severe diseases with weights ranging from 0.65 to 0.8 include cancer, severe depression, and brain injury. These specific weights have been derived by Stouthardt et al. \[[@B13]\] and it should be recognized that other authorities might assign different weights to these effects.
Additionally, DALY calculations can include age weights and discounting factors. Age weighting involves valuing life years lost at a certain age more than life years lost at other ages. Discount factors are used to value present years of life saved more than future years. The usual annual discount rate is 3%, implying that a year of healthy life gained in 10 years time is valued at 26 percent less than one gained now. The use of age weights and discount factors has been discussed -- and heavily debated -- elsewhere (for example \[[@B14],[@B15]\]).
Burden of disease calculations using DALYs were first published in the World Development Report \[[@B16]\]. Subsequently, Murray and Lopez \[[@B17]\] used DALYs in their extensive Global Burden of Disease project in order to introduce morbidity into the predominantly mortality-based health discussions. Since then, the World Health Organization (WHO) has endorsed the DALY approach, and it has been used in various studies on global, national and regional levels \[[@B3],[@B18]-[@B26]\]. Burden of disease calculations are now increasingly being asked for in order to develop, evaluate and prioritize health-related policy measures. As well as DALYs, various other summary measures exist to express population health or disease states, such as QALYs (Quality Adjusted Life Years), HALYs (Health Adjusted Life Years), DALEs (Disability Adjusted Life Expectancy), HALEs (Health Adjusted Life Expectancy), and various monetary valuation measures. Even though this paper focuses on DALYs, most of the uncertainties identified play a similar role for these alternative indicators.
Typology of uncertainty
-----------------------
Uncertainties in assessments about a complex world can take many forms. A typology of uncertainty can help to structure the different types of uncertainties. This can in turn help to identify useful methods and techniques to deal with the uncertainties, ranging from stakeholder discussion to sensitivity analysis.
We have adapted existing uncertainty typologies \[[@B27]-[@B33]\] to fit our purpose of identifying, further characterizing and dealing with the uncertainties that arise in eBoD assessments. Our typology (Table [1](#T1){ref-type="table"}) distinguishes between location, nature, range, recognized ignorance, methodological unreliability and value diversity among analysts, as six characteristics of uncertainty. These characteristics apply simultaneously to a piece of uncertain information. This typology differs from the one presented by Petersen \[[@B30]\] in the way in which the location of uncertainty and ontic uncertainty are specified. These changes have been made in order to make the typology more applicable to burden of disease assessments. How the typology presented by Petersen differs from previous typologies such as \[[@B27],[@B28],[@B32]\] is described in \[[@B30]\]. It is not claimed that our typology is the best typology (cf \[[@B30]\]); for other purposes, other typologies might be more useful.
######
Typology of uncertainty
Uncertainty characterizations *Categories*
-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- --------------------------------------------------------------------------------------------------------------------------------------- -------------------------------------------------------------------------------------------------------------------
**Location**: the location at which the uncertainty manifests itself in the assessment ***Model structure***: Structure and form of the relationships between the variables that describe the system
***Parameters***: Constants in functions that define the relationships between variables (such as relative risks or severity weights)
***Input data***: Input data sets (such as concentrations, demographic data, and incidence data)
**Nature**: the underlying cause of the uncertainty ***Epistemic***: resulting from incomplete knowledge
***Ontic*** ***Process variability***: resulting from natural and social variability in the system
***Normative uncertainty***: resulting from a plurality of socio-ethico-normative considerations within a society
**Range**: expression of the uncertainty ***Statistical (range + chance)***: specified probabilities and specified outcomes
***Scenario (range + \"what if\")***: specified outcomes, but unspecified probabilities
**Recognized ignorance**: unknown outcomes, unknown probabilities -- uncertainties are present, but no useful estimate can be given
**Methodological unreliability**: Methodological quality of all different elements of the assessment; a qualitative judgment of the assessment process which can based on e.g. its theoretical foundation, empirical basis, reproducibility and acceptance within the peer community
**Value diversity among analysts**: Potential value-ladenness of assumptions which inevitably involve -- to some degree -- arbitrary judgments by the analysts.
First, the location of uncertainty indicates where the uncertainty manifests itself among the main elements of the assessment. Distinction is made here between the context, model structure, parameters and input data. These locations will be further described below.
Second, the nature of uncertainty expresses whether uncertainty is primarily a consequence of the incompleteness and fallibility of knowledge, epistemic uncertainty, or primarily due to intrinsic properties of the system under study, ontic uncertainty -- ontic meaning pertaining to the object. In other contexts and disciplines, ontic uncertainty is often referred to as variability. The present study distinguishes between two types of ontic uncertainty: process variability and normative uncertainty. Process uncertainty relates to variability in natural or social processes, such as the inherent variability of the weather. Normative uncertainty relates to the existence of a fundamental plurality of social, ethical or normative considerations. An example of the latter is that individuals have fundamentally different views on wellbeing and the severity of illnesses.
Third, the range of uncertainty relates to the way uncertainty can be expressed, either as a statistical uncertainty or as a scenario uncertainty. A statistical uncertainty range is appropriate when uncertainties can be adequately expressed in statistical terms, for example, as a central estimate and an interval around it. However, deeper forms of uncertainty are frequently at play. These can often not be adequately described in terms of chances or probabilities, but can only be specified in terms of a range of possible events (scenarios). In absence of information on the relative likelihood of each scenario, they are usually treated as being equally plausible. Scenario uncertainties are often construed in terms of what-if statements.
Fourth, recognized ignorance concerns those aspects of uncertainty for which we cannot establish any useful estimate, for example due to processes that have been identified but that are yet poorly understood. Unrecognized ignorance is excluded from the typology, because it concerns pure ignorance about which we cannot say anything knowledgeable: we do not know what we do not know. However, experts may acknowledge that they are ignorant about particular sources of uncertainty and that this limits the reliability of the conclusions of their studies.
Fifth, the methodological unreliability of an element of an assessment reflects weaknesses in methodological quality. It is often not possible to quantitatively establish the accuracy of a model. In those cases, one may instead use qualitative judgments to express in what ways scientific knowledge is limited. Scientific peers may judge the methodological rigor of the procedures followed. This methodological rigor can, for instance, be determined by looking at the theoretical and empirical basis, the reproducibility of the assessment and its acceptance in the peer community.
Sixth, there can be value diversity among analysts in scientific practice. Value here refers to personal values and normative judgments, instead of to numerical values. Value diversity is often reflected in the existence of alternative assumptions in the assessment. Also, assumptions made by one expert can be contested by another expert. Assessors often have considerable freedom in making choices about the design of their assessment and the interpretation of data. These choices may be influenced by different underlying epistemic, socio-cultural and practical values held by the assessors. An example of a socio-cultural value is to base the assessment on worst-case assumptions, reflecting a risk-avoiding attitude. Experts with a risk-seeking attitude may find worst-case scenarios less relevant and might prefer best-case assumptions to inform a decision \[[@B34]\].
The different categories of the location of uncertainty are used to structure this paper. The other characteristics are discussed in each of the sections. The various types of uncertainty will be illustrated using examples from the eBoD literature. In Table [2](#T2){ref-type="table"}, some of these illustrations have been characterized according to the uncertainty typology. These examples are referred to with (Table [2](#T2){ref-type="table"}, section \#) in the remainder of this manuscript, in which \'section \#\' refers to the corresponding number in Table [2](#T2){ref-type="table"}. We will also point towards methods that can be used to deal with different types of uncertainties.
######
Illustrations of characterizations of uncertainties in environmental burden of disease assessments
----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
Source of uncertainty *Nature*\ *Range*\ *Recognized ignorance* *Methodological unreliability* *Value diversity among analysts*
**E**pistemic/Ontic (**Pro**cess Variability/**Nor**mative Uncertainty) **St**atistical/**Sc**enario
----------------------------- ------------------------------------------------------------------------------------------------------- ------------------------------ ------------------------ -------------------------------- ---------------------------------- ----
CONTEXTUAL UNCERTAINTY
1 Multiple ways of defining the \'total environment\' E/Nor Sc \- \+ ++
2 Only including diseases that cause at least 1% of the global burden of disease Nor Sc \-- \-- \+
MODEL STRUCTURE UNCERTAINTY
3 Specific form of the exposure-response relationship is unknown E Sc \+ \+ \+
4 Evidence for causality (environmental factor leading to health effect) is weak and contradicting E Sc ++ ++ \+
5 Incomplete understanding of the joint effect of smoking and radon in relation to lung cancer E Sc \+ \+ \+
6 Accounting for susceptible groups if the available relative risk is not representative for this group Pro/E St \+ \+ \+
PARAMETER UNCERTAINTY
7 Determining a relative risk (RR) for long-term exposure to PM~10~ E St \+ \+ \-
8 Applying an American RR for PM~10~to the Netherlands E Sc ++ \+ \+
9 Use of severity weights Nor/E Sc \+ \+ ++
INPUT DATA UNCERTAINTY
10 Extrapolating non-assessment-specific exposure measurements E Sc ++ \+ \+
11 Measuring population exposure E St \+ \+ \-
----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
Contextual uncertainty
----------------------
Contextual uncertainty stems from choices made about system boundaries and definitions used in an assessment. In eBoD studies, the definition of the environmental factor(s) considered, the associated health outcomes, the links between these, and the scenarios used in the study (including the study area, affected population, and time frame) have to be agreed upon.
### Examples of contextual uncertainty in environmental burden of disease assessments
#### Defining environment
Defining \'environment\' is not always straightforward. Whereas assessments on single risk factors can generally define exposure relatively easy, broader multiple factor analyses (about for example transport, agriculture or the total environment) need to define these boundaries more carefully.
\'Environment\' has been defined to exclude genetics, diet and smoking behavior, but include for instance effects of the natural environment such as dust exposure and natural disasters \[[@B2]\]; include physical, chemical and biological human-made or influenced exposures, but exclude occupational health and safety, the majority of traffic, war, and life-style factors \[[@B3]\]; include all the physical, chemical and biological factors external to the human host and all related behaviors, but exclude those natural environments that cannot reasonably be modified \[[@B5],[@B24],[@B35]\], etc. These definitions can have a significant influence on the outcome of an assessment. In technical terms, this uncertainty can be addressed by thoroughly defining the terms and scope of an assessment. However, this does not change the fact that different scopes and definitions are theoretically possible. Therefore, this contextual uncertainty (Table [2](#T2){ref-type="table"}, section 1), has an epistemic component, because we can not yet gauge the complete extent of the environment; and a normative component, because different researchers hold different normative views on what the environment consists of.
#### Defining health
Many assessments define health quite clinically, including only adverse health effects that have a medical diagnosis. However, a broader definition -- such as used by WHO \[[@B6]\] stating that health is a state of complete physical, mental and social well-being and not merely the absence of disease or infirmity -- also includes less severe health effects. For instance, the burden of disease related to noise varies significantly depending on whether noise annoyance and sleep disturbance are considered health effects.
A pragmatic and normative approach to define which diseases to include in an assessment has been employed by Smith et al. \[[@B2]\]. They included only disease categories that cause at least 1% of the global burden of disease (Table [2](#T2){ref-type="table"}, section 2). Other cut-off percentages could also have been adopted, making this source of uncertainty a form of scenario uncertainty with a normative nature and a degree of value diversity among analysts. Since it is known which diseases are excluded and why, recognized ignorance and methodological unreliability do not play a substantial role.
### Dealing with contextual uncertainty
Results of an assessment can be very sensitive to the definitions and system boundaries chosen. Most of these definitions cannot be harmonized across assessments, because they are dependent upon the purpose of a specific assessment \[[@B36]\]. There is not one single way to deal with contextual uncertainties, but a few general guidelines can be given. In summary, the chosen definitions and boundaries need to be discussed, reported and consistently used \[[@B2],[@B37]\]. This process may often need to involve relevant stakeholders. Even though stakeholder discussions may not reduce the uncertainties, they at least help to reveal them \[[@B1]\]. If more than one sensible definition can be made about an element in the assessment, thus leaving room for value diversity, multiple analyses can be run using different sets of definitions. This is especially useful when there is controversy about which definitions are most appropriate, or when the differences between definitions are considerable. Sensitivity and decision analyses can help to identify which sources of uncertainty mostly affect the final results \[[@B8]-[@B10]\].
Model structure uncertainty
---------------------------
Model structure uncertainty relates to uncertainty about the causal structure of the modeled system: uncertainty within the boundaries chosen. Various interpretations might prevail about the dominant variables and their causal relationships. Because of the many difficulties in studying the large scale low exposure environmental health risks that are so typical of the modern Western world, different views about the model structure often exist.
### Examples of model structure uncertainty in environmental burden of disease assessments
Even when the assessment context is agreed upon, that does not automatically mean that all potentially relevant variables are included in the assessment. For example, climate change may affect health in ways that are as yet unexpected and which can therefore not be assessed \[[@B38]\]. Model structure uncertainty can also relate to the applicability and form of exposure-response relationships (for instance, threshold versus no threshold, or linear versus nonlinear) (Table [2](#T2){ref-type="table"}, section 3) \[[@B39],[@B40]\]. Additionally, evidence for causality may not always be available or in agreement (Table [2](#T2){ref-type="table"}, section 4). An example is the inconclusiveness of the evidence for an association between noise exposure and cardiovascular impacts. Some reviewers find the evidence for this relationship sufficient, whereas others state that it is limited \[[@B41]\]. Similarly, a WHO eBoD study \[[@B42]\] based their estimate of the asthma-related burden of disease for children on a relative risk that Smith \[[@B43]\] considered not sufficiently robust. A comparable debate runs for the long-term health impacts of air pollution, which are as yet rather uncertain. Since different decisions can be made about whether sufficient evidence for causality exists \[[@B44],[@B45]\], such uncertainty is characterized as scenario uncertainty. The recognized ignorance about the existence of causality, which is partly due to methodological unreliability, could be reduced by further research, which indicates its epistemic nature.
Other examples of model structure uncertainty relate to partly unknown patho-physiological mechanisms, the use of proxies, potential latency times, vulnerable groups, co-morbidity and multi-causality. Multiple risk factors can simultaneously affect multiple health outcomes. Environmental stressors can cause health effects through intermediate factors and feedback systems. There may also be other correlated risk factors with common social and behavioral determinants \[[@B7],[@B46]-[@B48]\]. When assessing such interacting risk factors, it is necessary to know whether their effects are additive (separate effects added), synergistic (separate effects multiplied), or antagonistic (separate effects reduced) \[[@B49]\]. Incomplete understanding of the joint effect of smoking and radon with regard to lung cancer remains a key uncertainty in assessing the risk of indoor radon \[[@B50]\] (Table [2](#T2){ref-type="table"}, section 5). Similarly, different methods exist to account for specific susceptible population sub-groups when no representative exposure-response functions exist (Table [2](#T2){ref-type="table"}, section 6).
### Dealing with model structure uncertainty
Model structure uncertainty is often predominantly epistemic -- relating to incomplete or contradictory knowledge -- and more research can increase understanding and possibly reduce uncertainty.
Refsgaard et al. \[[@B51]\] reviewed strategies for assessing model structure uncertainty and present a framework for assessing the uncertainties of predictive models. It involves the use of multiple conceptual models, assessment of their pedigree, and reflection on the extent to which the sampled models adequately represent the space of plausible models. Additionally, sensitivity and decision analyses can provide information about the relative importance of variation between different alternative assumptions \[[@B8]-[@B10]\]. Widely used are also Bayesian belief networks, which can be used to assess multiple model structures \[[@B52],[@B53]\].
However, resources often limit the possibility of running extensive alternative calculations, and pragmatic choices need to be made. Therefore it is most important to document the assumed conceptual and technical model structure in a transparent way, to explore and document which limitations or other viewpoints exist, and to reflect on what this means for the robustness of the results. A graphical representation of the model showing which variables and linkages are included and excluded in the assessment increases the understanding of the model structure \[[@B54]\]. Standardization of the way such causal diagrams are presented -- all using the same convention for what certain shapes of boxes and types of arrows precisely mean -- is to be recommended. Finally, if very large disagreement or ambiguity about the model structure exists, one might also consider not to carry out an eBoD study in the first place.
Parameter uncertainty
---------------------
Parameters are used to describe a relationship between variables. They can be descriptive (such as relative risks, duration estimates, or attributable fractions) or normative (such as maximum life expectancy, severity weights, policy norms, age weights and discount factors).
### Examples of parameter uncertainty in environmental burden of disease assessments
#### Relative risks and attributable fractions
The most common descriptive parameter used in eBoD calculations is the relative risk (RR), which indicates the ratio of the risk of a disease or death among those exposed to the specified factor to those not exposed. The RR is usually derived from an epidemiological study or a meta-analysis of such studies, and subsequently applied to the specific study context. The epidemiological studies from which the RR stems can in themselves form a source of uncertainty (Table [2](#T2){ref-type="table"}, section 7). The methods used to derive RRs are fairly common, which limits value diversity, at least among epidemiologists. However, uncertainty can relate to differences in study design or measurement errors, giving rise to potential methodological unreliability \[[@B1]\]. For some environmental risks, especially new and emerging risks such as electromagnetic fields or genetically modified foods, RR estimates are available only to a limited degree or not at all. Depending on the assessment context, it can be debated whether specific RRs can be extrapolated to other regions, time periods, substance mixtures, or population sub-groups \[[@B7],[@B39],[@B55]\]. An example is the use of RRs for the long-term effects of PM~10~, which are currently mainly available from studies in the United States. The validity of the use of such risk measures in burden of disease studies in other countries is disputable, since air pollution mixtures (for which PM~10~is an indicator) and average population susceptibility may vary between countries \[[@B11],[@B56]\] (Table [2](#T2){ref-type="table"}, section 8). Similarly, in assessing the health impacts of climate change, problems have been encountered when long-term effects have had to be extrapolated from short-term associations \[[@B38]\].
#### Severity weights, age weights and discounting
Normative parameters such as severity weights for diseases or age weights and discounting factors for future health gains are generally based on the judgments of clinicians and economists respectively. They are therefore subjective interpretations of a number for which no \'true value\' exists (Table [2](#T2){ref-type="table"}, section 9) \[[@B57]\]. Different values will prevail depending on who is being asked, their age, gender, occupation, socio-economic status, cultural background and education level, amongst other things. This raises the question of the transferability of these weights to other situations. Estimates of severity weights also depend on the way health effects are presented to the people who are asked to make the valuation, the range of health effects to be valued in the same session, and the valuation methods \[[@B58]\]. As an extreme example, the severity weight for severe noise-related sleep disturbance has been estimated at 0.01 (as used in \[[@B20],[@B59]\]) up to as high as 0.12 \[[@B59]\]. Alternative but still realistic assumptions for all normative parameters mentioned above (severity weights, age weights and discounting) can lead to major differences in DALYs, by up to a factor of four \[[@B57]\]. This large variation should be a reminder of the need for caution in the use of such indicators for policy purposes. If the choice of policy is sensitive to the precise value of the indicator, then an indicator with a large concealed uncertainty may be worse than none at all.
### Dealing with parameter uncertainty
Parameter uncertainty (together with input data uncertainty, discussed next) is commonly quantitatively assessed in eBoD studies, through the use of statistical analyses. Confidence intervals (CI) of parameters such as the RR are used to calculate overall CIs for DALYs. This approach is only suitable for statistical uncertainty. However, scenario uncertainty (Table [2](#T2){ref-type="table"}, sections 8 and 10) on issues for which various interpretations exist (i.e. value diversity among analysts) is more difficult to represent in CIs. For these uncertainties, similar approaches as described for contextual and model structure uncertainty can be useful: sensitivity and decision analyses in combination with transparent reporting.
Input data uncertainty
----------------------
Uncertainty in input data may relate to a range of factors, including a lack of data, inaccurate measurements, or extrapolated data. Exposure data and disease data (incidence, prevalence or mortality data) are the most common input data sets needed for eBoD calculations.
### Examples of input data uncertainty in environmental burden of disease assessments
The greatest source of uncertainty pertaining to input data in eBoD assessments generally relates to a lack of assessment-specific measurements. For exposure data, one frequently has to rely on proxies for exposure, such as modeled environmental concentrations. Furthermore, for many risk factors, data on exposure or concentration distributions are available for only a limited number of years, regions, countries or demographic groups \[[@B5]\]. If no further monitoring can take place, assessors might need to extrapolate non-assessment specific data (Table [2](#T2){ref-type="table"}, section 10). In climate change research, for example, impacts often relate to future exposures which cannot be measured and hence need to be modeled \[[@B38]\]. In a WHO study on solid fuel use across 181 countries \[[@B5],[@B60]\], a combination of survey data and modeled data were used. Whether the modeled data can be meaningfully used may be judged differently by various scientists, leading to potential value diversity among analysts. Overall, Prüss-Üstün et al. \[[@B5]\] concluded that only for three environmental risk factors -- water sanitation and hygiene, solid fuel use, and outdoor air pollution -- were the necessary methodology and enough exposure data available to make sensible global estimates at country level.
Even when exposure can be measured, different measuring methods may lead to different results (Table [2](#T2){ref-type="table"}, section 11). An example is the measurement of noise exposure levels, which can differ by up to 10 dB(A) depending on the methodologies used \[[@B61]\]. In studies related to UV radiation and skin cancer, sun exposure of many years before is often estimated using recalled sunburns or time spent in the sun. Such exposure estimates based on self-reporting can differ significantly from measurement data of ambient UV radiation levels \[[@B40]\].
Similar issues play a role for background morbidity and mortality data, which are needed to calculate the estimated number of attributable cases. Such data should ideally stem from empirical research or adequate monitoring in the target population \[[@B62]\]. However, these data are often only available at highly aggregated levels. More specific data, for example on a local scale, suffer from the small-number problem, in that estimates for rare outcomes may be highly unstable. Research shows that indicative uncertainty ranges for regional prevalence rates of 16 important diseases may range from +/- 10 percent to +/- 90 percent \[[@B39]\]. The common solution -- modeling missing data, or extrapolating data of one country to another country -- yields epistemic uncertainty.
### Dealing with input data uncertainty
For morbidity data, models can be employed to calculate missing data and check for consistency in existing datasets. However, past trends in incidence and data inaccuracies can lead to large discrepancies between measurements and model calculations, and their use requires both caution and expert knowledge \[[@B47],[@B62],[@B63]\].
A data quality assessment can be used to evaluate whether input data are suitable for the intended purpose. Such an assessment involves \"the scientific and statistical evaluation of data to determine whether they meet the objectives of the project, and thus are of the right type, quality, and quantity to support their intended use\" \[[@B64]\]. The Numerical, Unit, Spread, Assessment and Pedigree (NUSAP) system \[[@B27],[@B65],[@B66]\] is another method to assess data quality. In addition to the more standard quantitative uncertainty assessment (number, units and the spread of those numbers), the NUSAP approach also includes an evaluation of the reliability of the information (assessment) and its the scientific basis (pedigree).
Using the uncertainty typology in practice
------------------------------------------
In the preceding paragraphs we have outlined ways to deal with various types of uncertainties. In practice, eBoD assessments do not only have to deal with uncertainties, but also with time and budget constraints. It might often not be possible to employ all possible methods to deal with all the uncertainties inherent in the assessment. Therefore, it is necessary to prioritize uncertainties and the work needed to assess or reduce them. Here we shortly describe how to 1. identify and characterize sources of uncertainty; 2. prioritize sources of uncertainties; and 3. select and apply methods for dealing with uncertainties. We will describe how, in all these steps, the uncertainty typology can be used to support the process. Subsequent communication of the results to policy makers will be discussed in the following paragraph.
1\. First, the different sources of uncertainty are to be identified. The generation of this longlist of uncertainty sources can be done using two different approaches: 1) by analyzing each step of the eBoD assessment at hand and subsequently characterizing each source according to the typology, and 2) by considering each possible type from the uncertainty typology and discussing where in the assessment this type of uncertainty may occur. Reasoning from both angles may help to minimize the chance that uncertainty sources are overlooked. The resulting list of uncertainties can be further characterized using the uncertainty typology.
2\. The relative importance of each uncertain element can subsequently be weighted, based on its potential impact on the outcome of the eBoD assessment in question. Where some form of quantification is possible, the relative importance can be assessed by means of sensitivity analysis \[[@B8]-[@B10]\]. However, for many sources of uncertainty, such quantification is not feasible. In that case, the relative importance can be assessed using expert judgment. Two possible approaches include coding and card sorting. In the coding approach \[[@B67]\], experts are asked to go over the longlist of uncertainty sources and code each source as being either of a) a crucial importance; b) an average importance; c) a medium importance or d) a low importance. This is a quick and dirty technique and, to avoid errors and biases, several experts should do this independently and discuss potential differences in their judgments. The card sorting approach (used by e.g \[[@B68]\]) is more advanced and involves organizing an expert workshop. Experts are asked to independently select the top 20% (or another percentage) sources of uncertainty that they consider most important in view of their impact on the eBoD calculation at hand, and sort these according to importance. The uncertainties are displayed on cards to facilitate the sorting -- hence the name. Results from individual experts are combined to arrive at a group ranking of the items on the longlist. Arguments used by the experts to defend their ranking need to be documented and special attention should be given to reasons for any substantial disagreement on the importance of a particular uncertainty source.
3\. Once the prioritization has been done, suitable tools can be selected for further analysis of the key uncertainties identified. Each uncertainty type may require a different method to address it, and to gauge its impact on decision making. The uncertainty tool catalogue by Van der Sluijs et al. \[[@B69]\] provides guidance for selecting appropriate methods that match the characterization of the uncertainty in the typology. Refsgaard et al. \[[@B70]\] also describe various methods for dealing with uncertainties, and explain which purposes they may serve.
It may not be possible to correctly identify, characterize and prioritize all sources of uncertainty in the beginning of an assessment. The typology may thus need to be reassessed throughout the project. New sources of uncertainty may be added or their weights may be adjusted. The uncertainty typology should therefore be used interactively throughout the study. As such, it also provides a framework to keep track of all sources of uncertainty, so that sources identified early in the project -- especially those that cannot be quantified -- are not forgotten at the end of the study, when results are reported.
Communicating uncertain results to policy makers
------------------------------------------------
Most policy makers will feel more comfortable when making decisions based on single, undisputed numbers with small uncertainty ranges, than on ambiguous or controversial estimates and scenario analyses. However, unfortunately that is often not the way complex processes can be described. On the other hand, giving policy makers a lengthy report listing all the possible uncertainties will not necessarily lead to informed policy making either. Scientists can help policy makers by assessing which uncertainties are most relevant for the policy decisions to be made. They can identify policy options that are robust given these uncertainties. If no single best policy option for all scenarios can be determined, all reasonable options can be discussed in a democratic process including scientists, stakeholders, policy makers and politicians \[[@B71]\]. As the communication needs of all these parties can vary greatly, a single mode of risk communication is rarely sufficient.
Uncertainties can be communicated linguistically, numerically, or graphically. Confidence intervals can be provided reflecting uncertainty in parameters and input data. For uncertainties that cannot be expressed in statistical intervals, other characterizations of likelihood can be used. Risbey et al. \[[@B72]\] have proposed expressions for different levels of precision, ranging from full well defended probability density functions, to percentile bounds, first order estimates, expected signs or trends, ambiguous signs or trends and, finally, effective ignorance. Many of the uncertainties identified in our study cannot be captured quantitatively, but some can be expressed in these latter characterizations of precision. Additionally, if any policy recommendations are made, the strength of these recommendations and the quality of the underlying evidence can be expressed using a uniform grading system \[[@B73],[@B74]\]. Such a systematic and explicit approach to judging the quality of evidence and the strength of recommendations can facilitate appraisal of these judgments, and improve communication \[[@B73],[@B74]\]. Providing a graphical representation of the underlying model in a standardized way can support further understanding of the assessment context and model structure.
In order not to overwhelm the user of the assessment results with uncertainties, the concept of progressive disclosure of information can be employed \[[@B75],[@B76]\]. This involves tailoring the information about uncertainty to the target audience. In a press release or a project summary, for example, the uncertainties that are most relevant to the final policy decisions need to be described, without any technical details. As such, a policy maker using the results of an eBoD assessment will not be directly confronted with a typology of all uncertainties, but will be provided with the information needed to properly interpret the results. The main assessment report may subsequently contain more detailed information, with emphasis on the nature, extent and sources of uncertainties. Ideally, it presents all methods, assumptions, parameters and input data, thereby providing maximum transparency of the assessment approach. Even though DALYs are made to reduce complex information to single numbers, it is essential to allow readers to unravel the DALYs and, when desired, reproduce them \[[@B11],[@B57]\] or recalculate any estimates using their own data or assumptions \[[@B77]\].
The assertion that burden of disease figures can only be properly interpreted when presented with assessment-specific, informative and complete background information leads to a second consideration related to the usability of assessment results. They can only be used for the specific purpose for which they were derived, and should not be used in other assessments or for other policy purposes.
Discussion
==========
Disability Adjusted Life Years -- or other forms of aggregated health measures -- can be very attractive indicators for policy makers. The measure combines information about the magnitude, severity and duration of adverse health effects into one number, thereby providing a means to compare otherwise incomparable environmental health problems. This simplification of the complex underlying reality is the defining advantage of the measure, but it also presents pitfalls. We have shown that various types of uncertainty can influence environmental burden of disease (eBoD) assessments and their output, thereby potentially influencing policy decisions based on these assessments. Statistically quantifiable uncertainty in parameters and input data -- the type of uncertainty that is usually well communicated in eBoD assessments -- is far from the only type of uncertainty, or even the most important. Variations in definitions of the environment, the health effects, and the scenarios assessed, unknown impacts of multi-causality and co-morbidity, lacking consensus about causality, controversial views about model structures, and many other sources of uncertainty may affect eBoD assessments, but cannot be easily quantified, and are usually not fully addressed.
The use of a typology to characterize and structure uncertainties can help to deal with them. Dealing with uncertainties does not necessarily mean reducing them. Much of the time, mere identification and proper communication of uncertainties along with systematic reflection on their policy implications is most important, or -- more practicably -- the only feasible thing to do. Ideally, policies should be robust under the uncertainties that are identified.
The potential extent of uncertainties presented here should not be interpreted as criticism of the DALY approach as such, or as an argument for not using the method. Instead, for some assessments, DALYs can be a very valuable way of presenting the possible extent of environmental health effects to policy makers. Uncertainties do not halt eBoD assessments, but do affect the assessment process and the interpretation and communication of its results. Scientists have the responsibility to assess and communicate assessments in such a way that underlying uncertainties are reflected in the outcomes. Results should not be presented as being more robust than can be inferred from the underlying knowledge base. And policy makers, for their part, have the responsibility to take information about uncertainty seriously and deal with it sensibly \[[@B78]\]. If not, then the interface between science and policy needs to be re-designed, lest misdirected policies be based on a false precision of scientific inputs.
In the meantime, on a meta-level, the methodology for calculating the eBoD and dealing with uncertainties needs to be improved \[[@B58]\]. For example, there should be a study of the disproportionate way in which the uncertainty in small severity weights (such as severe noise-related sleep disruption \[[@B20],[@B59]\]) affects overall assessment uncertainty.
The main limitation of the typology presented here is that it strongly relies on expert judgment and mainly yields qualitative insights. Its main application should be to precede and supplement quantitative uncertainty analysis, and not to replace it. In addition, further research on summary measures can perhaps in the future lead to better measures than DALYs, an area already explored by Murray \[[@B58]\]. It is not possible to completely harmonize methods and knowledge or to standardize datasets. These are often highly assessment-specific, and should stay that way. However, the criteria used for determining which methods and datasets to use should be harmonized as much as possible. Overall, it would be useful to further study the pitfalls of these types of aggregated indicators, and to develop methods to identify and prevent the hyper-sensitivity of policy decisions to overly precise indicators.
Conclusion
==========
Increased awareness of the issue of uncertainty and a well-structured approach towards assessing and communicating uncertainties can help to bring about a more balanced interpretation of the results of eBoD assessments. A typology of uncertainties such as presented in this paper can be used to systematically identify and map key uncertainties. As such, it precedes and complements quantitative uncertainty assessment. The use of a typology may facilitate a structured dialogue between scientists and stakeholders on possible sources and types of uncertainty. This may help the key actors to achieve a common understanding of the uncertainties and their importance.
Abbreviations
=============
CI: Confidence Interval; DALE: Disability Adjusted Life Expectancy; DALYs: Disability Adjusted Life Years; eBoD: Environmental burden of disease; HALE: Health Adjusted Life Expectancy; HALYs: Health Adjusted Life Years; PM: Particulate Matter; QALYs: Quality Adjusted Life Years; RR: Relative Risk; UV: Ultraviolet; WHO: World Health Organization
Competing interests
===================
The authors declare that they have no competing interests.
Authors\' contributions
=======================
AK initiated the research and studied the environmental burden of disease literature. AP and JPS designed and described the uncertainty typology. All authors discussed the application of the uncertainty typology to the case of environmental burden of disease assessment. AK drafted the main manuscript using input from ACP, JS and EL. All authors read and approved the final manuscript.
Acknowledgements
================
This work was partly financed through RIVMs strategic research budget. Thanks go out to David Briggs and participants of the EU Intarese project for providing a useful context for this work. Furthermore, we are grateful to Bert Brunekreef, as well as to the reviewers, for useful comments and suggestions.
| {
"pile_set_name": "PubMed Central"
} |
Background
==========
Innovations in genomic technologies provide new tools for enhancing productivity and wellbeing of domestic animals. Genomic selection, where genetic merit is predicted from genome-wide single nucleotide polymorphism (SNP) genotypes \[[@B1],[@B2]\], is used in the dairy industries in a number of countries \[[@B3],[@B4]\]. The rapid uptake of this technology has been driven by both the availability of commercial high-density SNP chips, and increased genetic gain over traditional progeny testing largely as a consequence of reduced generation interval and increased accuracy of selection at a younger age \[[@B5]-[@B7]\].
A number of SNP chips from Illumina ( <http://www.illumina.com>) and Affymetrix ( <http://www.affymetrix.com>) are available for cattle. These include 3K \[[@B8]\], 7K \[[@B9]\], 15K \[[@B10]\], 25K \[[@B11]\], 50K \[[@B12]\] and more recently 800K from Illumina, and 650K and 3 million SNP panels from Affymetrix. In addition next generation sequencing technologies for low-cost sequencing of whole genomes are now available \[[@B13]\]. Use of genotypic data from high-density SNPs potentially can increase accuracy of genomic selection but also the total cost of genotyping/sequencing. As new higher density chips are developed, re-genotyping previously genotyped samples or new samples with new chips or whole genome sequencing is expensive. For some applications, such as selection of heifers to be retained in the dairy herd or selection in beef production systems, low-density SNP panels e.g. 3-7K may be the only cost effective option (e.g. \[[@B14]\]). If low-cost genotyping could be useful, very large numbers of animals can be genotyped on a routine basis.
Accuracy of genomic predictions based on different subsets of low-density SNP panels up to 50K have been compared in a number of studies \[[@B15]-[@B18]\]. A common finding is that accuracy of genomic prediction for young animals increased as the number of markers increased from a few hundred up to all SNPs from 50K SNP chip. There are several possible strategies how to select loci for low-density panels \[[@B17]\]. However, instead of using lower density SNP in genomic prediction, a promising approach is to genotype a small proportion of the population with a high-density SNP panel and then employ genotype imputation methods for predicting high-density genotypes for the rest of the population genotyped with a lower density SNP panel (e.g. \[[@B8],[@B9]\]). Genotypic imputation is defined as the prediction of genotypes at the SNP locations in a sample of individuals for which assays are not directly available. These *in silico* genotypes obtained by imputation, *albeit* with some uncertainty, can then be used in genome-wide association and genomic selection analyses (e.g. \[[@B19],[@B20]\]). Such strategies are likely to result in more accurate predictions of genomic breeding values \[[@B21]\], improved ability to resolve or fine-map QTL or QTN, and integration and meta-analysis across large datasets with heterogeneous SNP information \[[@B22]\].
A number of imputation software programs (fastPHASE \[[@B23]\], MACH \[[@B24]\], IMPUTE \[[@B25]\], Beagle \[[@B19]\], PLINK \[[@B26]\], DualPhase \[[@B27]\]) have been used to infer missing or untyped genotypes based on known information derived from flanking markers. A number of studies on imputing genotypes have been published in dairy cattle \[[@B21],[@B28]-[@B33]\] using 50K data and more recently high-density SNP panels \[[@B34]-[@B36]\] reporting accuracies of imputation from lower SNP panels to 50K and up to high-density SNP panels examining different methods of imputation, often using limited number of scenarios and strategies of using test and reference panels. The direct comparisons across such studies are thus often difficult. Various factors affecting the accuracy of imputation require further systematic investigation. The accuracy of imputation can be improved by increasing the size of the reference population \[[@B37]\]. For some resource population the animals genotyped with different SNP panels are available. Such genotype resources can be better utilised by imputing in a tiered framework, utilising multiple reference panels, which might result in improved accuracy of imputation in the study samples \[[@B38]\].
The objectives of this study were to evaluate the accuracies of imputation using three different population based methods of imputation, different size of reference and test panels, different imputation strategies, different SNP array platforms, effect of relationship between reference and test animals and finally examine the effect of using imputed genotypes on the accuracy of genomic selection.
Methods
=======
Data
----
In total four datasets genotyped with four different SNP chips (Table [1](#T1){ref-type="table"}) were used. The largest dataset consisted of 2,727 (2,205 bulls and 522 cows) Australian Holstein-Friesian cattle \[[@B17]\] genotyped with Illumina BovineSNP50 BeadChip \[[@B12]\]. A second more recent dataset consisted of 845 Australian Holstein-Friesian heifers genotyped with Illumina 800K BovineHD beadChip (Illumina Inc., San Diego, CA). After applying quality control (minor allele frequencies (MAF) \>0.01, call rate\>0.9, Hardy Weinberg Equilibrium (HWE) P\>0.0001) a total of 42,136 and 610,879 autosomal SNPs from the 50K and the 800K chips, respectively, were used in the present study (Table [1](#T1){ref-type="table"}). In addition any genotype showing Mendelian inconsistencies was set to missing.
######
Description of different SNP chips and SNP subset panels
**Label used for SNP panel in this study** **SNP chip** **Number of SNPs on chip** **Filtered SNPs used in this study** **Remarks**
-------------------------------------------- --------------------------------- ---------------------------- -------------------------------------- --------------------------------------
15K 15K (ParAllele/Affymatrix) 15,036 205 SNPs from BTA20
25K 25K (Affymatrix) 25,068 328 SNPs from BTA20
50K Illumina BovineSNP50 BeadChip 54,001 42,136
3K Illumina BovineSNP50 BeadChip 3,000 3,000 Evenly spaced Subset of 50K
5K Illumina BovineSNP50 BeadChip 5,000 5,000 Evenly spaced Subset of 50K
10K Illumina BovineSNP50 BeadChip 10,000 10,000 Evenly spaced Subset of 50K
20K Illumina BovineSNP50 BeadChip 20,000 20,000 Evenly spaced Subset of 50K
35K Illumina BovineSNP50 BeadChip 35,000 35,000 Evenly spaced Subset of 50K
BovineLD 7K Illumina BovineLD BeadChip 6,909 6,662
Bovine3K Illumina Bovine3K BeadChip 2,900 2,500
800K Illumina 800K BovineHD beadChip 786,799 610,879
800K-imputed Illumina 800K BovineHD beadChip 786,799 610,879 Imputed best guess genotypes
800K-dosage Illumina 800K BovineHD beadChip 786,799 610,879 Imputed dosage for B-allele
49K Illumina BovineSNP50 BeadChip 54,001 49,394 Common SNP between 800K and 50K chip
Of the 2,205 bulls with 50K genotypic information, 1,419 were previously genotyped for 15K \[[@B10]\], and 431 for 25K ( \[[@B11]\], <http://www.affymetrix.com>). These datasets were used to test the accuracies of imputing SNP genotypes between different chips. The animals in all these datasets are related in a complex pedigree structure. The distributions of relatedness in the form of boxplots of pedigree kinship among animals in different datasets are given in Additional file [1](#S1){ref-type="supplementary-material"}.
Imputation Scenarios
--------------------
Animals were divided into reference and test sets for evaluating the accuracies of imputation. The animals included in the reference set have genotypes derived from the high-density SNP panel and the animals in the test set have genotypes from the lower density SNP panel. The lower density SNP panels of the test sets were created by using a subset of the genotyped SNPs. The rest of the genotypes of the test sets were masked and used to compute the accuracy of imputation. A number of imputation scenarios were generated by combining different reference and test sets and SNP densities. The animals (2,727) genotyped with 50K were divided into 8 different combinations of reference and test sets as presented in Table [2](#T2){ref-type="table"}. Reference animals in reference-test-ID 1--4 are a random sample of older bulls born before 2001. The 27 bulls for reference-test-ID 5 are key ancestors of the Australia Holstein-Friesian population. In reference-test-ID 8, younger bulls born between 2001 and 2004 are in the test set and all older bulls born before 2001 in the reference set.
######
Composition of reference and test sets for evaluating imputation accuracy up to 50K
**Reference-test-ID** **Data** **Reference set** **Test set** **Total (animals)**
----------------------- ---------- ------------------- -------------- --------------------- ------ ---- ---------------------- ------
1 50K 1363 50 bulls 1364 50 bulls+cows 2727
2 50K 681 25 bulls 2046 75 bulls+cows 2727
3 50K 272 10 bulls 2455 90 bulls+cows 2727
4 50K 136 5 bulls 2591 95 bulls+cows 2727
5 50K 27 1 key bulls 2700 99 bulls+cows 2727
6 50K 2205 81 all bulls 522 19 all cows 2727
7 50K 522 19 all cows 2205 81 all bulls 2727
8 50K 1753 80 training set bulls 452 20 test set young bulls 2205
The total number of animals (2,727) consisted of 2,205 bulls and 522 cows.
To examine the effect of pedigree relatedness between test and reference animals on the accuracy of imputation, the test animals with sire and without sires in the reference set were compared. In addition the highest value of pedigree kinship for each test animal with reference animals was computed. The test animals were classified into four interval categories with respect to their highest pedigree kinship viz. 0.0-0.01, 0.01-0.1, 0.1-0.2 and 0.2-0.4. The accuracy of imputation of the test animals in these four categories was compared using IMPUTE2.
For the 800K dataset, the 845 heifers were randomly divided in two subsets of approximately equal size *i.e.* 425 in the reference and 420 in the test set. This framework of imputation is referred here as a '2-tiered' framework. This was extended to a '3-tiered' framework by including an additional panel of 2,205 bulls with 50K SNP genotypes as a middle tier (Figure [1](#F1){ref-type="fig"}). An additional scenario using fewer animals in the top-tier was generated by randomly selecting 45 out of 425 reference heifers. The imputation for the tiered framework was performed with IMPUTE2 using the two reference panels in the same run.
{#F1}
Generating low-density SNP panels
---------------------------------
To mimic various low-density SNP panels, different subset of 50K SNPs were selected for the test sets. The SNP densities equivalent to 3000, 5000, 10000, 20000 and 35000 evenly spaced autosomal SNPs were generated by iterative thinning the SNPs based on spacing and MAF of SNPs (Table [1](#T1){ref-type="table"}). In each iteration, a SNP pair with the smallest interval was identified and the SNP with lower MAF was removed from the pair. A total of 1,324 SNPs on chromosome 20 from the 50K panel were used for the initial analyses to compare the imputation programs for different scenarios. The best method of imputation identified was then used for analysing all the autosomal SNPs from the Illumina Bovine3K and Illumina BovineLD 7K BeadChip (Illumina Inc., San Diego, CA) for assessing the comparative utility of imputed genotypes from these commercial panels up to 50K for genomic prediction.
Most of the SNPs on the 50K chip are present on the 800K chip. For the scenarios using the 800K panel the lower density SNP panels for the test set consisted of common SNPs between 800K and 50K as well as between 800K and Illumina Bovine3K and Illumina BovineLD 7K, respectively (Table [1](#T1){ref-type="table"}).
Imputation methods
------------------
Population based imputation methods rely on linkage disequilibrium relationship between SNPs, and essentially consist of two steps viz. inference of haplotypes and imputing untyped genotypes in the test set using information from the best fit haplotypes derived from the reference panel. We compared three commonly used population-based programs for imputing missing genotypes which don't rely on pedigree information viz. IMPUTE2, fastPhase and Beagle.
We used IMPUTE2 version 2.1.2 in this study which implements a Hidden Markov Model (HMM). The details of the algorithm are given in \[[@B25]\].The algorithm involves estimating haplotypes using all the SNP in reference set and then imputing the alleles at untyped SNPs in the test set based on the best fit haplotypes estimated from the reference. IMPUTE2 requires to specify the effective population size as an input parameter. This was set to 100 which is within the range of the effective population size reported for Holstein-Friesian dairy cattle \[[@B39],[@B40]\].
We used fastPHASE version 1.2.3 \[[@B23]\]. fastPhase uses a haplotype clustering algorithm which is based on the observation that haplotypes in a population tend to cluster into groups of closely related or similar haplotypes over a short region. fastPhase requires the number of clusters K as input and was set to 20 in this study.
Beagle version 3.3 is also based on a local haplotype-clustering model (as detailed in \[[@B19]\], \[[@B37]\]), similar to fastPHASE, but allows for a variable number of clusters across a region. Beagle uses a localized haplotype cluster-model to cluster haplotypes at each marker and then defines a HMM to find the most likely haplotype pairs based on the individual's known genotypes. The most likely genotype at untyped loci is generated from defined haplotype pairs. We used the option where reference and test panel are defined separately. Imputation was performed for each chromosome separately for all the three methods. Except the above mentioned parameters, programs were run with default parameters.
Accuracy of imputation
----------------------
All the three imputation methods provide the probability of the three possible genotypes at each missing genotype. We used the most likely genotype as the predicted genotype. For incorrectly imputed genotypes it is possible to impute one or both alleles incorrectly. To distinguish between these two cases, we computed the accuracy of imputing alleles as the percentage of correctly predicted alleles, and the allelic error rate of imputation as the percentage of incorrectly predicted alleles *i.e.* mean allelic error rate (%) = number of incorrectly predicted alleles / total number of alleles imputed in the test set × 100. In general allelic error rates are just slightly more than half of genotypic error rates. Accuracy of imputation was also computed as the percentage of correctly predicted genotypes for the masked genotypes.
800K imputed dataset for genomic prediction
-------------------------------------------
The data on 2,205 bulls genotyped with 50K were imputed, with IMPUTE2, up to 800K using 845 heifers genotyped with 800K as reference and using most likely genotype as the predicted genotype ('800K-imputed', Table [1](#T1){ref-type="table"}). In addition the dosage/copies of the B allele for each genotype was computed as *p*~*AB*~+2×*p*~*BB*~, where *p*~*AB*~ and *p*~*BB*~ are imputed probabilities of AB and BB genotypes, respectively. This measure takes into account the uncertainty of imputation and is an appropriate measure when using an additive model in genomic prediction and genome-wide association studies. These two datasets of 2,205 bulls with imputed genotypes ('800K-imputed') and imputed dosage ('800K-dosage') for 610,879 autosomal SNPs were used to compute genomic prediction.
Accuracy of genomic prediction
------------------------------
Accuracy of direct genetic values (DGV) using imputed and actual genotypes was investigated by dividing the data on 2,205 bulls in a training set of 1,753 bulls born between 1955 and 2000 and a validation/test set of 452 young bulls born between 2001 and 2004. SNP effects were obtained from the solution of the following mixed model equations \[[@B41],[@B16]\]
$$\left\lbrack {\begin{matrix}
{1'\mathbf{R}^{- 1}1} \\
{\mathbf{X}'\mathbf{R}^{- 1}1} \\
\end{matrix}\mspace{9mu}\begin{matrix}
{1'\mathbf{R}^{- 1}\mathbf{X}} \\
{\mathbf{X}'\mathbf{R}^{- 1}\mathbf{X} + \lambda\mathbf{I}} \\
\end{matrix}} \right\rbrack\begin{bmatrix}
\hat{\mu} \\
\hat{g} \\
\end{bmatrix} = \begin{bmatrix}
{1'\mathbf{R}^{- 1}y} \\
{\mathbf{X}'\mathbf{R}^{- 1}y} \\
\end{bmatrix}$$
where **y** is a vector of twice the daughter trait deviations (DTD) of bulls, 1 is a column vector of ones of size *N*~*Anim*~ , $\hat{\mu}$ is the general mean, *ĝ* is a vector of the estimated SNP effect, X is an *N*~*Anim*~ × *N*~SNP~ matrix of SNP genotypes coded as 0 (homozygote), 1 (heterozygote), or 2 (other homozygote), or SNP allele dosage. I is an identity matrix of size *N*~SNP~ × *N*~SNP~ , *λ* is a shrinkage parameter derived by cross-validation. R is a diagonal matrix with elements *R*~*ii*~ = (1/*rel*~*i*~)-1, where *rel*~*i*~ is the reliability of the DTD of ith bull. DGV were calculated as $\overset{\frown}{m} = \hat{\mu} + X\hat{g}$.
Five traits were analysed viz. milk yield, fat yield, protein yield, survival and daughter fertility which reflect a range of heritabilities (*i.e.* 0.25, 0.25, 0.25, 0.04 and 0.04, respectively). Phenotype information was provided by the Australian Dairy Herd Improvement Scheme (ADHIS, <http://www.adhis.com.au>). The phenotypes used were daughter trait deviations (DTD) for the bulls. The accuracy of the DGV prediction using subsets of SNP genotypes, and imputed SNP genotypes were compared to the DGV prediction obtained with the all 50K SNP genotypes. The accuracy of DGV prediction was computed as Pearson's correlation coefficient between DGV and DTD of the young bulls in the test data.
Results
=======
Imputation up to 50K
--------------------
### Comparison of imputation methods
The allelic error rates of imputing genotypes on BTA20 by the three imputation methods across different scenarios using evenly spaced SNP subsets in the test sets and different proportion of animals in the reference sets are presented in Figure [2](#F2){ref-type="fig"}. Detailed results on all the 42 scenarios are given in Additional file [2](#S2){ref-type="supplementary-material"}. In general IMPUTE2 has the lowest mean allelic error compared to Beagle and fastPhase, however, the difference between methods varies over different scenarios (Figure [2](#F2){ref-type="fig"}). The difference in error rate of IMPUTE2 and Beagle decreases with increasing size of the reference set and increasing SNP density in the test set (Figure [2](#F2){ref-type="fig"}). fastphase outperformed the other two methods in only one scenario where a higher SNP density (35K) was used in the test set and very few animals (27) were used as reference *i.e.* scenario 29 (Additional file [2](#S2){ref-type="supplementary-material"}). The accuracies of imputation of all the three haplotype based methods are much higher compared to imputation based on a simple sampling strategy using the allele frequencies of SNP in the reference set. The mean allelic error rates obtained from such sampling strategies are in the range of 22.5 to 26.8% for the different scenarios (Additional file [2](#S2){ref-type="supplementary-material"}).
{#F2}
### Effect of SNP density
The accuracy of imputation increases with the number of SNPs in the test set (Figure [2](#F2){ref-type="fig"}, Additional file [2](#S2){ref-type="supplementary-material"}) for all the scenarios and the methods examined here. The mean allelic error rate decreases from 2.80% for the evenly spaced 3K SNP panel to 0.76% for the 35K panel in the scenario where 50% animals are in the reference set (Figure [2](#F2){ref-type="fig"}a). The mean allelic error rate of imputation is lower for the evenly spaced 3K SNP panel (2.80%) compared to the Bovine3K panel (3.34%). There is a large reduction in the mean allelic error rate of imputation when using the 5K evenly spaced SNP panel (1.97%) in the test set (Additional file [2](#S2){ref-type="supplementary-material"}). Further reductions in error rate of imputation by increasing SNP density in the test set to 10K (1.36%), 20K (1.00%) and 35K (0.76%) are relatively smaller (Figure [2](#F2){ref-type="fig"}a).
### Effect of size of reference panel
The mean allelic error rate increases as the number of animals in the reference set decreases (Figure [2](#F2){ref-type="fig"}, Additional file [2](#S2){ref-type="supplementary-material"}). The lowest allelic error rate is obtained when 1,363 (50%) animals are in the reference and the rest in the test set. The mean allelic error rate ranges from 0.76% for the 35K SNP panel to 2.80% for evenly spaced 3K SNP panel using IMPUTE2. The mean imputation error rate for the cows using the bulls as reference ranges from 1.21 to 4.65% and for the bulls using the cows as reference ranges from 0.73 to 3.47% for different SNP densities using IMPUTE2 (Additional file [2](#S2){ref-type="supplementary-material"}).
### Effect of relatedness between test and reference animals
The mean allelic error rates for the test animals with sire and without sire in the reference for all the 42 scenarios using IMPUTE2 are given in Additional file [2](#S2){ref-type="supplementary-material"}. In general test animals with sire in the reference have slightly lower allelic error rate of imputation (2.61% for with sire vs. 3.34% without sire averaged across all the scenarios). We further compared the error rate with kinship estimates of the test animals with the reference animals. The results for the 42 scenarios presented in Additional file [3](#S3){ref-type="supplementary-material"} show that, in general, the mean allelic error rate decreases with the increase in the highest kinship of the test animals with the reference animals. This is more pronounced when the SNP panels in the test set are small and also when the reference size is small.
Imputation between SNP chips
----------------------------
The mean allelic error rates of imputing SNP genotypes between different SNP chips obtained with IMPUTE2 are presented in Table [3](#T3){ref-type="table"}. The results from BTA20 are given as an example. The mean allelic error rates of imputing 15K specific (205 SNPs) genotypes are 0.80%, 0.95% and 1.40% when 25%, 50% and 75% of the animals, respectively, are in the test set and the remainder of the animals with genotypes on 1529 SNPs (15K+50K) in the reference set. The mean allelic error rates of imputing 50K specific (1324 SNPs) genotypes are 2.85%, 3.15% and 4.25% when 25%, 50% and 75% of animals, respectively, are in the test set.
######
Mean allelic error rate of imputing SNP genotypes between different SNP chips obtained with IMPUTE2
**Scenario** **Animals masked (%)** **N animals total** **N animal reference** **N animals test** **N SNP** **N** **%** **Mean allelic error rate (%)**
-------------- ------------------------ --------------------- ------------------------ -------------------- ----------- ------- ------- ---------------------------------
15K by 50K 25 1419 1065 354 1529 205 13 0.80
50 1419 710 709 1529 205 13 0.95
75 1419 355 1064 1529 205 13 1.40
50K by 15K 25 1419 1065 354 1529 1324 87 2.85
50 1419 710 709 1529 1324 87 3.15
75 1419 355 1064 1529 1324 87 4.25
25K by 50K 25 431 324 107 1652 328 20 1.50
50 431 216 215 1652 328 20 1.85
75 431 108 323 1652 328 20 2.75
50K by 25K 25 431 324 107 1652 1324 80 2.75
50 431 216 215 1652 1324 80 2.75
75 431 108 323 1652 1324 80 4.55
The results are shown for three SNP chips viz. 15K, 25K and 50K and chromosome 20.
Similarly the mean allelic error rates of imputing of 25K specific (328 SNPs) genotypes are 1.50%, 1.85% and 2.75% when 25%, 50% and 75% of the animals, respectively, are in the test set. The respective mean allelic error rates of imputing 50K specific (1324 SNPs) genotypes are 2.75%, 2.75% and 4.55%. The error rates in these scenarios are slightly higher compared to the above mentioned corresponding scenarios including 15K, possibly due to a lower number of animals in the reference and the test sets. Overall the results indicate that a reasonable accuracy of imputation for untyped SNP genotypes can be achieved when combining datasets genotyped with these SNP chips.
Comparison of methods for imputation up to 800K
-----------------------------------------------
Only two methods (Beagle and IMPUTE2) were compared for imputing genotypes up to 800K using 50K. We did not include fastPhase in these comparisons because of the long computation time and the lower accuracy of fastPhase observed in the previous analyses within the 50K dataset. The chromosome-wise comparisons of the accuracies of the two methods are presented in Figure [3](#F3){ref-type="fig"}. The mean allelic error for imputing genotypes across different chromosomes ranges from 0.67% for BTA14 to 0.97% for BTA21 using IMPUTE2 and 0.84% for BTA14 to 1.28% for BTA27 when using Beagle. The mean error rates are slightly higher for smaller chromosomes (21--29) compared to larger chromosomes for both the methods (Figure [3](#F3){ref-type="fig"}). Genome-wide mean allelic error rate is less than 1% for both the methods (0.79% for IMPUTE2 and 0.99% for Beagle). Since IMPUTE2 outperformed Beagle for all the autosomes, this method was used for the analyses presented in the following sections.
{#F3}
Comparison of 2-tiered and 3-tiered approaches for imputation up to 800K
------------------------------------------------------------------------
Accuracies of imputation using a 2-tiered and 3-tiered approach (Figure [1](#F1){ref-type="fig"}) to impute up to 800K SNP genotypes with IMPUTE2 are shown in Figure [4](#F4){ref-type="fig"}. The results presented are for BTA 20 as an example. Across all the scenarios examined, the mean allelic error rate of imputation is lower in the 3-tiered approach compared to the 2-tiered (Figure [4](#F4){ref-type="fig"}). The mean allelic error rate of imputing up to 800K decreases from 4.78% in the 2-tiered approach to 4.62% in the 3-tiered when Bovine3K SNP panel are used in the test animals (Figure [4](#F4){ref-type="fig"}a). A similar decrease in the mean allelic error rate is observed for BovineLD 7K panel (2.00% to 1.84% for 2-tierd and 3-tiered approaches, respectively). However, the relative improvement in allelic error rate from 2-tiered to 3-tiered are marginal for imputing up to 800K genotypes from 49K genotypes (0.689% to 0.688% for 2-tierd and 3 tiered approaches, respectively).
{#F4}
We further tested the accuracy of imputation using a smaller number of animals in the top tier. The mean allelic error rates for all scenarios are much higher when a small number of animals (41 animals, 5% of 825 cows) is included in the top tier (Figure [4](#F4){ref-type="fig"}b). The mean allelic error rates for the 2-tiered approach ranges from 5.55% using 49K to 14.43% for using the Bovine3K panel in the test set. However, there are larger decrease in the error rates of imputation using the Bovine3K (14.43% to 9.58%), BovineLD 7K (10.01% to 6.03%) and 49K (5.55% to 3.41%), by including a middle tier of 2205 bulls with 50K genotypes when the top reference tier is small.
To further test the potential of using 800K for imputing even higher density genotypes (e.g. up to 3 million or whole genome sequence) we tested accuracy of imputing every 10^th^ SNP and 100^th^ SNP by masking these SNP genotypes in 50% of the 825 cows genotyped with 800K using BTA20 as an example. The imputation accuracies for masked genotypes were 99.78% and 99.80% for every 10^th^ and 100^th^ SNP, respectively. However, such a large number of animals genotyped with very high-density SNP arrays or whole genome sequence may not be available in immediate future. We also tested a scenario when a smaller reference set (41 animals) was used and the accuracies of imputed genotypes were 98.00% and 98.44% for imputing every 10^th^ and 100^th^ SNP, respectively suggestive that ultra high-density and whole genome sequence may also be imputed with a very high level of accuracy from a commercial high-density SNP array.
Accuracy of DGV prediction based on actual and imputed genotypes using 50K dataset
----------------------------------------------------------------------------------
Accuracy of DGV prediction of five dairy traits using actual 50K, Bovine3K and BovineLD 7K genotypes are compared with DGV predictions using imputed genotypes up to 50K in Table [4](#T4){ref-type="table"}. Accuracy of DGV predictions based on imputed genotypes are very close (within 2.4%) to those obtained using the actual 50K genotypes when all the training set bulls are used in the reference set for imputation (scenario A Table [4](#T4){ref-type="table"}). Accuracies of DGV using imputed genotypes are slightly lower when smaller reference set is used for imputation (scenario B Table [4](#T4){ref-type="table"}). In scenario B all the test bulls and most of the training bulls have imputed genotypes. The lower accuracies under scenario B are more evident for Bovine3K which has much higher mean allelic error rate (5.52%). In all the scenarios the accuracies of DGV from imputed genotypes are higher than from the actual smaller subset of SNPs on which the imputation is based. These results indicate that imputed genotypes for both training and test set can be used without any loss of accuracies of DGV prediction especially when BovineLD 7K is used.
######
Accuracy of prediction of direct genomic value (DGV) for 5 dairy traits based on Bovine3K, BovineLD 7K, 50K, imputed up to 50K, imputed up to 800K and imputed 800K-dosage
**Genotypes used** **Mean allelic error rate (%) of imputation** **Milk volume** **Fat yield** **Protein yield** **Survival** **Daughter fertility**
------------------------------------------------------ ----------------------------------------------- ----------------- --------------- ------------------- -------------- ------------------------
50K \- 0.540 0.527 0.499 0.224 0.251
Subset Bovine3K \- 0.444 0.464 0.429 0.187 0.200
Subset Bovine LD 7K \- 0.481 0.516 0.443 0.186 0.232
50K-imputed (Test imputedA^A^ using Bovine3K) 3.86 0.533 0.523 0.496 0.200 0.244
50K-imputed (Test imputed^A^ with BovineLD) 2.30 0.546 0.531 0.507 0.214 0.246
50K-imputed (Train & Test imputed^B^ using Bovine3K) 5.52 0.505 0.515 0.481 0.207 0.245
50K-imputed (Train & Test imputed^B^ using BovineLD) 3.06 0.530 0.524 0.492 0.209 0.248
800K-imputed^C^ \- 0.558 0.530 0.526 0.232 0.256
800K-dosage^C^ \- 0.554 0.525 0.520 0.229 0.253
^A^Genotypes of 452 young bulls with subset of original SNPs were imputed (using IMPUTE2) up to 50K using 1753 bulls as reference set. Hence for DGV prediction entire test set (452 young bulls) had imputed genotypes and all the training bulls (1753) had actual 50K genotypes.
^B^Genotypes of 2055 bulls with subset of original SNPs were imputed (using IMPUTE2) up to 50K using 136 bulls as reference set. Hence for DGV prediction the entire test set (452 young bulls) and 1617 bulls out of the training set of 1753 bulls had imputed genotypes.
^C^Data on 2205 bulls genotyped for 50K were imputed using IMPUTE2 up to 800K using 845 cows genotyped on 800K as reference.
Accuracy of DGV prediction based on 800K imputed data
-----------------------------------------------------
Table [4](#T4){ref-type="table"} further presents the results on accuracies of DGV prediction using imputed genotypes up to 800K. The accuracies of DGV prediction using the most likely genotype (800K-imputed) and allele dosage (800K-dosage) are quite similar viz. 0.558 and 0.554 for milk yield, 0.530 and 0.525 for protein yield, 0.526 and 0.520 for fat yield, 0.232 and 0.229 for survival and 0.256 to 0.253 for daughter fertility, respectively. Overall there is only a small improvement in DGV prediction using the imputed 800K genotypes over the actual 50K genotypes.
Discussion
==========
With the rapid development of higher density SNP chips for cattle, it is now common to have population samples genotyped with different SNP chips. We have presented different strategies for utilising such heterogenous SNP datasets efficiently. We compared accuracies of imputation within and across SNP chips and the accuracy of genomic prediction using imputed genotypes.
IMPUTE2 gave higher accuracies of imputation compared to Beagle and fastPhase. fastPhase may provide comparable accuracy when the reference panel is small and the SNP densities used in the test set is high. However fastPhase required more computing time compared to Beagle and IMPUTE2. For example for scenario 1 (Additional File [2](#S2){ref-type="supplementary-material"}), using a Linux machine with AMD Opteron Processor 6136, IMPUTE2, Beagle and fastPhase took 2.36, 6.19 and 20.7 hours of computing time and used 100MB, 807MB, 112MB RAM, respectively. Computation time on a multiprocessor machine can be reduced by dividing the chromosome into smaller segments. However, using IMPUTE2, we observed that accuracy was slightly higher when the whole chromosome was imputed in a single run (not shown). This may possibly be due to the extended linkage disequilibrium present in the bovine genome \[[@B42]\] which allows for better definition of long-range haplotypes when the whole chromosome is used.
Our estimates of mean allelic error of imputing up to 50K from evenly spaced 3K panel (2.8%) were lower compared to Bovine3K (3.3%) which may be because of the higher number of SNPs with higher MAF in evenly spaced 3K SNP panel. These estimates are comparable to the range of 2.1 to 5.5% reported by Dassonneville et al. \[[@B32]\] for Bovine3K and 3 to 4% obtained by Zhang et al. \[[@B30]\] for evenly spaced 3000 SNPs using DAGPHASE. We found an increase in the accuracy of imputation with an increase in the number of animals in the reference set. However, we tested only up to 1,363 animals in the reference. Larger reference sets might further improve accuracy of imputation.
We showed that 800K genotypes could be imputed with low allelic error using 50K genotypes (0.79% for all autosomes). Most of the SNPs had low error rate. However, we noted a very small proportion of the SNPs with higher imputation error than expected. For example we found 12 SNPs on BTA20 which had an allelic error rate of larger than 5%. We suspect that these SNPs may have incorrect positions on UMD3.1 assembly or contain errors in genotyping call itself. The mean error rates reported throughout this study include all such SNPs. If wrong map assignment and genotypic error of SNPs have a significant effect on the accuracy of imputation process is not known, but should be considered in future studies.
We showed that using additional reference panel genotyped with medium-density SNP chip in a 3-tiered framework increased the accuracy of imputation especially when the main reference panel was small. The additional gain in the accuracy of imputation in the 3-tiered approach may be due to better definition of haplotypes with the availability of large number of samples in the combined reference \[[@B38]\]. Our results suggest that increasing the size of the reference panel by including animals genotyped with different SNP chips in a tiered framework can improve the accuracy of imputation. We used population based methods for imputation and showed that these used relationship information indirectly. The degree of kinship between animals in test and reference set has a significant effect on the accuracy of imputation and as such can be strategically optimised in selecting animals to be genotyped if pedigree information is available. A number of other programs have been used for imputation ( \[[@B43]-[@B45]\], \[[@B33]\]) which use pedigree information directly along with haplotype data and these can be more efficient when required family information is available. Johnston et al. \[[@B44]\] suggested a blending approach that combined the strength of various programs available. Development of multi-tiered imputation strategies that utilises pedigree information seems promising when the animals genotyped with heterogenous SNP panels and up to whole genome sequences are available.
Using imputed genotypes up to 50K increased the accuracy of genomic selection compared to just using the smaller SNP subsets used for imputation. Similar observations were made by Johnston et al. \[[@B44]\] and Weigel et al. \[[@B46]\]. Therefore, using genotype imputation would increase return on investment when a larger proportion of the population is genotyped with lower density SNP panels.
By testing the utility of imputed 800K genotypes *i.e.* best guess genotypes and dosages of the B-allele, we showed that the accuracy of genomic prediction from imputed 800K genotypes was only marginally better compared to using 50K genotypes. Although we cannot compare these accuracies with the actual 800K genotypes in this study, however, mean allelic error rate of imputation up to 800K using 50K in the test samples was very small (0.79%). These error rates were obtained by using 425 cows in the reference set. The results of imputing up to 50K (Figure [2](#F2){ref-type="fig"}) show that using larger reference can improve accuracy of imputation even further. Moreover additional analyses within the 50K dataset indicate that small error rates of the imputed genotypes will have no notable effect on the accuracy of genomic selection. Hence we believe that presented accuracies of genomic prediction with imputed 800K genotypes are comparable to the actual 800K genotypes. However, we have only used one method for genomic prediction and it is possible that other methods may utilise higher density genotype more efficiently (e.g. \[[@B31]\], \[[@B47]\]). High-density SNP genotypes are likely to be useful for genome-wide association studies and across study meta-analysis of SNP-trait relationships. Further studies are required to see the utility of imputed genotypes to discover and map the casual mutation affecting phenotypes in dairy cattle.
Conclusions
===========
IMPUTE2 had the highest accuracy of the three imputation methods examined. Accuracy of imputation increases with the number of SNPs in the test set, increase in the number of samples in the reference set and presence of closely related animals in the reference. 800K SNP genotypes can be imputed with very high accuracies from 50K SNP genotypes and with slightly lower accuracies from lower density SNP panels (e.g. 3K, 7K). The accuracy of imputation is improved using a 3-tiered approach, which used an additional middle tier of 50K, compared to 2-tiered approach, especially when the top panel of animals genotyped with 800K SNPs is small. There is no appreciable loss in accuracy of genomic prediction using imputed 50K SNP genotypes derived from the commercial 3K or 7K panels compared to using the actual 50K SNP genotypes and both perform substantially higher than using 3K or 7K genotypes. Our results show that imputation from lower density SNP panels is a cost effective strategy for genomic selection. There is only a small gain in the accuracy of genomic prediction when using imputed 800K genotypes compared to actual 50K genotypes.
Abbreviations
=============
(GS): Genomic selection; (DGV): Direct genomic values; (SNP): Single nucleotide polymorphism; (LD): Linkage disequilibrium; (HMM): Hidden Markov Model; (HWE): Hardy-Weinberg Equilibrium; (MAF): Minor allele frequency; (DTD): Daughter trait deviations.
Competing interests
===================
The authors declare that they have no competing interests.
Authors\' contributions
=======================
MSK conceived study, contributed in its design, data collection, analyses and was the primary author for assembling the manuscript. GM contributed in the analysis and preparation of the manuscript. BJH contributed in the design, data acquisition, QC and preparation of the manuscript. HWR contributed in project concept, design, interpretation and manuscript preparation. All authors read and approved the final manuscript.
Supplementary Material
======================
###### Additional file 1
**Figure S1.** Distribution of pedigree kinship among animals within different datasets shown as boxplots.
######
Click here for file
###### Additional file 2
**Accuracy of imputation of genotypes (%) and mean allelic error rate (%) up to 50K using three imputation methods.**This file presents the results from different scenarios of imputation up to 50K. These scenarios were generated by using different proportion of animals in reference and test sets for varying SNP density (3K, 5K, 10K, 20K and 35K evenly spaced and Illumina Bovine3K) in the test set. The scenarios 1--6 used 1363 (50%) bulls, scenarios 7--12 used 681 (25%) bulls, scenarios 13--18 used 272 (10%) bulls, scenarios 19--24 used 136 (5%) bulls and scenarios 25--30 used 27 (1%) bulls in the reference set and the rest of animals in the test set. The scenarios 31--36 used all the bulls in the reference and all the cows in the test set. The scenarios 37--42 used all the cows in the reference and all the bulls in the test set. The results shown are for chromosome 20.
######
Click here for file
###### Additional file 3
**Effect of pedigree kinship between test and reference animals on the mean allelic error rate (%) of imputation.** This file presents the results of association of kinship with error rate of imputation in the form of bar charts from 42 scenarios of imputation up to 50K as given in Additional file [2](#S2){ref-type="supplementary-material"}. On Y-axis, is highest kinship estimate of a test animal with any of the reference animals and is presented as four interval categories viz. 0.0-0.01, 0.01-0.1, 0.1-0.2 and 0.2-0.4. On X-axis is the mean allelic error rate (%) on imputation.
######
Click here for file
Acknowledgements
================
The authors wish to thank Genetics Australia for semen samples, the Australian Dairy Herd Improvement Scheme (ADHIS) for providing phenotype and pedigree data. The study was supported by the Dairy Futures Cooperative Research Centre (CRC). The authors are grateful to Professors Chris Moran and Frank Nicholas for editorial suggestions in review of the manuscript.
| {
"pile_set_name": "PubMed Central"
} |
In fish, the sex reversal of reproductive organs can be induced by treatment with sex steroids during sexual differentiation in juveniles. This experimentally induced sex change was first described in medaka[@b1], and artificial sex change has been intensively studied since[@b2]. Although these experiments demonstrated that gonadal cells and somatic cells in the gonads possess cell type plasticity, it was not possible to induce sex change using sex steroids following sex differentiation. Thus, it was thought that sexual plasticity is lost after sex differentiation.
Female-to-male sex change is associated with a decrease in estrogen levels, followed by an increase in androgen levels[@b3]. Estrogens are produced by the conversion of aromatizable androgens by cytochrome P450 aromatase (P450arom), and the actions of P450arom are essential for sex differentiation and ovarian development in fish and other vertebrates[@b4][@b5]. Recently, it has become possible to reduce estrogen synthesis by inhibiting this aromatase activity using non-steroidal aromatase inhibitors (AIs), such as fadrozole hydrochloride, which is a reversible competitive inhibitor[@b6]. Further studies showed that sex change could be induced in many types of fish by aromatase inhibitor (AI) treatment during sex differentiation. In Japanese flounder (*Paralichthys olivaceus*) and tilapia (*Oreochromis niloticus*), brief treatment with AI during sex differentiation causes a sex reversal in which genetic females develop into phenotypically normal males[@b7][@b8].
Undifferentiated ovary-like gonads are initially developed during gonadal development in juvenile zebrafish, regardless of genotypic sex[@b9]. In genotypic male zebrafish, all oocytes disappear from the gonad by 30 days post-hatching, and spermatocytes develop concomitant with testicular differentiation[@b10]. In contrast, oocytes in the female ovaries continue to grow to maturation. The phenomenon of the presence of undifferentiated ovary-like gonads during the juvenile period is known as juvenile hermaphroditism. Recently, it has been reported that the massive, male-specific disappearance of oocytes from the gonad during the transition from ovary-like tissue to testis tissue is caused by apoptosis[@b11]. Therefore, it was suggested that oocyte apoptosis plays a significant role in the testicular differentiation of juvenile zebrafish. In zebrafish, the gonadal masculinization of genetic juvenile females can be induced by the dietary administration of an AI (fadrozole)[@b12]. Recently, two independent studies provided evidence for a population of undifferentiated stem cells in the ovary of adult zebrafish that can produce both types of gametes[@b13][@b14].
In this study, we examined whether AI (fadrozole) treatment induces a sex change in adult zebrafish. Our results support the hypothesis that sexual plasticity persists in adult zebrafish following sex differentiation, indicating that undifferentiated stem cells are maintained in adult fish that do not undergo sex change under natural conditions.
Results
=======
Ovarian fluorescent and transparent transgenic zebrafish lines
--------------------------------------------------------------
To allow us to monitor the changes in the ovaries of living fish during fadrozole administration, we used ovarian fluorescent and transparent transgenic zebrafish lines. The transparent strain of zebrafish, *roy*, is deficient in the production of iridescent color by iridophores, and the viscera and reproductive organs are thus visible from the outside of the fish body ([Figure 1A](#f1){ref-type="fig"}). In the transgenic (*TG*) line with ovarian fluorescence, GFP is loosely bound to the nucleus (germinal vesicles); thus, the germinal vesicles fluoresce bright green. As this fluorescence is relatively strong in small oocytes at stages 1 and 2 ([Figure 1B and C](#f1){ref-type="fig"}), this property was well suited for monitoring ovarian retraction.
Morphological changes in ovaries
--------------------------------
Although ovary size was reduced within several months, even in fish under the control conditions in a group of females alone (27%, n = 11), the ovaries underwent oscillatory changes in size during the treatment. The oscillatory intervals were approximately two weeks but were not constant among individuals. The ovaries of the fadrozole-treated fish gradually reduced in size. After five months, the ovaries retracted completely, and the fluorescence-expressing oocytes had disappeared in all fish (100%, n = 9) ([Figure 2](#f2){ref-type="fig"}). Testis-like white tissues were observed in several fish ([Supplementary Fig. S1](#s1){ref-type="supplementary-material"}), and mating and/or *in vitro* fertilization experiments were conducted using these fish to examine whether normal testis had developed in these fish. Although this trial was conducted more than three times for each fish, no fertilized egg was obtained. The tissues in the control and experimental groups were then subjected to histochemical examination ([Supplementary Fig. S2](#s1){ref-type="supplementary-material"} and [Figure 3](#f3){ref-type="fig"}). The morphology of ovarian tissues in the control fish showed no significant change compared to the ovaries from untreated females ([Figure 3](#f3){ref-type="fig"}). However, cysts filled with spermatozoa-like cells were observed in the fadrozole-treated fish; the cells were a consistent size but larger than the sperm cells observed in normal males. The cells were extracted from the testis-like tissues, and sperm formation was examined by fluorescence microscopy or electron microscopy. A morphology of spermatozoa-like cells without tails was observed under fluorescence microscopy ([Figure 4A](#f4){ref-type="fig"}), and electron microscopic observations revealed large sperm heads without tails. A midpiece-like structure was observed in some of the spermatozoa-like cells ([Figure 4B](#f4){ref-type="fig"}). We then evaluated the effect of discontinuing the AI treatment after the formation of the testis-like tissues and re-examined sperm formation. At two weeks after treatment discontinuation, broken cyst structures and dispersed spermatozoa-like cells were observed ([Figure 5A](#f5){ref-type="fig"}). Then, cyst structures filled with different stages of spermatozoa-like cells, similar to those observed in the normal testis, were observed in sections from fadrozole-treated fish at 8 weeks after replacing the AI food with the control treatment. As shown in [Figure 5B](#f5){ref-type="fig"}, a large number of normal sperm was extracted from the AI-treated testis-like tissues. Artificial fertilization using these sperm and eggs from a non-transgenic female was successful, and juveniles developed normally ([Figure 6](#f6){ref-type="fig"}). Fluorescence microscopy analysis confirmed that the juveniles had developed from the sperm formed in the sex-changed female. As expected, the juveniles were all females ([Figure 6B](#f6){ref-type="fig"}), and ovaries with green fluorescence were observed, except one in which the reproductive tissue had not yet developed at one month (one of 18). Again, this result indicates that the sperm from the sex-changed female were capable of fertilization.
Discussion
==========
One of the amazing reproduction system phenomena, sex change, has been reported in such teleost fish as cichlids and tropical fish. Sex change is a purely ontogenetic event in some species but can be triggered in others species by environmental stimuli, such as interactions with conspecifics. The size advantage model for sequential hermaphroditism remains the most widely accepted evolutionary explanation of the adaptive significance of sex change[@b15][@b16][@b17]. This model posits that, if an individual can reproduce more effectively as one sex when small or young and as the other sex when larger or older, it should change sex at some point in its life history. For example, sex change follows the size-advantage model in Nile tilapia: these fish can change sex from female to male before forming the nest.
Additionally, sex reversal, i.e., the development of an ovary or testis independent of the genetic sex, can be induced in juvenile fish by treatment with sex steroids. The experimental induction of sex reversal in species that do not undergo sex change in nature was first reported in medaka[@b1]. Indeed, the administration of sex steroids prior to gonadal formation induced all-female or all-male gonads according to the type of steroids administered[@b2]. The same results have been reported in other species, including zebrafish[@b12].
In this study, we tested the possibility that the capacity for sex change remained in the adult fish of species that do not undergo sex change in nature. Specifically, we examined whether undifferentiated germ and/or somatic stem cells are present in the differentiated ovary in such species and examined the effect of AI treatment in sexually mature female zebrafish. In all the females tested, AI treatment caused ovarian retraction, followed by the development of testes-like organs. Cyst structures filled with spermatocyte-like cells were observed in the sections of these organs, and electron microscopic observation revealed that the cysts contained large sperm heads without tails.
It has been reported that estradiol-17β (E2) plays an important role in the final differentiation of sperm[@b18]. Furthermore, zebrafish treated with low concentrations of tributyltin developed abnormal sperm without tails[@b19], an effect was thought to be caused by the disruption of E2 activity. The same effect has been reported for aromatase inhibitor treatment[@b20]. Based on these findings, the results in the present study can be interpreted as follows: a significant decrease in endogenous E2 levels triggered the retraction of the ovary and induced testis formation, though the final differentiation of the sperm was inhibited due to the extremely low concentration of E2. The plasma E2 levels were 1.35 ng/ml (±0.6, n = 3) in the control group but under the limit of detection in the AI-treated fish. Thus, functional sperm were not produced by fish that were exposed to continuous AI treatment. However, normal and fertilization-competent sperm were formed after replacing the AI-containing food with an AI-free food. Our results demonstrated that undifferentiated germ stem cells persist in adult fish, similar to the results obtained in tilapia and medaka[@b21]. The testis appeared to have developed from the area near the cloaca rather than the ovarian tissue ([Supplementary Fig. S1](#s1){ref-type="supplementary-material"}). Indeed, it is thought that a small number of undifferentiated germ stem cells, which serve as the source of the spermatogenic cells in the sex-reversed fish ([Supplementary Fig. S1](#s1){ref-type="supplementary-material"}), may remain in this area. Future detailed analyses of the retraction of the ovary and formation of the testis should lead to the identification of stem cells. Ample evidence for the presence of undifferentiated stem cells in the adult female zebrafish has been reported previously. Dranow *et al* showed that adult zebrafish could be sex reversed to fully functional males following the depletion of most of their germ cells[@b14]. Wong *et al* also reported sex reversal in the zebrafish by the transplantation of female germ cells into a male[@b13]. Fertile sperm were produced from the transplanted germ cells. These results suggest that there is a population of undifferentiated stem cells remaining in the adult female zebrafish.
We were also interested in investigating whether the sex-changed females would engage in reproductive behavior with normal females. Thus, we attempted to pair sex-changed and normal females. Although we used females with ovulated oocytes, the sex-changed female did not show any behavior toward the normal female. This result suggested that the sex change induced in this study was restricted to the gonadal tissues and did not cause changes in the brain. However, as normal males from the same batch as the females treated in this study also failed to show any mating behavior, this result may be attributed to the age of the fish (2 years old at the end of the AI treatment). To address the possibility of changes in the brain, these experiments should be repeated with younger fish.
Methods
=======
Animals
-------
The transgenic line *TG (β-actin:EGFP)* was established by Hsiao et al[@b22][@b23]. Although the cDNA integrated in this strain was constructed for the expression of *EGFP* driven by the medaka *β-actin* promoter, the expression of *EGFP* is restricted to the oocytes and gills in adult fish. A *roy* mutant zebrafish, which is deficient in the production of iridescent color exhibited by iridophores, resulting in a transparent body, was isolated. We crossed the *roy* and *TG* strains to establish a strain that enables the direct observation of oocytes in living fish. The resulting strain, *TG* (*β-actin:EGFP);roy*, is highly transparent, and its oocytes are easily observed by fluorescence in living fish. The *TG* zebrafish were bred and maintained at 28.5°C and a 14-h light/10-h dark cycle[@b24]. All experiments using zebrafish were carried out with the approval from the institutional ethics committee of Shizuoka University, Japan, strictly following the guidelines set for the usage of animals by this committee.
Reagents
--------
17,20β-DHP and DES were purchased from Sigma Chemical Co. (St. Louis, MO, USA). Fadrozole was obtained from Wako Pure Chemical Industries (Osaka, Japan).
Hormone preparation and treatments
----------------------------------
Fadrozole was mixed with powdered fish food grains (Chroma, Kobe, Japan) and kept at room temperature in a dark box. The sexually mature females used for treatments were selected after checking for maturity, as confirmed by the observation of normal mating behavior or by artificial fertilization using a recently developed technique for the induction of ovulation[@b25]. The fish were divided into control and fadrozole treatment groups of 7--10 fish each. Each group was housed in separate small aquarium with a filtration system placed in an incubator set at 28.5°C and a 14-h light/10-h dark cycle. The fish were fed food containing fadrozole or not containing fadrozole twice a day. The amount of fadrozole used (0.2 mg/g diet) was selected based on the results of pilot experiments.
Ovarian morphology was monitored by fluorescence microscopy observations at 2-week intervals during the treatment period. Following anaesthetization with 0.5% tricaine, the ovaries were photographed under a binocular microscope under both brightfield and fluorescent lighting conditions.
Fertility check
---------------
The fadrozole-treated individuals were allowed to mate with normal females. Each mating pair was housed in a plastic case paved with glass beads from evening until the morning of the next day. If no egg was obtained, ovulation was induced by adding a maturation-inducing steroid into the water[@b25]. Artificial fertilization was then conducted using squeezed eggs according to standard methods[@b24].
Sample collection
-----------------
When testes-like organs were observed after the retraction of the ovaries, fish from the fadrozole-treated and control groups were sacrificed, and blood and gonad samples were collected after dissection. Blood samples were obtained from the heart using a 10-μl heparinized glass needle (tapered using a needle puller; Terumo, Fujinomiya, Japan) kept on ice[@b26]. The blood samples were centrifuged at 5,000 rpm for 10 minutes, and the plasma was collected from each sample separately and stored at −20°C until further analysis. The gonads were then carefully removed, fixed in Bouin\'s solution for 18 hours and then preserved in 70% ethanol for further analysis. The fixed samples were dehydrated in a graded ethanol series and embedded in paraffin. Ten-micrometer sections were prepared, and standard histological techniques were used to stain the gonad sections with hematoxylin and eosin.
ELISA
-----
The E2 levels in the plasma were measured using an ELISA kit according to the manufacturer\'s instructions (Cayman Chemical Company, Ann Arbor, MI, USA).
Author Contributions
====================
T.T., K.T. and K.M. designed the biological research, and K.T., K.M., S.R. and Y.M. performed the biological research. T.S. and H.I. performed the electron microscopic observations. T.T. and M.N. wrote the paper.
Supplementary Material {#s1}
======================
###### Supplementary Information
Dataset 1
This work was supported by Grants-in-Aid for Scientific Research in Priority Areas from the Ministry of Education, Culture, Sports, Science and Technology of Japan and a grant from Japanese environmental agency (ExTEND2005 FS to T.T.). The funders had no role in the study design, data collection and analysis, decision to publish, or preparation of the manuscript. We thank S. Ii, T. Yamaguchi, S. Takayasu, H. Ikuma, C. Kamiya, A. Ando, J. Dake, T. Fukuda, Y. Ohira, T. Oshima, and So Tokumoto for zebrafish maintenance.
{#f1}
{#f2}
{#f3}
{#f4}
{#f5}
{#f6}
| {
"pile_set_name": "PubMed Central"
} |
Introduction {#Sec1}
============
Antimicrobial resistance (AMR) is a significant threat to health and to healthcare system sustainability (O'Neill [@CR14]). An increase in AMR could threaten the safety of surgery, cancer treatment and immunosuppressive strategies and could cost trillions by 2050 (O'Neill [@CR15]). Development of new antibiotics and alternatives is part of the solution, but very few new compounds are currently coming to market. The World Health Organization (WHO) has recommended other areas for action that have been adopted by many nation states (World Health Organization [@CR19]). These include improving surveillance of AMR and antimicrobial use, enhancing efforts in infection prevention and control and investing in antimicrobial stewardship.
Most efforts in antibiotic stewardship have focused on hospital and institutional care, but there are few examples of programs aimed at addressing the 80--90% of human antibiotic use that results from community prescriptions. A systematic review has concluded that public-facing programs that address both general public and clinicians can reduce antibiotic use (Cross et al. [@CR7]).
Given the strong links between antimicrobial use and resistance at the population level (Goossens et al. [@CR12]), the Do Bugs Need Drugs (DBND) program was imported into British Columbia (BC), Canada, from the neighbouring province of Alberta in 2005. DBND is an innovative program of community and provider education aimed at reducing unnecessary antibiotic use. Program components include freely accessible guidelines and continuing health education for prescribers, direct outreach through schools, daycares and community care facilities using a workforce of volunteer health sciences students and pharmacists, and public campaigns ranging from transit ads to social media (Table [1](#Tab1){ref-type="table"}). The province of BC has 4.6 million people in an area nearly three times that of Germany and more than twice the size of California. Initiation of various program components varied but was more or less complete by 2009.Table 1Do Bugs Need Drugs program activities with duration of implementationActivityDuration^a^Program media launch2005--2006TV advertising2006--2014 Typically January/FebruaryTransit advertising2007--2014 Typically September--NovemberRadio advertising2007--2010Children's outreach and advertising2008--2010 Including children's magazine educational inserts, Vancouver International Children's Festival sponsorship, children's website advertisingGrade 2 DBND curriculum2006--2014Daycare DBND curriculum2006--2014General teaching DBND curriculum2006--2014Older adult DBND curriculum2008--2014Kindergarten--Grade 3 DBND teaching resources2010--2014Educational lectures and workshops for healthcare practitioners and students2006--2014Information print material distribution2005--2014Do Bugs Need Drugs website2005--2014Antibiotic Wise website2014Bugs & Drugs reference guide distribution2006--2014 Began as hardcopy distribution, then mobile app, now websiteEducational packages for pharmacists and physicians2006Parent-targeted advertising2005, 2009Movie theatre advertising2005--2006Peer-reviewed journal publication2010--2014Surveillance and reporting of AMR and AMU trends2010--2014Online and social media advertising2014*AMR* antimicrobial resistance, *AMU* antimicrobial use^a^Some of the activities were implemented during different times of a year, not necessarily the whole year indicated within the table
After observing a decline in antibiotic use and crude costs since the program onset, we set out to better estimate the impact of the program itself taking into account key covariates such as population size and unit cost of drugs. Previous studies have focused on evaluation of institution-based antibiotic stewardship or clinical and microbiological outcomes after stewardship program implementation (Dik et al. [@CR9]; Dik et al. [@CR10]). But this study is different as it evaluates the effect of implementing a community-focused antibiotic stewardship program. The objective of this study was to examine the strength and magnitude of changes in antibiotic use and cost associated with DBND program implementation using interrupted time series analysis. Increasingly, time series analyses are advocated as a method to evaluate community interventions (Biglan et al. [@CR3]).
Methods {#Sec2}
=======
We obtained data on antibiotic use and costs from BC PharmaNet, a database that captures information on all outpatient prescriptions dispensed by community pharmacies in the Canadian province of British Columbia, with the exception of some drugs used for HIV and STI. Antibiotic prescription data for the period 1996 through 2014 were extracted by a third party within the BC Ministry of Health; patient identifiers were removed and we received an anonymized dataset. Records were line-listed by prescription and included fields on patient demographics, prescriber profession, date of prescription, and the name, dose, frequency and duration of therapy. We classified antibiotics according to the Anatomical Therapeutic Chemical (ATC) classification system developed by WHO (World Health Organization [@CR18]). Annual provincial population estimates were obtained from the Government of BC Vital Statistics (BC Statistics [@CR1]).
For the purposes of the time series analysis, we report antibiotic use in prescriptions per 1000 population per month. We analyzed the data first with respect to overall rate of antibiotic use, then according to use of the seven major ATC classes and the cost in Canadian dollars (CAD) monthly per 1000 population. For the monthly cost of the antibiotic, we calculated the total cost (i.e., cost of the medication and pharmacist dispensing fee); this cost was paid by either the government of BC (through PharmaCare insurance) and/or the patient (directly or through an additional extended benefits insurance plan).
Because community guidelines for many indications emphasize beta-lactam (J01C) and tetracycline (J01A) antibiotics as first-line agents, we looked for a change in the proportion of use attributed to these classes as evidence of class switching in alignment with the DBND program goals.
Over the course of the study period, a number of factors external to the DBND program could have influenced cost. We adjusted for these as follows: population adjustment was achieved by our focus on rates per population in our models; we adjusted for unit drug price change by setting the average price per unit for each drug (both generic and brand) in each year to their corresponding average price in 2014. For those drugs not present in 2014, we calculated the average price per unit of drug *x* in year *y* using the formula: (average drug price per unit of all drugs in 2014)/(average drug price per unit of all drugs in year *y*) × (average drug price per unit of drug *x* of that drug in year *y*). Since one of the objectives of this study was to examine the changes in antibiotic costs associated with DBND program implementation, adjusting for unit drug price took care of both inflation and changes in real prices over time.
To assess the impact of the DBND program, we conducted an interrupted time series analysis based on changes in monthly total cost for antibiotics per 1000 population. We first removed seasonal trends from the raw cost data, and then determined the time series components to include in our regression model, and finally fit a regression model including terms for the DBND program. We began by examining the data prior to the program (January 1, 1996 to August 31, 2005) and after its initiation (September 1, 2005 to December 31, 2014), looking for overall and seasonal trends in the yearly and monthly rates. We removed any seasonal trends using time series decomposition techniques (Electronic Statistics Textbook, StatSoft, Tulsa, OK, 2013) and focused on the seasonally adjusted data for all further analysis. An autoregressive moving average (ARMA) model was fit to determine appropriate autoregressive and moving average terms to include in our final analysis. We assessed stationarity and serial autocorrelation in the ARMA model using standard methods (i.e., augmented Dickey-Fuller test, autocorrelation and partial autocorrelation function plots of the seasonality adjusted data) (Chatfield [@CR4]). The best time series components were determined using the model that minimized AIC. We also plotted the fitted model's residuals to see the distribution of the observed error in order to check whether the fitted model is systematically correct or needs to be improved.
Our final analysis was a generalized least squares regression model of the seasonally adjusted cost data, including covariates for the time series (ARMA) components, the initiation of the DBND program and the duration of time since program initiation (Cowpertwait and Metcalfe [@CR6]; Chatfield [@CR4]). In addition to our main analysis, we also considered a more conservative measure of the program's effect. We compared the observed mean monthly cost for antibiotics in 2014 to the mean monthly cost for antibiotics in the year just before program implementation (2005). This mean monthly conservative estimation of change in cost was then applied to the population of BC in 2014 to estimate the conservative cost reduction between the program start time and 2014. We also applied the final regression model separately for each of the five health authorities of BC to estimate regional variation in this conservative cost reduction. The total expenditures of the DBND program were calculated until December 2014 and were compared to the extrapolated conservative total cost reduction for antibiotics, to estimate the reduction/saved against a dollar spent for the DBND program (cost-benefit analysis).
All analyses were performed using SAS 9.4 (SAS Institute, Cary, North Carolina) and R version 3.4.1, with a *p*-value less than 0.05 defined as statistically significant.
Results {#Sec3}
=======
The overall average monthly prescription rates ranged from 54.3 per 1000 population in 2005 to 46.4 per 1000 population in 2014, a reduction of 14.5%. In line with the overall decrease in prescription rates, mean monthly prescription rates decreased for quinolones, macrolides and sulfonamides and trimethoprim antibiotics (Fig. [1](#Fig1){ref-type="fig"}). Decreases were also observed in the yearly proportion that each drug comprised of the total prescriptions with the exception of tetracyclines and penicillins, which increased in their prescription rates by about 15% from 2005 to 2014 and in their yearly proportions by 35.6% and 0.4%, respectively (Fig. [2](#Fig2){ref-type="fig"}).Fig. 1Mean monthly prescriptions per thousand population in BC, 1996--2014Fig. 2Yearly proportion of prescriptions in BC, 1996--2014
The mean monthly costs of antibiotics (adjusted for drug unit cost) per 1000 population both before and after seasonal adjustment are displayed in Fig. [3](#Fig3){ref-type="fig"}. As expected, unadjusted mean monthly costs were highest between December and March.Fig. 3Boxplot showing median cost of antibiotics in BC by calendar month before and after seasonality adjustment, 1996--2014
Before modeling, the crude reduction in average monthly cost of antibiotics (adjusted for drug unit cost) fell by \$1,465 CAD between 2005 and 2014, representing a reduction in total annual expenditures on antibiotics of \$80.8 million CAD. Sixty-six percent of this cost savings was realized by patients or their third-party benefits insurer and the remainder was realized by PharmaCare (BC's government-funded pharmaceutical plan).
Interrupted time series analysis {#FPar2}
--------------------------------
The total monthly cost for antibiotics was increasing before the start of the DBND program, but then declined steadily following program implementation (Fig. [4](#Fig4){ref-type="fig"}). Table [2](#Tab2){ref-type="table"} shows the parameter estimates from the segmented regression model of the effects of the DBND program on mean monthly total cost of antibiotics per 1000 population. The results of the regression indicate that before implementation of the program, the mean monthly total cost for antibiotics was \$5845.39 CAD/1000 population and there was a significant month to month increase (\$8.12 CAD/1000 population) in the total cost (*p* \< 0.0001). There was a significant drop in the estimated mean monthly total cost (coefficient = − 384.98, *p* = 0.001) after the DBND program began (Table [2](#Tab2){ref-type="table"}). In addition to this level change, there was a significant change in the trend, which began to decline after the program by a monthly average of \$18.20 CAD in total cost of antibiotics per 1000 population (*p* \< 0.0001). The regression models also show a statistically significant month to month change in total cost of antibiotics both for the level and trend for each of the five health authorities in BC (Table [3](#Tab3){ref-type="table"}).Fig. 4Trends in monthly total cost of antibiotics before and after the DBND program in BC, 1996--2014Table 2Segmented regression model estimating the parameters and predicting mean monthly total cost per 1000 population over time in BC, 1996--2014Components/variablesEstimate (\$)Standard error (\$)t-statistic*P* valueIntercept5845.3982.7970.60\< 0.0001Baseline trend8.121.226.64\< 0.0001Level change after intervention− 384.98115.60− 3.330.001Trend change after intervention− 18.191.36− 10.15\< 0.0001Table 3Segmented regression model estimating the parameters for health authorities in BC, 1996--2014Fraser Health AuthorityVancouver Coastal Health AuthorityVancouver Island Health AuthorityInterior Health AuthorityNorthern Health AuthorityComponentsEstimate (\$)Estimate (\$)Estimate (\$)Estimate (\$)Estimate (\$)Intercept^a^5845.35981.65651.65492.035495.4Baseline trend^b^8.14.2913.013.42− 3.93Level change after intervention^c^− 384.9− 269.7− 802.2− 372.10− 313.26Trend change after intervention^d^− 24.6− 16.8− 21.3− 9.4− 4.4^a^Estimates are significant (*p* \< 0.001), ^b^ estimates are significant (*p* \< 0.001)^c^Estimates are significant (*p* \< 0.01), ^d^ estimates are significant (*p* \< 0.01)
Program effects {#FPar3}
---------------
Using Table [2](#Tab2){ref-type="table"}, we estimated that in December 2014, the mean monthly total cost of antibiotics was \$5284.90 CAD per 1000 persons. Had the program not been introduced, the mean monthly total cost of antibiotics was on a course to reaching \$7689.80 CAD/1000 persons by December 2014. This means that the mean monthly total cost of antibiotics decreased by a maximum of \$2404.90 CAD, a 31% reduction after the DBND program implementation.
We also estimated the reduction in mean monthly cost of antibiotics more conservatively: as the difference between the observed mean monthly cost of antibiotics in 2014 and the model estimated mean monthly cost of antibiotics in the last month before program implementation (August 2005). This conservative estimation of the mean monthly total cost of antibiotics reduction per 1000 population in December 2014 was \$1503 CAD (21% of mean monthly cost of antibiotics in August 2005) (Table [4](#Tab4){ref-type="table"}). Of this reduction, the patient portion represented \$933.80 CAD/month/1000 population, whereas the BC PharmaCare portion represented the remainder (Table [4](#Tab4){ref-type="table"}). Based on the parameters in Table [3](#Tab3){ref-type="table"}, we also estimated the conservative impact of the DBND program on 2014 antibiotic cost for each of the health authorities---ranging from \$4.27 million CAD to \$33.46 million CAD (Table [3](#Tab3){ref-type="table"}).Table 4Changes in mean monthly costs per 1000 population after implementing the DBND program in BC, 2014CostsChanges in slopeMaximum^a^ estimated reductionConservative^b^ estimated reductionConservative impact on 2014 drug costs in millions in BCMean95% CI*P* valueMean (%)Mean (%)Total (\$)− 18.2− 21.7, − 14.6\< 0.001− 2404.87 (− 31.3)− 1503 (− 21.1)83.6PharmaCare cost (\$)− 3.3− 8.0, 1.30.16− 398.26 (− 29.4)− 568.9 (− 37.3)31.7Patient cost (\$)− 15.4− 19, − 11.8\< 0.001− 1898.13 (− 30.8)− 933.83 (− 18)51.9^a^Difference between the observed mean monthly cost of antibiotics and the expected mean monthly cost of antibiotics had the intervention not taken place^b^Difference between the observed mean monthly cost of antibiotics in December 2014 and the mean monthly cost of antibiotics just before program implementation
Cost analysis {#FPar4}
-------------
Based on the conservative estimate of total monthly cost of antibiotics per 1000 population, we estimated a reduction of \$83.6 million CAD for antibiotics in 2014. In the same way, we also estimated the reduction in PharmaCare cost and patient cost (\$31.7 million CAD and \$51.9 million CAD, respectively) for the year 2014. A total of \$5.9 million CAD has been invested to implement the DBND program components in BC between the program start date and end of 2014. However, we estimated a total of \$449.7 million CAD reduction in total cost of antibiotics during this time period. Using the total DBND program cost and the conservative estimation of total cost reduction for antibiotics, we estimated that every \$1 CAD spent for the program saved a total of \$76.20 CAD.
Discussion {#Sec4}
==========
Our findings shed light on the potential value of population-based antimicrobial stewardship programs. This 19-year time series analysis indicates that the DBND program was successful in reducing overall antibiotic utilization and bringing about targeted class switching and that it was associated with significant antibiotic cost savings for BC, which does not necessarily determine causality. During 2006--2014, a total of \$449.7 million CAD was saved, with patients and their third--party insurers realizing almost two thirds of the savings.
The estimated cost reduction for antibiotics reported in this study was conservative because we assumed that the total costs would have remained at the level observed just before program implementation. In fact, our analyses showed that costs were trending steadily upward prior to the program, in part due to undesirable class switching to new macrolides and fluoroquinolones. Our analysis did not include other potential societal cost savings, such as those attributable to adverse events due to antibiotic use and subsequent hospital visits and stays. Where possible, taking into account a more complete societal perspective would have given a more robust and greater cost savings in this study (Oppong et al. [@CR16]; Coulter et al. [@CR5]; Dik et al. [@CR9]; Drummond et al. [@CR11]). The effect of a stewardship program on costs aligns with studies reporting on institutional stewardship efforts (Coulter et al. [@CR5]). However, this study represents one of the first evaluations of cost savings for community antibiotic stewardship at the population level.
In this study, we observed a gradual and continuous impact of the program on average total monthly cost reduction for antibiotics as evidenced by the slope/trend change after intervention. This is expected with a multifaceted community program rolled out gradually, as not all program components began simultaneously across all the health authorities in BC. We observed the total monthly cost reduction of antibiotics in individual health authorities within BC ranging from \$4.2 to \$33.4 million CAD in 2014. We estimated that one Canadian dollar spent for the DBND program implementation saved \$76.20 CAD. Given this return on investment, most health jurisdictions or organizations should consider the value of community-based antibiotic stewardship efforts, as they show the promise of achieving better practice while also saving costs.
Observed cost reductions were driven both by a significant drop in monthly prescription rate and by antibiotic class switching toward drugs of first choice. A significant portion of total cost reduction of antibiotics can be attributable to the change in the usage of antibiotic classes; that is, increasing the usage of less costly penicillins and tetracyclines and reducing the usage of more expensive fluoroquinolones and newer macrolides.
There are some important limitations for our study. First, it is ideal to have a clear start and end point of an intervention in a time series analysis (Bernal et al. [@CR2]). In this study, the program components rolled out at different time points and not simultaneously in all the health authorities of BC. In our case, the program had a clear start point but continued to run until the end of the study period. Had we included a "wash-in" period to represent program start-up, we would have elevated our estimate of cost savings, so have preferred to present the more conservative case.
While we reported the impact of the program, we assumed that the observed temporal changes in mean monthly cost of antibiotics and associated cost savings are due to the program. This is an assumption of the modeling approach and we acknowledge that the regression model cannot determine causality. We could not measure and adjust for other co-interventions that were in place during the study period which might have had an impact on the cost reductions in this study. For instance, introduction of pneumococcal conjugate vaccine (PCV) 7 in 2003 and PCV 13 in 2010 for infant and childhood vaccination could have reduced the burden of pneumococcal infections (Low [@CR13]). However, one Canadian study reported that while incidence of invasive pneumococcal disease declined for children \< 5 years, incidence was relatively unchanged in ages ≥ 5 years (Demczuk et al. [@CR8]).
A common problem with population-level public health interventions is the inability to assign a control. However, because the time series data in this study had a pre-intervention segment that served as a control for the post-intervention segment, this should address the threat to internal validity (Wagner et al. [@CR17]).
Programs with multifaceted educational components like DBND often diffuse gradually through the target population and the time to see an impact of such programs is related to the diffusion process and/or rate. This is why it is important to have the knowledge of diffusion process and diffusion rate after the onset of the program/intervention (Wagner et al. [@CR17]). For this study, we hypothesized based on the history and maturation that the diffusion process will be gradual but we could not measure the diffusion rate of the program through the related population.
We did not measure the individual program component effect and so we are not certain which component of the program contributed most to the observed changes in this study. Often, it is difficult to calculate this type of measure where a series of efforts are involved in implementing a multifaceted community intervention (Biglan et al. [@CR3]). However, based on the assumptions from the history and experiences during the program implementation, the combined DBND component(s) are presumed to be effective.
Conclusion {#Sec5}
==========
Our findings showed that significant cost savings have been observed in association with a community antimicrobial stewardship program focused on both public and prescribers, even after conservative estimation. Such programs are an effective strategy in cost-benefit terms and should therefore be considered for universal adoption in Canadian healthcare systems. However, additional work needs to be done in regard to the appropriateness of prescribing antibiotics and any unintentional consequences of prescriptions reduction.
**Publisher's note**
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
For this study, we received Institutional Research Board approval through the University of British Columbia (certificate H09-00650).
The authors declare that they have no conflict of interest.
| {
"pile_set_name": "PubMed Central"
} |
INTRODUCTION {#s1}
============
Resident and infiltrating leukocytes have an important role in sight-threatening diseases of the eye. Substantial evidence exists to confirm not only that activation and altered immune responses happen in uveitic syndromes ([@DMM019018C12]) but that they may also occur in the form of parainflammation ([@DMM019018C33]) and initiate or exacerbate age-related macular degeneration and diabetic retinopathy ([@DMM019018C31]). Here, the cellular participants of the immune system act as drivers for pathological angiogenesis, forming the two main causes of blindness in Western industrialized countries, namely, age-related macular degeneration and diabetic retinopathy ([@DMM019018C25]; [@DMM019018C14]; [@DMM019018C15]; [@DMM019018C10]). A better understanding of the role of inflammation in the development of pathological angiogenesis is essential for the development of disease-modifying strategies not only for monitoring disease progression but also to assess the efficacy of treatments.
The development of clinically viable imaging tools for *in vivo* visualization of inflammation is a crucial step in this process. To date, live imaging of inflammation has been restricted to experimental animal models, with examples such as magnetic particles (iron oxide and gadolinium chelates) for magnetic resonance imaging, fluorescent nanoparticles which can be tagged with aptamers or peptides targeted against cell surface biomarkers, circulating factors or nucleic acid structures, or dyes such as acridine orange, which is a known human carcinogen ([@DMM019018C10]; [@DMM019018C26]; [@DMM019018C19]; [@DMM019018C27]; [@DMM019018C4]; [@DMM019018C22]; [@DMM019018C9]). The only application that is currently used to image inflammatory cells in humans is based on *in vitro* radionucleotide labelling of leukocytes from the patient\'s own blood. Although this allows direct visualization of cell migration patterns, current imaging techniques do not allow sufficient resolution to track single cells ([@DMM019018C30]).
Here, we present the use of indocyanine green (ICG) for *in vivo* visualization of myeloid cells in the mouse. ICG is a near-infrared (NIR) fluorescence tricarbocyanine dye with a peak spectral absorption of 800-810 nm (in blood) that is approved by the US Food and Drug Administration for clinical use. The introduction of ICG in ophthalmic angiography in the late 1960s was largely because of its minimal toxicity and favourable optical and biophysical properties ([@DMM019018C11]). NIR light could penetrate the ocular pigments of the eye, such as melanin and xanthophyll, thereby allowing visualization of the deep choroidal vasculature (not possible with conventional fluorescein angiography). In addition, its tendency for conjugation to plasma proteins meant that the dye did not readily leak from the fenestrated choriocapillaris ([@DMM019018C2]).
Despite the wide use of intravenous (i.v.) ICG, US Food and Drug Administration approval and a good safety profile, there is limited evidence for its use in cellular imaging. This is probably because of its pharmacokinetic properties; ICG is rapidly removed from the circulation via hepatic clearance and it has a half-life of approximately 3-4 min, which precludes *in vivo* labelling of cells. However, we observed that by administering ICG as a depot injection, we could obtain reproducible labelling of peripheral CD11b^+^ circulating myeloid cells, establishing a novel method for *in vivo* tracking of these cells at near single-cell resolution as they invade the eye in response to inflammation and injury. Furthermore, we found *in vitro* evidence that human myeloid cells stain in a similar manner, which strongly supports the translation of this promising technique into clinical practice. TRANSLATIONAL IMPACT**Clinical issue**Inflammation underpins most of the pathological consequences of human disease. Until now, measurement of inflammation has been confined to surrogate markers, such as the size of a lesion, its appearance, or the levels of inflammatory mediators in the blood. Resident and infiltrating leukocytes are important players in eye diseases, and a better understanding of their role in these conditions is essential not only for monitoring disease progression but also to assess the efficacy of treatments. The development of clinically viable imaging tools for *in vivo* visualization of inflammation is a crucial step in this process, but current imaging techniques do not allow sufficient resolution to track single cells.**Results**In this work, the authors describe a simple method by which cellular effectors of inflammation -- namely, white-blood cells -- can be visualized, measured, and monitored over time in the mouse retina. They use a single depot injection of indocyanine green (a dye that has been used safely in humans for blood flow measurements for more than 60 years) to label inflammatory cells in the periphery. This allowed them to track the invasion of those cells into the retina in eye disease models. Retinal infiltration of inflammatory cells was detected in mice in three models of retinal inflammation and angiogenesis, after a single depot injection of indocyanine green dye. Cells could be monitored in the retina for a period of 14 days, and the large majority were identified as infiltrating lymphatic and myeloid cells.**Implications and future directions**If translated successfully into humans, this method will allow direct monitoring of inflammation and assessment of treatment strategies in a variety of diseases, including not only human ocular diseases but also neurodegenerative disorders, cardiovascular diseases and systemic immune-mediated disorders.
RESULTS {#s2}
=======
*In vitro* labelling of peripheral blood mononuclear cells (PBMCs) and splenocytes with ICG {#s2a}
-------------------------------------------------------------------------------------------
To establish whether ICG can label circulating leukocytes, we incubated whole blood with the dye *in vitro* for 30 min at room temperature. Visual inspection of a blood smear from human blood by fluorescent microscopy revealed a small population of fluorescent cells in the NIR channel ([Fig. 1](#DMM019018F1){ref-type="fig"}A,B). Next, we exposed PBMCs isolated from human and mouse blood, and mouse splenocytes to ICG (30 min at room temperature) and then analysed dye uptake and cell identity by flow cytometry. In human PBMCs, whilst all cells stained slightly with the dye, around 2-5% of all cells were strongly stained with ICG ([Fig. 1](#DMM019018F1){ref-type="fig"}C,D). In mouse PBMCs, \<1% of all cells were strongly stained with ICG ([Fig. 1](#DMM019018F1){ref-type="fig"}E,F) compared with 7-10% of mouse splenocytes, which consist of both splenic reservoir monocytes and resident macrophages ([Fig. 1](#DMM019018F1){ref-type="fig"}G,H). Fig. 1.***In vitro* labelling of peripheral blood mononuclear cells (PBMCs) and splenocytes in humans and mice.** (A,B) Blood smear of human blood incubated in a concentration of 6.25 µg/ml indocyanine green dye (ICG) for 30 min at room temperature. A small proportion of cells stained with ICG were visualized on the near-infrared channel at 10× magnification (A) and 20× magnification (B). (C,D) Detection of ICG-stained human PBMCs by flow cytometry in cells incubated in PBS and 6.25 µg/ml of ICG reveals that 4.9% of all cells were labelled and, of these, 4.4% were CD45-high. (E,F) In mouse PBMCs, although ICG labelling was also observed in the same *in vitro* conditions, a smaller proportion of ICG-labelled cells were detected; 0.8% of total cells. (G,H) Mouse splenocytes were more readily labelled with ICG, with 9.5% of total cells staining with ICG.
We observed that although the proportion of ICG-stained cells increased when incubated at higher temperatures and in higher concentrations of ICG, there was an increase in non-specific staining. Furthermore, the staining of cells varied with the number of washes and volume of media used. We identified that a 30 min incubation of ICG at a concentration of 6.25 µg/ml at room temperature, with two washes in 5 ml PBS, achieved reproducible staining of cells. Furthermore, we identified that the minimal dose at which stained cells could be detected was 1.5 µg/ml, although 5 µg/ml was required for reliable and reproducible staining. *In vitro*, the specificity of ICG binding to PBMCs was dependent upon the concentration, ambient temperature and period of incubation (data not shown).
Assessment of the efficiency of ICG cell labelling and its effects on macrophages and lymphocytes {#s2b}
-------------------------------------------------------------------------------------------------
To assess the efficiency of ICG uptake of different populations of cells, we used cultured mouse bone-marrow-derived macrophages (BMDMs) and magnetic-activated cell sorted CD4^+^ mouse lymphocytes from the spleen. We observed, using flow cytometry, that whilst both populations were stained with ICG, BMDMs were more strongly stained than CD4^+^ cells ([Fig. 2](#DMM019018F2){ref-type="fig"}A). A 2 h exposure of BMDMs to ICG (200 ng/ml) led to a green stain of BMDMs that was visible in bright-field microscopy ([Fig. 2](#DMM019018F2){ref-type="fig"}B). The uptake of ICG in BMDMs was not affected by activation of the cells with lipopolysaccharide (LPS; [Fig. 2](#DMM019018F2){ref-type="fig"}C,D). Conversely, ICG did not activate BMDMs, as assessed by interleukin-6 (IL-6) production ([Fig. 2](#DMM019018F2){ref-type="fig"}E). In order to assess whether ICG was internalized by the cells or whether it only stained the cell membrane, we measured ICG release of stained cells after lysis by spectrophotometry. Increasing absorption at 790 nm suggested the release of internalized dye ([Fig. 2](#DMM019018F2){ref-type="fig"}F). Fig. 2.**ICG labels macrophages by internalization of the dye but does not appear to cause activation.** (A) *In vitro*-labelled bone-marrow-derived macrophages (BMDMs, red) appear in flow cytometry more strongly labelled than CD4^+^ lymphocytes (blue; unlabelled cells are indicated by stippled lines). (B) Bright-field microscopy of BMDMs cultured for 2 h with 0.2 mg/ml ICG identifies regions of visible green dye apparently within cells. (C,D) Applying LPS, 5 ng/ml LPS only (C) and LPS+ICG (D) did not lead to a marked difference in ICG uptake. (E) There is no evidence of classical activation, as measured by the production of interleukin-6 (IL-6) from supernatants taken at different time points. Data were combined from two separate experiments (means+s.d. are shown). (F) BMDMs were incubated with 0.2 mg/ml ICG. Following several washes, PBS was incubated with the cells for 5 min, then tested by spectrophotometry and compared with supernatants from the same cells after lysis with 2% Triton. Increasing the duration of ICG incubation led to increased absorption, consistent with the release of progressive internalized ICG.
*In vivo* labelling of infiltrating leukocytes by ICG {#s2c}
-----------------------------------------------------
First, we examined whether ICG labelling was possible *in vivo*, and second, whether these labelled cells could also be imaged *in vivo*. To this end, we administered 1 mg of ICG by intraperitoneal (i.p.) injection to C57BL/6 mice 3 days before they were imaged in three disease models in which there is substantial leukocytic (including monocyte/macrophage) retinal infiltration.
The first model was endotoxin-induced uveitis induced by systemic delivery of LPS from *Escherichia coli*. A previous study detected Acridine Orange-labelled leukocytes in the subretinal space/deep retina, 2 days after LPS injection ([@DMM019018C18]). We found that ICG-labelled cells were detectable in the retina *in vivo* using fluorescence scanning laser ophthalmoscopy, 24 h after LPS injection, with a peak at 2 days after LPS administration (and 5 days after i.p. ICG; [Fig. 3](#DMM019018F3){ref-type="fig"}A). No signal was seen in the fluorescein channel (488 nm solid-state excitation laser and 500 nm barrier filter), indicating that the ICG signal did not derive from autofluorescence ([Fig. 3](#DMM019018F3){ref-type="fig"}B). Control animals that were injected only with ICG but not with LPS showed only a few sporadic ICG^+^ cells ([Fig. 3](#DMM019018F3){ref-type="fig"}C,D). This suggests low levels of peripheral cell trafficking in the normal retina, as previously inferred via flow cytometric analysis ([@DMM019018C1]; [@DMM019018C3]), which dramatically increases after LPS stimulation. Fig. 3.***In vivo* labelling of infiltrating leukocytes in two murine models of ocular inflammation.** (A) Inflammatory infiltration of the deep retina/choroid of an eye with endotoxin-induced uveitis, imaged using a scanning laser ophthalmoscope with a near-infrared filter (790-nm diode excitation laser and 800-nm long-pass filter). ICG-labelled cells were visualized as white dots throughout the 55° field of view after an intraperitoneal (i.p.) injection of ICG (5 days before imaging) and induction of systemic inflammation with an i.p. injection of lipopolysaccharide (2 days before imaging). (B) An image of the deep retina/choroid of the same mouse was taken in the fluorescein angiography (AF) channel, using a blue-light filter (488-nm solid-state excitation laser and 500-nm long-pass filter). No white dots are present, demonstrating that white dots imaged in A are not a consequence of autofluorescence but ICG-labelled cells. (C) A control mouse that only received i.p. ICG (3 days before imaging). Imaging with a near-infrared filter showed only a few sporadic ICG-labelled cells, suggesting a low-level circulation of myeloid cells into the retina. (D) An image of the deep retina/choroid of the same mouse was obtained using the 488-nm channel, illustrating that the identified cells were not autofluorescent in this range. (E) Inflammation of a retinal vein (vasculitis) is visualized using the near-infrared filter in an eye at peak experimental autoimmune uveitis. ICG-labelled cells were visualized as white dots clustering around a segment of vasculitis. Mice with experimental autoimmune uveitis received an injection of i.p. ICG 3 days before imaging peak disease on day 26. (F) An infrared-reflectance (IR) image (820-diode excitation laser, no barrier filter) of the same mouse was obtained, which demonstrates the segment of retinal vein affected by vasculitis with increased reflectance (higher white intensity) of the vein itself, and surrounding tissues. Of note, no white dots were observed in this image, indicating that again, the white dots observed in (C) are not the result of autofluorescence but of ICG-labelled cells.
The second model was experimental autoimmune uveoretinitis, an antigen-dependent CD4^+^ T-cell-initiated ocular autoimmune disease, which shares features with human uveitis. Unlike the endotoxin-induced uveitis model, which produces a diffuse, short-lived inflammation in the deep retina and choroid, the prominent features of the experimental autoimmune uveoretinitis model include marked infiltration of myeloid and T-cells around inner retinal blood vessels (vasculitis) and a more sustained inflammation with retinal infiltration and damage and remodelling developing over many weeks ([@DMM019018C6]). We found that 26 days after immunization of these animals, ICG-labelled cells could be visualized in the retina, choroid and particularly in areas of active vasculitis ([Fig. 3](#DMM019018F3){ref-type="fig"}E,F), correlating with published histology of activated myeloid cells in this model ([@DMM019018C5]).
The third model was the laser-induced choroidal neovascularization (CNV) model, a model widely used to study pathological angiogenesis in the retina ([@DMM019018C20]). The vascular response in this model is accompanied by an accumulation of infiltrating leukocytes in the laser lesion from the circulation and from within the retina ([@DMM019018C8]; [@DMM019018C13]). We hypothesized that the infiltrating leukocytes can be visualized by labelling cells with ICG in the periphery. Given that laser injury can induce autofluorescence, we first imaged animals without ICG. In the NIR channel, no fluorescence was detected in the lesion ([Fig. 4](#DMM019018F4){ref-type="fig"}A). Next, we tested whether administration (i.p.) of ICG immediately after the laser would result in ICG leakage into the retina. Although ICG could be visualized in the retinal vasculature and laser lesions, overt leakage of ICG into the retina was not apparent ([Fig. 4](#DMM019018F4){ref-type="fig"}B). As before, in order to ensure that there was sufficient time for ICG labelling of cells in the circulation and the clearance of ICG from the circulation, we injected ICG 3 days before laser induction of CNV. Given that ICG is rapidly cleared from the circulation, we reasoned that a 3-day delay would be sufficient to minimize any ICG in the circulation that might leak directly into the retina from the bloodstream and label resident retinal macrophages. Consistent with this, ICG could no longer be detected in the circulation by NIR imaging at this stage (not shown). Seven days after induction of the laser lesion (10 days after ICG), we observed a marked aggregation of clearly labelled cells in and around the laser lesions ([Fig. 4](#DMM019018F4){ref-type="fig"}C,C′). After a further 3 days (10 days after laser), the signal started to subside ([Fig. 4](#DMM019018F4){ref-type="fig"}D,D′). ICG^+^ cells were observed in laser lesions that developed CNV and also in those that had no fluorescein angiographic evidence of CNV after 7 days. This is illustrated by an example shown in [Fig. 4](#DMM019018F4){ref-type="fig"}E. Here, six laser lesions were applied. Although all of them accumulated ICG^+^ cells ([Fig. 4](#DMM019018F4){ref-type="fig"}E), only two lesions showed signs of neovascularization based on fluorescein angiography ([Fig. 4](#DMM019018F4){ref-type="fig"}F,G). Fig. 4.***In vivo* labelling of infiltrating leukocytes in a laser-induced choroidal neovascularization (CNV) murine model.** The scanning laser ophthalmoscope with a near-infrared filter (790-nm diode excitation laser and 800-nm long-pass filter) was used to image the retina and choroid. (A) A deep retinal/choroidal image of a control animal that did not receive an i.p. injection of ICG. No fluorescence was detected using the near-infrared filter. (B) A deep retinal/choroidal image of an animal that did receive i.p. ICG and laser induction of CNV, showing fluorescence in the retinal vessels and laser lesions. No obvious leakage of ICG could be seen in the surrounding retinal tissues. (C) A deep retinal/choroidal image of an animal that received i.p. ICG (10 days before imaging) and laser induction of CNV (7 days before imaging), showing an accumulation of ICG-labelled cells in and around the laser lesions. (C′) Magnified image of laser lesion and surrounding cells. (D) A deep retinal/choroidal image of the same animal shown in C 3 days later (13 days after i.p. ICG), showing that the intensity of the ICG signal has reduced. (D′) Magnified image of laser lesion and surrounding cells. (E) A deep retinal/choroidal image of an animal that received i.p. ICG (10 days before imaging) with six laser-induced CNV lesions (7 days before imaging), showing accumulation of ICG-labelled cells in and around all six laser lesions. (F,G) Corresponding fluorescein angiography images of the superficial retina and deep retina/choroid obtained 10 min after i.p. injection with 100 µl of fluorescein dye and imaged with a blue-light filter (488-nm solid-state excitation laser and 500-nm long-pass filter). (G) Deep retinal/choroidal images show that only two of six laser lesions subsequently developed choroidal neovascularization, and none of these 1-week-old lesions showed obvious fluorescein dye leakage into the surrounding tissues, suggesting that inflammatory cellular infiltration occurred independently of vascular leakage.
Furthermore, dissection of the mice 7 days after administration of ICG i.p. revealed green staining of lymphatic tissue in the thoracic cavity ([Fig. S1A,B](Fig. S1A,B)), mediastinal lymph nodes ([Fig. S1C](Fig. S1C)), thymus (not shown) and the greater omentum in the abdominal cavity ([Fig. S1D-F](Fig. S1D-F)). This suggests that inflammatory cells in the circulation, most probably both monocytes and lymphocytes that continuously circulate between the bloodstream and lymphoid organs, are labelled by the ICG. To check whether ICG might be toxic to the cells that take it up, we used flow cytometry to measure cell death in inflammatory cells that had been stained *in vitro* (after 30 min) and *in vivo* (10 days after ICG injection) in the CNV model. This showed no noticeable differences in cell counts, indicating that ICG has no gross effects on inflammatory cell populations in the spleen, the retina and the choroid ([Fig. S2](Fig. S2)) within the time frame of our experiments (10 days).
We next tested whether cellular infiltration around the laser-induced CNV lesion could be monitored over time. As before, ICG was administered i.p. 3 days before the laser, and animals were imaged sequentially 2, 5 and 8 days after laser CNV induction ([Fig. 5](#DMM019018F5){ref-type="fig"}). The number of ICG-labelled cells in and surrounding the CNV lesion could be observed qualitatively to increase over time ([Fig. 5](#DMM019018F5){ref-type="fig"}C-E). Next, we assessed the feasibility with which this could be quantified reproducibly. First, we attempted a simple pixel count by thresholding individual CNV lesions ([Fig. S3](Fig. S3)); a method we have used successfully for quantifying ICG-labelled cells in endotoxin-induced uveitis models (data not shown). However, because of the narrow range of intensity values within each lesion, it was not possible to visualize individual cells overlying the CNV lesion. Nonetheless, we found that by assessing the mean intensity values for a given pixel area using histograms, we were able to quantify cellular infiltration over CNV lesions ([Fig. S3G](Fig. S3G)). Using this method, we demonstrated that ICG-labelled cellular infiltration can be quantified and monitored over time ([Fig. 5](#DMM019018F5){ref-type="fig"}B,C′-E′). Fig. 5.***In vivo* quantification of laser-induced CNV-related infiltration.** (A) ICG was injected i.p. at day 0 before laser induction of CNV at day 3 in C57BL/6 mice. Imaging with the scanning laser ophthalmoscope using a near-infrared filter (790-nm diode excitation laser and 800-nm barrier filter) was performed on days 5, 7 and 10. (C-E) Increasing cellular infiltration in and surrounding the laser-CNV lesion was observed over time. (B,C′-E′) We demonstrate that the mean intensity values and standard deviations can be quantified over a given area and show an increase in inflammation over time. The mean intensity value of the lesion at day 5 is represented by the black histogram (C′), day 7 the superimposed light grey histogram (D′) and day 10 the superimposed medium grey histogram (E′).
Given that ICG is routinely used in clinical practice, we compared the dosage and delivery routes normally used in humans (5 mg i.v. bolus injection) with our mouse protocol (1 mg i.p. bolus injection). We reduced the ICG in a stepwise fashion from 1 mg to 0.0125 mg, which roughly equates to the human dose in ophthalmic use. Invading cells could still be detected readily after administration of 0.5 mg ICG i.p., but at 0.25 mg they were fainter. With progressively reduced amounts of ICG (0.125, 0.05, 0.025 and 0.0125 mg), the lesions were faintly fluorescent, but individual cells could no longer be detected ([Fig. S4](Fig. S4)). Out of the different delivery routes tested, i.p. was the most efficient at labelling cells. A subcutaneous depot (1 mg) produced weaker labelling, but individual cells could still be detected, whereas i.v. (1 mg) delivery produced only very faint staining. In addition, i.v. delivery of ICG shortly before the laser produced only staining of the retinal and choroidal vasculature, with diffuse leakage of ICG into the laser-induced CNV lesion, and individual cells could not be detected (not shown). Oral administration (by gavage or in drinking water) did not lead to any labelling ([Fig. S5](Fig. S5)).
Characterization of *in vivo* ICG-labelled cells {#s2d}
------------------------------------------------
To test the identity of ICG^+^ cells invading the retina in the laser-induced CNV model, we used an *in vivo* staining protocol. After ICG imaging ([Fig. 6](#DMM019018F6){ref-type="fig"}A-C), a fluorescently labelled (fluorescein isothiocyanate) antibody against CD11b was injected. Over the course of 30 min, this labelled a population of cells that matched the ICG-labelled cells spatially ([Fig. 6](#DMM019018F6){ref-type="fig"}D-F). The CD11b signal was weaker than the ICG signal and there was not a perfect overlap. Nevertheless, all CD11b-positive cells were also ICG positive. The same approach was also taken with an anti-CD45 antibody, with the same outcome (data not shown). This suggests the identity of most of the ICG-labelled cells in the retina to be infiltrating myeloid cells. Fig. 6.**Characterization of *in vivo* ICG-labelled cells in a laser-induced CNV murine model.** The animal was injected i.p. with ICG 10 days before imaging and laser CNV induced 7 days before imaging. Fifty microlitres of a CD11b antibody conjugated to fluorescein isothiocyanate (FITC) was injected into the tail vein before being imaged at 15 and 30 min post-injection to assess the identity of ICG-labelled cells. A blue-light filter (488-nm solid-state excitation laser and 500-nm barrier filter) was used for the detection of cell labelling by CD11b-FITC. A scanning laser ophthalmoscope was used to acquire all images. (A) An infrared-reflectance (IR) image (820-diode excitation laser, no barrier filter) demonstrating the presence of laser-induced CNV lesions in the deep retina/choroid. (B) A deep retinal/choroidal image obtained with a near-infrared filter (790-nm diode excitation laser and 800-nm barrier filter), showing ICG-labelled cells surrounding two laser-induced CNV lesions. (C) A magnified view of ICG-labelled cells surrounding the CNV lesion from the top lesion in (B). (D,E) A deep retinal/choroidal image of the same mouse shown in (B) at 15 min (D) and 30 min (E) after a tail vein injection of CD11b-FITC. (F) A magnified view of CD11b-FITC-labelled cells surrounding the CNV lesion from (E), showing co-labelling of most but not all ICG-labelled cells.
DISCUSSION {#s3}
==========
We describe here a novel technique for *in vivo* labelling of a subpopulation of circulating cells with ICG dye. We found that in order to label these cells reproducibly *in vivo*, ICG must be administered as a depot preparation. Using this approach, we showed in three disease models of retinal inflammation that infiltrating leukocytes can be detected, tracked over time and quantified in the murine retina.
We found that the commonly used i.v. route of administration of ICG failed to label infiltrating leukocytes. However, by administering ICG as a depot, we have shown that infiltrating cells were clearly labelled in our disease models. A possible explanation for this is the increase in the period of ICG exposure that circulating cells have, in order to be labelled. In keeping with this, we observed *in vitro* that an increased time of exposure to ICG resulted in increased cell labelling. Furthermore, we observed both *in vivo* and *in vitro* that the majority of labelled cells were of myeloid origin.
Our technique will be useful in the above-mentioned mouse models. In particular, the laser-induced neovascularization model is widely used to study pathological vessel growth in the retina. This model also contains an important inflammatory component, which has so far been assessed mainly by immunohistochemistry on postmortem tissue. With our ICG technique, it will be possible to monitor inflammatory cell invasion longitudinally in these animals.
More generally, in the field of inflammation, the ability to monitor a sequence of events over time is particularly important in order to establish cause-and-effect relationships. As such, recent developments have been focused on the ability to visualize inflammation *in vivo* using molecular imaging techniques, such as intravital microscopy using fluorescent antibodies, nanoparticles and transgenic animals that express fluorescent proteins ([@DMM019018C18]; [@DMM019018C28]). One study has used intravitreal ICG in murine CNV models ([@DMM019018C24]). The vitreous cavity acts as a reservoir, thereby allowing prolonged exposure of cells to ICG (similar to that of a depot application). Intravitreal ICG was observed to label microglial cells for up to several weeks, allowing time-lapse observation of migratory behaviour after injury. In a similar fashion, our ICG method labels peripheral circulating cells but avoids potential neurotoxic effects of prolonged exposure of ICG to the retina, as observed by vitreoretinal surgeons who use ICG as a vital stain to visualize the internal limiting membrane during macular hole surgery ([@DMM019018C29]). Although depot ICG does not target specific cell types or epitopes like transgenic animals or antibodies, nor have we identified the identities or proportion of infiltrating cells stained, it has the advantage of having a simple application technique (depot injection), does not depend on specific mouse strains and, most importantly, has the potential for rapid translation to human use.
Current strategies for *in vivo* molecular imaging in humans include conjugation of NIR dyes, such as ICG, with ligands such as small molecules, antibodies, peptides, DNA and nanoparticles. Nevertheless, none of these probes has been approved for human use because of their so far insufficiently characterized safety profiles. In particular, nanoparticles suffer from poorly characterized distribution, accumulation and clearance from the human body, and potential cytotoxicity of heavy metal ingredients ([@DMM019018C23]; [@DMM019018C16]). In contrast, ICG is an inert, water-soluble organic dye that is rapidly bound to plasma proteins and solely removed from the circulation through the liver via a specific carrier-mediated transport system ([@DMM019018C7]). Furthermore, ICG is known not to provoke inflammation even when injected directly into tissues.
The ICG dye was first developed in the mid-1950s to determine cardiac output and hepatic function. The dye is rapidly cleared, low in toxicity and well tolerated by subjects, even at doses that exceed what is now routinely given by tenfold ([@DMM019018C17]; [@DMM019018C32]). This has led to its application to a variety of clinical uses, with examples including sentinel node biopsy, tumour demarcation surgery and lymphatic vessel assessment. The main advantage of ICG is its track record of safety and tolerance, more than 60 years of application in clinical practice, and the ability of an NIR dye to be visualized without the need for tissue windows, i.e. for non-invasive imaging.
Techniques for *in vivo* cellular imaging require a biocompatible, non-toxic agent that allows reproducible quantification of these infiltrating leukocytes. Alternative methods that have been described for *in vivo* imaging are dyes which are conjugated to antibodies, allowing more specificity of binding or *ex-vivo* labelling (labelling cells outside the circulation and reintroduction into the circulation). Our results clearly demonstrate that this simple method of using ICG as a depot injection to label cells *in vivo* can be achieved in mouse models. Translation of this insight into the clinic will involve the development of a formulation suitable for depot ICG application in humans. This might be useful not only for monitoring ocular inflammation but might also have applications further afield, for example, the detection of an immune response to new treatment strategies, such as stem cell therapies or retinal implants, or serve as an imaging biomarker for predicting the onset of angiogenesis in age-related macular degeneration or diabetic retinopathy.
MATERIALS AND METHODS {#s4}
=====================
*In vitro* labelling of peripheral blood mononuclear cells and splenocytes with ICG {#s4a}
-----------------------------------------------------------------------------------
Mouse blood was drawn via cardiac puncture with a 0.5 M EDTA-coated 23-gauge needle, before either red cell lysis or Ficoll-gradient separation using Histopaque-1077 (Sigma-Aldrich, UK) according to the manufacturer\'s guidelines. Cells were stained with ICG, then incubated with Fc-block (BD Biosciences, UK) before primary antibody staining at the manufacturer\'s recommended concentrations at 4°C for 20 min. All antibodies were from BD Biosciences.
Human PBMCs were isolated from 20 ml of whole blood using a Ficoll gradient.
PBMCs were isolated from murine whole blood using Percoll-gradient separation, and splenocytes were mechanically dissociated before being incubated in 6.25 µg/ml ICG for 30 min at room temperature. PBMCs and splenocytes were washed and co-stained with CD45, CD11b and CD3 fluorescent antibodies (Miltenyi Biotech, Bisley, UK).
Flow cytometric analysis of ICG-labelled cells {#s4b}
----------------------------------------------
ICG-stained PBMCs, whole blood and splenocytes were analysed using a BD Bioscience LSRII flow cytometer because no commercial machine was available with a near-infrared laser for ideal excitation of ICG. Suboptimal excitation by the 633 nm red laser nonetheless resulted in a reliable signal using a 780/60 bandpass filter. A minimum of 10,000 events was collected for each sample, and fluorescence-minus-one controls were used to determine the placement of gates. Data were processed using FlowJo v10.1 (TreeStar, Ashton, OR, USA). For the toxicity assessment, gating on the following populations was performed (after exclusion of debris and cellular aggregates): lymphocytes (CD4^+^/CD8^+^/B220^+^); total myeloid populations (CD11b^+^); neutrophils (Ly6G^+^); natural killer cells (NK1.1^+^); dendritic cells (CD11c^high^); and monocytes/macrophages (CD11b^+^/CD11c^low^/Ly6G^--^/NK1.1^+^). Cell death within each population was then analysed using a live:dead stain.
Generation of bone-marrow-derived macrophages {#s4c}
---------------------------------------------
Generation of bone-marrow-derived macrophages (BMDMs) followed the protocol from the original report ([@DMM019018C21]). In brief, mouse femurs and tibias were collected and bone marrow cells were flushed out, followed by 8 days of maturation in Teflon bags with Dulbecco\'s modified Eagle\'s medium containing 10% heat-inactivated fetal calf serum, 5% normal horse serum, 1 mM sodium pyruvate, 2 mM L-glutamine, 100 U/ml penicillin-streptomycin, 50 mM 2-mercaptoethanol (all from Life Technologies) and 100 pg/ml colony-stimulating factor 1 generated from L929 fibroblast-conditioned media.
Mouse interleukin-6 ELISA {#s4d}
-------------------------
BMDMs were plated at a density of 5×10^5^ cells/well in a 96-well flat-bottomed plate. Four hours later, the medium was exchanged for 0.2 mg/ml ICG dye, with either medium alone or 5 ng/ml LPS. After the indicated times, supernatants were collected and frozen at −80°C. Mouse IL-6 sandwich ELISA was performed in technical triplicate according to the manufacturer\'s instructions (BD Biosciences, UK) using rat anti-mouse IL-6, 554400 and biotin rat anti-mouse IL-6, 554402.
Spectrophotometry of ICG {#s4e}
------------------------
BMDMs were plated in flat-bottomed 96-well plates for 4 h in 10% fetal calf serum and Dulbecco\'s modified Eagle\'s medium, before the addition of ICG to a concentration of 0.2 mg/ml. Cells were incubated for the specified amount of time before washing five times with PBS. After this, 200 μl of PBS was added to the well for 5 min and then removed for spectrophotometry. Triton (2%) in 200 μl PBS was then added to each well and the supernatant also removed after 5 min. Spectrophotometry was performed using a SpectraMax 190 (Molecular Devices, Wokingham, UK), with SoftMax Pro (v6.0) software, set to an absorbance of 790 nm.
Bright-field microscopy {#s4f}
-----------------------
BMDMs were imaged in flat-bottomed 96-well plates with a Leica DMIRB microscope using Leica QFluoro software v3.1. No filters were used, but light intensity was adjusted.
Magnetic-activated cell sorted separation of CD4^+^ cells {#s4g}
---------------------------------------------------------
Single-cell spleen suspensions from C57BL/6 mice were erythrocyte lysed by ammonium-calcium-potassium (ACK) buffer. Splenic CD4 cells were prepared by enriching CD4^+^ cells using anti-CD4 microbeads (Miltenyi Biotech, UK) according to the manufacturer\'s instructions.
Animals {#s4h}
-------
All animals were handled in accordance with the UK Animals (Scientific Procedures) Act 1986. Female C57BL/6J mice (Harlan, UK) at 7-8 weeks of age were used. For *in vivo* procedures, the mice were anaesthetized with an i.p. injection of medetomindine hydrochloride (1 mg/kg body weight; Domitor; Pfizer Animal Health, New York, NY, USA) and ketamine (60 mg/kg body weight) in water. Pupillary dilatation was achieved with one drop of 1% Tropicamide (Bausch and Lomb, Kingston-upon-Thames, UK).
Induction of endotoxin-induced uveitis {#s4i}
--------------------------------------
Female C57BL/6J mice (Harlan, UK) at 8 weeks of age received a single i.p. injection of 0.2 mg of LPS from *E. coli* (Sigma-Aldrich, St Louis, MO, USA) in PBS. A 1 mg i.p. bolus injection of ICG was given 24 h before the LPS treatment. Imaging was performed 48 h later, at a previously determined time point when there is peak infiltration by myeloid cells.
Induction of experimental autoimmune uveoretinitis {#s4j}
--------------------------------------------------
Female C57BL/6J mice (Harlan, UK) at 7 weeks of age received 500 μg of human RBP-1-20 peptide subcutaneously, emulsified in complete Freund\'s adjuvant (Sigma-Aldrich, UK) supplemented with 1.5 mg/ml *Mycobacterium tuberculosis* H37RA (Difco Laboratories, BD, Oxford, UK). Pertussis toxin (1.5 μg) was simultaneously administered into the peritoneal space (Tocris Bioscience, Bristol, UK). Imaging was performed 26 days later at the time point determined to be the disease peak in our facility using this protocol.
Induction of laser choroidal neovascularization {#s4k}
-----------------------------------------------
In female C57BL/6J mice (Harlan, UK) at 7-8 weeks of age, laser CNV was induced using a slit-lamp-mounted diode laser system (wavelength 680 nm; Keeler, Windsor, UK). The laser settings used were as follows: 200 mW power, 100 ms duration and 100 µm spot diameter. Laser CNV lesions were applied at a distance of three disc diameters from the optic nerve, avoiding any blood vessels.
*In vivo* imaging {#s4l}
-----------------
Ocular imaging was performed using a scanning laser ophthalmoscope (Spectralis™ HRA; Heidelberg Engineering, Heidelberg, Germany). A lens with a 55° field of view was used, and a mean of 100 consecutive frames was taken for each image. In order to visualize ICG-labelled cells, a near-infrared filter (790 nm diode excitation laser and 800 nm long-pass filter) was used. Cell labelling was achieved with various doses (from 1 to 0.0125 mg; [Fig. S4](Fig. S4)) of 5 mg ICG (Pulsion Medical Systems AG) dissolved in 5 ml of water and administered 3 days before laser induction of CNV, i.p. injection of LPS or imaging of the experimental autoimmune uveoretinitis model. The different routes of administration are summarized in [Fig. S5](Fig. S5). To assist the identification of laser CNV lesions, infrared-reflectance imaging was performed with an 820 diode excitation laser and no barrier filter. For autofluorescence imaging and fluorescein angiography, a blue-light filter (488 nm solid-state excitation laser and 500 nm long-pass filter) was used. Fluorescein angiography was performed 1 week after laser CNV induction with an i.p. injection of 0.2 ml fluorescein sodium (2%). Images were acquired at 90 s and 7 min after injection.
Quantification of ICG^+^ cells in the retina {#s4m}
--------------------------------------------
Images were exported from the Heidelberg eye explorer version 1.7.1.0 and processed in Adobe Photoshop CS5 (Adobe Systems, San Jose, CA, USA). Details of image processing and quantification can be found in the Results section and [Fig. S5](Fig. S5).
**Competing interests**
The authors declare no competing or financial interests.
**Author contributions**
D.A.S., C.J.C., S.L., D.A.C., J.W.B.B. and M.F. designed the research. D.A.S., C.J.C., S.S., M.B.P. and S.L. generated and collected data. D.A.S., C.J.C., S.S., M.B.P., S.L., D.A.C., P.A.K, A.T., C.A.E., J.W.B.B., R.W.L., A.D.D. and M.F. analysed and/or interpreted data. D.A.S., C.J.C., S.L., D.A.C., P.A.K., A.T., C.A.E., J.W.B.B., R.W.L., A.D.D. and M.F. wrote and/or revised the manuscript.
**Funding**
D.A.S. was supported by a grant from Fight for Sight, UK (grant number 1987) and the Special Trustees of Moorfields Eye Hospital and the National Institute for Health Research (NIHR) Biomedical Research Centre based at Moorfields Eye Hospital NHS Foundation Trust and UCL Institute of Ophthalmology.
**Supplementary information**
Supplementary information available online at <http://dmm.biologists.org/lookup/suppl/doi:10.1242/dmm.019018/-/DC1>
| {
"pile_set_name": "PubMed Central"
} |
1. Introduction {#sec1}
===============
Diabetes mellitus is a metabolic disease characterized by hyperglycemia, which results from defects in insulin secretion from pancreatic beta cells, insulin resistance in peripheral tissues, and increased glucose production by the liver \[[@B1]--[@B3]\]. The liver plays a critical role in the maintenance of glucose homeostasis by balancing the uptake, storage, and release of glucose \[[@B4]\]. Fasting or starvation induces glucose synthesis in the liver through glycogenolysis and gluconeogenesis \[[@B5]\]. However, elevated hepatic glucose production is associated with the pathogenesis of type 2 diabetes \[[@B6], [@B7]\]. In this process, glucose-6-phosphatase (G6Pase) catalyzes the terminal step in the glycogenolytic and gluconeogenic pathways, and phosphoenolpyruvate carboxykinase (PEPCK) is a key regulatory enzyme driving gluconeogenesis \[[@B8], [@B9]\]. Insulin suppresses gluconeogenesis by inhibiting the transcription of PEPCK and G6Pase \[[@B10], [@B11]\].
Arsenic trioxide has a long history as biomedical interest and is approved by the Food and Drug Administration (FDA) for treatment of certain leukemias \[[@B12], [@B13]\]. Sodium meta-arsenite (NaAsO~2~) is produced by dissolving arsenic trioxide. Sodium meta-arsenite (SA, KML001) has entered phase II clinical trials for the treatment of solid tumors and hematopoietic malignancies. In addition, sodium meta-arsenite (SA) is reported to have insulin-mimetic effects on glucose homeostasis. SA inhibits forskolin/dexamethasone-induced PEPCK and G6Pase gene expression in hepatic cell lines and rat primary hepatocytes \[[@B14], [@B15]\]. SA activates AMP-activated protein kinase \[[@B15], [@B16]\], which in turn induces small heterodimer partner (SHP) which inhibits the expression of hepatic gluconeogenic genes, and this repression is abolished by SHP inhibition \[[@B15]\]. SA also suppresses dexamethasone-induced PEPCK transcription in 14-day chick embryo livers*in vivo* \[[@B17]\].
Despite the demonstrated effects of SA in reducing the expression of gluconeogenesis genes, the antidiabetic effect of SA in type 2 diabetes has not yet been evaluated*in vivo*. In this study, we examined the therapeutic effect of SA in diabetic*db/db* mice, an animal model of human type 2 diabetes, as well as the mechanisms involved in the improvement of hepatic gluconeogenesis.
2. Materials and Methods {#sec2}
========================
2.1. Animals {#sec2.1}
------------
*db/db* mice were obtained from the Korea Research Institute of Bioscience and Biotechnology (Daejeon, Korea) and*C57BL/6*mice were obtained from the Orient Bio Inc. (Gyeonggi, Korea) and maintained in specific pathogen-free conditions at the animal facility at Gachon University of Medicine and Science under a 12 h light : 12 h dark photoperiod. Animals were fed*ad libitum* on a standard rodent diet. The*db/db* male mice (aged 6--8 weeks) were monitored for the development of hyperglycemia using a glucometer (One Touch Ultra; LifeScan Inc., Milpitas, CA, USA). Pair-fed diabetic*db/db* mice were given the same daily amount of food as that eaten by the corresponding SA-treated group during the previous day. All animal experiments were carried out under a protocol approved by the Institutional Animal Care and Use Committee at the Gachon University of Medicine and Science. A total of 68 animals were used in the experiments described here.
2.2. Treatment with SA {#sec2.2}
----------------------
Six- to eight-week-old diabetic male*db/db* mice (random blood glucose levels \> 300 mg dL^−1^ for 3 consecutive days) were orally intubated with SA (10 mg kg^−1^ body weight/day; Komipharm, Seoul, Korea) or phosphate buffered saline (PBS) for 8 weeks. Food consumption was measured weekly. After 4 and 8 weeks of treatment, glucose levels were measured following the removal of food for 14 h. All animal groups were body weight-matched with the SA-treated group at the beginning of each experiment.
2.3. Blood Analysis {#sec2.3}
-------------------
After 8 weeks of SA treatment, mice were not fed for 4 h, and blood samples were drawn into heparinized capillary tubes from the periorbital veins. The whole blood was used for HbA1c measurements, and serum was used for alanine aminotransferase (ALT) and aspartate aminotransferase (AST) measurements using the AU480 Chemistry System (Beckman Coulter Life Sciences, California, USA).
2.4. Glucose, Insulin, and Pyruvate Tolerance Tests {#sec2.4}
---------------------------------------------------
After 4 and 8 weeks of SA treatment, mice were not fed for 14 h, and then a glucose solution (2 g kg^−1^ body weight) was injected intraperitoneally. Blood glucose levels were measured at 0, 30, 60, 90, 150, and 180 min after glucose injection at 9:00 a.m. For insulin tolerance tests, mice were not fed for 4 h and were injected with insulin (2 units kg^−1^ body weight, i.p.), and blood glucose levels were measured at 0, 30, 60, and 90 min after insulin injection at 1:00 p.m. For pyruvate tolerance tests, C57BL/6 mice were orally intubated with SA (10 mg kg^−1^ body weight) or PBS and then fasted overnight. Fifteen hours after oral intubation, mice were injected i.p. with 1 g kg^−1^ body weight of sodium pyruvate in PBS, and blood glucose levels were measured at 0, 15, 30, 60, 90, and 120 min after pyruvate injection at 9:00 a.m.
2.5. Real-Time Quantitative PCR (RT-qPCR) {#sec2.5}
-----------------------------------------
Total RNA was isolated from the liver of SA-treated mice or hepatocytes from C57BL/6 mice, and cDNA was synthesized using the PrimeScript First-Strand cDNA Synthesis Kit (TaKaRa Bio, Inc., Otsu, Japan). PCR was carried out in a 7900HT fast real-time PCR system (Applied Biosystems, Carlsbad, CA) at 95°C for 10 min, followed by 40 cycles at 95°C for 15 s, 60°C for 1 min. As an internal control, cyclophilin mRNA was amplified. The specific PCR primers were G6Pase: sense 5′-GTGTTGACATCGGCCC-3′, antisense 5′-AACTGAAGCCGGTTAG-3′; PEPCK: sense 5′-CGCAAGCTGAAGAAATATGACAA3′, antisense 5′-TCGATCCTGGCCACATCTC-3′; hepatocyte nuclear factor-4*α* (HNF-4*α*): sense 5′-CCAACCTCAATTCATCCAACA-3′, antisense 5′-CCCGGTCCGCCACAGAT-3′; SHP: sense 5′-AGG AACCTGCCGTTCCTTCTG-3, antisense 5′-TGG CTT CCT CTA GCA GGA TC-3′; Sirt1: sense 5′-TTGGTGGTACAAACAGGTATTGA-3′, antisense 5′-CAGTGAGAAAATGCTGGCCTA-3′; AgRP: sense 5′-TGCTACTGCCGCTTCTTCAA-3′, antisense 5′-CTTTGCCCAAACAACATCCA-3′; POMC: sense 5′-GAGGCCACTGAACATCTTTGTC-3′, antisense 5′-GCAGAGGCAAACAAGATTGG; MC4R: sense 5′-TGCTGGTGAGCGTTTCGA-3′, antisense 5′-GGCATCCGTATCCGTACT-3′; and cyclophilin: sense 5′-TGGAGAGCACCAAGACAGACA-3′, antisense 5′-TGCCGGAGTCGACAATGAT-3′. The relative copy number was calculated using the threshold crossing point (Ct) as calculated by the 7900HT fast real-time PCR software combined with the delta delta Ct calculations.
2.6. Glucose Production Assay {#sec2.6}
-----------------------------
Primary hepatocytes were isolated from C57BL/6 mice by collagenase perfusion. For the hepatic glucose production assay, hepatocytes were seeded in 6-well plates at a density of 2 × 10^5^ cells 2 mL^−1^/well and cultured in Hepatozyme SFM media (Gibco) for 24 h; unattached cells were discarded. Hepatocytes were then treated for 24 h with SA (5 or 10 *μ*M). After 24 h of culture, cells were washed with PBS and then cultured in glucose-free DMEM supplemented with 20 mM sodium lactate and 2 mM sodium pyruvate for 2 h. Glucose concentration in the media was measured by a glucose assay kit (BioVision Research Products, Mountain View, CA).
2.7. Western Blot Analysis {#sec2.7}
--------------------------
Hepatocyte cell lysates were preparedand subjected to sodium dodecyl sulfate-polyacrylamide gel electrophoresis. Proteins were electrotransferred onto nitrocellulose membranes. Membranes were incubated with anti-Sirt1, anti-acetyl-FoxO1, or anti-FoxO1 antibodies (Santa Cruz Biotechnology, Santa Cruz, CA) overnight at 4°C. After washing, membranes were incubated with a horseradish peroxidase-conjugated secondary antibody (anti-rabbit IgG, Chemicon International, Temecula, CA) for 1 h at room temperature. Reactive bands were detected by enhanced chemiluminescence (Thermo Fisher Scientific, Rockford, IL).
2.8. Indirect Calorimetry {#sec2.8}
-------------------------
A comprehensive animal metabolic monitoring system (CLAMS; Columbus Instruments, Columbus, OH) was used to evaluate energy expenditure, respiratory exchange ratio, and locomotor activity. Energy expenditure was measured by assessing oxygen consumption with indirect calorimetry. The respiratory exchange ratio was computed as carbon dioxide output (VCO~2~) divided by oxygen consumption (VO~2~). Locomotor activity was measured on *x*- and *z*-axis by using infrared beams to count the beam breaks during 72 h.
2.9. Cytotoxicity and Proliferation Assays {#sec2.9}
------------------------------------------
For cytotoxicity and proliferation assays, hepatocytes were seeded in 96-well plates at a density of 2 × 10^4^ cells 100 *μ*L^−1^/well and incubated for 24 h. The cells were then incubated in culture medium containing SA (5 or 10 *μ*M). Following 24 h of incubation, cell viability was determined using a Cell Counting Kit-8 (Dojindo Laboratories, Kumamoto, Japan) according to the manufacturer\'s protocol. For the ^3^H-thymidine incorporation assay, hepatocytes were seeded and incubated as described above with the addition of ^3^H-thymidine (1 *μ*Ci/well). After 24 h of incubation, the cells were washed with PBS and then analyzed for ^3^H-thymidine incorporation using a scintillation beta-counter, 1450 LSC and Luminescence Counter MicroBeta TriLux (Perkin Elmer).
2.10. Glucose-Stimulated Insulin Secretion {#sec2.10}
------------------------------------------
After 8 weeks of SA treatment, mice were not fed for 14 h, and then a glucose solution (2 g kg^−1^ body weight) was injected intraperitoneally. Blood glucose levels were measured at 0 and 30 min after glucose injection. The concentration of serum insulin was measured by an ultrasensitive mouse insulin enzyme immunosorbent assay kit (ALPCO, Windham, NH).
2.11. Statistical Analysis {#sec2.11}
--------------------------
Data are presented as mean ± SE. Statistical significance of the difference between two groups was analyzed by unpaired Student\'s *t*-test for comparison of two groups or ANOVA followed by Fisher\'s protected least significant difference test for multiple groups. *P* \< 0.05 was accepted as significant.
3. Results {#sec3}
==========
3.1. Reduction of Blood Glucose Level and Food Intake in SA-Treated*db/db* Mice {#sec3.1}
-------------------------------------------------------------------------------
To investigate the effects of SA on a type 2 diabetic mouse model, we orally intubated diabetic*db/db* mice with SA daily and measured HbA1c levels, glucose levels, food intake, and body weight. We found that HbA1c levels of SA-treated mice were significantly lower compared with the untreated, PBS-treated, and pair-fed control groups at 8 weeks of treatment ([Figure 1(a)](#fig1){ref-type="fig"}). Blood glucose levels were unchanged among groups at 4 weeks of treatment but were significantly lower in the SA-treated group at 8 weeks, compared with the control groups ([Figure 1(b)](#fig1){ref-type="fig"}). Fasting blood glucose levels were also significantly decreased in SA-treated mice compared with untreated, PBS-treated mice, and pair-fed mice at 8 weeks of treatment ([Figure 1(c)](#fig1){ref-type="fig"}). Interestingly, food intake in SA-treated mice was significantly decreased compared with PBS-treated mice ([Figure 1(d)](#fig1){ref-type="fig"}), but body weight gain was significantly increased at 8 weeks of SA treatment ([Figure 1(e)](#fig1){ref-type="fig"}). In contrast, the pair-fed group had significantly lower body weights as compared with the PBS-treated group ([Figure 1(e)](#fig1){ref-type="fig"}). In addition, the blood concentrations of ALT and AST, indicators of liver damage, were not different between SA- and PBS-treated*db/db* mice (see Supplementary Figure 1 in Supplementary Material available online at <http://dx.doi.org/10.1155/2014/961732>). These results suggest that treatment with 10 mg kg^−1^ of SA for 8 weeks may not have toxic effects in*db/db* mice. In low concentrations, SA has been reported to cause transient stimulation of cell growth \[[@B18]\]. In our study, we also found that SA in low doses (5 *μ*M) stimulated hepatocyte cell growth and did not show cell toxicity, even at 10 *μ*M (Supplementary Figure 2).
3.2. Improvement of Glucose Tolerance in SA-Treated*db/db* Mice {#sec3.2}
---------------------------------------------------------------
To determine whether blood glucose levels are properly controlled in SA-treated*db/db* mice, we performed intraperitoneal glucose tolerance tests at 4 and 8 weeks of SA treatment. Blood glucose levels in SA-treated mice were significantly lower at all time points following glucose injection compared with pair-fed mice and significantly lower after 90 min compared with PBS-treated mice at 4 weeks ([Figure 2(a)](#fig2){ref-type="fig"}). At 8 weeks of treatment, blood glucose levels of SA-treated mice were significantly reduced at all time points compared with PBS-treated and pair-fed mice ([Figure 2(b)](#fig2){ref-type="fig"}). Blood glucose levels in the pair-fed group were not significantly different from the untreated or PBS-treated groups at any time point, except at 60 min after glucose loading ([Figure 2(b)](#fig2){ref-type="fig"}). The area under the curve was significantly reduced in SA-treated mice compared with PBS-treated and pair-fed mice at both 4 and 8 weeks of treatment (Figures [2(c)](#fig2){ref-type="fig"} and [2(d)](#fig2){ref-type="fig"}).
To address whether SA treatment improves insulin sensitivity, we performed insulin tolerance tests at 4 and 8 weeks of SA treatment. Glucose reduction in SA-treated mice was not different from untreated, PBS-treated, or pair-fed mice at 4 weeks or 8 weeks of treatment (Figures [3(a)](#fig3){ref-type="fig"} and [3(b)](#fig3){ref-type="fig"}). Similarly, the area under the curves was not different among groups (Figures [3(c)](#fig3){ref-type="fig"} and [3(d)](#fig3){ref-type="fig"}). To measure insulin secretion in SA-treated mice, we analyzed glucose-stimulated insulin secretion after 8 weeks of SA treatment. Insulin secretion was not different between the SA- and PBS-treated groups (Supplementary Figure 3).
3.3. Decreased Expression of Gluconeogenic Genes in SA-Treated*db/db* Mice and Reduction of Glucose Production and Expression of Gluconeogenesis-Related Genes in SA-Treated Hepatocytes {#sec3.3}
----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
To determine whether SA treatment affects the expression of genes involved in glucose production, we examined the expression of G6Pase and PEPCK mRNA, which are involved in gluconeogenesis in the liver. The expression of both G6Pase and PEPCK mRNA was significantly decreased in the liver of SA-treated*db/db* mice compared with PBS-treated mice at 8 weeks (Figures [4(a)](#fig4){ref-type="fig"} and [4(b)](#fig4){ref-type="fig"}). To investigate gluconeogenic fluxes*in vivo*, C57BL/6 mice were orally intubated with SA or PBS. After 15 h later from oral intubation, mice were injected i.p. with sodium pyruvate, and blood glucose levels were measured. SA treatment significantly inhibited blood glucose level at 15 min after pyruvate injection ([Figure 4(c)](#fig4){ref-type="fig"}). To investigate the direct effects of SA on hepatic gluconeogenesis*in vitro*, we treated hepatocytes isolated from C57BL/6 mice with SA and then measured glucose production. SA treatment significantly decreased glucose production approximately 70% and 90% at 5 and 10 *μ*M, respectively, as compared with untreated hepatocytes ([Figure 4(d)](#fig4){ref-type="fig"}). In addition, the expression of G6Pase and PEPCK mRNA was significantly decreased in SA-treated hepatocytes (Figures [4(e)](#fig4){ref-type="fig"} and [4(f)](#fig4){ref-type="fig"}). The promoter activity of G6Pase and PEPCK is regulated by the transcription factors HNF-4*α*, HNF-3*β*, and FoxO1 \[[@B9], [@B19]\]. Small heterodimer partner (SHP) has been shown to downregulate G6Pase and PEPCK via the repression of HNF-4*α*, HNF-3*β*, and FoxO1 \[[@B20], [@B21]\]. Thus, we measured the expression of HNF-4*α* and SHP mRNA in SA-treated hepatocytes. SA treatment significantly decreased the expression of HNF-4*α* mRNA and significantly increased SHP mRNA (Figures [4(g)](#fig4){ref-type="fig"} and [4(h)](#fig4){ref-type="fig"}).
3.4. Decreased Expression of Sirt1 mRNA and Protein and Increased Acetylation of FoxO1 in SA-Treated Hepatocytes {#sec3.4}
----------------------------------------------------------------------------------------------------------------
FoxO1 activity is known to be regulated by phosphorylation and acetylation \[[@B22], [@B23]\]. Sirt1 increases FoxO1 DNA-binding ability by deacetylating FoxO1 and potentiating its transcription activity \[[@B24]\]. Thus, we determined the expression of Sirt1 and acetylated FoxO1 in SA-treated hepatocytes. The expression of Sirt1 mRNA and protein in SA-treated hepatocytes was significantly decreased compared with untreated control hepatocytes (Figures [5(a)](#fig5){ref-type="fig"} and [5(b)](#fig5){ref-type="fig"}). In addition, acetylated FoxO1 was increased in SA-treated hepatocytes compared with untreated cells ([Figure 5(b)](#fig5){ref-type="fig"}).
3.5. Decreased Energy Expenditure in SA-Treated*db/db* Mice {#sec3.5}
-----------------------------------------------------------
SA-treated mice showed a significant increase in body weight in spite of a significant reduction in food intake (Figures [1(d)](#fig1){ref-type="fig"} and [1(e)](#fig1){ref-type="fig"}). To investigate whether the body weight gain is due to metabolic changes induced by SA treatment, we measured food intake, oxygen consumption, carbon dioxide output, energy expenditure, and respiratory exchange ratio by a metabolic monitoring system over 72 h at 8 weeks of SA treatment. Food intake was significantly decreased in SA-treated mice compared with untreated mice ([Figure 6(a)](#fig6){ref-type="fig"}). Oxygen consumption ([Figure 6(b)](#fig6){ref-type="fig"}), carbon dioxide output ([Figure 6(c)](#fig6){ref-type="fig"}), and energy expenditure ([Figure 6(d)](#fig6){ref-type="fig"}) were significantly decreased in SA-treated mice compared with untreated or pair-fed groups. The respiratory exchange ratio was not different in SA-treated or pair-fed mice compared with untreated mice ([Figure 6(e)](#fig6){ref-type="fig"}).
Feeding behavior is regulated by a system with the hypothalamus at the centre, and energy homeostasis is maintained through regulation of food intake and energy expenditure \[[@B25], [@B26]\]. Therefore, we investigated the expression of hypothalamic molecules known to be involved in appetite regulation, AgRP, proopiomelanocortin (POMC), and a POMC receptor, melanocortin 4 receptor (MC4R), in SA-treated mice. Interestingly, the expression of AgRP mRNA, but not POMC and MC4R mRNA, was significantly increased in SA-treated mice compared with PBS-treated controls and pair-fed group (Figures [7(a)](#fig7){ref-type="fig"}--[7(c)](#fig7){ref-type="fig"}).
4. Discussion {#sec4}
=============
The liver is an important organ in the regulation of glucose homeostasis in fed and fasting conditions \[[@B4]\]. Excess gluconeogenesis and glycogenolysis in the liver lead to elevated hepatic glucose production, which is a major cause of hyperglycemia in type 2 diabetes \[[@B6], [@B7]\]. SA has been previously shown to downregulate the expression of the hepatic gluconeogenic genes, G6Pase and PEPCK, in hepatic cell lines, rat hepatocytes, and embryonic chick liver \[[@B14], [@B15], [@B17]\]. In this study, we report that SA reduces hepatic gluconeogenesis in diabetic*db/db* mice, an animal model of human type 2 diabetes. SA-treated diabetic*db/db* mice showed decreased blood glucose levels under fed and fasting conditions and improved glucose tolerance. In addition, the expression of gluconeogenesis-related genes was downregulated in SA-treated*db/db* mice liver and mouse hepatocytes.
Gluconeogenesis is a metabolic pathway that generates glucose from noncarbohydrate carbon substrates. This pathway is catalyzed by several enzymes, first and last ones being PEPCK and G6Pase. SA suppressed the expression of PEPCK promoter-driven luciferase constructs in a rat hepatoma H4IIE cell line \[[@B14]\]. SA also repressed forskolin/dexamethasone-stimulated PEPCK and G6Pase gene expression and induced SHP gene expression via AMP-activated protein kinase to inhibit gluconeogenic enzyme gene expression in hepatic cell lines \[[@B15]\]. In our study, G6Pase and PEPCK mRNA levels were significantly reduced in SA-treated diabetic*db/db* mice liver. In addition, HbA1c and fed and fasting glucose levels were significantly decreased in SA-treated diabetic*db/db* mice. These results suggest that SA could improve hyperglycemia in type 2 diabetes through the reduction of gluconeogenesis.
SHP is a member of the nuclear receptor family of intracellular transcription factors \[[@B27]\]. SHP has been shown to inhibit numerous nuclear receptors and transcription factors as transcriptional corepressor \[[@B28]\]. The induction of SHP has been shown to downregulate G6Pase and PEPCK through the repression of HNF-4*α*-, HNF-3*β*-, and FoxO1-mediated transcriptional activity \[[@B20], [@B21]\]. In our study, the expression of SHP mRNA was significantly increased in SA-treated hepatocytes, and HNF-4*α* mRNA was decreased by SA in a dose-dependent manner in mouse hepatocytes. This suggests that SA-induced increases of SHP work as repressor to inhibit HNF-4*α* expression and subsequently G6Pase and PEPCK production.
FoxO1 is a transcription factor that has an important function in the regulation of gluconeogenesis \[[@B29]\]. Its transcriptional activity is regulated through phosphorylation or acetylation of multiple residues. When FoxO1 is phosphorylated, it is excluded from the nucleus, resulting in decreased transcription of G6Pase and decreased hepatic gluconeogenesis \[[@B22]\]. FoxO1 transcriptional activity is also regulated by acetylation, which inhibits the ability of FoxO1 to interact with the G6Pase promoter by decreasing the stability of the FoxO1-DNA complex \[[@B23]\].
Sirt1 reverses this acetylation process of FoxO1 and has been shown to play critical roles in glucose homeostasis in liver. However, Sirt1 has been shown to confer both positive and negative effects on hepatic gluconeogenesis. The transcriptional activity of FoxO1 is increased by Sirt1 through deacetylation \[[@B24]\]; on the other hand, Sirt1 suppresses the ability of HNF-4*α* to downregulate gluconeogenic gene expression. In our study, the expressions of Sirt1 mRNA and protein levels were significantly decreased depending on SA concentration. It was reported that SA induces NF-*κ*B activation \[[@B30]\], which enhances the expression of miR-34a, a tumor suppressor gene \[[@B31]\]. And SA treatment increases miR-34a expression in TK-6 cells \[[@B32]\], which downregulates the expression of Sirt1 \[[@B33]\]. Thus, the decrease of Sirt1 mRNA by SA in our study may be mediated by the NF-*κ*B/miR-34a pathway. Taken together, all these results suggest that SA decreases Sirt1, thus reducing the deacetylation of FoxO1 and reducing its transcriptional activity, which contributes to the reduced expression of hepatic gluconeogenic genes.
Interestingly, SA-treated*db/db* mice showed reduced food intake over the 8 weeks of the experiment; however, SA-treated mice gained more body weight compared with PBS-treated and pair-fed mice. In addition, SA-treated mice showed decreased consumption of oxygen and production of carbon dioxide, energy expenditure, and locomotor activity compared with pair-fed mice. In the brain, energy balance is regulated by neural and hormonal integrated signals. The hypothalamus is a primary center in which feeding behavior and energy metabolism are regulated \[[@B34], [@B35]\]. AgRP is a neuropeptide produced in the hypothalamus by the AgRP/neuropeptide Y (NPY) neuron. AgRP increases appetite and decreases metabolism and energy expenditure \[[@B36]--[@B38]\]. AgRP treatment for 7 days\' intracerebroventricular administration in rat decreased oxygen consumption and energy expenditure \[[@B39]\]. In our study, the expression of AgRP mRNA was significantly increased in SA-treated mice compared with PBS-treated mice, whereas the pair-fed mice were not different from the PBS group. These results suggest that SA decreases energy expenditure and increases weight gain via the induction of AgRP expression in hypothalamus. It is known that FoxO1 regulates AgRP transcription in the hypothalamus, and decreased FoxO1 activity is associated with decreased AgRP expression \[[@B40]\]. However, we found that mRNA expression of AgRP was increased although FoxO1 transcriptional activity was decreased by SA treatment. The expression of AgRP may be regulated by several molecules as well as FoxO1 and further studies are needed for the detailed molecular mechanisms.
It is paradoxical that food intake decreased in SA-treated*db/db* mice in spite of an increase in AgRP, which might be expected to stimulate appetite. In addition, the appetite-stimulating effects of AgRP are inhibited by leptin \[[@B41]\], but*db/db* mice are deficient for the leptin receptor gene \[[@B42]\]. Therefore, the increase of AgRP expression may not affect the appetite in SA-treated*db/db* mice.
In conclusion, our studies show that SA improves glucose metabolism in a type 2 diabetic mouse model. SA directly inhibits hepatic glucose production via regulation of gluconeogenesis-related genes and protein expression, such as PEPCK, G6Pase, SHP, Sirt1, and acetylated FoxO1. In addition, SA increases body weight via decreasing energy expenditure due to induction of AgRP gene expression in hypothalamus. This may benefit end-state lean diabetes patients and increase their longevity.
Supplementary Material {#supplementary-material-sec}
======================
######
Supplemental Figure 1: The concentration of ALT and AST in blood of SA-treated db/db mice.
Supplemental Figure 2: Proliferation induced in SA-treated mouse hepatocytes.
Supplemental Figure 3: Glucose-stimulated insulin secretion test in SA-treated db/db mice.
This study was supported by grants from Komipharm International Co. Ltd. and in part by a grant from the Korean Health Technology R&D Project, Ministry of Health & Welfare, Republic of Korea (A102060). The authors thank Dr. Ann Kyle for editorial assistance.
Conflict of Interests
=====================
The authors declare that this work was partially supported by a grant from the Komipharm International Co. and that Sujong Kim is an employee of Komipharm International Co.
Authors' Contribution
=====================
Young-Sun Lee and Eun-Kyu Lee contributed equally to this research.
{#fig1}
{#fig2}
{#fig3}
{#fig4}
{#fig5}
{#fig6}
{#fig7}
[^1]: Academic Editor: Bernard Portha
| {
"pile_set_name": "PubMed Central"
} |
Introduction {#Sec1}
============
Early life is a sensitive time period for brain development where adverse experiences can have far-reaching consequences. Nearly one million cases of child maltreatment are reported in the United States each year, which includes reports of neglect, physical abuse, and sexual abuse \[[@CR1]\]. Although not all stress or even all early life stress (ELS) is maladaptive, ELS has been reliably associated with poor life outcomes including heart disease \[[@CR2]\], cancer \[[@CR3]\], and various psychopathologies including major depression and suicidal behavior \[[@CR4]\]. Other types of early life trauma, such as parental loss and natural disaster have also been linked to increased rates of adult-onset depression \[[@CR5], [@CR6]\]. Although major depressive disorder (MDD) is not homogenous, most often it is associated with stress as a predisposing or precipitating event. Globally, the lifetime prevalence of MDD varies by country with an overall rate of 3% \[[@CR7]\] and 15% in the U.S. adult population \[[@CR8]\], costing approximately \$100 billion annually \[[@CR9]\]. Though antidepressant (AD) use has increased over time \[[@CR10]\], both response to and remission rates after using ADs remain low \[[@CR11]--[@CR13]\]. Although our understanding of the neurobiological underpinnings of depression and other psychiatric disorders has substantially improved over the last 2--3 decades, the etiopathology of MDD, and subsequently the optimal therapeutic solution, remains obscure. Moreover, ELS and MDD are both linked to an increased risk of suicide \[[@CR14], [@CR15]\]. It is evident that early life adversity poses a significant public health risk.
Recently, the contribution of early life experiences to stress and depression vulnerability has received tremendous attention \[[@CR16]\]. In a broad context, stress elicits a host of biological responses including neurochemical cascades in the hypothalamic--pituitary--adrenal (HPA) axis \[[@CR17]\] and immune reactions \[[@CR18]\], which can subsequently alter neuronal connectivity and signaling \[[@CR19]\] as well as brain matter density \[[@CR20]\]. In resilient adults, chronic stress promotes adaptation over time through dendritic remodeling in the hippocampus, amygdala, and prefrontal cortex (PFC) \[[@CR21]\], yet in the case of susceptible individuals, these adaptive changes may not be reversible after stress is removed \[[@CR22]\]. In order to enact these changes, the environment interacts with gene function without directly affecting the genome itself through epigenetic modifications like methylation, histone modifications, and noncoding RNAs \[[@CR23]\]. These changes have all been identified in patients with major depression and those who have committed suicide (see \[[@CR24], [@CR25]\] for review) with some focus on ELS \[[@CR23]\]. The adolescent brain is significantly more structurally plastic and readily encodes environmental experience into structural and functional changes through epigenetic programming \[[@CR26]\]. Because development is cumulative, ELS has the potential to cause widespread alterations in brain function that persist over the lifetime \[[@CR27]\]. It is still not clear which factors associated with ELS are the most detrimental for later life outcomes or exactly how these events can have such long-lasting effects on brain functions. Furthermore, it has been proposed that depression, subsequent to ELS, comprises a subset of patients with unique etiopathology who would benefit from different treatment or preventative medicine as compared with other patients \[[@CR28], [@CR29]\]. Further understanding of the molecular neurobiology associated with ELS will bring us closer to identifying vulnerable populations and developing effective treatments.
Adverse prenatal and postnatal experiences have both been associated with later onset depression \[[@CR30]--[@CR32]\]. However, the neurobiological and epigenetic mechanisms by which these stressors lead to adult stress-susceptibility seems uniquely linked to their developmental timing. Physical and psychological stressors experienced by the mother can indirectly affect the fetus prenatally as a direct neurochemical stressor via the placenta. Provencal and Binder \[[@CR33]\] reported that alterations in gene function and epigenetic changes in the placenta which lead to increased placental permeability to glucocorticoids (GCs) are specific outcomes of prenatal stress. While both pre- and postnatal stress have been associated with epigenetic modifications via upstream signaling mechanisms including GC receptor (GR) activity, with prenatal stress these effects are restricted to the uterine environment during gestation \[[@CR33]\]. In contrast, postnatal ELS constitutes a variety of stressor types (i.e. physical or psychological) that are directly experienced until adulthood.
The current review aims to identify noncoding RNAs, specifically microRNAs (miRNAs) as key to stress susceptibility as a result of these directly experienced postnatal stressors. In addition, this review will critically examine the role of ELS-associated miRNAs across the existing depression and suicide literature.
Overview of miRNAs {#Sec2}
==================
The environment alters a host of cellular functions throughout the body via a set of well programmed molecular mechanisms known collectively as epigenetics. Being an integral part of the epigenetic machinery, various modifiers (including noncoding RNAs, methylated cytosine, and chromatin remodelers) have gained recognition in the past decades for their role in development, synaptic plasticity, and cell death and proliferation \[[@CR34], [@CR35]\]. MiRNAs, a class of small noncoding RNAs, have recently been implicated in ELS \[[@CR36], [@CR37]\] and several mental illnesses \[[@CR38]--[@CR41]\]. Their role in neuronal development and brain physiology \[[@CR35]\] has made them a strong candidate for the study of psychiatric disorders affected by ELS. MiRNAs are short sequences of nucleotides (\~22 nucleotides long) that have the direct ability to post-transcriptionally modify the cellular stability of messenger RNA (mRNA) thereby altering subsequent protein production. Regardless of their association with any disease or disorder, individual miRNAs can have hundreds of mRNA targets with varying functions making them global modulators of gene expression \[[@CR34]\]. Within cell nuclei, right after transcription, primary miRNAs (pri-miRNAs) are cleaved by Drosha ribonuclease III (DROSHA) and microprocessor complex subunit, DGCR8, into precursor miRNAs (pre-miRNAs), Fig. [1a](#Fig1){ref-type="fig"} \[[@CR34]\]. Exportin 5 translocates pre-miRNAs to the cytoplasm where they are converted into mature miRNAs by the endoribonuclease, Dicer, and TAR RNA-binding protein (TRBP). The double-stranded miRNA: miRNA\* complex is bound by an Argonaute protein. Argonaute selects one strand as the mature miRNA and the miRNA\* strand is degraded. The mature miRNA within the Argonaute protein is known as the RNA-induced silencing complex (RISC) and can readily pair with specific mRNAs \[[@CR42]\]. The nucleotide sequence of mature miRNA is complementary to one---or sometimes many---mRNAs and typically bind to the 3′ untranslated region (UTR) of the target mRNA. MiRNA most often block the translation of mRNA into proteins by a process of repression involving the deadenylation of target mRNA. Alternatively, when a miRNA sequence is highly complementary to its target mRNA, it can slice the mRNA causing degradation \[[@CR34]\].Fig. 1**a** First RNA polymerase II transcribes the miRNA gene resulting in a pri-miRNA with a hairpin loop structure. This structure is cleaved by DROSHA and DGCR8 (blue arrows) into a pre-miRNA and transported out of the cell by EXPO-5. Dicer and TRBP cleave away the loop structure (gray arrows) leaving a miRNA-miRNA\* duplex. AGO 2 loads the mature miRNA (red), forming the RISC complex, and the miRNA\* strand (black) is degraded. RISC can bind to specific gene targets and lead to translational repression. **b** Methylation at the miRNA gene promoter region can reduce transcription of pri-miRNAs by RNA Pol II. This results in decreased production of mature miRNAs and altered downstream repression of their target genes. **c** In the presence of compatible circRNAs, there is competition for miRNA binding. Each circRNA can have multiple binding sites for a single miRNA effectively reducing miRNA-target gene interactions and their associated translational repression. As a result, both methylation and circRNAs can promote protein production. Abbreviations: miRNA (microRNA), RNA Pol II (RNA Polymerase II), pri-miRNA (primary miRNA), pre-miRNA (precursor miRNA), EXPO-5 (exportin-5), TRBP (Tar RNA-binding protein), RISC (RNA-induced silencing complex), AGO 2 (argonaute), tRNA (transfer RNA), CH~3~ (methyl group), mRNA (messenger RNA), circRNA (circular RNA)
Epigenetic modification of miRNAs {#Sec3}
---------------------------------
More recently, epigenetic modifications of miRNAs have been investigated, especially via miRNA promoter region methylation \[[@CR43]--[@CR46]\]. This adds another layer of complexity to understanding epigenetic mechanisms of disease etiopathology. Canonical DNA methyltransferases add a methyl group to cytosines in the promoter region of miRNA coding genes in the same way that conventional genes are methylated (Fig. [1b](#Fig1){ref-type="fig"}) \[[@CR45]\]. Functionally, methylation-induced conformational changes in chromatin structure make DNA inaccessible for gene transcription---in this case the gene codes for a pri-miRNA \[[@CR47]\]. In this way, methylation at the promoter region of a miRNA can repress miRNA expression. Another epigenetic modifier, long noncoding RNA (lncRNA), has been gaining attention in the study of psychiatric disorders \[[@CR38]\]. Preclinical studies have identified transcriptome-wide changes in lncRNAs after repeated social defeat stress \[[@CR48]\] and learned helpless depression models \[[@CR49], [@CR50]\]. A subset of lncRNAs, circular RNAs (circRNAs), have been described as the "RNA sponge" and serve as master regulators of miRNA expression by binding and inhibiting their function, Fig. [1c](#Fig1){ref-type="fig"} \[[@CR51]\]. Similarly to miRNA promoter region methylation, circRNAs are inversely related to miRNA expression, subsequently increasing miRNA-target gene expression. Whereas methylation can alter miRNA expression by inhibiting its synthesis, circRNAs regulate miRNA bioavailability by binding to mature miRNAs in the cytoplasm.
miRNAs and ELS {#Sec4}
==============
Preclinical studies {#Sec5}
-------------------
Animal models have been used to assess causal effects of ELS on the epigenome. By far, the most popular animal model of ELS is maternal separation. In this procedure, neonatal pups are separated from dams for some time each day for the first 2--3 weeks of life. Many studies report behavioral changes after maternal separation \[[@CR36], [@CR52]--[@CR54]\], although there are some mixed findings across specific behavioral tests, sex, and rodent strain \[[@CR55]\]. To date, only a handful of studies have explored miRNA changes after maternal separation. Zhang et al. reported that 6 h of daily maternal separation from postnatal days (PND) 1--14 resulted in increased expression of miR-504 in the nucleus accumbens \[[@CR53]\]. In animals that also experienced chronic unpredictable stress in adulthood, miR-504 expression was further increased. MiR-504 directly targets the 3′UTR of the dopamine D1 receptor gene (*DRD1*) \[[@CR56]\] and *DRD1*-containing neurons have been shown to play an important role in the development of anhedonic behavior in rodents \[[@CR57]\]. Furthermore, knockdown of DRD1 in mouse medial prefrontal cortex (PFC) causes an increase in avoidant behavior after social defeat stress \[[@CR58]\]. Later, Zhang et al. also reported decreased nucleus accumbens expression of miR-9 in animals who received a combination of maternal separation and chronic unpredictable stress \[[@CR59]\]. Likewise, an interaction between both stressors caused the greatest changes in miR-326 expression in both nucleus accumbens and striatum (increased and decreased expression, respectively) \[[@CR59]\]. Notably, the 3′UTR of the dopamine D2 receptor (*DRD2*) is a predicted target of miR-326, yet Zhang et al. found a positive relationship between miR-326 and *DRD2* expression \[[@CR59]\]. Clearly, further studies are needed to elucidate the mechanism by which miR-326 alters *DRD2* after ELS. Nonetheless, resilience after adult social defeat stress in rats has been inversely correlated with miR-326 expression in the amygdala \[[@CR60]\]. Bai et al. \[[@CR36]\] reported that the same 6 h maternal separation paradigm increased miR-16 expression in the hippocampus as compared with controls and animals who received chronic unpredictable stress. Although only a few changes in miRNA expression were reported after maternal separation, these studies support the notion that ELS induces susceptibility to later life stress at the epigenome level. Uchida et al. \[[@CR52]\] maternally separated rodents for 180 min per day (half of the separation time in Zhang et al. \[[@CR53], [@CR59]\]) and found significant increases in depression-like behaviors such as anhedonia in the sucrose preference test and immobility in the forced swim test as well as increases in miR-132, miR-124, miR-9, and miR-29a expression. MiR-124 and −132 are mostly restricted to the nervous system and are key to brain development through their roles in neuronal differentiation (miR-124) \[[@CR61]\] and morphogenesis (miR-132) \[[@CR62]\]. MiR-9 regulates microglia function through its target HECT domain E3 ubiquitin protein ligase 1 (*HECTD1)* \[[@CR63]\] and miR-29a has been implicated in apoptotic pathways following endoplasmic reticulum stress via its target, an apoptosis regulator: myeloid leukemia cell differentiation protein (*Mcl-1)* \[[@CR64]\]. In addition, RE1 silencing transcription factor (REST), a transcription factor involved in neuronal differentiation, was upregulated after maternal separation \[[@CR52]\]. Overexpression of REST 4 in mice caused increased expression of miR-132, miR-121, and miR-9-3 \[[@CR52]\]. REST can also repress expression by binding to RE1 sites, which can be found on the regulatory elements of the corticotropin-releasing hormone gene, *CRH* \[[@CR65]\] and brain-derived neurotrophic factor (*BDNF)* \[[@CR66]\] genes, among others; both genes are important in stress and depression \[[@CR67]\]. The promoter region of miRNAs miR-132, miR-124, miR-9, and miR-29a are each relatively close to an RE1 binding site \[[@CR65]\]. Bahi \[[@CR54]\] employed an alternative paradigm, where half of the pups in each litter were maternally separated in isolation, while the rest of the pups remained with the dam. In the pups who had been separated, a stereotaxic injection of miR-124 lentivirus into the dentate gyrus increased anxiety-like behaviors in the elevated plus maze and reduced social interaction behaviors along with decreased BDNF mRNA expression. Interestingly, although *BDNF* is a known target of miR-124a and both controls and maternally separated animals received a miR-124 injection, only animals who experienced isolated maternal separation exhibited these changes in *BDNF*. It is possible that ELS causes methylation changes, which may make miRNA targets more available for posttranscriptional modification. Otherwise, these miRNAs, though prevalent throughout the brain, cannot affect mRNA translation.
In contrast to ELS, maternal separation in short bouts of 15 min has also been used to induce increased maternal care behaviors \[[@CR68], [@CR69]\]. A recent study found that this augmented maternal care (AMC) decreased expression of DGCR8, part of the miRNA processing machinery, in the hypothalamus and increased levels of miR-488, −144, and −542-5p \[[@CR69]\]. AMC also decreased expression of miR-421 and miR-376b-5p. These differentially expressed miRNAs are predicted to target a number of genes relevant to stress signaling pathways and neuronal regulation \[[@CR69]\]. In particular, miR-144 is predicted to target Galanin \[[@CR69]\], a protein important to the noradrenergic system and responsive to restraint stress in rodents \[[@CR70]\]. In panic and anxiety disorder, miR-488 has been shown to regulate proopiomelanocortin, a precursor to the HPA hormone, adrenocorticotropin \[[@CR71]\]. Another predicted target of miR-488, arginine vasopressin receptor 1a was recently shown to be downregulated in mice after ELS \[[@CR72]\]. In a study by Uchida et al. \[[@CR52]\], a similar 15-min maternal separation paradigm was used as a handled control and compared with 180-min maternal separation. Compared with animal-facility-reared controls, 15-min maternal separation did not significantly alter behavior or miRNA expression. No other studies have explored the effects of AMC on miRNA expression.
Other animal models of postnatal ELS are mostly applied in the peri-pubertal time period, near PND-28, in order to approximate teen or adolescent stress. Two independent studies have investigated the effects of chronic variable stress (CVS) on miRNA expression in rodent PFC and basolateral amygdala, respectively \[[@CR73], [@CR74]\]. Xu et al. \[[@CR73]\] found that CVS increased miR-18a and −124a expression in the basolateral amygdala by PND-55, but at PND-90, miR-18a expression was the same in CVS and control animals. MiR-124a, however, remained significantly increased in adult rats at PND-90. In order to exogenously activate the HPA response via GR, a separate sample of animals without CVS were dexamethasone-treated and exhibited the same patterns of miRNA expression as CVS animals. Using the same CVS paradigm, Xu et al. \[[@CR75]\] found similarly increased levels of miR-124a and miR-18a expression in both PFC and hippocampus. This further corroborates the hypothesis that ELS acts to sensitize the HPA axis to later life stress. In addition, administration of RU486, a non-selective GR antagonist, negated the effects of both ELS and dexamethasone administration across miRNA expression, gene expression, and behavioral measures \[[@CR73]\] thereby showing that these effects are dependent on GR signaling. Specifically, RU486 returned rearing, crossing, and grooming behaviors in the open field test, percent time in the open arm of elevated plus maze, and sucrose preference to the same level as controls. Morrison et al. \[[@CR74]\] compared isolated and social housing in adulthood after adolescent CVS. Contrary to previous findings in animal models of ELS, they found that CVS alone did not significantly alter any miRNAs across the transcriptome as compared with controls. However, CVS animals housed socially exhibited downregulation of 23 miRNAs \[[@CR74]\]. These findings are consistent with studies that house animals in groups after ELS. However, it is still not clear why social isolation concurrent with CVS did not affect behavioral outcomes or miRNA expression. Finally, Liu et al. \[[@CR76]\] applied an inescapable shock to chronically stressed adolescent rats and found decreased miR-135a expression in PFC and increased miR-16 in the hippocampus. To date, changes in miRNA expression in preclinical studies of ELS are not completely consistent. It seems clear that these changes are not only brain region specific, but they are also dependent on the type and timing of stress paradigm used.
Clinical studies {#Sec6}
----------------
The human literature on ELS as it relates to miRNAs is still quite limited and mostly overlaps with various psychiatric disorders. In an effort to explore the contribution of ELS to schizophrenia onset, Cattane et al. \[[@CR77]\] collected blood samples from 32 control participants (11 with and 22 without early trauma history) and used microarray to assess miRNA expression levels in whole blood. Participants with any psychiatric diagnosis were excluded. Altogether, 80 miRNAs were found to be significantly differentially expressed in the early trauma group compared with participants without trauma \[[@CR77]\]. Specifically, miR-29b-3p, miR-29c-3p, and miR-16-5p were significantly upregulated, while miR-200b-5p and miR-125b-1-3p were significantly downregulated. However, when they studied hippocampal miRNA expression in prenatally stressed rodents, they found decreases only in miR-125-1-3p, which was also present in cortisol-treated hippocampal progenitor cells \[[@CR77]\]. This shows that miR-125-1-3p is specifically responsive to ELS and the effects are lasting and consistent across species. It also implicates hippocampal miR-125-1-3p in the corticosteroid signaling relevant to stress because of its response to cortisol treatment. In fact, miR-125-1-3p has been shown to target aldosterone synthase (*CYP11B2*) \[[@CR78]\], the final enzyme in the conversion of cholesterol into the mineralocorticoid, aldosterone \[[@CR79]\].
Since miRNAs themselves can be regulated through epigenetic modifications, two studies have explored methylation patterns on miRNA promoter regions in relation to ELS. In an all-male sample from low socio-economic status and high child abuse backgrounds, Suderman et al. \[[@CR80]\] found patterns of altered methylation in promoter regions of 31 miRNAs. After confirmation using methylated DNA immunoprecipitation, they found that miR-514, let-7d, miR-520c, miR-215, miR-519a, and miR-519e were hypermethylated, whereas miR-203 was hypomethylated. Another study explored methylation patterns on the promoter region of miR124-3 in blood leukocytes from patients with borderline personality disorder (BPD) \[[@CR81]\]. The results showed that a methylated region near the gene coding miR-124-3p was associated with severity of childhood trauma as well as BPD symptom severity. It should be noted that depressed patients without ELS were used as a comparison control because there was not an accessible BPD cohort without ELS exposure. These two studies are clearly limited in scope by their sample characteristics, nevertheless, they provide insight into the complex epigenetic changes induced by ELS exposure.
From the above described studies, it is clear that miRNA expression varies widely between different ELS paradigms and again across different species, with little overlap between studies. Still, seven miRNAs were identified by more than one study as relevant to ELS experience: miR-16, miR-18a, miR-9, miR-29, miR-200, miR-125, and miR-124. MiR-16 is upregulated in the hippocampus after both maternal separation \[[@CR36]\] and inescapable shock \[[@CR76]\] in rodents as well as in blood from healthy individuals with an ELS history \[[@CR77]\]. In rodents, miR-16 has previously been implicated in serotonin transmission systems \[[@CR82]\] as well as resilience to altered behavior after chronic adult stress \[[@CR83]\]. Conversely, in a recent meta-analysis of human literature, Yuan et al. \[[@CR84]\] found no significant differences in miR-16 expression between depressed patients and controls. Without comparing patients based on ELS exposure, it is unclear whether miR-16 contributes to depression after ELS. In addition, because miR-16 has been implicated in ELS, adult stress, and AD effects, it is not yet clear if miR-16 is a general marker for stress or if it is more specific to ELS.
MiR-124 has been widely studied in relation to MDD \[[@CR40], [@CR85]\] as well as GC function \[[@CR86]--[@CR88]\]. As previously described, stereotaxic injection of a miR-124 lentivirus into the hippocampus downregulates its target, *BDNF*, only in maternally separated animals \[[@CR54]\]. Also, after adolescent CVS, miR-124 shows lasting increases in expression beginning in early adulthood \[[@CR73]\]. Dexamethasone treatment yielded similar changes and RU486, a GR antagonist, negated changes caused by dexamethasone as well as CVS. MiR-124 expression after chronic unpredictable stress in later adolescence also correlates positively with severity of depression-like behaviors and inversely with GR expression in the amygdala \[[@CR73]\]. A 180-min maternal separation also increased miR-124 expression as well as REST4, a key regulator of *BDNF* \[[@CR52]\]. Lastly, in patients with BPD and a history of ELS, Prados et al. \[[@CR81]\] found hyper-methylation of the promoter region of miR-124 was correlated with ELS history and symptom severity as compared with a sample of depressed patients with no trauma history. Together, this evidence supports the premise that ELS sensitizes different brain regions to miR-124 and alters GC pathway signaling, thereby causing depression- or anxiety-related behavioral outcomes. Furthermore, miR-124 interaction with *BDNF* may be mediated by REST4 during adolescent development. Though, it is well understood that miRNAs are responsive to early life environment, it is not clear how these external events precipitate change in miRNAs. Methylation is one candidate mechanism whereby miRNAs may be environmentally altered leading to subsequent changes in gene expression. In cortisol-treated rats, the Dwivedi group \[[@CR40]\] found decreased methylation in the promoter region of miR-124 on chromosome 3 as well as decreased expression of DNA methyltransferase 3a concurrent with increased miR-124 expression levels and decreased target gene expression. Bearing in mind the sample limitations in Prados et al.---i.e., comparison of a no ELS MDD group to an ELS group with BPD---\[[@CR81]\], further studies are necessary to explore miRNA methylation specific to ELS independent of psychiatric diagnosis. Thus far, miR-124 has been implicated in ELS \[[@CR54], [@CR73]\], acute stress \[[@CR87]\], and MDD \[[@CR40]\] as well as BPD \[[@CR81]\], among others. Similar to miR-16, it is not yet clear how ELS and miR-124 uniquely contribute such different psychiatric disorders.
Currently, clinical findings in ELS are limited to studies of peripherally circulating miRNAs. Until the last 5 years, there have been almost no direct comparisons of miRNA profiles between the central and peripheral nervous system. In one study of patients with Alzheimer's disease, it was estimated that 73% of 312 tested miRNAs were detectable in both cerebrospinal fluid and blood, but only 36 of these miRNAs were equally expressed \[[@CR89]\]. In rodents that exhibited behavioral resilience to chronic mild stress, out of ten preselected miRNAs, miR-34a-5p was the only one found to be significantly upregulated in both serum and ventral tegmental area (VTA), but not prefrontal cortex \[[@CR90]\]. Though the use of circulating blood miRNAs may be useful for biomarker detection, future studies will need to apply a systems neuroscience approach to identify viable therapeutic targets for psychiatric disorders.
miRNAs in ELS-induced depression and schizophrenia {#Sec7}
==================================================
Although, miRNA changes relevant to depression are well documented in the literature \[[@CR91], [@CR92]\], no studies in MDD patients have explored the contribution of ELS experience to miRNA expression alterations. A recent meta-analysis showed that ELS interacts with an allele for FK506 binding protein 51 (FKBP5) to confer risk for depression or post-traumatic stress disorder \[[@CR93]\]. Moreover, a SNP in the FKBP5 gene has been associated with susceptibility to develop depression after childhood physical abuse \[[@CR94]\]. FKBP5 is a co-chaperone of GR, binding to GR in the absence of GC \[[@CR95], [@CR96]\] and, in increased concentrations, has been shown to compete for GR binding with corticosteroids \[[@CR97], [@CR98]\]. Normally, FKBP5 acts within cell nuclei to desensitize GRs after a stress response, thereby opposing the HPA response \[[@CR95]\]. MiR-124 may play a role in altered FKBP5 binding and function via its interaction with GR. MiR-124 has been shown to target GR \[[@CR40], [@CR86]\] and relate to depression-like behaviors in animals models \[[@CR73], [@CR85]\]. Xu et al. \[[@CR73], [@CR75]\] not only found behavioral changes after adolescent CVS in rats, but also found increased miR-124 expression, decreased GR expression, and increased FKBP5 in the PFC, hippocampus and basolateral amygdala. Altered methylation patterns at the promoter region of miR-124 have also been found in patients with BPD and a history of early life adversity \[[@CR81]\]. It is possible that ELS confers susceptibility to depression through these epigenetic changes involving miR-124.
Most often, animal models of ELS, including maternal separation and CVS, are intended as a depression model. Studies using animal models of ELS report increased behavioral phenotypes related to depression and anxiety, such as immobility in the forced swim test \[[@CR52], [@CR53]\] or anhedonia as measured by the sucrose preference test \[[@CR52], [@CR53], [@CR73]\]. These behaviors are also reported to correlate with miRNA expression. O'Connor et al. \[[@CR37]\] found that ketamine, electroconvulsive shock therapy (ECT), and chronic fluoxetine administration modified several miRNAs after early maternal separation, including miR-598-5p and miR-451; miR-9 was also downregulated, but only in ketamine and ECT treated animals. Clearly miRNAs are responsive to AD treatment after ELS, but it is still not certain what mechanisms are responsible for this effect. No studies have directly investigated the effect of ADs on miRNA promoter region methylation after ELS, but there is a strong relationship between AD response and global methylation patterns \[[@CR99]\]. Specific miRNA expression alterations have been associated with resistance to ADs treatment, but little more is known about the mechanism of this effect \[[@CR100], [@CR101]\]. Further research is needed to determine if miRNA promoter methylation is a leading mechanism for AD efficacy, particularly in ELS.
Other psychiatric disorders, including schizophrenia, have been associated with ELS history \[[@CR4]\] but little preclinical work involving miRNAs has been done in these disorders. A primary challenge for the field is to develop ELS-based animal models that approximate disorders other than depression or anxiety. Overall, less variety of miRNAs have been implicated in schizophrenia as compared with MDD \[[@CR38]\] or even ELS. MiR-137, miR-181b, and miR-219-5p have consistently been identified in patients as peripheral signatures of schizophrenia regardless of ELS history \[[@CR38]\]. In one clinical study of patients with schizophrenia and self-reported ELS history, miR-125-1-3p was significantly downregulated compared with patients without ELS \[[@CR77]\]. Future studies in patients should utilize self-report measures like the Childhood Trauma Questionnaire \[[@CR102]\] to conduct analyses both on diagnostic status and ELS history; similar methods have been applied to examine differences in brain function via fMRI \[[@CR103]\].
miRNAs in ELS and suicidal behavior {#Sec8}
===================================
Suicide-related deaths are the most significant consequence of such prevalent MDD. Worldwide, an estimated 1 million people commit suicide annually \[[@CR104]\]. Not only has suicide been attributed to depressed mood \[[@CR105]\] or impulsivity \[[@CR106]\], but early-life adversity has also been found to contribute \[[@CR15]\] to increased suicide risk. Though several studies have investigated miRNAs relevant to suicide, no studies to date have assessed the contribution of ELS to miRNA profiles in suicide.
Although suicide is most often associated with depression, impulsivity, or a lack of self-control has also been found to predict suicide attempts \[[@CR107]\] though this relationship is under debate \[[@CR108], [@CR109]\]. Teens who experienced earlier life stress report higher rates of impulsive behavior \[[@CR110]\], and suicide rates among teens are higher than the general population \[[@CR111]\]. Using an in-silico approach, Pietrzykowski and Spijker \[[@CR112]\] identified several putative candidate miRNAs in the amygdala important for impulsive behaviors in mice, such as miR--190b, −28a, −340, −219a, and −49. A SNP in the binding regions of miR-641 on the *SNAP* gene has also been found associated with trait impulsivity \[[@CR113]\]. In healthy individuals with an early adversity history, miR-641 was found to be significantly upregulated \[[@CR77]\]. An interaction between this SNP and miR-641 may be involved in the relationship between ELS and later impulsivity. Several studies have shown changes in miRNA expression in depressed-suicide brain \[[@CR39], [@CR114], [@CR115]\]. Specifically, miRNAs 497 \[[@CR114]\], 146b-5p \[[@CR115]\], and 330-3p \[[@CR39]\] have all been reported in both depressed-suicide victims and healthy individuals with ELS history \[[@CR77]\]. Further studies will be necessary to concretely link miRNA changes to both ELS experience and suicide or suicidal ideation.
Conclusion and future directions {#Sec9}
================================
Epigenetic modifiers, especially miRNAs, have received increasing attention for their role in stress susceptibility after early stress. In the current review, we have presented evidence of altered miRNA expression after ELS experience. Table [1](#Tab1){ref-type="table"} lists all peer reviewed studies of miRNA expression after ELS in both animal models and human participants. Although, there was not much overlap in brain region of interest across the animal studies of ELS, some similar miRNAs were repeatedly found to be significantly altered. These miRNAs were also detected in blood of human participants with ELS history. Namely, miRNAs in the 124, 125, 29, 16, and 200 families were altered both in rodent brain as well as human blood. Rodent studies also reported a correlation between these miRNA expression levels and depressive or anxiety-like phenotypes. Collectively, the evidence suggests that ELS induces changes in miRNA function via a complex interaction of genes relevant to HPA axis as well as other neuroendocrine signaling systems.Table 1Preclinical and clinical studies showing involvement of miRNAs in ELS and associated depression and suicidalityStudySpecies/model or human sampleBrain areas or sourceParticipating miRNAsReference1**Rat**, 180-min maternal separation combined with chronic restraintMedial PFCmiR- 132, miR-124, miR-9, miR-29aUchida et al. \[[@CR52]\]2**Rat**, 360-min maternal separationHippocampusmiR-16Bai et al. \[[@CR36]\]3**Rat**, 180-min maternal separation, fluoxetine, ECT, and ketamine treatmentHippocampusFluoxetine: 2 miRNAsO'Connor et al. \[[@CR37], [@CR33]\]ECT: 10 miRNAsKetamine: 14 miRNAsAll: miR-598-5p and miR-4514**Rat**, 360-min maternal separation and adult chronic variable stressNucleus accumbensmiR-504Zhang et al. \[[@CR53]\]5**Rat**, 360-min maternal separation and adult chronic variable stressNucleus accumbens and striatummiR-9, miR-326Zhang et al. \[[@CR59]\]6**Rat**, 90-min maternal separation and miR-124 lentiviral injectionDentate gyrusmiR-124Bahi \[[@CR54]\]7**Mouse**, chronic variable stressPFC23 downregulated miRNAs including miR-200cMorrison et al. \[[@CR74]\]8**Rat**, inescapable shockPFC and hippocampusPFC: miR-125Liu et al. \[[@CR76]\]Hipp: miR-169**Rat**, chronic variable stress or dexamethasone treatmentBasolateral amygdalamiR-124a, miR-18aXu et al. \[[@CR73]\]10**Rat**, 15-min maternal separation (augmented maternal care)HypothalamusmiR-488, miR-144, miR-542-5p, miR-421, miR-376b-5pVogel Ciernia et al. \[[@CR69]\]11**Rat**, chronic variable stress or dexamethasone treatmentPFC and hippocampusmiR-124a, miR18aXu et al. \[[@CR75]\]12**Rat**, prenatal restraint stressHippocampusmiR-125-1-3pCattane et al. \[[@CR77]\]13**Human**, 40 men with high reported childhood abuseWhole bloodMethylation at promoter region of miR-514, miR-520c, miR-215, miR-519a, miR-519e, mir-203, and let7dSuderman et al. \[[@CR80]\]14**Human**, patients with borderline personality disorderBlood leukocytesMethylation near promoter for miR-124-3Prados et al. \[[@CR81]\]12**Human**, 52 healthy adults characterized by childhood trauma scoreWhole blood80 miRNAs including miR-125-1-3p, miR-29b-3p, miR-29c-3p, miR-16-5p, miR-200b-5p, miR-641, miR-146b-5p, miR-497, and miR-330-3pCattane et al. \[[@CR77]\]
Almost all brain miRNAs are coexpressed at varying levels across different brain regions \[[@CR116]\] presumably according to a region's functional needs. There is also evidence for cell-type (neuron vs glia) specific miRNAs in the central nervous system \[[@CR117]\], which play a role in functions like neuronal differentiation and synaptic plasticity. Because ELS can have a profound effect on brain development (as compared with stress experienced in adulthood) and miRNAs have been strongly implicated in the process of neuronal development, ELS has the potential to significantly modify the anatomical distribution of miRNAs over time. For example, in ELS, neuron-specific miR-124 \[[@CR117]\] was significantly altered in medial PFC, amygdala, and hippocampus \[[@CR52], [@CR54], [@CR73], [@CR75]\], whereas microglia-specific miR-200c \[[@CR117]\] was altered only in PFC \[[@CR74]\]. In the extant literature, animal models of ELS vary widely in terms of stressor type, duration, and developmental timing. Furthermore, few studies have explored the same brain region of interest, with the most studied brain region being the hippocampus. With such limited overlap across the literature, the connection between ELS experience and differences in miRNA expression---and, to a greater extent, function---is still not clear, although some of these studies have suggested that these miRNAs are involved in synaptic plasticity \[[@CR52]\], dopamine receptor attenuation \[[@CR59]\], and HPA-axis regulation \[[@CR73]\]. Future replication is greatly needed in regions like the PFC, amygdala, and hypothalamus, which have also been shown to play a major role in the overall stress response as well as depression symptoms and suicide risk. In addition, it will be interesting to know how these brain areas coordinately regulate miRNA expression. MiR-124 may be of particular importance in ELS because of its role in neurodevelopment \[[@CR118], [@CR119]\]. Our lab has previously reported changes in miR-124 expression in PFC of corticosterone-administered rats \[[@CR40]\]. Changes in miR-124 expression have also been reported after early postnatal stress \[[@CR52]\], peri-adolescent stress \[[@CR73], [@CR75]\], and adult stress \[[@CR87]\]. However, it has not yet been systematically tested whether miR-124 expression covaries with the postnatal developmental timing of stress exposure or other factors like chronicity, duration, or accumulation of stress. In particular, miR-124 has been implicated in synaptic plasticity along with miR-125b, −132, and −485 in dendritic morphology \[[@CR120]\]. Although there is moderate overlap between miRNA affected by adult stress and those altered by ELS, unique changes will become clearer with further study. Rani et al. found that, in healthy adults, miRNA expression is positively correlated with age \[[@CR121]\]. Going forward, it will be important to carefully consider the age of participants in miRNA expression studies, especially those examining the long-term effects of ELS. Considering the majority of the extant research has been limited to studying expression changes, it is also critical to examine the unique mechanisms by which miRNAs are altered after different types of stressors; here, we have presented miRNA promoter region methylation and circRNAs as candidate mechanisms, but the paucity of research on these mechanisms in ELS requires expansion and replication. Lastly, it would be valuable to explore differences in functions of miRNAs as a result of variation in stress. Several groups have used lentiviral injection to selectively augment and reduce individual miRNAs in living rodent brain \[[@CR85], [@CR122]\], including one ELS study \[[@CR54]\]. Knockout mice have also been used extensively to study the importance of specific (and clusters of) miRNAs \[[@CR123]\] and more recent technological advances propose non-invasive means of delivering brain-specific antagomirs to study miRNA knockdown \[[@CR124]\]. In combination with genome-wide expression and in-vitro studies, these techniques can help us to elucidate the importance of miRNAs in stress susceptibility. Understanding miRNA characteristics specific to ELS as opposed to other life stresses will aid in designing more personalized treatment plans for those having depression or suicidal behavior.
Thus far, the human literature on ELS and miRNAs is limited. Future human patient and postmortem studies need to utilize available health history data or family report to determine if the patient experienced ELS. This could become increasingly possible as private health information is digitized and stored in national database. In addition, miRNAs isolated from blood present a challenge because these miRNAs represent a systemic expression profile, rather than brain-specific expression. Exosomes may prove to be the best candidate for isolating brain miRNAs from patient blood. Exosomes are small vesicles, packaged intracellularly in various tissues to move molecules, including miRNAs, extracellularly \[[@CR125]\] and even across the blood--brain barrier \[[@CR126]\]. These vesicle membranes contain proteins exclusive to their cellular origin. Fiandaca et al. \[[@CR127]\] have used neural cell adhesion molecule L1 (NCAM-L1) antibody to immuno-precipitate neuron-specific exosomes from patient plasma. Although, a specific brain region cannot be identified for these peripheral circulating exosomes, this technology may help to identify brain biomarkers for disease susceptibility in living patients and may prove useful in prevention of disease onset after ELS. Ultimately, though, these methods still require validation in psychiatric patient populations. Initially, large scale and genome-wide studies of brain-derived exosomal noncoding RNAs will be particularly important for characterizing patients and their treatment outcomes. Because of their ability to contain and transport miRNAs and cross the blood--brain barrier, exosomes have even been proposed as a treatment vehicle \[[@CR128]\]. In addition, it has been widely shown that current AD therapies can alter miRNA expression \[[@CR129]\]. In the near future, providers may be able to use predictive algorithms to select AD treatments such that they target miRNAs either based on a diagnosed psychiatric disorder or even the individual's circulating miRNA landscape as assessed by genetic testing. Although current research on ELS and miRNAs is limited, the exponential growth in both scientific technology and the storage of all types of health information makes it a promising avenue for neuropsychiatry study.
**Publisher's note** Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
The research was partly supported by grants from National Institute of Mental Health (R01MH082802; 1R01MH101890; R01MH100616; 1R01MH107183; R21MH112014) and American Foundation for Suicide Prevention (DIG-0-041-18) to D. Dwivedi.
The authors declare that they have no conflict of interest.
| {
"pile_set_name": "PubMed Central"
} |
Introduction {#s1}
============
Thrombotic microangiopathy (TMA) encompasses a group of disorders presenting with microangiopathic hemolytic anemia (MAHA), thrombocytopenia, and ischemic organ damage, most frequently of the kidneys and the central nervous system \[[@b1], [@b2]\].
Atypical hemolytic uremic syndrome (aHUS) is a complement-mediated TMA caused by dysregulation of the alternative complement pathway. At least 50% of patients have an underlying inherited or acquired complement abnormality, exacerbated by complement-activating conditions, like infections, drugs, pregnancy, or cancer \[[@b3], [@b4]\].
Drug-induced TMA has been reported with antineoplastic agents including gemcitabine, docetaxel, and doxorubicin \[[@b5], [@b6], [@b7]\]. Direct cytotoxic and immune-mediated endothelial damage have been proposed as underlying pathologies \[[@b5]\]. While immune-mediated damage generally shows acute onset within 2 -- 3 weeks of drug exposure, the clinical manifestation of cytotoxic damage is either acute or slowly progressive with cumulative dose-dependent toxicity \[[@b5], [@b8], [@b9]\]. Distinguishing cancer-related TMA as a consequence of cancer itself from cases of chemotherapy-induced TMA can be challenging. However, metastatic disease is more common in cancer-related TMA, whereas in chemotherapy-induced TMA, little or no active malignancy is detectable \[[@b10]\]. While discontinuation of the offending drug and supportive care are the primary treatment options in drug-induced TMA, in some cases this intervention is unable to limit the already dysregulated complement activity and requires therapeutic complement inhibition.
Eculizumab, approved as therapy for aHUS in 2011, is a humanized monoclonal antibody that binds to the complement component C5, preventing its cleavage into C5a and ultimately the formation of the membrane attack complex (SC5b-9) \[[@b11]\].
While the contributory role of complement dysregulation in drug-induced TMA is increasingly acknowledged, data on the efficacy of eculizumab, the duration of such therapy, and the incidence and type of detected complement abnormalities are sparse.
Here, we report on two patients with chemotherapy-induced TMA, who were successfully managed with temporary eculizumab therapy and remained relapse free for a follow-up of 47 and 15 months, respectively. [](#Table1){ref-type="table"}
Case reports {#s2}
============
Patient 1 {#sub-s1}
---------
A 52-year-old woman admitted with acute onset of altered mental status, bloody diarrhea, and anuric acute kidney injury. Six months prior to admission, chemotherapy with docetaxel, doxorubicin, and cyclophosphamide was started for invasive ductal breast cancer. The last dose of chemotherapy was administered 3 days before symptoms started. Admission laboratory showed serum creatinine of 480 µmol/L, Coombs-negative hemolytic anemia with schistocytes on peripheral blood smear and thrombocytopenia of 64/µL. Lactate dehydrogenase (LDH) was 1,658 IU/mL and haptoglobin \< 0.10 g/L. Coagulation tests were within the normal range, international normalized ratio (INR) 1.3 and partial thromboplastin time (PTT) 35 seconds, ruling out disseminated intravascular coagulation. Shiga toxin producing *E. coli*associated hemolytic uremic syndrome (STEC-HUS) was excluded by negative stool cultures for Shiga toxin-producing *E. coli* strains. Stool cultures for shigella, salmonella, campylobacter, and yersiania as well as PCR for *Clostridium difficile* were negative. Thrombotic thrombocytopenic purpura (TTP) was excluded by ADAMTS13 activity of 37% of control values. No ADAMTS13 antibodies were detected.
Shortly after admission, she developed seizures with respiratory failure requiring intubation. With the diagnosis of TMA, therapeutic plasma exchanges (TPE) were started. Daily TPE over 12 days and steroid therapy showed no effect on clinical symptoms and hemolysis. She had persistent seizures and required renal replacement therapy. Eculizumab was eventually initiated 20 days after admission. Immediately after the first dose of eculizumab, we observed a rapid and dramatic improvement of neurological symptoms. Renal replacement therapy could be discontinued 2 weeks later. Eculizumab was administered 6 times over a period of 5 weeks ([Figure 1](#Figure1){ref-type="fig"}A). Breast-conserving surgery was performed 7 weeks after termination of eculizumab, followed by radiation therapy. During 47 months of follow-up, renal function continued to improve (eGFR 68 mL/min), and no relapse of TMA has occurred ([Figure 1](#Figure1){ref-type="fig"}A) ([Table 2](#Table2){ref-type="table"}).
Next-generation sequencing identified a homozygous polymorphism in complement factor H (CFH) gene (synonymous variant c.1419 G\>A, p.Ala473Ala). CFH autoantibodies were not detected.
Patient 2 {#sub-s2}
---------
A 57-year-old woman admitted with hypertensive urgency, progressive decline of renal function, and MAHA. She was diagnosed with pancreatic cancer 30 months prior to admission and since then treated with gemcitabine and Nab-paclitaxel. Chemotherapy had been discontinued 6 weeks earlier when a decline in renal function, hemolytic anemia, and mild thrombocytopenia (130/µL) was first noted.
Admission laboratory showed serum creatinine of 363 µmol/L (eGFR 11 mL/min), Coombs negative hemolytic anemia with schistocytes on peripheral blood smear, and thrombocytopenia of 108/µL. LDH was 760 IU/mL, haptoglobin \< 0.10 g/L, and coagulation tests were within the normal range (INR 1, PTT 31 seconds). Stool cultures were negative for Shiga toxin-producing *E. coli*, and ADAMTS13 activity was 61% of control values. Urinalysis showed proteinuria with a protein/creatinine ratio of 0.8 g/g. Several plasma infusions were given without improvement of hemolysis or renal function. The patient received eculizumab 7 days after admission, followed by prompt resolution of hemolysis and improvement of renal function. Eculizumab was discontinued after a total of 8 doses over a period of 10 weeks. During 15 months of follow-up, renal function remained stable (eGFR 36 mL/min), and no relapse of TMA occurred ([Figure 1](#Figure1){ref-type="fig"}B) ([Table 2](#Table2){ref-type="table"}). The patient died of pancreatic cancer 18 months after initial hospital admission.
Next-generation sequencing identified a heterozygous polymorphism in CFH gene (synonymous variant c.1419 G\>A, p.Ala473 Ala).
Discussion {#s3}
==========
We report on two patients with chemotherapy-induced TMA, persistent after discontinuation of the culprit drug and TPE/plasma infusion. In both patients, the clinical response to therapy with eculizumab was prompt and remission stable after cessation of treatment. Treatment with eculizumab was well tolerated, and no adverse events were reported.
Due to limited clinical experience, the optimal strategy for treatment of chemotherapy-induced TMA, especially the role of eculizumab, is not yet clear. Discontinuation of the offending drug and supportive care are the primary treatment options. Due to the long turnaround time for us to receive the results of the ADAMTS13 activity plasma infusions (case \#2) and therapeutic plasma exchange (case \#1) was initially used.
In agreement with others we found that TPE/plasma infusion was not effective in patients with chemotherapy-induced TMA \[[@b12], [@b13], [@b14], [@b15], [@b16], [@b17], [@b18]\].
Patient 1 presented, among other symptoms, with bloody diarrhea. However, the presence of diarrhea is not sufficient to exclude other forms of TMA as \~ 30 -- 40% of aHUS and TTP cases involve gastrointestinal symptoms, including bloody diarrhea \[[@b19], [@b20]\].
Genetic mutations leading to dysregulation of the alternative complement pathway or autoantibodies against complement regulatory proteins are identified in \~ 50% of aHUS patients \[[@b4]\]. However, genetic variants in complement regulatory proteins are also detected in patients with secondary, e.g., chemotherapy-induced, TMA \[[@b3]\]. In such patients, the underlying dysregulation of the alternative complement system may be unmasked by the applied drug, acting as a complement-activating trigger. Given the incomplete penetrance of the genetic defects, complement-activating conditions play an important role for the development of TMA \[[@b3], [@b21], [@b22]\].
In most patients with TMA, a complement-activating trigger can be identified, and in 28% of patients with an activating trigger, a genetic risk mutation can be found \[[@b3]\].
The largest group of aHUS-associated mutations occurs in the CFH gene, and more than 60% of these mutations are clustered within the C-terminal recognition region \[[@b23], [@b24], [@b25]\]. CFH is a central regulator of the alternative pathway of complement by acting as a cofactor to factor I in the breakdown and inactivation of C3b \[[@b26]\].
In both of our patients, we identified a polymorphism in the CFH gene (c.1419 G\>A). This variant has a high allele frequency in the general population, and its isolated occurrence seems not to be associated with an increased aHUS risk \[[@b27]\].
The mechanism of gemcitabine-related TMA appears to be dose-related with a reported incidence of 1%. Both immune-mediated and cytotoxic injury have been proposed as underlying pathophysiology \[[@b12], [@b28]\]. Cytotoxic damage is the assumed mechanism in the few described cases of doxorubicin- and docetaxel-related TMA \[[@b5], [@b6], [@b29]\].
To date, several cases of the use of eculizumab in chemotherapy-induced TMA have been reported, most of them regarding gemcitabine (summarized in [Table 1](#Table1){ref-type="table"}). Before the availability of eculizumab, a case series reported \~ 29 patients with suspected gemcitabine-related TMA. Despite discontinuation of gemcitabine, 7 (24%) patients progressed to end-stage renal disease (ESRD), and 3 (10%) patients developed chronic renal failure \[[@b30]\]. These reports and our observation support induction therapy with eculizumab in cases of persisting TMA.
Eculizumab is approved for lifelong therapy of aHUS. However, the possible side effects, especially the risk of meningococcal infection, the inconvenience of a bi-monthly application, and the significant costs have prompted interest in alternative dosing schedules and complete discontinuation. A recent review analyzed data from unpublished cases, published case reports, clinical trials, and the Global aHUS Registry regarding patient outcomes after eculizumab discontinuation \[[@b31]\]. Of the case reports, a subsequent TMA manifestation was observed in 31% (16/52) of patients after eculizumab discontinuation. Data from five clinical trials documented a relapse in 20% (12/61) of patients after cessation of therapy with eculizumab with a median follow-up of 24 weeks. Terminal renal failure occurred in 5% (3/61) of the patients. Of note, relapse risk was independent of an identified genetic mutation, high-risk polymorphism, or autoantibody status. Data from the Global aHUS Registry found a relapse in 16% (12/76) of patients. In the cases described above, disease recurrence was unpredictable in both timing and severity \[[@b31]\].
The French aHUS Registry described a relapse rate after eculizumab discontinuation in 31% (12/38) of the patients \[[@b32]\]. The risk of recurrence was higher in the presence of complement gene variants. The highest risk was associated with CFH variants, whereas no relapse was seen in patients without identified mutations or negative CFH autoantibodies. In case of relapse, early reinstitution (≤ 48 hours) of eculizumab resulted in rapid hematologic remission and a return of serum creatinine to baseline level \[[@b32]\].
While current evidence suggests a relapse rate after eculizumab discontinuation of \~ 30%, there is little available clinical data for estimating the risk of relapse in chemotherapy-induced TMA \[[@b29]\].
In 2017, the KDIGO controversies conference published recommendations for best treatment strategies in aHUS. No evidence was currently seen to support lifelong therapy in all aHUS patients. The consensus suggested that eculizumab withdrawal could be considered on an individual and risk-stratified basis after a minimum treatment duration of 6 -- 12 months to ensure recovery of endothelial damage \[[@b33]\]. Important risk factors for TMA relapses constitute an identified genetic mutation, former TMA episodes, or concomitant permanent or likely recurrent complement-activating condition. Close monitoring of renal function and hematological parameters after eculizumab withdrawal is mandatory; however, there are no evidence-based data about the reliability of a specific parameter and the optimal frequency of testing \[[@b33]\].
In summary, our report supports the role of complement-directed therapy with eculizumab as an effective therapeutic option in the management of refractory chemotherapy-induced TMA. In our opinion, eculizumab discontinuation is feasible in carefully selected patients after permanent removal of the complement-activating condition. Further studies are needed to elucidate the role of genetic variants in complement-regulatory proteins in chemotherapy-induced TMA and to define parameters predictive of complement activation and likely TMA recurrence. Until then, the decision to withdraw eculizumab has to be made on an individual basis.
Funding {#s4}
=======
None.
Conflict of interest {#s5}
====================
B.S.: Consultant/Speaker Honoraria from Alexion, Amgen, Novartis, Astellas, Boehringer Ingelheim, Vifor Pharma, Astra-Zeneka, Janssen. Grants from Alexion, Sanofi, Pfizer.
C.B. is an employee of Limbach and holds a part-time faculty appointment at the University of Freiburg. His research lab receives support from the Deutsche Forschungsgemeinschaft (DFG) DFG BE 3910/8-1 and DFG BE 3910/9-1, the Collaborative Research Center (SFB) KIDGEM 1140 and from the Federal Ministry of Education and Research (BMBF, 01GM1903I and 01GM1903G). He received speaker honoraria from Alexion and PTC Therapeutics.
The other authors do not have a conflict of interest.
{#Figure1}
Table 1.Studies of chemotherapy-induced thrombotic microangiopathy treated with eculizumab.Patients\
(n)DrugPrevious therapyDoses of eculizumab (range) / duration of treatmentMedian follow-up (range)Improved renal outcomeGenetic analysisReference1GemcitabineDW + TPE + Steroids + RTX4 / 3 weeks17.5 weeksYesNTStarck \[[@b34]\] 20141Mitomycin CDW + TPE8 / 3 months18 monthsYesNTFaguer \[[@b35]\] 20131CisplatinDWNot reported / 4 monthsNot reportedYes, relapse 2 months after stop of eculizumabCD46 mutationGilbert \[[@b36]\] 20134GemcitabineDW + TPE in 1 patient\
DW + steroids in 1 patient\
DW in 2 patients6.25 (5 -- 8) / not reportedNot reportedYesNTAl-Ustwani \[[@b37]\] 20141GemcitabineDW + TPE4 / not reported11 weeksNoNDTsai \[[@b38]\] 20141GemcitabineDW + Steroids6 / 7 weeks3 monthsNoNTKarkowsky \[[@b39]\] 20151GemcitabineDW + TPE7 / 10 weeksNot reportedYesNTRogier \[[@b16]\] 20161GemcitabineDW + TPE7 / 8 weeks3 monthsYesNTLopez \[[@b40]\] 20178GemcitabineDW4.5 (3 -- 22) / not reportedNot reportedYesNTGrall \[[@b41]\] 20167Gemcitabine\
Dasatinib\
Bevacizumab\
BleomycinDW\
DW + TPE (2 patients)Not reported / 14 weeks\
(2 -- 24 weeks), 1 ongoingNot reportedYesNTWeitz/Deloughery \[[@b42]\] 20182Gemcitabine\
CarfilzomibDW + TPENot reported42.5 weeks (33-52)YesNTGosain \[[@b43]\] 20171GemcitabineDW + TPE20 / 9 months17 monthsYesNTKrishnappa \[[@b44]\] 2018[^1]
Table 2.Summary case report 1 and 2.CaseAge/ GenderCACOrgan involvement/ PresentationGeneticEculizumab duration/dosesCreatinine (µmol/L) at the end of eculizumabCreatinine (µmol/L) at last follow-upRelapseFollow-up (months)152 y/FDocetaxel DoxorubicinKidney/acute\
CNS/acute\
Lungs/acuteCFH polymorphism synonymous variant\
c.1419 G\>A, p.Ala4773 Ala5 weeks/612083No47257 y/FGemcitabineKidney/chronic\
Severe hypertensionCFH polymorphism\
Synonymous variant\
c.1419 G\>A\
p.Ala473 Ala10 weeks/8154140No15[^2]
[^1]: DW = offending drug withdrawn; TPE = therapeutic plasma exchange; RTX = rituximab; NT = not tested; ND = not detected.
[^2]: CAC = complement activating condition; CNS = central nervous system; CFH = complement factor H.
| {
"pile_set_name": "PubMed Central"
} |
Introduction {#Sec1}
============
Growing amounts of evidence are showing the influence of sex (i.e., biological determinants) and/or gender (that includes socio-cultural matter) on pathological conditions and clinical outcomes \[[@CR1], [@CR2]\]. Differences have been detected in a number of either transmissible or non-transmissible diseases. In particular, epidemiology clearly suggests that, apart from reproductive organs, several forms of cancers of great relevance, e.g., melanoma or leukemias, clearly display a gender disparity in terms of incidence, prevalence, or response to therapy \[[@CR3]\]. Although a general female advantage has been observed by these epidemiological studies, few data have emerged so far explaining this trend \[[@CR4]\]. Recent results also suggest that sex-specific differences could take place in cell death programs, representing critical features for the identification of sex-specific chemotherapeutic targets \[[@CR5]\]. As a general rule, cells from males and females (here called male cells, XY, and female cells, XX) respond differently to injuries possibly because of their different capability to face cellular stress \[[@CR6]\]. In particular, the same stressor can preferentially induce apoptosis in male cells, and survival by induction of autophagy in female cells. Of notice, these results come from studies carried out with non-transformed cells, essentially vessel cells (such as vascular smooth muscle cells or endothelial cells and fibroblasts) \[[@CR7]--[@CR9]\]. This difference is probably due to the greater capacity of the XX cells to prevent and repair the damage than XY ones \[[@CR8]\]. In addition, XY and XX cells are differently susceptible to various cytotoxic agents. For instance, XY neurons were reported to be more sensitive to excitotoxicity than XX neurons, which appeared more prone to staurosporine-induced apoptosis \[[@CR9]\]. These gender differences in cell susceptibility to an exogenous stress, that could be applied to non-cancer cells of different histotype, seem to be related to the inability of XY cells to maintain intracellular levels of reduced glutathione, paralleled by an increased activity of Superoxide Dismutase (SOD) and catalase activity in XX cells as observed either in vitro or in in vivo studies \[[@CR2], [@CR6]--[@CR13]\]. Both genetic and hormonal differences have been hypothesized to be at the basis of all these disparities contributing to sex-specific phenotypes \[[@CR14]\].
Discussion of literature data {#Sec2}
=============================
The genetic issue {#Sec3}
-----------------
Many observations on the differences between XX and XY cells are derived from studies carried out with in vitro models. This suggests that at least part of the observed disparities are independent from the effects of sexual hormones and could be directly imputable to genetic differences, including X and Y sex chromosomes. However, the study of the roles of these chromosomes in the maintenance of cell homeostasis and death is still at the beginning. One key point deals with the presence of the two X chromosomes in cells from females and one X chromosome in cells from males. As a general rule, the phenomenon of X-inactivation should theoretically re-equilibrate female and male gene expression. However, this is not always the case, as up to 15% of X-linked genes escape X-chromosome inactivation (XCI) leading, in females, to the presence of a second, functional copy of the genes located in the XCI-escaped regions \[[@CR15], [@CR16]\]. A systematic analysis of these genes has not been performed, due to the extremely heterogeneous scenario of XCI (escaper genes may display different levels of expression between tissues and subjects \[[@CR17]\]), thus making it difficult to understand which role, if any, these escaper genes may play in cell-sex disparity. However, Dunford and coauthors recently reported that, among escaper genes, there are tumor suppressors (ATRX, CNKSR2, DDX3X, KDM5C, KDM6A, and MAGEC3) which, when mutated, are significantly associated with cancer in males \[[@CR18]\]. On this basis, it has been suggested that the presence of a second functional copy of a gene in an XCI-escaper region could protect females from the negative effects of the mutated copy \[[@CR18]\]. In few words, the presence of two X chromosomes, i.e., different alleles of the same gene, could provide a significant advantage to cells from females since they could counteract gene mutations leading to cancer \[[@CR19]\]. However, it is possible that this sort of "XCI-mediated protection" does not last for the whole life. Actually, it is reported that from middle age (around 55 years) onward, a phenomenon of age-related preferential inactivation of one X chromosome takes place \[[@CR20]--[@CR23]\]. This XCI skewing can lead to the expression of deleterious alleles and thus to an increased risk of morbidity. Actually, women that are offsprings of long-lived parents have a lower XCI skewing and lower prevalence of several diseases such as cardiovascular, skeletal, respiratory, neurologic diseases, and cancer with respect to the age-matched counterparts that were born from non-long-lived parents \[[@CR24]\]. The reasons for this XCI skewing toward the chromosome containing the deleterious alleles are not known. However, this phenomenon has been observed since 1995 for many diseases including cancer \[[@CR25]--[@CR30]\].
On the other side, the presence in women of a number of non-inactivated X-linked alleles is associated not only to the protection against diseases but also to some immunological disorders, including autoimmune diseases \[[@CR31]\]. This phenomenon is likely accounted by the fact that many XCI-escaper genes are involved in immune response. This issue will be further discussed later.
The epigenetic issue {#Sec4}
--------------------
Besides the genetic determinants mentioned above, a number of epigenetic factors, including microRNAs (miRs), have been hypothesized to play a role in cancer onset and progression. MiRs are in fact emerging as critical factors in post-transcriptional regulation of gene expression by influencing the main cellular activities, including cell proliferation and dissemination capabilities as well as cell death. These short non-coding RNAs (21--25 nucleotides) negatively modulate their target mRNAs by direct binding to the 3ʹuntranslated regions (UTRs) with consequent translational repression or mRNA degradation, depending on sequence complementarity \[[@CR32]\]. Notably, X chromosome contains an unexpected high number of miRs, at present 118, in comparison with only 2 miRs localized on chromosome Y, and an average of 40--50 on the autosomes \[[@CR33]\]. The regulatory power of these small non-coding RNAs is well recognized, as 30--50% of all protein-coding genes are targeted by miRs according to complex interconnected circuitries as each gene might be regulated by many miRs and each single miR has up to hundreds of direct targets.
The role of miRs in cancer has been deeply analyzed demonstrating their tumorigenic as well as tumor-suppressor functions. As concerns sex chromosomes, it is important to highlight that the X-chromosome not only displays a high density of miRs, but also that several of them are involved in immunity regulation \[[@CR34]\]. Moreover, the contemporary presence, on this chromosome, of genes coding for proteins playing immunological roles together with a number of miRs impacting on immune system integrity and function may suggest the existence of sex-related loops functional to immunosurveillance (or failure of immunosurveillance) against tumor onset and progression \[[@CR33]\]. Since the development of immunotherapeutic approaches recently gained the attention of physicians in the fight against cancer, the scenario depicted above appears to be of great interest. In particular, inhibitors of different immune checkpoints have been introduced for the treatment of different, otherwise non-responsive, tumors as well as, more recently, for first line treatment. This strategy stems from the ability of cancer cells to escape the surveillance normally exerted by the immune system via a series of key molecules, e.g., cytotoxic T-lymphocyte antigen-4 (CTLA-4) or programmed cell death protein 1 (PD-1), capable of inhibiting immune cells \[[@CR35], [@CR36]\]. In particular, PD-L1 and PD-L2 are transmembrane proteins that, by binding to their receptor PD-1, activate the PD-1/PD-L1 or PD-L2 pathway and induce T-cells to undergo apoptosis suppressing their activation. Hence, the immunosurveillance against tumors is hindered and tumor progression can take place \[[@CR36]\].
Finally, it is important to mention that further possible epigenetic mechanisms could give rise to sex disparities. In particular, it has been observed that DNA methylation pattern of autosomes is different between males and females \[[@CR37], [@CR38]\]. In the first study \[[@CR37]\], carried out in saliva samples, 580 autosomal sites showing strong differences between males and females have been detected. In the second study \[[@CR38]\], performed on three independent cohorts of European subjects, 1184 CpG sites with different methylation levels between sexes were identified in blood cells. The same study also showed that two genes, CISH and RAB23, displayed a significant association between DNA methylation and expression in men and women. These two genes are involved in Hedgehog pathway \[[@CR39]\] and inhibition of the JAK-STAT pathway \[[@CR40]\]. Hedgehog pathway is crucial for response to injury, tissue stress, healing, and regeneration, whereas JAK-STAT pathway is fundamental for regulatory T-cell function \[[@CR40]\]. Therefore, a sex-specific epigenetic control over these two pathways may in part account for the observed differences between men and women in immune and autoimmune responses \[[@CR41]\].
Sex and immunity {#Sec5}
----------------
A significant difference between the male and female immune system function has recently been described. As reviewed by Klein and Flanagan, the female immune system appears more efficient in a number of species, including humans \[[@CR41]\]. Importantly, this different efficiency is present all along the entire human lifespan. From childhood to old age, the female immune system appears more powerful and able to better counteract infectious and non-infectious diseases, including cancer \[[@CR41]\]. This fact may be related to the different evolution of the two sexes in relation to their different and complementary biological functions. It means that, independent from the hormone-related functions, immunity is per se sex-biased, i.e., it could depend upon genetic or epigenetic matters. In addition, in females, this advantage can also become detrimental since the sex-biased nature of the immune system function can result in autoimmune diseases \[[@CR31], [@CR41], [@CR42]\]. In this scenario, a role of sex has also been linked to PD-1 modulation \[[@CR43], [@CR44]\]. In patients with melanoma, the inhibition of PD-1/PD-L1 interaction resulted significantly associated with sex as the median objective response rate (ORR) was 54.6% among men and just 33.1% among women, and median progression-free survival (PFS) was 18 months vs. 5.5 months, respectively \[[@CR43], [@CR45]\]. The underlying reasons are not known, but they possibly involve the immune system sex disparity obviously based on sex hormones and other factors differently expressed in women and also on genetic, e.g., sex chromosomal-associated issues, and epigenetic signals, such as miRs.
Sex chromosomes and miRs {#Sec6}
------------------------
Based on the above findings, an miR-dependent regulation of molecules of relevance in modern immunotherapy such as PD-L1 should merit particular attention. Indeed, PD-L1 expression appears to be directly or indirectly controlled by several X-linked miRs (Fig. [1](#Fig1){ref-type="fig"} and Table [1](#Tab1){ref-type="table"}). According to TargetScan 7.1 (www. targetscan.org), miR-106b, miR-20b, and miR-513, all three localized on chromosome X, are putative repressors of PD-L1 by direct binding to their 3ʹUTRs. Specifically, miR-106b and miR-20b, which are part of the miR-106--363 cluster, including miR-106a, 18b, 20b, 19b2, 92a2, and miR-363, were reported to play an oncogenic role in different tumors \[[@CR46]\]. Furthermore, miR106a was reported to downregulate the anti-inflammatory cytokine IL-10 \[[@CR47]\] and the whole cluster was suggested to play a role in both innate and adaptive immunity \[[@CR48]\]. Interestingly, miR-106b and miR-20a, together with miR-221, were proposed as biomarkers for early detection of gastric cancer \[[@CR46]\]. Other miRs, such as miR-513 and miR-514, members of the X-linked primate-specific miR-506--514 cluster, have been associated with cancer, particularly with melanocyte transformation, melanoma promotion, and sensitivity to BRAF inhibitors \[[@CR49], [@CR50]\]. In addition, in biliary epithelial cells, miR-513 is downregulated by interferon- γ (IFN-γ) and regulates PD-L1 translation by direct targeting, thus suggesting an miR-mediated gene regulation of responses to IFN-γ \[[@CR51]\].Fig. 1X-linked microRNAs regulating cancer cell homeostasis. Schematic representation of miR-based, direct or indirect, regulation of PD-L1, as an example of immune checkpoint key molecule. Additional relevant genes targeted by the indicated miRs and involved in cell cycle and cell death modulation are also shown. See text for details. PD-L1 programmed cell death protein ligand, HIF1α hypoxia-inducible factor-1α, STAT3 signal transducer and activation of transcription-3, ER α estrogen receptor α, IFNγ interferon γ, PTEN phosphatase and tensin homolog, PUMA p53 upregulated modulator of apoptosis, APAF1 apoptotic protease activating factor-1, CASP3 caspase 3, BCL2L11 BCL2-like11, PPP2R2A protein phosphatase 2, regulatory subunit b, α, TP53INP1 tumor protein p53-inducible nuclear protein 1, PGC-1α peroxisome proliferator-activated receptor-gamma coactivator 1-alpha, CDKN1B -p27kip1 cyclin-dependent kinase inhibitor 1b, CDKN1C-p57kip2 cyclin-dependent kinase inhibitor 1cTable 1X-linked miRNAs putatively involved in direct or indirect regulation of the PD-1/PD-L1 axismicroRNA IDX-chromosome positionRegulationRole in cancer3′UTR targetGender disparitiesReferences**hsa-miR-106a**chrX: 134170198-134170278 \[-\]ERProliferation, invasion, migration and drug resistancePD-L1; PD-L2; HIF-1; STAT3F\>M in systemic lupus erythematosus\[[@CR31], [@CR46], [@CR86]\]**hsa-miR-18b**chrX: 134170041-134170111 \[-\]ERProliferation, metabolismHIF-1--\[[@CR53], [@CR86]\]**hsa-miR-19b**chrX: 134169671-134169766 \[-\]ERProliferation, invasion, drug resistanceHIF-1--\[[@CR53], [@CR86]\]**hsa-miR-20b**chrX: 134169809-134169877 \[-\]ERProliferation, invasion, autophagy, drug resistancePD-L1; PD-L2; HIF-1; STAT3--\[[@CR46], [@CR56], [@CR86]\]**hsa-miR-221**chrX: 45746157-45746266 \[-\]ER, ARProliferation, migration, invasionSTAT3F\>M in metabolic syndrome\[[@CR52], [@CR63], [@CR87], [@CR88]\]**hsa-miR-222**chrX: 45747015-45747124 \[-\]ER, ARProliferation, migration, invasionSTAT3F\<M in whole heart or isolated cardiomyocytes\[[@CR52], [@CR63], [@CR64], [@CR87]\]**hsa-miR-424**chrX: 134546614-134546711 \[-\]--Proliferation, invasion, apoptosisPD-L1F\>M in systemic lupus erythematosus\[[@CR57]\]**hsa-miR-513**chrX: 147213463-147213591 \[-\]--Proliferation, migration, apoptosisPD-L1--\[[@CR51]\]This table summarizes microRNA target genes and/or sex-hormonal regulation and, when described, their sex-biased (F\>M or F\<M) expression in different pathophysiological conditions. F\>M and F\<M indicate higher or lower expression in females respect to males*ER* estrogen receptor, *AR* androgen receptor, *PD-L1/2* programmed cell death protein ligand 1/2, *HIF-1* hypoxia-inducible factor-1, *STAT3* signal transducer and activation of transcription-3
PD-L1 transcription is induced by hypoxia-inducible factor-1α (HIFα) and signal transducer and activation of transcription-3 (STAT3) factors, directly acting on its promoter \[[@CR52]\]. These transcription factors are regulated by miR-221&222 and by miR-18 and miR-19, the former couple specifically localized on chromosome X and the latter encoded on both chromosome 13 and chromosome X. As also miR-20a/b and miR-106a/b are produced from different miR gene clusters, apparently derived from genetic duplications, it is important to discriminate the contribution of X-chromosome-encoded miRs, looking for their possible role as functional modulators of female immunity \[[@CR53], [@CR54]\]. Both miR-18 and -19 are able to potentiate the nuclear factor-κB (NF-κB) activity, in turn favoring inflammation \[[@CR53], [@CR55]\]. Likewise, miR-20b appears to target both HIF-1a and STAT3 transcription factors \[[@CR56]\], thus repressing PD-L1, either by direct pairing to its 3′UTR or indirectly through the downregulation of these two activating transcription factors. It is important to point out that, according to Targetscan, miR-20 and miR-106 are putative controllers also of PD-L2, in agreement with its predominant post-transcriptional regulation. Finally, another X-linked miR, miR-424, directly targets either PD-L1 or CD80, thus possibly regulating both the PD-L1/PD-1 and CD80/CTLA-4 pathways further supporting the option of X-linked disparities associated with immune checkpoint responses \[[@CR57]\].
How these X-linked miRs could provide a protection to female cells but not to male ones is still a matter of debate. A possibility that has been envisaged some years ago is that some of these X-linked miRs lie in the chromosomal regions that escape XCI and therefore are expressed at higher levels in female cells \[[@CR33]\]. A precise map of XCI-escaper miRs is still missing and this should be the objective of specific research. However, it has been observed that in patients with active lupus 18, X-linked miRs are more expressed in CD4+ T-cells from women with respect to men, and five of them are regulated by demethylation \[[@CR31]\]. This could suggest that these five miRs are XCI escapers, or, alternatively, that they could be demethylated by an XCI-escaper gene.
Among the X-linked miRs, miR-221&222 are the most extensively studied in tumors of different origins where they act as oncomirs controlling the development and progression of the tumor through the down-modulation of several key targets \[[@CR58]\]. However, their possible impact on sex differences detected in tumor incidence and progression is still neglected. Literature on this argument appears to be focused on the study of the impact of hormones on cancer. For instance, it has been suggested that melanoma could be classified among the hormone-sensitive tumors according to complex, overlapping actions played by estrogens and androgens, particularly by the opposite effects of α and β estrogen receptors (ER) \[[@CR59]\]. In particular, ERβ has been reported as capable of inducing autophagy-mediated cell death both in post-mitotic cells and proliferating cells, whereas ERα has been suggested to induce proliferation in transformed cells and autophagy in post-mitotic cells \[[@CR60], [@CR61]\]. Interestingly, among the number of X-linked miRs, six of them putatively bind and regulate ERα, including miR-221&222 \[[@CR62]\]. Specifically, miR-221&222 inhibit ERα mRNA translation by direct binding to its 3ʹUTR, being in turn repressed by ERα according to a negative feedback loop \[[@CR63]\].
Conversely, some insights are derived from studies carried out on miR-221&222 in cardiac cells. In fact, these two miRs display lower expression levels in females with respect to male murine cardiomyocytes contributing to sex-dimorphic cardiac phenotypes \[[@CR64]\]. The mechanism underlying this disparity appears to involve endothelial nitric oxide synthase (eNOS) modulation via the unblocked expression of miR-221&222 direct target Ets-1 \[[@CR65]\]. These data could be of relevance in view of the different cardiotoxicity of cancer chemotherapy in males and females \[[@CR66]\].
Programmed cell death and miRs {#Sec7}
------------------------------
Once again, cells from males and females seem to act differently when headed for death if subjected to the same exogenous stress, as female non-tumor cells are more prone to the autophagic protective effects, whereas male cells more frequently undergo apoptosis and/or necrosis \[[@CR7], [@CR8], [@CR11], [@CR12], [@CR67], [@CR68]\]. Looking for factors underlying this disparity, one option to consider is the involvement of the X-chromosome-linked inhibitor of apoptosis (XIAP) whose reduction can promote apoptosis and/or autophagy besides restoring sensitivity to chemotherapeutic drugs \[[@CR69]\]. A sex-related loop would possibly involve 17β-estradiol (E2) that via ERα activates the miR-23 family and p53, in turn decreasing XIAP and inducing apoptosis \[[@CR70]\]. Interestingly, miR-23a was reported to be differentially expressed by cells from male and female murine brains and XIAP was indicated as a mediator of sex-related responses after stroke \[[@CR71]\]. Notably, also miR-23c, belonging to the same miR family, is localized on chromosome X.
An additional Armadillo family member was recently identified as localized on chromosome X (Xq21.33-q22.2). This tumor-suppressor gene, called ALEX1 (Arm protein lost in epithelial cancer), was downregulated along with progression in several different solid tumors originating from epithelial tissues and its restored expression resulted able to inhibit proliferation and induce apoptosis \[[@CR72]\]. For this gene, a differential tissue-specific sex hormone regulation and a critical role of its chromosomal localization has recently been hypothesized \[[@CR73]\].
Finally, numerous miRs have been reported to regulate apoptosis, autophagy, and necrosis, also connecting the crosstalk between these types of cell death. Here again, some of these miRs are located on chromosome X. Examples are miR-374a, which is involved in the autophagic process through the inhibition of autophagy-related 5 (ATG5) and Ultraviolet Radiation resistance associated (UVRAG) proteins, and miR-504, which acts on the expression of the key tumor suppressor p53 \[[@CR74]\]. Of interest are the X-linked oncomir-221&222, as among their direct targets include a relevant number of proapoptotic proteins such as phosphatase and tensin homolog (PTEN), p53 upregulated modulator of apoptosis (PUMA), apoptotic protease activating factor-1 (APAF1), and caspase 3 (CASP-3) \[[@CR58], [@CR75], [@CR76]\], as well as the BCL2-like11 **(**Bcl2L11)-Bax/Bak axis \[[@CR77]\]. Additional targets of miR-221&222 are TP53INP1, PPP2R2A, and PGC-1a, recently described for their participation in cell death, either through apoptosis or autophagy \[[@CR78], [@CR79]\]. Last but not the least, beclin-1, a key player in autophagy, was demonstrated as a new target of miR-221 \[[@CR80]\] (see Fig. [1](#Fig1){ref-type="fig"}).
Conclusions {#Sec8}
===========
Human genome studies evidenced the presence on the X-chromosome of an unexpectedly high number of genes and miRs. This apparently non-casual localization might suggest the existence of X-linked functional circuitries, possibly contributing to sex-associated specificities, e.g., in immune responses. This fact, together with the presence of oncogenes apparently escaping XCI, could account at least in part for the sex disparity observed in several pathological settings. Furthermore, different possible mechanisms may account for the different sex-related expression levels of miRs, including X-linked transcription factors or cross-regulation by other miRs localized on X chromosome \[[@CR81], [@CR82]\].
A special case of miR-dependent, sex-specific regulation of immune responses and cancer immunosurveillance discussed in this review is that of the PD-1/PD-L1 pathway, whose targeting with monoclonal antibodies (mAbs) has given really impressive therapeutic results \[[@CR83], [@CR84]\]. As mentioned, it has been observed that several miRs, either involved in the programmed cell-death processes or specifically targeting PD-L1, are localized on chromosome X. Therefore, an miR-based modulation of such pathway could be at the basis of many sex disparities observed between men and women in terms of stronger immunological responses and immunosurveillance \[[@CR41], [@CR44]\].
In this perspective, the need for further gender-specific research emerges in order to fill the gap between clinical data and our knowledge on the mechanisms underlying the detected gender disparity in the onset and response to therapy of different forms of cancer.
Finally, as very recently suggested by an Editorial appeared in Nature Medicine "the failure to assess the influence of sex chromosomes in studies of the genome doesn't necessarily boil down to a lack of tools: there is also a challenge of a lack of will. It takes a bit more effort to include sex chromosomes in certain genomic analyses, and so this step is sometimes skipped" \[[@CR85]\]. We think the same stands for epigenetics, and time has come to deal with both issues in order to develop a real first-stage personalized approach to a number of life-threatening diseases.
Partially supported by the Italian Association for Cancer Research (IG 18526) to PM, (IG18815) to AC, and (IG11610) to LG and by Ministry of Health (RF-2011-02346986) to PM.
The authors declare that they have no conflict of interest.
| {
"pile_set_name": "PubMed Central"
} |
1. Introduction {#sec1}
===============
Heart failure (HF) is defined as the inability of the heart to pump sufficient blood to meet the body demands. Heart failure is a major public health problem especially in the aging population (≥65 years old), affecting nearly 5 million Americans and 15 million European people. HF is associated with high morbidity and reduced life expectancy, with 5-year mortality of newly diagnosed HF as high as 50% and 10-year survival of 26.7% \[[@B1], [@B2]\]. Systolic and diastolic dysfunctions are considered the most important underlying pathology of HF and it mainly occurred due to ischemic heart disease (coronary artery disease, myocardial infarction), uncontrolled hypertension, idiopathic dilated cardiomyopathy (IDCM), hypertrophic cardiomyopathy (HCM), chronic myocarditis, and valvular heart diseases \[[@B3], [@B4]\].
Effective management of HF depends on a correct and rapid diagnosis. Presently, BNP (brain natriuretic peptide) or N-terminal pro-brain natriuretic peptide (NT-proBNP) assay is generally accepted by the international community for diagnostic evaluation and risk stratification of patients with HF. However, regardless of its widespread clinical use, BNP is still encumbered by reduced specificity. As a result, diagnosis of HF remains challenging. Although significant improvement happened in the clinical management of HF over the last 2 decades, traditional treatments are ultimately ineffective in many patients who progress to advanced HF. Therefore, a novel diagnostic, prognostic biomarker and new therapeutic approach are required for clinical management of HF patients. Circulating miRNAs seem to be the right choice for novel noninvasive biomarkers as well as new treatment strategies for HF \[[@B1], [@B5]\].
Microribonucleic acids (miRNAs) are highly specific, endogenous, small (\~22 nucleotides), single-stranded, noncoding RNAs that regulate the rate of protein synthesis at the posttranscriptional level by binding to the 3′-untranslated region (3′-UTR) through altering the stability of the targeted mRNAs. Recently, it has been well established that miRNAs are critically regulating all the biological functions such as development, metabolism, cellular differentiation, proliferation, migration, secretion, excitation, conduction, aging, apoptosis, stem cell regulation, and immune function of the cell types relevant to the cardiovascular system such as endothelial cells and cardiac myocytes and conductive cells and smooth muscle and inflammatory cells and fibroblasts. Moreover, miRNAs are directly involved in many cardiovascular disease conditions \[[@B6], [@B7]\].
However, presence of miRNAs in body fluid is called circulating miRNAs. Circulating miRNAs have some useful characteristics as biomarkers, since they are highly constant and easily detectable in the peripheral circulation. Several research groups established that altered circulating miRNAs levels have been associated with different forms of heart disease, including ventricular septal defect (miR-155-5p, miR-222-3p, and miR-498) \[[@B8]\], atrial and ventricular arrhythmia (miR-1, miR-26, and miR-328) \[[@B9]--[@B13]\], hypertension (miR-296, miR-133b, and miR-625) \[[@B14]\], coronary artery disease (miR-133a, miR-208a, and miR-126) \[[@B15], [@B16]\], as well as, in our previous study, miR-149, miR-424, and miR-765 \[[@B17], [@B18]\], acute coronary syndrome (miR-1, miR-133a, and miR-208a) \[[@B19]\], acute myocardial infarction (miR-208b, miR-1, and miR-133) \[[@B20]--[@B23]\], and heart failure (miR-423-5p, miR-320a, and miR-22) \[[@B24], [@B25]\].
It has been well known that oxidative stress induced cardiac cell injury through expression changes of multiple genes plays a critical role in the pathogenesis of various types of heart diseases and there is a strong correlation between oxidative stress markers and HF. Several factors are associated with oxidative stress; among them increased reactive oxygen species (ROS) levels or decreased antioxidant defenses are responsible for disease progression \[[@B26], [@B27]\]. ROS is a collective term which includes hydrogen peroxide (H~2~O~2~), superoxide anion O~2~ ^−•^, and hydroxyl radicals (^•^OH). ROS can be produced within living cells by several potential sources such as mitochondria, plasma membrane bound NADPH oxidases (NOXs), endoplasmic reticulum (ER), and different enzymes involved in redox reactions such as xanthine oxidases (XOs), lipoxygenases, peroxidase, cytochromes, mono- and dioxygenases, and uncoupled nitric oxide synthases (NOSs) \[[@B28]--[@B31]\].
In experimental study, it has been established that mitochondrial DNA deletions contribute to the phenotype of systolic heart failure through increased mitochondrial ROS \[[@B32]\]. Congestive heart failure is related to intrinsic alterations of mitochondrial oxidative phosphorylation which lead to elevated myocardial cytosolic free ADP. ATP sensitive K^+^ (KATP) channels act as metabolic sensors that are essential for maintaining coronary blood flow and for mediating the response of the myocardium to oxidative stress. As a result, in the failing heart the balance between myocardial ATP demands and oxygen supply is significantly dependent on functioning KATP channels. However, KATP channels\' function is critically regulated by miR-9a-3p \[[@B33], [@B34]\].
A recent study demonstrated that miRNAs exist in the heart mitochondria which significantly regulates mitochondrial gene expression and prevents ROS induced cardiomyocyte injury \[[@B35]\]. Nicotinamide adenine dinucleotide phosphate oxidase (NOX) proteins produce ROS involved in redox signaling pathway and directly linked with development of HF. Accumulating evidence supports mainly beneficial effects of Nox4 in the cardiovascular system. Overexpression of Nox4 reduced angiotensin II-induced high blood pressure, protects endothelial dysfunction and atherosclerosis, also regulating oxidative stress and apoptosis in cardiac myocytes, and prevents the development of HF \[[@B36]--[@B39]\]. However, miR-188, miR-146a, and miR-25 significantly regulated the expression of NOX4 and protected oxidative/nitrative stress induced endothelial and myocardial dysfunction \[[@B40]--[@B42]\].
In this review, we mainly focus on more recent clinical data regarding circulating miRNAs and their potential value in diagnosis, prognosis, and therapeutic targets for heart failure patients. In addition, we also discuss our own technique of extraction of RNA and detection of circulating miRNAs from human plasma and oxidative stress associated miRNAs with HF.
2. Discovery of MicroRNA to Circulating MicroRNA {#sec2}
================================================
MicroRNA was first discovered in 1993 by Lee et al. \[[@B43]\] and Wightman et al. \[[@B44]\] from the nematode*Caenorhabditis elegans* and 7 years later it was identified in humans. In 2002, Calin et al. firstly recognized the pathological role of miR-15 and miR-16 in the development of chronic B-cell leukemia \[[@B45]\] ([Figure 1](#fig1){ref-type="fig"}). In 2005, Kwon et al. firstly established the role of miR-1 in cardiac development in*Drosophila* and it is critically regulated by the Notch signaling pathway (Notch 1 receptor) \[[@B46]\]. Just one year later, Van Rooij et al. through northern blot analysis from cardiac tissue demonstrated that several stress-responsive miRNAs were significantly dysregulated in cardiac hypertrophy and HF in both mice and humans. Moreover, they also reported that in particular miR-195 was remarkably upregulated in idiopathic end-stage failing human hearts and plays an important role in cardiac remodeling in transgenic mice \[[@B47]\].
According to Mitchell et al., in 2008, the presence of miRNA was detected in plasma and serum of prostate cancer patients and healthy subjects. They also reported that circulating miRNAs are remarkably stable to incubation at room temperature for up to 24 h or to subjecting them to up to eight cycles of freeze-thawing and are highly protected from endogenous RNase activity.
Very importantly, when synthetic miRNAs*(cel-miR-39, cel-miR-54, and cel-miR-238)* were added to plasma prior to RNase denaturing solution, they were rapidly degraded (\<2 min)---suggesting that exogenous miRNAs are not stable to RNase activity, but endogenous plasma miRNAs are resistant to RNase activity because they are normally attached to microvesicles or exosomes and also form protein-miRNA complexes in extracellular fluids \[[@B48]\]. Several other studies also established that circulating miRNAs remain stable after being subjected to severe conditions that would usually degrade most RNAs, such as boiling, very low, or high pH levels. Besides, recent studies have evaluated that miRNAs are preserved in archived 10-year-old human serum samples \[[@B49], [@B50]\].
More recently, miRNAs have been easily detected in a wide range of bodily fluids, including urine, saliva, tears, seminal fluid, cerebrospinal fluid, pleural fluid, peritoneal fluid, amniotic fluid, and breast milk. Therefore, the discovery of miRNAs in body fluids opens up the possibility of using them as noninvasive diagnostic and prognostic biomarkers in cardiovascular diseases including HF \[[@B51], [@B52]\]. Though last two decades huge research happened in microRNAs, the origins of circulating miRNAs are largely unknown. Some studies have proposed that they are secreted in membrane-bounded-vesicles (apoptotic bodies, microvesicles, and exosomes), while others mentioned that they are secreted in vesicle-free medium but associated with protein-miRNA complexes (AGO2, NPM1, and HDL). It has also been hypothesized that circulating miRNAs are released from various healthy or disease tissues such as heart, lung, liver, kidney, and brain. According to miRBase data line (<http://www.mirbase.org/>), up to date 2588 mature miRNAs were exploded in human; among them more than 200 miRNAs were identified as a circulating miRNAs \[[@B4], [@B5]\].
3. Biogenesis of Circulating miRNAs {#sec3}
===================================
In the nucleus, microRNA genes are transcribed by RNA polymerase II into many hundreds or thousands of nucleotides in length called primary miRNAs (pri-miRNAs), which are folded into characteristic hairpin structures. Then, pri-miRNAs are cleaved by Drosha, RNase III enzyme, in association with DGCR8, RNA-binding protein, to form smaller, 70- to 100-nucleotide-long, preliminary miRNAs (pre-miRNAs). Subsequently, pre-miRNAs are actively transported to the cytoplasm by the Ran-GTP dependent transporter in association with exportin-5 \[[@B7], [@B53]\].
In the cytoplasm, the pre-miRNAs are cleaved by Dicer (another RNase III enzyme) and processed into double-stranded, immature miRNA (miRNA:miRNA^*∗*^) duplexes which are approximately 22 nucleotides in length. Afterward, the duplex is unwound and assembled into the RNA-induced silencing complex (RISC) by associating with argonaute proteins; the mature miRNA negatively regulates gene expression through translational repression or mRNA degradation, according to the sequence complementarity between the miRNA and the target mRNA. Moreover, miRNA and mRNA interaction is thought to be an ideal Watson-Crick base-pairing of nucleotides 2 to 8 (seed region) at the 5′-end of the miRNA, whereas nucleation sites usually present on the 3′-untranslated region (3′-UTR) of the targeted mRNA. Pre-miRNAs can be attached to MVBs or other RNA-binding proteins (NPM1, AGO2, or HDL); after fusion with the plasma membrane, circulating miRNAs enter the bloodstream. MiRNAs can also be released by cells through exosomes, apoptotic bodies, and other vesicular-like bodies \[[@B7], [@B51], [@B53]\]. More details are in [Figure 2](#fig2){ref-type="fig"}.
4. Extraction of RNA and Detection of Circulating miRNAs from Human Plasma {#sec4}
==========================================================================
In our previous experiments \[[@B17], [@B18], [@B54]\], we enrolled 300 human subjects with the acute myocardial infarction (AMI), coronary artery disease (CAD), and AMI with HF and healthy controls and we used the following techniques and achieved the best result.
4.1. RNA Isolation {#sec4.1}
------------------
RNA isolation is as follows:Total RNA was isolated from human plasma and mice plasma by using a phenol-based TRIZOL reagent. In short, 250 *μ*L of plasma was mixed briefly with 750 *μ*L of TRIZOl, incubated for 5 min at room temperature (RT), and then mixed with 200 *μ*L chloroform, incubated for 3 min at RT.The aqueous phase, interphase, and organic phase were separated by centrifugation at 12,000 rmp for 15 min at 4°C.The upper aqueous phase was collected and subsequently mixed with 500 *μ*L of 100% isopropanol and incubated at −20°C for 16 h and after that centrifuged at 13000 rmp for 15 min at 4°C for precipitation.RNA samples were washed 3 times with 500 *μ*L of 75% ethanol and centrifuged again at 7500 rmp for 10 min at 4°C.Finally, supernatants were eliminated and dried for 5 min. Then, RNA samples were dissolved in 30 *μ*L of RNase-free (DEPC) water and incubated overnight at 4°C.Afterward, the concentration and purity of RNA were determined spectrophotometrically and stored at −80°C for future use ([Figure 3](#fig3){ref-type="fig"}).
4.2. Detection of Circulating miRNAs {#sec4.2}
------------------------------------
Real-time quantitative reverse-transcription PCR analysis was carried out to examine the expression of specific miRNAs in plasma. A total 4 *μ*L of pure RNA (OD, 1.8--2.2; nucleic acid concentration, 50--500 *μ*g) was reverse-transcribed to cDNA at 42°C for 30 minutes. Subsequently, 2 *μ*L of cDNA was used as the template in real-time quantitative PCR reaction. MiR-156a and U6 were used as the normalization control. The Ct values from qRT-PCR assays between 15 and 35 were considered to be expressed.
To decrease the possible errors resulting from qRT-PCR assays, in our previous study we used three potential endogenous control miRNAs (U6, cel-miR-39, and miR-156a); among them we had chosen miR-156a as a standard inner control. Depending on our own experience, the use of synthetic mimic miR-156 instead of commonly used cel-miR-39 during RNA extraction from plasma of AMI or CAD patients obtained better quality of total RNA (OD ratio: 1.8--2.2, nucleic acid concentration: 50--500 *μ*g). Furthermore, with the use of endogenous miR-156 instead of commonly used U6 for endogenous control during real-time PCR, we achieved more reliable results (Ct value: 18--25) \[[@B17], [@B18], [@B54]\].
5. Circulating miRNAs as a Potential Biomarker for HF Patients {#sec5}
==============================================================
Recently, several research groups suggested that circulating miRNAs might be useful as stable blood-based biomarkers in HF patients. Tijsen et al. found that 6 circulating miRNAs (miR-423-5p, miR-18b, miR-129-5p, miR-1254, miR-675, and miR-622) were upregulated in patients with HF, among which miR-423-5p was most significantly related to the clinical diagnosis of HF. Moreover, the predictive power of miR-423-5p was also high within the dyspnea population, upon comparing HF and non-HF cases. Furthermore, circulating miR-423-5p was significantly upregulated in systolic HF patients and rat HF model and closely associated with N-terminal pro-brain natriuretic peptide (NT-proBNP) and ejection fraction (EF) \[[@B24], [@B55]\].
Cardiac myocyte-associated plasma miR-499 was highly elevated in patients with acute heart failure (AHF) as compared with control subjects, whereas no significant changes were observed in diastolic dysfunction and it was not affected by a wide range of clinical parameters, including age, sex, body mass index, kidney function, systolic blood pressure, and white blood cell count. Furthermore, plasma level of the liver-specific miR-122 was not only increased in hepatic injury but also significantly increased in AHF patients, possibly reflecting hepatic venous congestion \[[@B20]\]. Goren et al. demonstrated that serum levels of miR-320a, miR-22, and miR-92b were significantly elevated in systolic HF patients and correlated with important clinical parameters such as elevated serum NT-proBNP levels, a wide QRS, and dilatation of the left ventricle and left atrium \[[@B24]\].
The plasma miRNA expression levels of chronic HF patients were compared to healthy subjects using a microRNA array, and it was noticed that plasma miR-361-5p levels were significantly decreased in all individuals with chronic HF. This result suggested plasma miR-361-5p as a potential, novel biomarker for chronic HF patients \[[@B56]\].
A clinical study examined the miRNA expressions in peripheral blood mononuclear cells (PBMCs) from chronic HF patients and found that miR-107, miR-139, and miR-142-5p were markedly downregulated in both non-ischemic dilated cardiomyopathy (NIDCM) and ischemic cardiomyopathy (ICM) patients compared to healthy subjects. Furthermore, miR-142-3p and miR-29b levels were highly upregulated in NIDCM while miR-125b and miR-497 levels were downregulated in ICM patients. Therefore, miRNAs in peripheral blood mononuclear cells may be used as biomarkers for diagnosis of chronic HF patients \[[@B57]\].
Circulating miR-155 levels were significantly decreased in patients with chronic HF and strongly associated with ventricular arrhythmia and angiotensin receptor type 1 single nucleotide polymorphism 1166A/C \[[@B58]\]. It has been demonstrated that muscle-specific miR-133 expression is associated with signs of HF in patients undergoing coronary artery bypass surgery. According to New York Heart Association (NYHA) functional classification of HF, miR-133 expression decreased significantly with increased severity of HF. Additionally, patients with NT-proBNP levels \>1,800 pg/mL showed a 25% decrease in miR-133 expression compared to patients with levels \<300 pg/mL, indicating its potential as a biomarker for HF \[[@B59]\].
A comprehensive microRNA (miRNA) and messenger RNA (mRNA) analysis was performed on plasma and myocardial specimens from HF patients, chronic obstructive pulmonary disease (COPD) patients, and healthy controls and found that miR-103, miR-142-3p, miR-30b, and miR-342-3p levels were differentially expressed between HF and controls and COPD and other breathless patients. Though individually NT-proBNP was most significant in predicting HF and exhibited greater sensitivity and specificity, combining microRNA levels with NT-proBNP may add diagnostic value. Furthermore, the expressions of circulating microRNAs (miR-30b, miR-103, miR-199a-3p, miR-23a, miR-27b, miR-324-5p, miR-342-3p, and miR-142-3p) were significantly altered in end-stage HF patients. Moreover, although the miRNAs are more sensitive than mRNAs for diagnosis of heart failure patients, combined miRNA and mRNA profiling may have better value for diagnosis and prognosis of end-stage cardiomyopathy patients \[[@B60], [@B61]\].
The plasma level of miR-126 was remarkably decreased in congestive heart failure (CHF) patients and negatively correlated with age, logBNP, and NYHA class and could be a useful biomarker for CHF \[[@B62]\]. The serum expression levels of miR-210 and miR-30a were significantly elevated in CHF patients compared to healthy subjects and correlated with NT-pro-BNP. In addition, these miRNA levels were higher in patients with EF \>40% than in those with EF \<40%, which might serve as a potential biomarker for HF \[[@B63]\].
Recent study identified that three cardiac fibroblast-derived miRNAs (miR-660-3p, miR-665, and miR-1285-3p) were found obviously elevated in heart and plasma during CHF and significantly correlated with left ventricular ejection fraction (LVEF), holding promise as potential diagnostic biomarker for CHF. These three circulating miRNAs may be used as potential biomarkers for diagnosis of CHF patients \[[@B64]\].
Serum level of miR-1 was markedly downregulated in patients with symptomatic HF and its expression reduced with severity of NYHA class, as well as being negatively correlated with NT-proBNP concentration in patients in NYHA class II/III. Upregulation of miR-21 was found in all patients, independent of HF severity, and significantly correlated with galectin-3 concentration. As a result, alteration of miR-1 and miR-21 expression may be essential for the development of HF; miR-1 might become a biomarker for diagnosis of HF \[[@B65]\]. The expression of baseline plasma miR-30d level is significantly related to response to cardiac resynchronization therapy (CRT) in HF patients with dyssynchrony (HFDYS). Upregulation of miR-30d in cultured cardiomyocytes is highly correlated with areas of increased wall stress in HFDYS and led to cardiomyocyte growth and protected against apoptosis by targeting the mitogen-associated kinase 4. In addition, miR-30d plasma level is negatively correlated with high-sensitivity troponin T in HFDYS patients. Therefore, plasma miR-30 is an essential biomarker for HF patients \[[@B66]\].
Very interestingly, myocardial biopsy specimens collected from Chinese patients presenting with recent HF were compared with a group of patients without HF undergoing routine cardiac surgery and it was found that miR-1, miR-21, miR-23, miR-29, miR-130, miR-195, and miR-199 expressions were evidently increased in the HF group as compared to those without HF. Furthermore, related mRNAs (casp3, coll I, coll III, and TGF) were also significantly upregulated in the HF group. These miRNAs may be used as an early diagnostic biomarker for CHF patients and possibly future therapeutic targets \[[@B67]\].
One study reported that the expressions of serum miR-21, miR-378, and miR-940 levels were considerably upregulated in response to an acute exhaustive exercise in CHF patients while the rest were not changed. However, no robust correlation was recognized between changes of these miRNAs and exercise capacity, muscle damage, or inflammation. Therefore, circulating miRNAs may be used as a diagnostic or prognostic biomarker of exercise adaptation in CHF patients \[[@B68]\]. More recently, it has been demonstrated that altered circulating miR-199a-3p levels are the best predictor of early worsening renal function (WRF) in AHF patients \[[@B69]\].
The expression of miR-214 was significantly increased in the serum of CHF patients, as well as in hypertrophic and failing hearts of humans and mice. On the contrary, upregulation of miR-214 markedly reduced angiogenesis of human umbilical vein endothelial cells (HUVECs) by targeting XBP1. Therefore, miR-214 plays an important role in the control/inhibition of cardiac angiogenesis which is associated with HF \[[@B70]\]. Plasma miRNAs (miR-16, miR-20b, miR-93, miR-106b, and miR-223) levels were most robustly changed in rats with hypertension-induced HF and also highly correlated with circulating BNP. These plasma microRNAs could potentially serve as biomarkers of therapeutic efficacy and disease progression in hypertension-induced HF \[[@B71]\].
Nair et al. demonstrated miRNA screening in isolated diastolic dysfunction with preserved systolic function to discover promising candidate miRNAs. They found that circulating miRNAs (miR-454, miR-500, miR-1246, and miR-142-3p) expressions levels were significantly altered in diastolic dysfunction patients. In addition, circulating miR-454 and miR-500 were decreased, while miR-1246 was increased in diastolic dysfunction. Furthermore, circulating miR-142-3p level was significantly reduced while miR-124-5p level was markedly elevated in stable compensated dilated cardiomyopathy patients. Therefore, these circulating miRNAs may serve as new diagnostic biomarkers and be suggested as therapeutic targets for drug discovery of HF patients \[[@B72]\].
Very recently, a translational perspective study was carried out on three patient cohorts, no heart failure (no-HF), HF with reduced ejection fraction (HFrEF), and HF with preserved ejection fraction (HFpEF), using Taqman miRNA arrays. The serum levels of miR-30c, miR-146a, miR-221, miR-328, and miR-375 were differentially expressed in HFpEF and HFrEF, and their expression levels were also different between HF and no-HF. It was summarized that combinations of two or more miRNAs with BNP are essential biomarkers for diagnosis of HF and also helpful in the differentiation of HFpEF from HFrEF compared with using BNP alone \[[@B1]\]. In a study group, pericardial fluid (PF) was collected from HF patients during open-heart surgery and more than 2 hundred microRNAs were detected in PF. But 5 miRNAs (miR-21-5p, miR-451a, miR-125b-5p, let-7b-5p, and miR-16-5p) were highly expressed in PF. However, the overall miRNA expressions of the cardiac patients were nearly similar despite the differences in disease aetiologies and HF stages. They concluded that miRNAs may act as endocrine and paracrine signaling factors by mediating the local crosstalk between cardiac cells \[[@B73]\].
The incidence of ischemic HF in post-AMI patients is increasing. The serum level of miR-192 was markedly elevated in AMI patients with development of ischemic HF. Moreover, miR-194 and miR-34a expression levels were not only upregulated but also significantly correlated with left ventricular end-diastolic dimension 1 year after AMI. Therefore, circulating p53-responsive miRNAs (miR-192, miR-194, and miR-34a) are useful predictive indicators of heart failure after AMI \[[@B74]\]. A prospective, nonrandomized study was performed in 81 patients with symptomatic (NYHA functional class III or IV) HF eligible for CRT. At 12-month follow-up, 55 patients (68%) were considered responders and 26 were considered nonresponders to CRT (32%). Among responders, five circulating miRNAs (miR-26b-5p, miR-145-5p, miR-92a-3p, miR-30e-5p, and miR-29a-3p) were significantly increased as compared with nonresponders, and these miRNAs inversely correlated with NT-proBNP and directly correlated with EF. Besides, in responders, reverse remodeling is associated with favourable changes in miRNAs that regulate cardiac fibrosis, apoptosis, and hypertrophy. This study indicated that serum miRNAs may be used as a potential prognostic biomarker and a new therapeutic target for HF \[[@B75]\].
In a large cohort, clinical study investigated circulating miRNAs expression patterns in patients with stable and advanced HF before and at different time points after a left ventricular assist device (LVAD) implantation. Circulating miR-1908 was 2.0- and 2.1-fold and miR-1180 was 4.5- and 4.0-fold upregulated in patients with advanced HF and in patients with stable HF compared with healthy controls (NFs), respectively. Moreover, the cardiac-specific circulating miR-208b, miR-208a, and miR-499 and the muscle-specific miR-1-1 and miR-133b were 143-fold, 78-fold, 28-fold, 18-fold, and 21-fold higher in advanced HF at LVAD implantation compared with NFs. Surprisingly, the miRNA changes in advanced HF reversed 3 and 6 months after LVAD support. It was suggested that circulating miRNAs serve as excellent diagnostic and prognostic biomarkers and also have therapeutic potential for HF patients \[[@B3]\].
Morley-Smith et al. retrospectively obtained 53 serial plasma and 20 ventricular myocardial samples from 19 patients with severe advanced heart failure who underwent HeartMate II LVAD implantation. By using microarray assay, they found that expressions of circulating miR-483-3p levels were markedly sustained upregulated with LVAD support, with median fold changes from baseline of 2.17, 2.27, 1.87, and 2.82 at 3, 6, 9, and 12 months, respectively, while baseline plasma miR-1202 identified good versus poor LVAD responders. In addition, NT-proBNP levels were inversely correlated with duration of LVAD support. As they suggest, these two circulating miRNAs are novel prognostic biomarkers for HF patients \[[@B76]\] ([Table 1](#tab1){ref-type="table"}).
6. Therapeutic Role of miRNAs in HF {#sec6}
===================================
Acute and chronic stress to the heart results in a pathological remodeling response accompanied by cardiomyocyte hypertrophy, fibrosis, pump failure, and myocyte degeneration and apoptosis, which leads to HF and sudden cardiac death. Therapeutic inhibition of cardiac-specific miR-208a via subcutaneous delivery of anti-miR-208a during hypertension-induced HF in Dahl hypertensive rats dose-dependently prevents pathological cardiac remodeling and improves cardiac function with survival rate \[[@B77]\]. Intravenous delivery of locked nucleic acid- (LNA-) modified anti-miR-15 dose-dependently reduces infarct size, cardiac fibrosis, and pathological cardiac remodeling and enhances cardiac function in both mice and pigs \[[@B78]\].
The level of miR-1 was downregulated in a chronic rat HF model and its expression was restored to normal levels during reverse remodeling by sarcoplasmic reticulum calcium ATPase 2a (SERCA2a) gene therapy through an Akt/FoxO3A-dependent pathway, which was also associated with normalized sodium-calcium exchanger 1 (NCX1, functional target of miR-1) expression and significantly improved cardiac function \[[@B79]\]. Left ventricular assist devices (LVADs) are being used in patients with HF. Plasma *α*-1-antichymotrypsin (ACT) levels were upregulated in HF patients as compared with healthy subjects and normalized by 6 months of LVAD support. MiR-137 directly targeted ACT, thereby indicating that ACT and miR-137 play an important role in the pathophysiology of HF and reverse remodeling during mechanical support in chronic HF \[[@B80]\]. Trimetazidine (TMZ) improves right ventricular (RV) function and decreases apoptosis and fibrosis in RV myocardial cells (RVMCs) by increasing miR-21 expression in vitro and in vivo \[[@B81]\]. Circulating levels of miR-16, miR-20b, miR-93, miR-106b, miR-223, and miR-423-5p were significantly upregulated in response to hypertension-induced HF, whereas this effect was blunted in response to treatment with anti-miR-208a as well as an ACE inhibitor and prevented progression of hypertension-induced heart HF \[[@B63]\].
The miR-24 is enriched in cardiac endothelial cells and considerably increased after cardiac ischemia. Besides, miR-24 significantly induced endothelial cell apoptosis and markedly impaired angiogenesis. Inhibition of endothelial miR-24 significantly reduced myocardial infarct size via prevention of endothelial apoptosis and enhancement of vascularity after AMI through targeting of the endothelium-enriched transcription factor GATA2 and the p21-activated kinase 4 (PAK4), which led to preserved cardiac function and was closely related to HF \[[@B82]\].
In a rat aortic stenosis model and end-stage HF patients, the expressions of miR-24 were significantly upregulated in failing cardiomyocytes, which is a suppressor of JP2 expression. Junctophilin-2 (JP2) connects the sarcoplasmic reticulum (SR) to the cell membrane, including T-tubules (TT), forming structural units for excitation-contraction (E-C) coupling in cardiomyocytes. Bioinformatic analysis predicted two prospective binding sites of miR-24 in the 3′-untranslated regions of JP2 mRNA. Therefore, miR-24 and JP2 have a strong relationship between the upstream hypertrophy/HF signals and defective E-C coupling, and suggests a new therapeutic option for the treatment of HF \[[@B83], [@B84]\].
Furthermore, yes-associated protein (YAP) promotes cardiomyocyte growth in postnatal hearts. The miR-206 regulates YAP-induced cardiac hypertrophy and survival during ischemia/reperfusion injury by silencing Forkhead box protein P1, which is related to HF \[[@B85]\]. Recently, it has been demonstrated that circulating miR-340 was significantly upregulated in failing human hearts because of dilated cardiomyopathy. On the contrary, knockdown of miR-340 using antagomir remarkably attenuated cardiac eccentric hypertrophy and HF via target gene dystrophin (DMD) \[[@B86]\].
The expressions of miR-212 and miR-132 were significantly upregulated in cardiomyocyte by hypertrophic stimuli through hyperactivation of prohypertrophic calcineurin/NFAT signaling pathway and an impaired autophagic response. However, pharmacological inhibition of miR-132 by intravenous antagomir injection rescues cardiac hypertrophy and pressure-overload-induced HF in mice, offering a novel therapeutic approach for cardiac failure \[[@B87]\].
Six members of the miR-17-92 cluster are strongly associated with aging, among which miR-18a, miR-19a, and miR-19b were significantly downregulated in aged cardiomyocytes and hearts of old failure-prone mice and also directly linked with elderly HF patients. In addition, miR-18 and miR-19 regulated their targets\' connective tissue growth factor (CTGF) and thrombospondin-1 (TSP-1) expression, respectively, and prevent age-related remodeling in the heart. It was suggested that these miRNAs may be used as potential new therapeutic targets for the modulation of aging-induced cardiac remodeling in geriatric patients \[[@B88]\].
MicroRNAs profiling and differential expressions analysis were performed in transplanted hearts of HF patients with LVAD implantation and found that miR-338-3p, miR-142-5p and miR-142-3p, miR-216a-5p, miR-223-3p, miR-27a-5p, and miR-378g were significantly correlated with off-pump cardiac index values. It was suggested that these miRNAs might contribute to molecular regulation of reverse remodeling and heart recovery mechanisms in HF patients \[[@B89]\]. Therefore, circulating microRNAs offer a promising novel diagnostic and prognostic biomarker as well as a new therapeutic target for HF patients.
7. Oxidative Stress Associated miRNAs with Heart Failure {#sec7}
========================================================
Oxidative stress plays a significant role in cardiovascular diseases, such as hypertension, cardiac hypertrophy, atherosclerosis, AMI, and HF. Oxidative stress represents a persistent imbalance between the production and the compensation of ROS. Excessive accumulation of ROS can damage proteins, lipids, and DNA of a cell which leads to cellular apoptosis and cell death. Studies have shown that plasma miR-1 level is significantly increased in HF and critically involved in the progression of HF through cardiomyocyte apoptosis and cell death. Administration of anti-miR-1 significantly decreased ROS and oxidative stress susceptibility through SOD1, Gclc, and G6PD and markedly increased heart function in mice model \[[@B90]\]. MiR-181c significantly improved heart function by suppression of ROS production through targeting the 3′-end of mt-COX-1 (cytochrome c oxidase subunit 1) in HF mice \[[@B35]\].
MiR-25 was markedly elevated in response to oxidative stimulation in cardiomyocytes. However, overexpression of miR-25 protected cardiomyocytes against oxidative damage by downregulating mitochondrial calcium uniporter (MCU) and enhanced cardiomyocyte activity \[[@B92]\]. Moreover, miR-21 protects against the oxidative stress induced injury on cardiac myocytes through its target gene programed cell death 4 (PDCD4) \[[@B26]\]. Furthermore, miR-499 protects cardiomyocytes from oxidative stress induced apoptosis via its effects on Pdcd4, Pacs2, and SOX6 \[[@B94], [@B95]\]. In addition, the expression of miR-200c is significantly elevated by oxidative stress and markedly inhibits endothelial cell growth and increased cell apoptosis and senescence through downregulation of the zinc finger e-box binding homeobox 1 (ZEB1) and endothelium-dependent relaxations (EDRs), while suppression of miR-200c by anti-miR-200c enhanced cell growth and significantly decreased cell apoptosis. Furthermore, holocytochrome c synthetase (HCCS), caspase-3, cyclooxygenase-2 (COX-2), and Slc25a3 protein expressions are essentially controlled by miR-200c and enormously reduced cardiomyocyte death \[[@B29], [@B96]--[@B99]\]. The expression of miR-144 was significantly downregulated in human failing hearts. However, upregulation of the miR-144 to protect against oxidative stress induced cardiomyocyte death via targeting the CUG triplet repeat-binding protein 2- (CUGBP2-) COX-2 signaling pathway \[[@B100], [@B101]\].
It has been mentioned that upregulated miR-466h inhibits antiapoptotic genes and induces apoptosis. In contrast, the inhibition of the miR-466h increased the expression levels of bcl2l2, dad1, birc6, stat5a, and smo genes and resulted in increased cell viability and significantly decreased caspase-3/7 activity \[[@B102]\]. Moreover, miR-466h-5p is a member of the miR-297-669 cluster located in intron 10 of Sfmbt2 gene on chromosome 2 and has an important proapoptotic role. Further study showed that the time-dependant activation of miR-466h-5p, miR-669c, and the Sfmbt2 gene followed the inhibition of histone deacetylation caused by glucose deprivation-induced oxidative stress \[[@B103]\]. Circulating miR-17-5p level was considerably altered in HF patients and its level was significantly upregulated in myocardium under oxidative stress. Overexpression of miR-17-5p provoked cardiomyocyte injury with decreased cell viability and enhanced apoptotic cell death induced by H~2~O~2~, while inhibition of miR-17-5p by its anti-miR-17-5p markedly reduced the infarct area and apoptosis and significantly increased cell viability through targeting Stat3 \[[@B73], [@B104], [@B105]\].
8. Limitations and Future Directions {#sec8}
====================================
Up to now, most of the studies evaluating that circulating miRNAs act as a biomarker for HF patients are single-center study using relatively small samples, which results in many divergences between different reports. Therefore, it is necessary to perform multicenter large-scale clinical studies to confirm the potential value of circulating miRNAs as noninvasive diagnostic or prognostic biomarkers for HF patients. However, measurement of circulating miRNAs requires qRT-PCR, which is expensive and time-consuming. Therefore, less expensive and newer techniques to detect circulating miRNA levels more rapidly can be expected in the near future.
So far, most of the studies focus on the role of circulating miRNAs as biomarkers for HF. The information regarding the molecular mechanism of circulating miRNAs in HF patients is largely unknown. Consequently, it is also essential to elucidate the mechanisms behind the change of circulating miRNAs before application in clinical practice. MicroRNAs are rapidly becoming an exciting pharmacological target in the treatment of cardiovascular disease including HF. However, there are several limitations overcome to promote miRNAs as a possible therapeutic target. Firstly, miRNAs often have hundreds or even thousands of predicted mRNA targets, but, at physiological expression levels, miRNA most likely targets only a little fraction of them. Secondly, because systemic delivery of antagomirs and miRNA mimics affects miRNA expression, generally, effects observed in the heart could also be secondary to effects in other tissues, for example, as a result of changed blood pressure or alteration in the level of circulating hormones. Finally, the most valuable model to study the function of miRNA is genetic deletion, which should exclusively derepress only those mRNAs that are physiologically repressed by the miRNA. Therefore, it is necessary to discover appropriate route of administration with fewer side effects and miRNA biology must be examined through the use of a combination of genetic animal models and pharmacological manipulation as well as making sure of the safeness of anti-miR or mimic before using it in clinical practice.
9. Conclusion {#sec9}
=============
In summary, we suggested that circulating miRNAs may have a potential value for the management of HF patients.
This work was supported by (1) National Natural Science Foundation of China (Grants no. 81370230 and no. 81570279 to Zhu Ping), (2) Natural Science Foundation of Guangdong Province, China (Grant no. 2014A030311041 to Zhu Ping), (3) Science and Technology Program of Guangzhou, China (Grant no. 201508020107 to Zhu Ping), and the social development Project of Zhen Jiang (SH2014028 to Baohai Zhang).
Competing Interests
===================
All authors declare that there is no conflict of interests.
Authors\' Contributions
=======================
Md Sayed Ali Sheikh designed the paper and wrote it. Umme Salma drawn the picture, Baohai Zhang collected the data, and Zhu Ping revised the paper.
{#fig1}
{#fig2}
{#fig3}
MicroRNAs Regulation Sources Association Methods Study population Number of samples Potential value Reference
---------------------------------------------------------------- ------------ ------------------ ------------- ----------------------- ------------------ ------------------- ----------------------- --------------------
miR-208b, miR-208a, and miR-499 Up Serum and plasma NT-proBNP A miRNA array/qRT-PCR Human *n* = 68 Diagnostic/prognostic \[[@B3], [@B20]\]
miR-423-5p Up Plasma NT-proBNP A miRNA array/qRT-PCR Human *n* = 113 Diagnostic \[[@B24], [@B55]\]
miR-133 Down Plasma NT-proBNP qRT-PCR Human *n* = 83 Diagnostic \[[@B59]\]
miR-103, miR-142-3p, miR-30b, and miR-342-3p Down Plasma NT-proBNP A miRNA array/qRT-PCR Human *n* = 150 Diagnostic \[[@B60]\]
miR-126 Down Plasma NT-proBNP qRT-PCR Human *n* = 60 Diagnostic \[[@B62]\]
miR-210 and miR-30a Up Serum NT-proBNP qRT-PCR Human *n* = 40 Diagnostic \[[@B63]\]
miR-192 Up Prognostic \[[@B74]\]
miR-26b-5p, miR-145-5p, miR-92a-3p, miR-30e-5p, and miR-29a-3p Up Plasma NT-proBNP A miRNA array/qRT-PCR Human *n* = 96 Prognostic \[[@B75]\]
miR-483-3p Up Plasma NT-proBNP Microarray assay Human *n* = 92 Prognostic \[[@B76]\]
[^1]: Academic Editor: Zhao Zhong Chong
| {
"pile_set_name": "PubMed Central"
} |
Introduction
============
No disease-modifying treatments are currently available for the treatment of osteoarthritis (OA) and affected end-stage joints are often surgically replaced ([@ref-14]). OA was once viewed as mechanical wear-and-tear of the cartilage, but in recent years there is increasing evidence that OA is a complex disease affecting the whole joint ([@ref-19]). Multiple proteins might interfere with OA pathology, including tumor necrosis factor-alpha (TNFα)-stimulated protein 6 (TSG-6) ([@ref-31]).
TSG-6, encoded by the TNFα-induced protein 6 (TNFAIP6) gene, is glycoprotein expressed by different cell types after inflammatory stimulation ([@ref-21]). While the protein is not detectable in healthy joint tissue, TSG-6 is present in the synovium, cartilage, and blood vessel walls of both rheumatoid arthritis (RA) and OA patients ([@ref-1]). In OA patients, the TSG-6 activity is associated with disease progression ([@ref-30]). TSG-6 protein consists of a Link and a CUB domain that can bind multiple proteins and extracellular matrix molecules ([@ref-22]) and has protective effects in experimental arthritis models. Transgenic mice with cartilage-specific constitutive overexpression of TSG-6 show chondroprotective, but not anti-inflammatory effects of TSG-6 in the antigen-induced arthritis model ([@ref-11]). In addition, recombinant TSG-6 also showed anti-inflammatory effects in collagen-induced arthritis ([@ref-23]). Correspondingly, TSG-6 deficient mice suffered increased neutrophil influx and increased arthritis severity after induction of proteoglycan-induced arthritis ([@ref-28]). Moreover, the therapeutic anti-inflammatory effects of mesenchymal stromal cells in a murine model for myocardial infarction were shown to be TSG-6-dependent ([@ref-15]).
Several mechanisms have been proposed for the chondroprotective and anti-inflammatory effects of TSG-6 ([@ref-22]). Firstly, the Link domain of TSG-6 can bind to many cytokines and glycosaminoglycans (GAGs), which might prevent the migration of leukocytes to inflammatory sites ([@ref-8], [@ref-7]). Secondly, TSG-6 can bind non-covalently to the inter-α-inhibitor (IαI) to increase its anti-plasmin activity ([@ref-10]). The inhibition of plasmin may prevent the cleavage and activation of pro-matrix metalloproteinases (pro-MMPs) to protect the cartilage. In addition, TSG-6 has been shown to bind to BMP-2 and RANKL which reduced osteoclast activity, but not osteoclast differentiation ([@ref-17], [@ref-18]).
Because cartilage damage, inflammation and bone remodeling have also been implicated in the pathogenesis of OA, we hypothesized that TSG-6 gene therapy might also have protective effects in experimental OA. In this study we first showed functionality of the TSG-6 gene therapy in vitro by inhibiting osteoclast activity. Subsequently, we tested the TSG-6 gene therapy in the collagenase-induced OA (CIOA). We did not observe therapeutic effects of TSG-6 overexpression on inflammation or cartilage damage. In contrast, we observed ectopic bone formation at the medial femur/tibia region of the joint.
Materials and Methods
=====================
Virus production
----------------
For cloning of the third generation self-inactivating lentiviral vector (sin), the full-length TSG-6 coding sequence was obtained from a cDNA sample of inflamed synovium of a C57BL/6 mouse by nested PCR. The first round of PCR was done by amplification using forward primer 5′-CGGCTCTGCAACCGAAGA-3′ and reverse primer 5′-ATCCAAAAGTATTTATTACAGCAAT-3′. For the second round of amplification, forward primer 5′-GTCGACGCCACCATGGTCGTCCTCC-3′ and reverse primer 5′-CATATGATCCAAAAGTATTTATTAC-3′ were used, introducing restriction sites for *Sal*I and *Nde*I, respectively, after which the TSG-6 gene was cloned in the PCR-script CAM cloning vector (Agilent Technologies, Amstelveen, Netherlands) according to the manufacturer's protocol. Subsequently, the TSG-6 gene was cloned in the pRRL-cPPT-PGK-luc-PRE-SIN vector (PGK-TSG6) using the strategy previously described ([@ref-3]). As a control vector (PGK-luc), the Trifusion reporter gene was digested from the pcDNA3.1 vector (kind gift from Robert E. Reeves, Stanford University) using *Nhe*I and *Xba*I (New England Biolabs, Ipswich, MA, USA). For adenovirus cloning, TSG-6 was digested from the PCR-script CAM vector using *Sal*I and ligated in the *Sal*I pre-digested pShuttle-CMV (Stratagene, La Jolla, CA, USA) (CMV-TSG6). The CMV-luciferase adenovirus (CMV-luc) was used as control vector. Lentiviral and adenoviral vector production were performed as described previously ([@ref-2]; [@ref-3]).
Bone marrow-derived cell culture
--------------------------------
Bone marrow-derived cells (BMDCs) of C57BL/6 mice were obtained from the femur and the tibia. Bone marrow cells were flushed from the bones using RPMI medium supplemented with 10% fetal bovine serum (FBS), 1 mM pyruvate and 1% penicillin/streptomycin (P/S) using a Microlance 3 needle (Becton Dickinson (BD), Breda, The Netherlands). Cells were passed trough a 70 μm cell strainer (Corning, NY, USA) and centrifuged for 5 min at 1,500 rpm/423 g in a Heraeus Megafuge 16R (Thermo Scientific, Waltham, MA, USA). In a 96-well plate (Greiner Bio-one, Alphen a/d Rijn, The Netherlands), 10^5^ cells/well were seeded on the surface of the plate for differentiation analysis or on an elephant dentin slice for the bone resoption assay. Osteoclast differentiation was induced in alpha-MEM medium, supplemented with 5% FBS, 1% P/S, 30 ng/ml murine M-CSF (R&D systems, Oxford, UK) and 20 ng/ml soluble murine RANKL (R&D Systems, Oxford, UK). For tartrate resistant acid phosphatase (TRAP) staining, after 72 h the the osteoclast medium was replaced to induce osteoclast fusion with or without 1 μg/ml recombinant murine TSG-6 (rmTSG-6) (R&D Systems, Oxford, UK) for 24 h. Subsequently, cells were fixed and stained with the TRAP staining kit (Sigma-Aldrich, Zwijndrecht, The Netherlands) according to the manufacturer's protocol. For the bone resorption assay, after 72 h the differentiation medium was replaced and cells were transduced with 166 ng lentivirus per 10^5^ cells in each well or left untreated until day 4, at which an optimal dose 1 μg/ml rmTSG-6 (based on [@ref-17]) was added to the rmTSG-6 group to align with protein production in the TSG6-transduced cells. At day 7, medium was removed from all dentin slices. Subsequently, the cells were lysed with H~2~O.
Bone resorption assay
---------------------
After incubation with the osteoclasts, dentin slices were incubated with 10% NH~3~ and sonicated for 20 cycles 30 s on/off on a Bioruptor Next-gen (Diagenode, Seraing, Belgium) at "high" intensity. After washing, the slices were incubated for 10 min with 10% KAl(SO~4~)~2~ and stained with PhastGel Blue R-350 coomassie tablets (GE Healthcare, Eindhoven, The Netherlands) according to the manufacturer's protocol. For the quantification, five pictures at 200× magnification were taken from every dentin slice using the Labovert FS (Leitz, Leica, Rijswijk, The Netherlands). The Leica application Suite (LAS) was used to analyse the average bone resorption percentage per dentin slice.
Synovial biopsy culture
-----------------------
Synovial biopsies were obtained from surplus C.B-17 mice after sacrifice. Using a 3 mm biopsy punch (Stiefel, Wachtersbach, Germany), synovial explants were obtained and combined from the lateral and medial synovium. The biopsies were kept in 200 μl RPMI medium supplemented with 1 mM pyruvate and 1% P/S and 10^7^ infectious units adenovirus CMV-luc or CMV-TSG6 was added. After 2 h, 20 μl FBS was added and after 24 h, total RNA was isolated using 500 μl Tri reagent (Sigma-Aldrich, Zwijndrecht, The Netherlands) according to the manufacturer's protocol and processed and analyzed as described below.
RNA isolation and qPCR
----------------------
Synovial biopsies in Trizol were first homogenated using Magnalyzer Green Beads (Roche Life Sciences, Almere, The Netherlands) and processed according to the provided protocol. DNA was removed by DNAse and cDNA was generated using Moloney murine leukemia virus reverse transcriptase, 0.5 μg/μl oligo(dT) primers and 12.5 mM dNTPs (Thermo Scientific). Quantative real-time PCR was perfomed as previously described ([@ref-3]) using forward primer 5′-CAACCCACATGCAAAGGAG-3′ and reverse primer 5′-TACTCATTTGGGAAGCCCG-3′.
Collagenase-induced OA
----------------------
A total of 30 female C57BL/6J mice (Janvier) of 10--12 weeks old were randomized and housed in groups of five in filtertop cages at DM-II level with 12 h light-dark cycles and water and standard diet (AB Diets, Woerden, The Netherlands) were provided ad libitum. The mice received an intra-articular (i.a.) injection of 10^7^ infectious units adenovirus CMV-luc or CMV-TSG6 dissolved in 6 μl 0.9% NaCl four days prior to the start of the induction of the model. All i.a. injections were performed during the day in the right knee using a BD microlance needle 30G 1/2′ (BD) under general anaesthesia using 2.5% isoflurane. The injections were performed by a technician without knowledge of the viral vector content in a laminar flow cabinet. CIOA was induced by two i.a. injection of one unit collagenase type VII (Sigma-Aldrich, Zwijndrecht, The Netherlands) in 6 μl 0.9% NaCl at day 1 and day 3 as described previously ([@ref-27]). At day 21, a second injection of 10^7^ infectious units adenovirus of the same viral vector was given and mice were sacrificed by cervical dislocation at day 42. All animal experiments were approved by the local authority Animal Care and Use Committee and local ethics committee (RU-DEC 2014-080). Animal care was in accordance with the institution guidelines.
Prosense measurement
--------------------
Six days after the first collagenase injection, five mice per group received i.v. injections of 1.33 nmol Prosense 680 probes (PerkinElmer, Groningen, The Netherlands) in 100 μl in the orbita plexus. At day 7, hair was removed from the knees and the mice were imaged with the IVIS Lumina (PerkinElmer, Groningen, The Netherlands) using the Cy5.5 filter. The data was analyzed using Living Image 3.0 (PerkinElmer, Groningen, The Netherlands). Regions of interest of the same size were drawn around the knees and the florescence intensity were determined for the experimental and contralateral knees.
X-ray imaging
-------------
After sacrifice, the mouse knees were removed and imaged using the Faxitron FX-20 (Faxitron, Tucson, AZ, USA) at 26 kV for 10 s. The images were blinded and randomized and the ectopic bone formation was given an arbitrary score of 0--5. 0 = No ectopic bone formation visible, 1 = ectopic bone formation just detectable, 2 = clear ectopic bone formation, 3 = bone formation stretching along femur and tibia, 4 = large ectopic bone formation, 5 = severe ectopic bone formation, similar to the most severe sample.
Histological analysis
---------------------
After X-ray imaging, knee joints were fixed in formalin, embedded in paraffin and cut in 7 μm sections. The sections were stained with Safranin-O/Fast Green, blinded and randomized. Cartilage damage was assessed using an arbitrary score of 0--30 based on the OARSI cartilage OA histopathology grading system ([@ref-26]), modified for the assessment of murine knee joints by scoring the staging 0--5. Several mice showed dislocation of the knee joint, but these were distributed equally between the groups and are incorporated in the analysis.
Statistical analysis
--------------------
Statistical comparisons were performed by one-way analysis of variance (ANOVA) and Mann--Whitney *U* test as indicated in the text using GraphPad Prism 5.03. Results are depicted as mean ± 95% confidence interval (CI) and *p*-values below 0.05 were regarded as significant.
Results
=======
TSG-6 inhibits osteoclast activity
----------------------------------
An increase in TRAP-positive osteoclasts is observed in the subchondral bone of OA patients ([@ref-25]). We therefore first tested the functionality of viral expression of TSG-6 in vitro by assessing its effects on osteoclast activity. BMDCs were differentiated to multinucleated osteoclasts using M-CSF and RANKL in the absence or presence of recombinant murine TSG-6. The formation of multi-nucleated osteoclasts from BMDCs was observed in conditions with and without TSG-6 ([Fig. S1](#supp-2){ref-type="supplementary-material"}).
We subsequently assessed the effects of TSG-6 on osteoclast resorption activity on dentin slices. Under positive control conditions with lentiviral control virus, resorption pits covered ∼20% of the dentin surface ([Fig. 1](#fig-1){ref-type="fig"}). The resorption by the multinucleated osteoclasts was significantly reduced by both recombinant TSG-6 (∼72%) and the lentiviral expression of TSG-6 (∼49%). No statistical difference was observed between the effects of recombinant TSG-6 and lentivirus-expressed TSG-6 This shows that viral overexpression of TSG-6 can result in functional levels of TSG-6 that can inhibit osteoclast activity.
{#fig-1}
Overexpression of TSG-6 in synovium was first investigated in synovial explants. After transduction with adenoviral CMV-TSG6, the expression of TSG-6 was significantly increased compared to CMV-luciferase ([Fig. 2A](#fig-2){ref-type="fig"}).
{ref-type="fig"}. Sections were stained with Safranin-O and counterstained with Fast Green. Ectopic bone formation is indicated with black arrows. Quantitative results are depicted as mean ± 95% confidence interval (CI) and statistical comparisons were performed by Mann--Whitney *U* test.](peerj-06-4771-g002){#fig-2}
TSG-6 effects in CIOA
---------------------
The protective effects of TSG-6 on inflammation and cartilage damage were tested in the CIOA model, an experimental OA that includes inflammation ([@ref-27]). Mice received an injection with adenovirus to provide high expression levels of TSG-6 or control luciferase in the right knee four days prior to the first collagenase injection. The effects of TSG-6 on inflammation-associated protease activity was determined in a subset of the mice after injection of Prosense 680 probes at day 6 after the first collagense injection. At day 7, the fluorescence intensity was measured and compared to the naïve contralateral knee. No significant differences in protease activity were found between the TSG-6 adenovirus (average 2.0-fold compared to contralateral joint) and the luciferase control virus (average 1.6-fold compared to contralateral joint) ([Fig. 2B](#fig-2){ref-type="fig"}). A second adenovirus injection was given at day 20 to provide TSG-6 expression for the second half of the model and mice were sacrificed at day 42. The cartilage damage was assessed in histological sections using the OARSI cartilage OA histopathology grading system. No significant differences were observed in cartilage damage between the control group and the TSG-6 treated group ([Figs. 2C](#fig-2){ref-type="fig"}--[2E](#fig-2){ref-type="fig"}).
Before the joints were processed for histological analysis, the bone structure was analyzed using X-ray imaging and scored using an arbitrary scoring method from 0 to 5. Surprisingly, mice treated with adenoviral TSG-6 showed significantly more ectopic bone formation (77%) compared to the control group (20%) located at the medial collateral ligament ([Figs. 2F](#fig-2){ref-type="fig"}--[2H](#fig-2){ref-type="fig"}). Safranin-O staining of histological sections of the joints shown in [Figs. 2F](#fig-2){ref-type="fig"} and [2G](#fig-2){ref-type="fig"} shows the presence of cartilage around the ectopic bone, indicating that the ectopic bone formation might be the result of endochondral ossification ([Figs. 2I](#fig-2){ref-type="fig"} and [2J](#fig-2){ref-type="fig"}).
Discussion
==========
In this study, we show that in vitro viral overexpression of TSG-6 in BMDCs has no effect on multinuclear osteoclast formation, but can reduce resorption activity of the osteoclasts. This has been observed in earlier studies ([@ref-17], [@ref-18]) and shows that functional TSG-6 can be expressed after viral gene transfer and reduce in vitro bone resorption. However, when tested in the CIOA model, TSG-6 could not reduce cartilage damage. In contrast, we observed increased ectopic bone formation in mice with TSG-6 overexpression.
The lack of protective effects from TSG-6 overexpression on the development of cartilage damage, despite the inhibitory effects on osteoclast activity, might be related to the indications that cartilage damage in the CIOA model is caused by mechanical stress resulting from joint instability ([@ref-29]). Mechanical stress is implicated as an important cause for cartilage damage in many OA patients ([@ref-12]; [@ref-4]), but is distinct from the mechanisms that cause cartilage damage in the RA models in which TSG-6 treatment was successful.
Although TSG-6 is associated with different functions that might be protective for OA, the expression and activity of TSG-6 correlate with progression of OA ([@ref-30]). The exact mechanisms by which TSG-6 might be involved OA progression is not completely understood and is difficult to study because of the ability of TSG-6 to bind many different proteins and matrix components. Thus, TSG-6 may have tissue-specific functions, resulting in protective effects in experimental RA, but detrimental effects in OA models. One potential mechanism could be related to a disturbance of damage repair mechanisms. TSG-6 has been shown to bind to fibronectin (FN), stimulating FN matrix assembly involved in damage repair ([@ref-13]). However, TSG-6 can also serve as a bridging molecule between FN and thrombospondin-1 (TSP-1). TSP-1 is increased in OA chondrocytes and osteophytes and has multiple functions, including the activation of latent transforming growth factor-β (TGF-β) ([@ref-5]; [@ref-24]). Increased TGF-β activity has previously been correlated with chondrocyte differentiaton and enthesophyte formation similar to this study in the medial collateral ligament in CIOA ([@ref-2]; [@ref-6]). The formation of osteophytes in these marginal locations is associated with OA progression in human OA ([@ref-9]).
An alternative mechanism for increased TGF-β activity could be related to hyaluronan (HA). In the context of fibrosis, studies have shown that the heavy chain transfer to HA by TSG-6 can stabilize the HA coat, which is essential for TGF-β-mediated myofibroblast differentiation ([@ref-20]). Possibly, TSG-6 can increase TGF-β-mediated chondrogenesis in a similar way. HA could also increase chondrogenesis in adipose-derived stem cells ([@ref-32]). This was dependent on the interaction between HA and CD44, which is influenced by TSG-6 ([@ref-16]). The formation of bone is dependent on the ratio between bone matrix deposition and bone resorption. The inhibiting effects on osteoclast activity observed in [Fig. 1](#fig-1){ref-type="fig"} might favor the anabolic activity and stimulate the ectopic bone formation. Although the exact mechanism by which TSG-6 can increase ectopic bone formation in vivo remains to be elucidated, no improvement in cartilage damage in was observed, indicating that intra-articular gene therapy with TSG-6 might not be a promising treatment for OA.
Supplemental Information
========================
10.7717/peerj.4771/supp-1
###### Raw data.
######
Click here for additional data file.
10.7717/peerj.4771/supp-2
###### Osteoclastogenesis and bone resorption by bone marrow-derived cells (BMDCs) after TSG-6 treatment.
**(A)** Representative TRAP staining of BMDCs differentiated to osteoclasts using M-CSF and RANKL. Multi-nucleated osteoclasts are indicated by black arrows. **(B)** Typical example of bone resorption on dentin slices by osteoclasts.
######
Click here for additional data file.
10.7717/peerj.4771/supp-3
###### NC3Rs ARRIVE guidelines checklist.
######
Click here for additional data file.
Additional Information and Declarations
=======================================
The authors declare that they have no competing interests.
[Mathijs G.A. Broeren](#author-1){ref-type="contrib"} conceived and designed the experiments, performed the experiments, analyzed the data, prepared figures and/or tables, approved the final draft.
[Irene Di Ceglie](#author-2){ref-type="contrib"} conceived and designed the experiments, performed the experiments, analyzed the data, authored or reviewed drafts of the paper, approved the final draft.
[Miranda B. Bennink](#author-3){ref-type="contrib"} conceived and designed the experiments, performed the experiments, analyzed the data, prepared figures and/or tables, authored or reviewed drafts of the paper, approved the final draft.
[Peter L.E.M. van Lent](#author-4){ref-type="contrib"} conceived and designed the experiments, contributed reagents/materials/analysis tools, authored or reviewed drafts of the paper, approved the final draft.
[Wim B. van den Berg](#author-5){ref-type="contrib"} conceived and designed the experiments, approved the final draft.
[Marije I. Koenders](#author-6){ref-type="contrib"} analyzed the data, contributed reagents/materials/analysis tools, authored or reviewed drafts of the paper, approved the final draft.
[Esmeralda N. Blaney Davidson](#author-7){ref-type="contrib"} performed the experiments, analyzed the data, contributed reagents/materials/analysis tools, authored or reviewed drafts of the paper, approved the final draft.
[Peter M. van der Kraan](#author-8){ref-type="contrib"} analyzed the data, authored or reviewed drafts of the paper, approved the final draft.
[Fons A.J. van de Loo](#author-9){ref-type="contrib"} conceived and designed the experiments, prepared figures and/or tables, authored or reviewed drafts of the paper, approved the final draft, obtaining of funding.
The following information was supplied relating to ethical approvals (i.e., approving body and any reference numbers):
All animal experiments were approved by the Radboud University DierExperimentele Commissie (RU-DEC).
The following information was supplied regarding data availability:
The raw data are provided in a [Supplemental File](#supplemental-information){ref-type="supplementary-material"}.
| {
"pile_set_name": "PubMed Central"
} |
Image en médecine {#sec1}
=================
Nous rapportons une observation rare d'un adénocarcinome du sinus sphénoïdal se présentant sous forme d'une tumeur extensive de la base du crâne. Il s'agissait d'une patiente de 42 ans qui a consulté pour une symptomatologie unilatérale droite faite d'une obstruction nasale, une diplopie et des névralgies de l'hémiface. L'examen clinique a montré une paralysie des V^ème^ et VI^ème^ paires crâniennes. Le scanner cérébral a montré une volumineuse masse hétérogène sphénoïdale et clivale atteignant le sinus caverneux droit, avec une composante tissulaire périphérique au niveau du sinus sphénoïdal. Cette dernière a été biopsiée sous anesthésie générale, par voie endonasale à travers une sphénoïdomie. L'examen histologique définitif a conclu à un adénocarcinome de type non intestinal. La patiente est décédée en altération de l'état général en cours d'exploration. Les adénocarcinomes de la base du crane prennent naissance le plus fréquemment au niveau de l'ethmoïde. L'origine sphénoïdale est exceptionnelle. L'aspect radiologique est aspécifique et évoque la malignité. C'est un diagnostic qui doit être évoqué devant une tumeur agressive, même de l'étage moyen de la base du crâne.([figure 1](#f0001){ref-type="fig"})
{#f0001}
| {
"pile_set_name": "PubMed Central"
} |
Supported by the National Institute of Diabetes, Digestive, and Kidney Diseases (U01 grants DK 62497 to J.A.B., DK 62470 to S.J.K., DK 62481 to K.M.L., DK 62456 to C.S., DK 62466 to V.V., DK 62453 to R.J.S., DK 84538 to K.W., DK 62436 to E.A., and DK 642453 to V.L.N.); the National Center for Advancing Translational Sciences, National Institutes of Health (UL1 TR001878 to the Children's Hospital of Philadelphia); Clinical Translational Science Awards **(**UL1 TR001082 to the University of Colorado, Denver); and the Cincinnati Center for Translational Science and Training (5KL2TR001426‐03 to the Cincinnati Children's Hospital).
Potential conflict of interest: Dr. Karpen consults for Albireo, Intercept, and Retrophin. Dr. Sokol consults for and received grants from Shire; he consults for Albireo, Alexion, and Retrophin. The other authors have nothing to report.
BA
: biliary atresia
CD
: clusters of differentiation
CI
: confidence interval
CS&T
: cytometer setup and tracking
FACS
: fluorescent activated cell sorted
FBS
: fetal bovine serum
Foxp3
: forkhead box P3
HLA‐DR
: human leukocyte antigen‐DR isotype
HPE
: hepatoportoenterostomy
HR
: hazard ratio
IFN
: interferon
IL
: interleukin
IQR
: interquartile range
IVIg
: intravenous immunoglobulin
MFI
: mean fluorescent intensity
MPO
: myeloperoxidase
NET
: neutrophil extracellular trap
NK
: natural killer cells
O.D.
: optical density
PBMC
: peripheral blood mononuclear cell
PRIME
: Safety Study of Intravenous Immunoglobulin Post‐Portoenterostomy in Infants With Biliary Atresia
Th
: T helper
TNF
: tumor necrosis factor
Treg
: regulatory T cell
Biliary atresia (BA) is a progressive fibroinflammatory cholangiopathy of infancy that results in obstruction of the biliary tree within 3‐4 months of age. If no therapy is implemented, portal hypertension and end‐stage liver disease ensue, leaving liver transplantation as the only therapeutic option for long‐term survival. Hepatoportoenterostomy (HPE) is the operative procedure used currently to improve bile drainage in infants with BA.[1](#hep41332-bib-0001){ref-type="ref"} Although prompt diagnosis and surgical intervention may restore bile flow, progression to end‐stage liver disease occurs in almost 80% of patients by age 20 years, with over 50% of the patients requiring liver transplantation by 2 years of age.[2](#hep41332-bib-0002){ref-type="ref"}
The biological basis for the progression of liver disease after HPE is not fully understood, but the presence of inflammation and proinflammatory cytokines in the liver and bile ducts at the time of diagnosis suggests that the host immune response, at least in part, mediates the progressive injury. Both innate and adaptive immune responses have been implicated in the pathogenesis of bile duct injury in BA.[3](#hep41332-bib-0003){ref-type="ref"}, [4](#hep41332-bib-0004){ref-type="ref"}, [5](#hep41332-bib-0005){ref-type="ref"}, [6](#hep41332-bib-0006){ref-type="ref"}, [7](#hep41332-bib-0007){ref-type="ref"}, [8](#hep41332-bib-0008){ref-type="ref"}, [9](#hep41332-bib-0009){ref-type="ref"}, [10](#hep41332-bib-0010){ref-type="ref"}, [11](#hep41332-bib-0011){ref-type="ref"}, [12](#hep41332-bib-0012){ref-type="ref"}, [13](#hep41332-bib-0013){ref-type="ref"}, [14](#hep41332-bib-0014){ref-type="ref"}, [15](#hep41332-bib-0015){ref-type="ref"}, [16](#hep41332-bib-0016){ref-type="ref"}, [17](#hep41332-bib-0017){ref-type="ref"}, [18](#hep41332-bib-0018){ref-type="ref"}, [19](#hep41332-bib-0019){ref-type="ref"}, [20](#hep41332-bib-0020){ref-type="ref"}, [21](#hep41332-bib-0021){ref-type="ref"}, [22](#hep41332-bib-0022){ref-type="ref"} Given these observations, the immunosuppressant intravenous immunoglobulin (IVIg) was recently tested in a phase I/IIa trial following HPE in infants with BA (Safety Study of Intravenous Immunoglobulin Post‐Portoenterostomy in Infants With Biliary Atresia \[PRIME\] study).[23](#hep41332-bib-0023){ref-type="ref"} IVIg has been used in a number of immune‐mediated and autoimmune diseases to attenuate the inflammatory response and reduce disease severity,[24](#hep41332-bib-0024){ref-type="ref"}, [25](#hep41332-bib-0025){ref-type="ref"}, [26](#hep41332-bib-0026){ref-type="ref"}, [27](#hep41332-bib-0027){ref-type="ref"}, [28](#hep41332-bib-0028){ref-type="ref"}, [29](#hep41332-bib-0029){ref-type="ref"} including in the mouse model of BA.[30](#hep41332-bib-0030){ref-type="ref"} IVIg has a multitude of effects on the immune system, including inhibition of T‐cell activation, antibody and cytokine production, dendritic cell maturation, natural killer (NK) cell trafficking, and neutrophil function.[28](#hep41332-bib-0028){ref-type="ref"}, [29](#hep41332-bib-0029){ref-type="ref"} Furthermore, IVIg is associated with expansion and activation of anti‐inflammatory regulatory T cells (Tregs).[29](#hep41332-bib-0029){ref-type="ref"}
In the PRIME study, IVIg was administered to 29 infants with BA at 3‐5, 30, and 60 days post‐HPE. In comparison to a historical cohort, IVIg therapy did not improve outcome in BA based on the outcome measures of total serum bilirubin levels of \<1.5 mg/dL at 90 days post‐HPE and transplant‐free survival at 360 days post‐HPE.[23](#hep41332-bib-0023){ref-type="ref"} A component of the PRIME study included immunophenotyping patients with BA over time in the setting of IVIg therapy. The aim of this study was to characterize the peripheral blood immunophenotype of patients with BA at diagnosis (baseline) and at 60, 90, 180, and 360 days post‐HPE in order to determine potential changes over time in response to IVIg and correlations of specific immune markers with outcomes (serum bilirubin, transplant‐free survival).
Participants and Methods {#hep41332-sec-0002}
========================
Prime Study {#hep41332-sec-0003}
-----------
PRIME[23](#hep41332-bib-0023){ref-type="ref"} was a multicenter, single‐arm, open‐label phase I/IIa trial of IVIg therapy following HPE in infants with BA. It was conducted at eight clinical sites in the Childhood Liver Disease Research Network (ChiLDReN) funded by the National Institute of Diabetes and Digestive and Kidney Diseases ([clinicaltrials.gov NCT01854827](https://clinicaltrials.gov/ct2/show/NCT01854827)). Ethical approval was obtained at each site and at the Data Coordinating Center; parents or legal guardians of the infants provided written informed consent. The study protocol conformed to the ethical guidelines of the 1975 Declaration of Helsinki as reflected in *a priori* approval by the appropriate institutional review committees. Enrollment began in October 2013 and ended in July 2015, with the 360‐day follow‐up completed in July 2016. We refer the reader to Mack et al.[23](#hep41332-bib-0023){ref-type="ref"} for complete information on the PRIME study. Inclusion criteria included age ≤120 days, enrollment within 3 days of diagnosis of BA and HPE, postconception age ≥36 weeks, and weight ≥2,000 g. Exclusion criteria included BA with extrahepatic congenital manifestations (i.e., BA splenic malformation syndrome) or another concurrent chronic condition prohibiting the use of IVIg. Participants received three infusions of 1 g/kg body weight of IVIg (Gamunex‐C; Grifols, Los Angeles, CA) at days 3‐5, 30, and 60 days after HPE. Routine clinical care guidelines for postoperative care included oral ursodeoxycholic acid for 360 days after HPE, trimethoprim‐sulfamethoxasole for 180 days after HPE, and vitamin supplementation. Corticosteroid treatment was prohibited.
Peripheral Blood Immunophenotyping {#hep41332-sec-0004}
----------------------------------
Whole blood and plasma were obtained at the time of diagnosis (baseline; prior to IVIg infusion \#1), 60 days post‐HPE (prior to IVIg infusion \#3), and 90, 180, and 360 days post‐HPE. Blood or plasma was not collected for immunophenotyping after a patient received a liver transplant. Whole blood was shipped overnight express at room temperature to the University of Colorado‐Anschutz Medical Campus for Ficoll‐gradient purification of peripheral blood mononuclear cells (PBMCs) and dextran‐gradient purification of neutrophils according to standard protocols (BD Biosciences, San Jose, CA). PBMCs and neutrophils were frozen in 90% fetal bovine serum (FBS)‐10% dimethyl sulfoxide in liquid nitrogen until the time of analysis. Plasma was aliquoted and frozen at −20°C at the local site and then shipped in batches to the Children's Hospital Colorado Clinical Translational Research Center (CTRC) Core Laboratory for analysis. The specific peripheral blood immune cells and plasma cytokines analyzed were chosen based on those immune markers described in the literature to be associated with BA or known to be altered by IVIg.
### Fluorescence‐Activated Cell Sorting Analysis {#hep41332-sec-0005}
Fluorescence‐activated cell sorting (FACS) analysis was performed by the ClinImmune Laboratories Flow Cytometry Core, University of Colorado‐Anschutz Medical Campus, according to standard protocol. All samples were processed over a 2‐week time period using identical conditions for staining and FACS analysis. Cryopreserved PBMCs were thawed in complete media (1% phosphosilicate glass, 10% FBS, 89% Roswell Park Memorial Institute media), washed with staining buffer (phosphate‐buffered saline containing 2% FBS), and \>1.0 × 10^6^ cells were surfaced stained with PBMC markers, including NK cell subsets (clusters of differentiation \[CD\]3, CD56), macrophage subsets (CD14, CD16, CD11b), T‐cell subsets (CD3, CD4, CD8), B cells (CD19), activation markers on macrophages, B cells and T cells (CD38, CD69, CD86, CD32, interferon \[IFN\]‐γR1 \[CD119\], human leukocyte antigen‐DR isotype \[HLA‐DR\]); Tregs (CD3, CD4, CD25, CD127, and forkhead box P3 \[Foxp3\]); and neutrophil markers (CD66b, CD15, CD14; activation markers CD62L, chemokine (C‐X‐C motif) receptor 4 \[CXCR4\], CD54). Cells were fixed with 1% paraformaldehyde and analyzed using a Canto II cytometer (BD Immunocytometry Systems). Similarly, for the Treg panel, cells were washed, fixed, surface stained, permeabilized (eBioscience), and then intracellularly stained with Foxp3, washed and resuspended in staining buffer, and analyzed using an LSR II cytometer (BD Immunocytometry Systems). Fluorescence minus one (FMO) or isotype controls were used in all experiments.
Between 0.25 million and 0.5 million events were collected. Electronic compensation was performed with antibody capture beads (BD Biosciences) stained separately with individual monoclonal antibodies used in the test samples. To ensure the accuracy and precision of the measurements taken from day to day, quality control was performed daily on the LSR‐II using the cytometer setup and tracking (CS&T) feature within BD FACSDiva software. The program uses standardized CS&T beads (BD Biosciences) to determine voltage, laser delays, and area scaling and to track these settings over time. A manual quality control using rainbow beads was also performed daily to verify the laser delay and area scaling determined by the CS&T.
The data files were analyzed using Diva software (BD Biosciences) and FlowJo Software (Treestar Inc.). Acquired data were analyzed using FlowJo software using appropriate templates created for each of the panels. Gates were adjusted on a patient‐specific basis using the FMO or isotype controls. The percentage of each cell type was determined and reported as a percentage of the parent cell type. For studies pertaining to activation markers, the mean fluorescent intensity (MFI) of the activation marker on the particular cell type was recorded. The frequencies and MFI for each gated population was exported into an excel spreadsheet for additional statistical analysis. The gating strategy for FACS analysis is provided in Supporting Fig. [S1](#hep41332-sup-0001){ref-type="supplementary-material"}.
### Plasma Protein Analysis {#hep41332-sec-0006}
Quantification of cytokines was performed by the Children's Hospital Colorado CTRC Core Laboratory using the Luminex T helper (Th)1/Th2 multiplex assay (interleukin \[IL\]‐1β, IL‐2, IL‐4, IL‐5, IL‐6, IL‐8, IL‐10, granulocyte‐macrophage colony‐stimulating factor, IFN‐γ, tumor necrosis factor \[TNF\]‐α) and the IL‐17 enzyme‐linked immunosorbent assay (Thermo Fisher Scientific, Waltham, MA) according to standard protocols. Quantification of neutrophil function was performed with assays available from Cayman Chemical (Ann Arbor, MI) for plasma elastase, myeloperoxidase (MPO), and neutrophil extracellular traps (NETs) according to protocols. Both positive and negative controls were performed with each assay, and individual samples were analyzed in duplicate.
Clinical Outcomes {#hep41332-sec-0007}
-----------------
In this study, measures of interest included the PBMC and neutrophil subsets, cell activation markers, plasma cytokines and neutrophil functional assays (detailed above), and clinical outcomes as defined by 1) changes in total bilirubin at 90 days and 180 days and 2) liver transplant or death. Considering these two endpoints together, a composite clinical outcome was defined as follows: If a patient had no liver transplant/death and serum bilirubin was \<1.5 mg/dL at 360 days post‐HPE, then the patient was grouped into the Good Outcome group; otherwise, if a patient had liver transplant/death or serum bilirubin persisting above 1.5 mg/dL up to 360 days post‐HPE, then the patient was grouped into the Poor Outcome group. Total bilirubin was determined directly using standard laboratory methods or calculated by the addition of direct plus indirect bilirubin.[31](#hep41332-bib-0031){ref-type="ref"} Total bilirubin levels after transplant were not used for the analysis.
Statistical Analysis {#hep41332-sec-0008}
--------------------
The analysis population is the modified intention‐to‐treat (mITT) population; all PRIME participants who maintained eligibility and received at least 80% of one dose of IVIg were considered. Participants with missing immune markers were excluded from the given analyses. If total bilirubin levels were missing at 360 days, good bile drainage was imputed if the total bilirubin value was \<1.5 mg/dL at 270 days, the visit immediately prior to the 360‐day time point.
Changes in immune biomarkers and total bilirubin at 90 days and 180 days were calculated relative to baseline. Spearman's correlation coefficients and accompanying *P* values are presented to describe the relationship between a change in biomarker and a change in total bilirubin at 90 and 180 days. To assess the impact of changes in markers to 60 days in relation to transplant/patient death, Cox proportional hazards models were fitted among the patients who survived up to 60 days (hence, completed all three doses of IVIg). Logistic regression was used to assess the dichotomous Poor Outcome at 360 days. Two models were assessed for each biomarker, one modeling Poor Outcome associated with a change from baseline to 60 days post‐HPE and the second modeling Poor Outcome associated with a change from baseline to 90 days post‐HPE. All analyses were performed using SAS version 9.3 (SAS Institute Inc.).
Results {#hep41332-sec-0009}
=======
Patient Cohort {#hep41332-sec-0010}
--------------
There were 29 participants in the PRIME study; 25 of these received all three IVIg infusions and 4 received only two infusions. The mITT population had a mean ± SD age at HPE of 60 ± 19 days. There were more female (62%) than male participants, and the population was predominantly white and non‐Hispanic (Table [1](#hep41332-tbl-0001){ref-type="table"}). Serum total bilirubin at baseline was 8.3 ± 3.4 mg/dL (range, 3‐17 mg/dL); bilirubin levels at all time points are shown in Table [2](#hep41332-tbl-0002){ref-type="table"}. Forty‐one percent of participants had a liver transplant by 360 days post‐HPE (median time to transplant,149.5 days post‐HPE), and 1 individual died after the transplant. As detailed in the recently published PRIME study,[23](#hep41332-bib-0023){ref-type="ref"} there was no significant increase in the proportion of IVIg participants with a serum total bilirubin \<1.5 mg/dL at 90, 180, or 360 days post‐HPE compared to a historical placebo‐arm group (Table [2](#hep41332-tbl-0002){ref-type="table"}). Survival with the native liver in the IVIg participants showed no significant benefit over that of the historical placebo‐arm participants, with a difference at 360 days of --11.9%.[23](#hep41332-bib-0023){ref-type="ref"} There were 17 participants (59%) in the Good Outcome group and 12 participants (41%) in the Poor Outcome group.
######
Patient Cohort[23](#hep41332-bib-0023){ref-type="ref"}
Baseline demographics (n = 29) Value
-------------------------------- ----------
Age at HPE, days (mean ± SD) 60 ± 19
Age at HPE, n (%)
≤30 days 2 (7%)
\>30 to ≤45 days 5 (17%)
\>45 to ≤60 days 7 (24%)
\>60 to ≤90 days 14 (48%)
\>90 to ≤120 days 1 (3%)
Female, n (%) 18 (62%)
Race, n (%)
White/Caucasian 22 (76%)
Black/African American 3 (10%)
Asian 2 (7%)
Refused/not reported 2 (7%)
Ethnicity, n (%)
Hispanic 9 (31%)
Non‐Hispanic 20 (69%)
John Wiley & Sons, Ltd
######
Patient Outcomes
Baseline Day 60 Day 90 Day 180 Day 360
------------------------------------- ------------- ---------------- ----------------- ---------------- -----------------
Total bilirubin, mg/dL
n 29 26 27 21 10
Mean (SD) 8.3 (3.4) 6.0 (7.5) 6.0 (7.0) 3.7 (6.4) 2.1 (3.9)
Median (IQR) 7.7 (6, 10) 3.3 (0.9, 9.7) 2.3 (0.6, 12.5) 0.9 (0.3, 2.4) 0.65 (0.3, 2.0)
Minimum, maximum 3, 17 0.4, 34.9 0.2, 21.4 0.2, 25.4 0.18, 12.9
Patient outcomes over time (n = 29)
Transplant/deaths, n (%) 0 (0%) 2 (6.9%) 5 (17.2%) 9 (31.0%) 12 (41.4%)
Alive with native liver:
Bilirubin \<1.5 mg/dL, n 0 9 11 14 7
Bilirubin ≥1.5 mg/dL, n 29 16 13 5 3
John Wiley & Sons, Ltd
Activated NK Cells, Decreased Tregs, and High IL‐8 and Neutrophil Activation Products Correlate With Change in Serum Bilirubin {#hep41332-sec-0011}
------------------------------------------------------------------------------------------------------------------------------
Quantification of immune markers was performed at baseline and at 60, 90, 180, and 360 days post‐HPE. Statistical analyses focused on changes in immune marker values relative to changes in serum total bilirubin in the first 90 days post‐HPE based on the fact that this time period overlaps with the time frame of IVIg infusions. Furthermore, 90 days post‐HPE has been shown to be a useful time point to analyze the strength of a biomarker (e.g., bilirubin) in predicting short‐term outcome in BA.[32](#hep41332-bib-0032){ref-type="ref"} Spearman correlation coefficient analyses revealed that the percentage of NK cells co‐expressing activation markers HLA‐DR and CD38 (HLADR^+^CD38^+^NK cells) as well as NK cell CD69 (activation marker) and HLA‐DR levels from baseline to 90 days post‐HPE positively correlated with a change in bilirubin over the same period (Fig. [1](#hep41332-fig-0001){ref-type="fig"}). In contrast, the change in the percentage of anti‐inflammatory Tregs (CD3^+^4^+^25^+^Foxp3^hi^CD127^low^) from baseline to 90 days post‐HPE negatively correlated with a change in serum bilirubin.
{#hep41332-fig-0001}
IL‐8 (chemokine \[C‐X‐C motif\] ligand 8 \[CXCL8\]) is a chemokine that recruits neutrophils and other leukocytes to sites of tissue injury. IL‐8 signals the initiation of the oxidative burst in activated neutrophils, resulting in increased production of reactive oxygen species and proteolytic enzymes, such as elastase and MPO.[33](#hep41332-bib-0033){ref-type="ref"} Furthermore, a byproduct of neutrophil activation is the formation of NETs that have been shown to have both antimicrobial properties as well as the potential for stimulating inflammatory and autoimmune responses.[34](#hep41332-bib-0034){ref-type="ref"}, [35](#hep41332-bib-0035){ref-type="ref"}, [36](#hep41332-bib-0036){ref-type="ref"}, [37](#hep41332-bib-0037){ref-type="ref"} The change in IL‐8, elastase, and NETs from baseline to day 90 post‐HPE positively correlated with a change in bilirubin (Table [3](#hep41332-tbl-0003){ref-type="table"}; *P *\< 0.05). Individual patient data of the changes in a specific immune marker with the change in bilirubin at 90 days post‐HPE are displayed in Fig. [1](#hep41332-fig-0001){ref-type="fig"}; these data are fitted with a regression line to show the linear trends over time. Descriptive statistics of the immune markers that were found to correlate with outcome are shown in Supporting Table [S1](#hep41332-sup-0002){ref-type="supplementary-material"}. Descriptive statistics of baseline values of all immune markers analyzed are shown in Supporting Table [S2](#hep41332-sup-0003){ref-type="supplementary-material"}. Only four cytokines were consistently detectable in the sera at all time points (IL‐4, IL‐6, IL‐8, and TNF‐α).
######
Spearman Correlation Between Change in Immune Marker and Change in Bilirubin From Baseline to 90 Days Post‐HPE and 180 Days Post‐HPE
Change From Baseline to Day 90 Change From Baseline to Day 180
---------------------- -------------------------------- --------------------------------- -------- -------------------------- --------- -------
Total bilirubin --1.66 (--2.55, 0.25) --1.96 (--2.85, --1.14)
\% HLADR^+^CD38^+^NK --0.92 (--9.49, 1.75) 0.723 0.003 --2.67 (--10.39, 8) 0.198 0.497
NK cells: CD69 MFI --52 (--156, 59) 0.749 0.002 --43.5 (--124, 46) 0.704 0.005
NK cells: HLADR MFI --1,567 (--6,839, 682) 0.530 0.051 --1,793 (--5,474, 2,069) 0.316 0.292
\% Tregs --0.69 (--2.19, 1.46) --0.653 0.011 0.43 (--1.18, 1.82) --0.716 0.004
IL‐8 (pg/mL) --27.7 (--51.8, 240.6) 0.699 0.001 --5.2 (--71.2, 39.6) 0.583 0.018
NET (O.D.) 0.32 (--0.02, 0.61) 0.604 0.0172 0.13 (--0.55, 0.51) 0 1.00
Elastase (ng/mL) 1.44 (--152.9, 337.5) 0.524 0.037 109.83 (--99.3, 247.12) 0.602 0.029
John Wiley & Sons, Ltd
NK cell CD69 expression, Treg, IL‐8, and elastase correlations with change in bilirubin persisted at 180 days post‐HPE. It should be noted that these analyses did not include 2 patients who received a transplant before 90 days and 8 patients who received a transplant before 180 days. It is possible that the correlation may have been stronger if these patients had been included. The other immune markers analyzed did not correlate with a change in bilirubin at 90 or 180 days post‐HPE (data not shown). In summary, NK cell activation, Treg deficiency, and high levels of IL‐8 were associated with increases in elastase and NETs and correlated with rising bilirubin. This suggests that these immune pathways may contribute to bile duct injury in BA.
Activated NK Cells and IL‐8 Associated With Need for Liver Transplant {#hep41332-sec-0012}
---------------------------------------------------------------------
We then asked the question as to whether or not the immune biomarkers that correlated with the change in bilirubin levels would also be associated with the risk for liver transplant/death by 360 days post‐HPE. All immune markers were analyzed over time based on the patient outcome group. The Good Outcome group had serum bilirubin \<1.5 mg/dL and no liver transplant/death, and the Poor Outcome group had serum bilirubin ≥1.5 mg/dL or liver transplant/death. This revealed trends in changes over time, especially in the first 90 days post‐HPE; however, there was no significant difference in the actual value of the immune marker at each time point between groups (Fig. [2](#hep41332-fig-0002){ref-type="fig"}). Cox modeling for the risk of transplant/death at 360 days post‐HPE was performed, demonstrating the hazard ratios (HRs) for the risk of transplant/death in relation to the immune biomarker change from baseline to 60 days post‐HPE. This analysis revealed that an increase in the percentage of HLA‐DR^+^CD38^+^ NK cells (HR, 1.15; 95% confidence interval \[CI\], 1.00‐1.32; *P* = 0.055) and in plasma IL‐8 levels (HR, 1.71; 95% CI, 1.24‐2.36; *P *= 0.001) at 60 days post‐HPE was significantly associated with an increased risk of transplant/death by 360 days post‐HPE (Table [4](#hep41332-tbl-0004){ref-type="table"}). In addition, Cox modeling with the actual IL‐8 level at 60 and 90 days post‐HPE as a predictor also indicated that IL‐8 was the strongest marker associated with outcome. The actual IL‐8 level at 60 and 90 days post‐HPE predicted survival with the native liver at 360 days, after adjusting for baseline level (60 days: HR, 1.57; 95% CI, 1.12‐2.21; *P = *0.009; 90 days: HR, 1.59; 95% CI, 1.03‐2.47; *P = *0.037). Logistic regression modeling revealed trends toward a marginal significance in the percentage of Treg change from baseline to 60 days post‐HPE in the Good Outcome group (odds ratio \[OR\], 0.31; 95% CI, 0.09‐1.12; *P = *0.074) and a change from baseline to 90 days post‐HPE in NK cell CD69 expression (OR, 7.87; 95% CI, 0.69‐90.04; *P = *0.097).
{#hep41332-fig-0002}
######
Cox Model HRs for Liver Transplant in Relation to Immune Marker Change from Baseline to 60 Days Post‐HPE
Biomarker HR[\*](#hep41332-note-0003){ref-type="fn"} (95% CI) *P*Value
------------------------------- ----------------------------------------------------- ----------
\% HLADR^+^CD38^+^NK cells 1.15 (1.00, 1.32) 0.055
NK cells: CD69 (per 100 MFI) 1.4 (0.79, 2.47) 0.246
NK cells: HLADR (per 100 MFI) 1.004 (0.99, 1.02) 0.686
\% Tregs 0.81 (0.58, 1.12) 0.196
IL‐8 (pg/mL) 1.71 (1.24, 2.36) 0.001
NET (O.D.) 2.79 (0.6, 12.95) 0.191
Elastase (ng/mL) 1.00 (1.00, 1.00) 0.205
Cox regression model was used, conditional on survival up to day 60.
John Wiley & Sons, Ltd
Discussion {#hep41332-sec-0013}
==========
This study provides comprehensive immunophenotyping in BA, identifying significant associations of specific immune markers with short‐term outcomes. Despite the anti‐inflammatory effects of IVIg in the first 90 days post‐HPE, activated NK cells, high IL‐8 and neutrophil byproducts, and decreased Tregs directly correlated with a poor outcome based on rising bilirubin and/or need for liver transplant. A plausible mechanism as to how these immune cells interact with each other, resulting in worsening biliary disease, is outlined in Fig. [3](#hep41332-fig-0003){ref-type="fig"}. Previous investigations in patients with BA and in the rotavirus‐induced mouse model of BA (murine BA) have shown decreased number and function of Tregs.[19](#hep41332-bib-0019){ref-type="ref"}, [20](#hep41332-bib-0020){ref-type="ref"}, [21](#hep41332-bib-0021){ref-type="ref"}, [22](#hep41332-bib-0022){ref-type="ref"} Here, we show for the first time that Treg deficiencies correlate with rising bilirubin levels and the need for liver transplant, suggesting that the lack of Treg control of inflammation results in increased bile duct damage. Downstream effects of the diminished regulation of inflammation include activation of NK cells and increased production of IL‐8. Similar to Tregs, there is abundant literature supporting the contribution of NK cells to biliary disease in both human and murine BA.[5](#hep41332-bib-0005){ref-type="ref"}, [18](#hep41332-bib-0018){ref-type="ref"}, [19](#hep41332-bib-0019){ref-type="ref"} Our study is the first to show that activated NK cells circulating in peripheral blood are associated with worse clinical outcome.
{#hep41332-fig-0003}
IL‐8 is produced by multiple immune cells, including macrophages and NK cells.[33](#hep41332-bib-0033){ref-type="ref"} The main function of IL‐8 is to act as a chemokine and promote neutrophil activation and migration to sites of inflammation (i.e., liver). Our finding of elevated IL‐8 as an immune marker for poor outcome is supported by studies showing that liver IL‐8 levels correlated with liver inflammation, fibrosis, and portal hypertension in BA.[38](#hep41332-bib-0038){ref-type="ref"}, [39](#hep41332-bib-0039){ref-type="ref"}, [40](#hep41332-bib-0040){ref-type="ref"} Activated neutrophils produce potent byproducts that can directly or indirectly damage cells and include elastase, MPO, and NETs.[34](#hep41332-bib-0034){ref-type="ref"}, [35](#hep41332-bib-0035){ref-type="ref"}, [36](#hep41332-bib-0036){ref-type="ref"}, [37](#hep41332-bib-0037){ref-type="ref"} IL‐8‐activated neutrophils have also been shown to promote fibrosis through activation of stellate cells. Activated hepatic stellate cells (myofibroblasts) are a major source of extracellular matrix production leading to progressive liver fibrosis (Fig. [3](#hep41332-fig-0003){ref-type="fig"}).[41](#hep41332-bib-0041){ref-type="ref"}, [42](#hep41332-bib-0042){ref-type="ref"} There is a paucity of data on the role of neutrophils in bile duct injury in BA that warrants further research based on the findings that elastase and NETs correlated with rising bilirubin post‐HPE. The small number of patients with BA analyzed in this study prohibits us from making any broad conclusions regarding the use of IL‐8 as a biomarker of outcome in BA, and future studies should be appropriately powered in order to determine the accuracy of IL‐8 as a biomarker of severity of disease. Currently, the only validated biomarker of outcome is the total bilirubin level at 3 months post‐HPE.[32](#hep41332-bib-0032){ref-type="ref"}
There were several limitations to this study. Most importantly, immunophenotyping was performed only on patients with BA who had received IVIg within 60 days post‐HPE. Whether or not IVIg had a direct effect on altering a patient's immunophenotype is unclear. However, based on the fact that there was no overall improvement in bilirubin at 90 days post‐HPE or in survival with the native liver compared to historical controls,[23](#hep41332-bib-0023){ref-type="ref"} it is doubtful that IVIg had any positive impact on disease progression or severity.
Second, because there were no control groups of other age‐matched liver diseases or normal infants, we were unable to definitively answer the question of whether or not the observed BA immune marker levels differed from controls. However, a review of the literature identified multiple studies that reported levels of IL‐8, elastase, and Tregs in healthy infants. Published literature on normal values of NK cell activation markers and NETs in infants was not found. The mean IL‐8 level of normal infants in the first week of life is \~13‐17 pg/mL (range, 0.4‐50) and remains at this low level throughout childhood.[43](#hep41332-bib-0043){ref-type="ref"}, [44](#hep41332-bib-0044){ref-type="ref"}, [45](#hep41332-bib-0045){ref-type="ref"} These published normal IL‐8 levels are much lower than the observed IL‐8 levels in patients with BA at all time points (means ranged from 163 pg/mL to 328 pg/mL; see Supporting Table [S1](#hep41332-sup-0002){ref-type="supplementary-material"}). The serum elastase levels of normal infants age 5 ± 0.65 months old was 64.1 ± 12.9 ng/mL.[46](#hep41332-bib-0046){ref-type="ref"} In our patients with BA, the serum elastase mean ranged from 300 ng/mL to 435.5 ng/mL at all time points (Supporting Table [S1](#hep41332-sup-0002){ref-type="supplementary-material"}). The almost 10‐fold increase in IL‐8 and elastase in patients with BA compared to historical normal controls warrants future research with appropriate comparison groups (age‐matched, other liver diseases, healthy children). Treg quantities in normal infants are similar to adults by 1 week of age, and Treg levels reported in normal infants are similar to those described here in BA.[47](#hep41332-bib-0047){ref-type="ref"} Despite similar levels of Tregs compared to historical controls, one could argue that Tregs should be higher than normal in the inflammatory state of BA; future research should focus on determining if Treg function is abnormal in BA.
Third, the immunophenotype in the peripheral blood may or may not reflect the inflammatory environment within the liver. Interestingly, published investigations on liver‐specific inflammation in BA have described increased and activated NK cells,[5](#hep41332-bib-0005){ref-type="ref"}, [17](#hep41332-bib-0017){ref-type="ref"}, [18](#hep41332-bib-0018){ref-type="ref"}, [19](#hep41332-bib-0019){ref-type="ref"} decreased Tregs,[19](#hep41332-bib-0019){ref-type="ref"}, [21](#hep41332-bib-0021){ref-type="ref"}, [22](#hep41332-bib-0022){ref-type="ref"}, [48](#hep41332-bib-0048){ref-type="ref"} and increased IL‐8[39](#hep41332-bib-0039){ref-type="ref"}, [40](#hep41332-bib-0040){ref-type="ref"} in the liver and/or biliary remnant in human and murine BA, mirroring our findings in peripheral blood. This study provides further evidence that many of the immune pathways that are altered at the time of diagnosis persist post‐HPE and may contribute to ongoing intrahepatic biliary injury and fibrosis. Inflammatory molecules associated with robust organ‐specific inflammation can "spill over" into the peripheral blood, altering the immunophenotype.[49](#hep41332-bib-0049){ref-type="ref"}, [50](#hep41332-bib-0050){ref-type="ref"} However, other immune markers identified as up‐regulated in the liver of patients with BA, such as the Th1 cytokines[4](#hep41332-bib-0004){ref-type="ref"}, [6](#hep41332-bib-0006){ref-type="ref"}, [7](#hep41332-bib-0007){ref-type="ref"}, [8](#hep41332-bib-0008){ref-type="ref"}, [9](#hep41332-bib-0009){ref-type="ref"} IL‐2, IFN‐γ, and TNF‐α, as well as IL‐17,[51](#hep41332-bib-0051){ref-type="ref"} were undetectable/low in the serum of patients with BA. This may be related to the level of assay detection for these cytokines or an effect from the IVIg. Future studies should determine correlations of serum and liver immune profiles to understand the significance of these inflammatory pathways over time.
In summary, loss of adequate regulation of inflammation and exaggerated NK cell, IL‐8, and neutrophil responses were associated with poor short‐term outcomes in BA. Future research should focus on these immune pathways in patients with BA in the absence of IVIg treatment and in comparison to control groups in order to determine the specific contributions of these immune markers to bile duct injury and fibrosis. Determination of key immune markers associated with poor outcome is necessary in order to develop targets for immunotherapy.
Supporting information
======================
######
######
Click here for additional data file.
######
######
Click here for additional data file.
######
######
Click here for additional data file.
FFF Enterprises (Temecula, CA) supplied and shipped the IVIg. Dr. Sharon Sen and Dr. Brent Palmer from the Department of Medicine Flow Cytometry Core at the University of Colorado contributed to the design and implementation of the flow cytometry experiments.
| {
"pile_set_name": "PubMed Central"
} |
The electronic version of this article is the complete one and can be found at: <http://f1000.com/prime/reports/b/5/2>
Introduction {#s01}
============
Flexibility and disorder are two different concepts. When it applies to a polypeptide chain that has hundreds of internal degrees of freedom, flexibility describes concerted changes that affect a few degrees of freedom, modifying the overall structure without destroying it. Disorder implies a lack of constraints on many or all the degrees of freedom of the chain and no permanent structure, but the flexibility of proteins is intrinsic, part of their function, and an essential feature of molecular recognition. Many X-ray structures, some going back to the early 1970s, illustrate how a protein can adjust its conformation while making specific interactions with a ligand. Disorder does occur in the test tube, as purified polypeptides are seen to lack a permanent structure. The concept of "intrinsically disordered proteins" (IDPs) assumes that the lack of structure also occurs in the cell, and that a disordered polypeptide is capable of specific molecular recognition and performs a viable biological function \[[@bib-001]-[@bib-007]\]. The evidence is currently scant for both assumptions. *In vivo*, most proteins are part of oligomeric assemblies and multi-component complexes, and the disorder observed with purified polypeptides *in vitro* may result from the absence of other components. On the other hand, disorder-order transitions are sometimes observed both in the crystal and in solution when two proteins form a complex. In such cases, accepted mechanisms of protein-protein recognition may account for observed kinetics of the association reaction, but they do not explain its specificity in the crowded environment of the cell. Nevertheless, disorder must occur *in vivo* when polypeptide chains are being synthesized, and it may represent a serious obstacle to the self-assembly of multi-component complexes. The concept of IDP provides no plausible model for that process, and we suggest that most, if not all, IDPs are in fact PWPs (proteins waiting for a partner) protected from promiscuous interactions by chaperones and subject to the quality control apparatus of the cell until they meet their cognate partners.
IDPs are (mostly) artifacts of current methods of protein production {#s02}
====================================================================
In the last twenty years, the great majority of proteins used in biophysical and structural studies have been over-expressed from cloned DNA fragments in *Escherichia coli* or another expression host. The procedure, standard in structural genomics, has obvious limitations in spite of its success. The target protein may be part of a hetero-complex or a multi-component assembly in the source organism, where it interacts with other polypeptide chains, nucleic acids, or prosthetic groups. These components are absent, or at least not over-expressed, in the expression host, and the target may not fold properly without making these interactions. The long tail segments present in many ribosomal proteins illustrate the case: they are disordered in the purified protein but fully ordered in the ribosome, where interactions with the RNA determine their conformation \[[@bib-008]-[@bib-009]\].
Genome-wide studies of protein-protein interactions by genetic (yeast two-hybrid) and analytical (tandem-affinity purification coupled to mass spectrometry) methods indicate that a majority of eukaryotic proteins are part of hetero-complexes coded by more than one gene. In the yeast *Saccharomyces cerevisiae*, at least 70% of the proteins involved in transcription and translation are known to be part of assemblies that contain an average of 4.7 components \[[@bib-010]\], and the list is still far from complete. When we launched the Orsay Yeast Structural Genomics pilot-program in 2001, we knew hetero-complexes to be a problem, though not to what extent. In fact, of the 208 *S. cerevisiae* open reading frames (ORFs) that we selected as targets, 75% were expressed at a satisfactory level in *E. coli*, but only 25% could be purified in a soluble form. Nearly half of those gave crystals of some sort, but few were suitable for structure determination \[[@bib-011]-[@bib-012]\]. The low yield of the purification procedure and the poor quality of the crystals suggested that many of our targets did not fold properly, so we deleted terminal segments that sequence-based procedures predicted to be disordered. The new constructs were often expressed at a higher level, but only one in four showed better solubility, and only seven yielded better crystals \[[@bib-013]\]. When the pilot-program was completed in 2005, it had produced a structure for 12 novel proteins, only 6% of the initial set. Yet, the ORF-by-ORF approach proved fruitful. The procedure developed for the pilot-program yielded many other X-ray structures in Orsay, and it could easily be adapted to prepare yeast hetero-complexes in the frame of the 3D-Repertoire and SPINE2-Complexes European programs \[[@bib-014]-[@bib-015]\].
Other structural genomics programs have had a similar experience on a much larger scale \[[@bib-016]\]. Expressing individual eubacterial or archaeal ORFs in *E. coli* often yields more soluble proteins (up to 50%) than we got with yeast. The great majority are homo-oligomers, and so were most of the yeast proteins we solved. Mammalian proteins, including human, do far worse: less than 10% express as soluble material. Moreover, very few of the mammalian structures determined by structural genomics programs (or in other labs for that matter) are of full-length proteins. Most are fragments, often single domains cut out of ORFs that are too large for expression in *E. coli* or *in vitro*. Splitting a mammalian ORF into putative domains yields many constructs that do not express into soluble proteins, and when some remain unfolded, it may just be due to all the intra- and inter-chain contacts that cannot be made.
Flexibility versus disorder in crystals and in macromolecular recognition {#s03}
=========================================================================
Although a crystal is definitely not the best place to find disorder, the first sequence-based methods to identify IDPs relied on features observed in crystal structures \[[@bib-017]-[@bib-018]\]. In a Protein Data Bank (PDB) entry, residues that are present in the sequence, but not the coordinate set, count as disordered, but to a crystallographer 'disorder' only means that the electron density is low, and its atomic interpretation uncertain. The corresponding atoms either have a high B-factor or are reported as 'missing'. The B-factor measures the mean-square fluctuation of the atomic position: an atom that moves by 1.25 Å has B≈120 Å^2^ and a weak electron density. In the PDB, only 3% of the protein atoms have such high B-factors, because when a side chain or a chain segment has a weak density it usually counts as 'missing', even though the amplitude of its movement may be less than the length of a covalent bond. A low electron density can also mean that the atom occupies several discrete positions, but this is rarely reported: in the PDB, alternate positions concern only 0.8% of all protein atoms, almost all of them side chain atoms. A chain segment with two conformations is likely to be 'missing', albeit far from a state of intrinsic disorder. Even when a whole domain is 'missing', there may be no actual disorder. An early example is Kol, an immunoglobulin that forms crystals in which only the antigen-binding Fab moieties are in contact. The Fc moieties are free to move in the empty space in between, and they lack electron density even though they are fully structured \[[@bib-019]\]. The linker peptide, a short polyproline II helix, is flexible but not disordered either \[[@bib-020]\].
Kol illustrates how flexibility has been part of protein crystallography almost from the beginning, and its functional importance was soon recognized \[[@bib-021]\]. Whereas this is now commonplace, other cases dating from the same early period are still worth citing: hemoglobin, where flexibility is required for the allosteric transition, and the NAD-dependent dehydrogenases. X-ray structures determined in the 1970s show how the dehydrogenases change from an open conformation in the absence of the coenzyme, to a closed one in its presence \[[@bib-022]-[@bib-024]\]. The transition, which involves movements of flexible loops and/or hinge rotations of domains and subunits, remodels the active site and allows the coenzyme to enter and leave. Thus, it must play an essential part in the catalytic cycle.
A decade before any 3D structure was known, Koshland \[[@bib-025]-[@bib-026]\] had predicted enzymes to be flexible and offered substrate-induced conformation changes as the answer to the question: how does hexokinase manage to transfer the gamma-phosphate of ATP to a sugar hydroxyl and not to water, equally reactive and much more abundant? X-ray structures have shown the prediction to be correct for hexokinase \[[@bib-027]\], and for many other enzymes. Lysozyme, ribonuclease A, and chymotrypsin, initially seemed to prove Koshland wrong, but we now know that their apparent rigidity is the exception, not the rule (and the requirement for excluding water does not hold for hydrolases). Moreover, the flexibility of chymotrypsin was soon established by a structure of its precursor chymotrypsinogen \[[@bib-028]\], from which it differs by the cleavage of a single peptide bond. The cleavage induces main chain movements throughout the molecule, including the active site and the substrate binding pocket. The related trypsin/trypsinogen system also displays a large change in conformation, and, interestingly, two competing sets of X-ray structures describe it in different ways, possibly as alternative interpretations of a weak density: Felhammer et al. \[[@bib-029]\] see disordered loops in trypsinogen becoming ordered in trypsin, where Kossiakoff et al. \[[@bib-030]\] describe movements between defined positions.
Trypsinogen also illustrates the role of flexibility (or disorder-order transitions) in macromolecular recognition: it becomes fully ordered and trypsin-like, when it binds the pancreatic trypsin inhibitor \[[@bib-031]\]. In general, flexibility shows up as conformation changes when comparing two X-ray structures obtained with and without a ligand; the ligand can be anything from H^+^ or a metal ion to DNA or another protein. Disorder-order transitions are less common, and the disorder may only be apparent. An early example concerns DNA recognition by the lactose operon repressor (LacR), a 154 kDa tetramer. One-dimensional proton nuclear magnetic resonance (NMR) spectra, albeit unresolved as expected for a protein this size, contained narrow lines that could be attributed to the DNA-binding 'headpiece' (residues 1-61) \[[@bib-032]-[@bib-033]\]. The headpiece is folded, but flexibly connected to the protein body. It is also mobile in crystals, and it takes a fixed position only in the presence of the cognate DNA \[[@bib-034]\]. As a result, the PDB reports it as 'missing' in the free repressor (entry 1LBI) but present in the DNA complex (entry 1EFA). Here again, flexibility implies no disorder, and its functional role is obvious: the headpiece can orient itself relative to the DNA double helix much faster than the rotational diffusion of the whole tetramer would allow. This would be useless if it was not properly folded.
Modeling rigid and flexible recognition {#s04}
=======================================
Rigid body macromolecular recognition accounts for the high stability and specificity of antigen-antibody, enzyme-inhibitor, and many other types of protein-protein complexes. Its mechanism is relatively well understood: two complementary protein surfaces come into contact to form an interface that typically involves 24 residues and buries 800 Å^2^ of protein surface on each component \[[@bib-035]-[@bib-036]\]. In such systems, docking algorithms that simulate the association of the free components generally yield good quality models of the assembly \[[@bib-037]-[@bib-039]\]. These algorithms take into account a number of properties, including electrostatics, but shape recognition is their essential criterion. Their performance degrades quickly when the molecules change conformation, and then flexibility must be simulated in order to generate acceptable solutions \[[@bib-040]-[@bib-041]\].
The kinetics of association are rather simple in the absence of conformation changes. A single bimolecular step is usually observed, and the rate constant (k~on~) is in the range 5.10^4^-5.10^8^ M^-1^s^-1^ \[[@bib-042]\], compatible with a simple diffusion-collision mechanism. The lower bound of the range corresponds to random collisions that yield a stable complex if, and only if, the proper regions of the two protein surfaces happen to face each other. The lock-and-key model requires in principle the two binding patches to be perfectly positioned and oriented. It effectively predicts k~on~=0 but with more reasonable assumptions on the geometry of the transition state, k~on~ evaluates to 10^5^-10^6^ M^-1^s^-1^ \[[@bib-043]-[@bib-044]\], and most enzyme-inhibitor and antigen-antibody complexes have binding rates in this range. LacR binds DNA much faster than this, but it undergoes facilitated one-dimensional diffusion along the double helix, a mechanism applicable only to DNA recognition \[[@bib-045]-[@bib-046]\]. Long-range electrostatic interactions modulate binding rates in a way that can be modeled from the charge distribution on the protein surfaces \[[@bib-044],[@bib-047]-[@bib-049]\], and that quantitatively explains most of the larger k~on~ values reported in \[[@bib-042]\].
Flexibility adds a level of complexity to the binding process. Conformation changes and disorder-to-order transitions are expected to make association slower, but in a way that is difficult to model. A plausible mechanism of flexible recognition is conformer selection: a fraction of the receptors pre-exist in the correct conformation, and only those can bind the ligand ('receptor' and 'ligand' are here for convenience only). An alternative is induced fit: most, if not all, of the receptor molecules are able to form a low affinity complex with the ligand, and the interaction promotes the conformation changes that yield the stable assembly. Both mechanisms predict the binding kinetics to be biphasic, but the first order step (the conformation change) is often fast on the time scale of the experiment, and only one phase is detected. Its rate should be proportional to the fraction of the receptors that have the correct conformation, if conformer selection is the dominant mechanism, and to the probability that the intermediate evolves into the product before it dissociates, if induced fit applies.
Kinetic data are available on many systems that involve conformation changes and a few that display disorder-to-order transitions. Conformer selection can often be excluded. For instance, the NAD-dependent dehydrogenases must be in an open conformation when they bind the coenzyme, which they do at nearly diffusion-limited rates. On the other hand, conformer selection certainly contributes to protein-protein recognition when the conformation changes are of limited amplitude \[[@bib-049]-[@bib-051]\]. However, the observed binding rates either imply that native-like conformations are highly populated to start with, or that induced fit coexists with conformer selection. Thus, induced fit must be the dominant mechanism when trypsinogen binds pancreatic trypsin inhibitor. The affinity of the precursor for pancreatic trypsin inhibitor is eight orders of magnitude less than for trypsin, due to k~off~ increasing by six orders while k~on~ decreases by only two \[[@bib-052]-[@bib-053]\]. Conformer selection would require 1% of the trypsinogen molecules to preexist in a trypsin-like conformation, whereas the actual fraction is estimated to be less than one in a million.
![The p27^Kip1^-Cdk2-cyclin A ternary complex\
In the crystal structure \[[@bib-055]\] (Protein Data Bank entry 1JSU), the kinase inhibitory domain of p27^Kip1^ (KID, in green) is observed to be partly helical and interact with both the kinase and the cyclin. When it is free in solution, KID is mostly disordered, but some of the helical structure is already present \[[@bib-054]\], making conformer selection a plausible alternative to induced fit as an explanation of the rapid association with the cyclin.](biolrep-05-02-g001){#fig-001}
Conformer selection may also involve (partly) disordered proteins. An example is the kinase inhibitory domain (KID) of the p27^Kip1^ cyclin-dependent inhibitor. In solution, KID contains a significant amount of α-helix detected by NMR and circular dichroism \[[@bib-054]\]. In crystals of the ternary complex with Cdk2 and cyclin A ([Figure 1](#fig-001){ref-type="fig"}), its N-terminal half is partly helical and interacts with the cyclin, whereas the C-terminal half forms an open loop in contact with the kinase \[[@bib-055]\]. KID has nanomolar affinity for either Cdk2 alone or cyclin A alone and remarkable binding kinetics: k~on~ is low (5.10^3^ M^-1^s^-1^) for Cdk2 alone, and high (1.6 to 3.10^6^ M^-1^s^-1^) for cyclin A alone. It is also high with the Cdk2-cyclin complex, but then the reaction is biphasic and the second step as slow as for Cdk2 alone \[[@bib-054]\]. Albeit compatible with induced folding, the kinetics suggest a conformer selection mechanism by which the cyclin quickly associates with the many KID molecules that contain a helical fragment, while the C-terminal conformation recognized by the kinase is very rare.
Disorder in vivo: does it exist, and how does the cell deal with it? {#s05}
====================================================================
But what is the actual state of KID in the living cell? There, the overall protein concentration reaches hundreds of grams per liter, orders of magnitude above the concentrations of the purified proteins in test tube experiments \[[@bib-056]\]. As a result, proteins disordered *in vitro* may become partially ordered, and this can be tested in the test tube by adding molecular crowding agents. These agents have little effect on KID \[[@bib-057]\], but FlgM, a 97-residue polypeptide that binds the transcription factor σ^28^, gains structure in their presence. In dilute solution, FlgM is disordered except for transient α-helices in its C-terminal half. This half becomes fully ordered upon binding to σ^28^, while the N-terminal half remains disordered \[[@bib-058]-[@bib-059]\]. Adding a high concentration of other proteins (bovine serum albumin or ovalbumin), or glucose, induces structure in the C-terminal, but not the N-terminal half. Remarkably, in-cell NMR shows that the polypeptide over-expressed in *E. coli* also has an ordered C-terminal and a disordered N-terminal half. As *E. coli* σ^28^ is not over-expressed, an interaction with it cannot explain that transition, and it may be induced by the crowded environment in the cell \[[@bib-060]\].
If the disorder seen *in vitro* for KID, FlgM and other putative IDPs effectively occurs in the cell, it must affect the stability, the kinetics and the specificity of the interactions that mediate the function of these polypeptides. A conformation change or a disorder-order transition costs free energy, and it should make the assembly less stable. It does in trypsinogen/pancreatic trypsin inhibitor relative to trypsin/pancreatic trypsin inhibitor, but in general flexible recognition is associated with the formation of large interfaces \[[@bib-035]\], and the additional interactions must offset that cost. Thus, KID loses over 2800 Å^2^ of accessible surface area in contact with Cdk2-cyclin A, four times as much as an antibody in contact with the cognate antigen. The kinetic constraints could be of more consequence: the low k~on~ of KID for binding Cdk2 (as opposed to Cdk2-cyclin A) predicts the binary complex to form in about an hour at a KID concentration of 10^-7^ M, and such a long lag is probably not compatible with its inhibitory function.
However, the most significant constraint in recognition is specificity. A disordered polypeptide chain has no defined shape, and it contains the same chemical groups as all other proteins, positioned more or less at random in space. How can it recognize, or be recognized by, another biomolecule? In a test tube experiment, the cognate interactions have no (or very few) competitors; *in vivo* they have thousands or millions. A linear sequence motif, or the presence of modified residues (phosphorylation, for instance), can serve as identification in some cases, but in general disorder must imply promiscuity, and be incompatible with all but a few cellular functions.
Another argument against this is that promiscuous interactions can be toxic \[[@bib-061]\], and cells have efficient quality control mechanisms designed to prevent them and to degrade or sequester misfolded polypeptides. Artificial conditions, such as over-expression in *E. coli*, may allow putative IDPs to escape quality control, but the problem of handling disorder in the cell is more general. It concerns all nascent polypeptide chains, and, most of all, those that will form oligomers. Nascent polypeptides are protected by chaperone proteins as they exit the ribosome, or sequestered by chaperonins, such as GroES/GroEL in bacteria, until their folding is completed. How they assemble to form oligomers is not understood at present. Their subunits are often unstable *in vitro*, and, whether folded or partially unfolded, they carry large hydrophobic surface patches that are prone to non-specific interactions. In the cell, their concentration must be kept low, and their assembly cannot be fast because it is a second or higher order reaction. With hetero-complexes, stoichiometry raises an additional question. It cannot be exact when subunits are independently synthesized on the ribosome, and the component in excess is a source of promiscuous interactions. Here again, the cell protects itself by using chaperones and protein degradation. Hemoglobin is an example: in beta-thalassemia, the alpha-chains are produced in excess of the beta-chains. They cannot form homo-tetramers (the beta-chains do) and are unstable, but a specialized chaperone prevents them from releasing heme and damaging the red blood cells \[[@bib-062]\]; in precursor cells, excess alpha-chains are polyubiquitinated and degraded by the proteasome \[[@bib-063]\].
Conclusion {#s06}
==========
The hemoglobin alpha-chain is not an IDP, it is a PWP, and, as such, it represents a very common situation. We contend that most, if not all, putative IDPs are in fact PWPs. They are unfolded in the test tube, but *in vivo* they are folded and part of a multi-component assembly (possibly of more than one). In general, molecular disorder is not compatible with function. A partly disordered polypeptide may be capable of specific recognition through a conformer selection mechanism, but then it is the ordered population that reacts, and the disorder is neither intrinsic nor functional. While disorder cannot be entirely avoided in the cell, it remains transient, and it is kept to a minimum by sophisticated mechanisms of biosynthesis and quality control. The mechanism that limits the damage an improper assembly of hemoglobin can cause in beta-thalassemia is probably one of many. Research in the field is very active and highly relevant to human health, and we may expect more to be discovered in coming years.
We thank Dr. Sameer Velankar (Hinxton, UK) for the statistics on atomic B-factors and occupancies in the Protein Data Bank.
Disclosure
==========
MJES is Director and Shareholder of Equinox Pharma Ltd, which is involved in the commercialization and exploitation of chemoinformatics and bioinformatics software.
IDP
: intrinsically disordered protein
KID
: kinase inhibitory domain
LacR
: lactose operon repressor
NMR
: nuclear magnetic resonance
ORF
: open reading frame
PDB
: Protein Data Bank
PWP
: proteins waiting for a partner
| {
"pile_set_name": "PubMed Central"
} |
Introduction {#hyz084s5}
============
Platinum-based combination chemotherapy is the mainstay of treatment for metastatic urothelial carcinoma (UC), and the gemcitabine plus cisplatin regimen has been the most widely used as the first-line treatment. In general, UC is a chemo-sensitive tumor. Good initial response rates of around 50--70% have been reported. However, the majority of patients show disease relapse during the follow-up after the completion of first-line chemotherapy, and salvage regimens have been tested, including combination regimens or single agents, worldwide ([@hyz084C1]--[@hyz084C3]). Although the recent development of immunotherapy with check-point inhibitors, such as atezolizumab ([@hyz084C4]), pembrolizumab ([@hyz084C5]), nivolumab ([@hyz084C6]), durvalumab ([@hyz084C7]), or avelumab ([@hyz084C8]), has changed the treatment paradigm for metastatic UC, optimizing the use of an effective regimen may be a key factor to improve outcomes.
Maintenance therapy for patients showing a good response or disease stabilization after systemic chemotherapy has been introduced for several tumors, such as non-small cell lung cancer ([@hyz084C9]). In our hospital and affiliated teaching hospitals, aiming at long-term disease control, systemic chemotherapy was continued with a 1--3-month drug-free interval for selected patients who achieved disease control. In the present study, we report our experience of this maintenance systemic chemotherapy (MSC) strategy for metastatic UC patients.
Materials and methods {#hyz084s6}
=====================
The present retrospective study was approved by the institutional review board. A total of 228 metastatic UC patients treated with at least two cycles of systemic chemotherapy between 2000 and 2013 at Hokkaido University Hospital and 6 affiliated teaching hospitals were included. Data on the patient characteristics, details of treatments such as chemotherapy regimens or numbers of chemotherapy cycles performed, and overall survival outcomes were retrospectively collected. Based on this cohort, we previously published a paper evaluating prognostic factors in real-world clinical practice in Japan ([@hyz084C10],[@hyz084C11]). Our general treatment strategy was reported in a previous study. Briefly, in the early study period, the MEC regimen (methotrexate, epirubicin, and cisplatin), which was accepted as an alternative to MVAC in Japan based on a prospective randomized study showing a similar response rate and incidence of adverse effects ([@hyz084C12]), was utilized as the first-line regimen. In the later period, the GC regimen (gemcitabine and cisplatin) was selected. In patients refractory to first-line chemotherapy, a salvage regimen such as a taxane-based combination regimen was considered. In patients with an impaired renal function, dose reduction was considered, as previously reported ([@hyz084C13]), or cisplatin was replaced with carboplatin. In selected patients with oligometastasis, which meant metastasis in a single organ with a small number of metastases (e.g. single pulmonary metastasis), a good performance status, and stabilization of disease, surgical consolidation was also considered. The objective response was evaluated by the treating physician according to the Response Evaluation Criteria in Solid Tumors, version 1.1, in most cases.
In selected patients showing disease control, systemic chemotherapy was intentionally continued while extending the interval of treatment, named 'maintenance systemic chemotherapy (MSC)', after discussion between patients and physicians. They were the main cohorts in the current study. The reasons for the discontinuation of MSC were newly collected for the present analysis.
Statistical methods {#hyz084s7}
===================
Patient characteristics between MSC and non-MSC cohorts were compared using the Mann--Whitney *U* test. Overall survival (OS) was estimated from the initiation of treatment for metastatic UC or the initiation of MSC until death or the last follow-up. The log-rank test was used to determine the significance of differences between survival estimates. The Cox proportional hazards model was also utilized to identify prognostic characteristics. The parameters analyzed were sex, age, ECOG-performance status (PS), primary site, histology of primary site, hemoglobin (Hb) level, lactate dehydrogenase level, C-reactive protein level, corrected calcium level, estimated glomerular filtration rate level, history of prior chemotherapy, resection of the primary site, each metastatic site (lymph node, lung, liver, bone, local recurrence, visceral metastasis \[lung, liver, or bone\]), and number of metastatic organs.
Because of the heterogeneity of patient backgrounds between MSC and non-MSC cohorts, propensity score matching was also utilized to adjust for the confounding factors in order to select patients for MSC. A logistic regression model, which included age (continuous), sex, ECOG PS, status of primary site (resected or not), metastatic sites (presence of lymph node, lung, bone, liver, local recurrence, or absence), number of metastatic organs (single or multiple), and baseline renal function (fit or unfit), was used to estimate each patient's probability of receiving MSC. Patients without MSC were matched on a one-to-one basis with patients with MSC based on nearest-neighbor matching. All calculations were performed using JMP version 12.2.0. A value of *P* \< 0.05 was considered significant.
Results {#hyz084s8}
=======
Table [1](#hyz084TB1){ref-type="table"} shows patient characteristics according to the receipt/non-receipt of MSC. The MSC group showed a younger age (median age, years: MSC 63, non-MSC 67.5, *P* = 0.044), more frequent resection of the primary site (MSC 67.5%, non-MSC 50%, *P* = 0.0418), and a better PS (PS0: MSC 87.5%, non-MSC 70.2%, *P* = 0.0844) at the time of initiating systemic chemotherapy. In terms of the response after first-line chemotherapy, the majority of patients in the MSC cohort showed at least stable disease, while 35.1% showed progressive disease in the non-MSC cohort.
######
Patient characteristics
MSC cohort, *n* = 40 Non MSC cohort, *n* = 188 p-value
-------------------------------------------------------- ---------------------------------- ---------------------------------- ---------
**Age, year** median 63 (range, 42--80) median 67.5 (range, 30--83) 0.044
**Sex male / female**
Male 28 (70%) 146 (77.7%) 0.3111
Female 12 (30%) 42 (22.3%)
**ECOG performance status** 0.0844
0 35 (87.5%) 132 (70.2%)
1 2 (5%) 36 (19.1%)
2 2 (5%) 7 (3.7%)
3 0 2 (1.1%)
Unknown 1 (2.5%) 11 (5.9%)
**Primary site**
Bladder 21 (52.5%) 90 (47.9%) 0.6108
Upper urinary tract 17 (42.5%) 81 (43.1%)
Both 1 (2.5%) 14 (7.4%)
Urethra/prostate 1 (2.5%) 3 (1.6%)
**Pathology of primary site**
Pure urothelial carcinoma 33 (82.5%) 142 (75.5%) 0.507
Others 4 (10%) 32 (17%)
Unknown (cytology positive) 3 (7.5%) 14 (7.4%)
**Baseline laboratory data**
Hemoglobin, g/dL (*n* = 224) median 12.25 (range, 9.2--15.2) median 12.1 (range, 7.3--17.8) 0.859
Lactic dehydrogenase, IU/L (*n* = 225) medain 178 (range, 124--996) medain198 (range, 105--1154) 0.0608
CRP, mg/dL (*n* = 223) median 0.38 (range, 0.02--9.43) median 0.5 (range, 0.01--19.87) 0.67
Corrected calcium, mg/dL (*n* = 209) median 9.4 (range, 4.4--10.8) median 9.5 (range, 4.1--11.7) 0.5197
Estimated GFR (eGFR), mL/ min./ 1.73 m^2^ (*n* = 224) median 61.0 (range, 34.3--122.7) median 56.2 (range, 21.2--130.1) 0.4093
eGFR (*n* = 224)
Fit (≥60 mL/min./1.73 m^2^) 22 (55%) 80 (42.6%) 0.1856
Cisplatin-unfit (\<60 mL/min./1.73 m^2^) 18 (45%) 104 (55.3%)
**Primary site at the initiation of chemotherapy**
Resected 27 (67.5%) 94 (50%) 0.0418
Not resected 13 (32.5%) 94 (50%)
**Metastatic site**
Lymph node 24 (60%) 127 (67.6%) 0.364
Lung 18 (30%) 68 (36.2%) 0.2994
Bone 11 (27.5%) 34 (18.1%) 0.1886
Liver 5 (12.5%) 15 (8.0%) 0.3793
Local recurrence 3 (7.5%) 16 (8.5%) 0.8316
**Visceral metastasis (lung, liver, or bone)**
Yes 25 (62.5%) 94 (50%) 0.1485
No 15 (37.5%) 94 (50%)
**Single organ metastasis** 20 (50%) 124 (66%) 0.061
**Response after first-line chemotherapy**
CR 5 (12.5%) 23 (12.2%) 0.0002
PR 18 (45%) 55 (29.3%)
SD 15 (37.5%) 39 (20.7%)
PD 2 (5%) 66 (35.1%)
Unknown 0 5 (2.7%)
MSC = maintenance systemic chemotherapy.
Table [2](#hyz084TB2){ref-type="table"} shows a summary of MSC. Thirty patients (75%, 30/40) underwent MSC following first-line chemotherapy, and 10 (25%, 10/40) patients following salvage chemotherapy. The median number of chemotherapy cycles was 6, and the responses were CR in 6 patients (15%, 6/40), PR in 19 patients (47.5%, 19/40), SD in 14 patients (35%, 14/40), and PD in 1 patient (2.5%, 1/40) before MSC introduction. Gemcitabine plus CDDP or carboplatin was mainly performed as MSC (70%, 28/40). MSC was repeated quarterly in 30 patients (75%, 30/40), every 2 months in 8 patients (20%, 8/40), and with other intervals in 2 patients (5%, 2/40). Overall, a median of 3 cycles (range: 1--29) of MSC were performed. The reason for the discontinuation of MSC was PD in 24 patients (60%, 24/40), favorable disease control in 9 patients (22.5%, 9/40), and myelosuppression in 3 patients (7.5%, 3/40), and for other reasons in 2 patients (5%, 2/40). MSC was ongoing in 2 patients (5%, 2/40). In 15 patients (37.5%, 15/40), salvage chemotherapy treatment was performed following MSC.
######
Summary of maintenance chemotherapy
--------------------------------------------------------------------- -------------------------
**MSC was started following**
First-line 30
Second-line 8
Third-line 2
**Chemotherapy cycles performed before MSC** Median 6 (range, 2--15)
**Response before induction of MSC**
CR 6
PR 19
SD 14
PD 1
**Regimens used as MSC**
Gemcitabine plus CDDP or carboplatin 28
Methotrexate plus epirubicin plus CDDP or nedaplatin (CDDP analog) 8
Paclitaxel plus ifosphamide plus nedaplatin 3
Gemcitabine monotherapy 1
**Interval of MSC**
Every 3 months 30
Every 2 months 8
2--4 months 2
**Chemotherapy cycles performed as MSC** Median 3 (range, 1--29)
**Reason for discontinuation of MSC**
PD 24
Disease stabilization 9
Myelosupression 3
Patient's wish 1
Death due to other cause 1
On MSC 2
**Treatment after cessation of MSC**
BSC/follow up 19
Salvage chemotherapy 15
Participation in clinical trial 2
Radiation 1
--------------------------------------------------------------------- -------------------------
Figure [1](#hyz084F1){ref-type="fig"} shows overall survival curves from the initiation of treatment for metastatic UC. Overall, the median OS was 39 months from the initiation of treatment for metastases in the MSC cohort, as compared with 14 months in the non-MSC cohort (Figure [1](#hyz084F1){ref-type="fig"}a, *P* \< 0.0001). As for the patients showing CR/PR/SD after first-line chemotherapy (*n* = 155), MSC was still associated with longer survival (median OS: MSC cohort; 39 months, non-MSC cohort; 20 months, *P* = 0.0195, Figure [1](#hyz084F1){ref-type="fig"}b). To further improve compatibility, propensity score matching was utilized in the patients showing CR/PR/SD after first-line chemotherapy. Table [3](#hyz084TB3){ref-type="table"} shows patients characteristics after propensity score adjustments. The patient distributions were closely balanced between the two cohorts. Median OS was 37 months in the MSC cohort and 19 months in the non-MSC cohort after propensity score matching (Figure [1](#hyz084F1){ref-type="fig"}c, *P* = 0.0573).
######
Characteristics after propensity score matching of the patients showing CR/PR/SD after first-line chemotherapy
*n* = 36 with MSC *n* = 36 without MSC p-value
----------------------------------------------------- --------------------------------- --------------------------------- ---------
**Age, year** median 64 (range, 42--80) median 67.5 (range, 45--78) 0.4463
**Sex male / female**
Male 26 23 0.4478
Female 10 13
**ECOG performance status**
0 32 29 0.4781
1 2 5
2 2 2
**Primary site**
Bladder 19 17 0.5803
Upper urinary tract 15 17
Both 1 0
Urethra/prostate 1 2
**Pathology of primary site**
Pure urothelial carcinoma 29 32 0.5124
Others 4 3
Unknown (cytology positive) 3 1
**Baseline laboratory data**
Hemoglobin, g/dL (*n* = 72) median 12.4 (range, 9.2--15.2) median 12.45 (range, 8.6--16.5) 0.4107
Lactic dehydrogenase, IU/L (*n* = 72) medain 178 (range, 124--699) medain 202 (range, 150--441) 0.0977
CRP, mg/dL (*n* = 71) median 0.46 (range, 0.02--9.43) median 0.43 (range, 0.01--7.18) 0.7423
Corrected calcium, mg/dL (*n* = 66) median 9.45 (range, 4.4--10.8) median 9.45 (range, 7.5--10.8) 0.9282
Estimated GFR (eGFR), ml/min./1.73 m^2^ (*n* = 72) median 60.4 (range, 34.3--85.7) median 61.3 (range, 28.5--99.3) 0.9686
eGFR (*n* = 72)
Fit (≥60 mL/min./1.73 m^2^) 19 19 1
Cisplatin-unfit (\<60 mL/min./1.73 m^2^) 17 17
**Primary site at the initiation of chemotherapy**
Resected 23 26 0.4478
Not resected 13 10
**Metastatic site**
Lymph node 22 21 0.8101
Lung 16 18 0.6367
Bone 10 8 0.5859
Liver 4 4 1
Local recurrence 3 5 0.4511
**Visceral metastasis (lung, liver, or bone)**
Yes 22 24 0.6235
No 14 12
**Single organ metastasis** 17 16 0.813
**Response after first-line chemotherapy**
CR 5 9 0.0875
PR 17 21
SD 14 6
{#hyz084F1}
Figure [2](#hyz084F2){ref-type="fig"}a shows an OS estimate from the initiation of MSC. The median OS was 27 months from the initiation of MSC. Regarding the survival impacts of baseline clinical characteristics, PS0 (*P* = 0.0169), the absence of lung metastasis (*P* = 0.0387), and resection of the primary site (*P* = 0.0495) were associated with long-term survival after the initiation of MSC (Table [4](#hyz084TB4){ref-type="table"}). None of these factors retained prognostic significance on multivariate analysis, although lung metastasis and the performance status showed marginal values (Table [5](#hyz084TB5){ref-type="table"}). Figure [2](#hyz084F2){ref-type="fig"}b shows the OS curves from the initiation of MSC divided by the timing of maintenance initiation. There was no significant difference in survival between the two cohorts (first-line vs. second /third --line, *P* = 0.8041).
{#hyz084F2}
######
Univariate analysis of prognostic factors after the initiation of MSC
n Median survival time (95% CI) P-value
---------------------------------------------------- ---- ------------------------------- ---------
**Age, year**
≥67 16 24 (9-NR) 0.9749
\<67 24 30 (17--58)
**Sex**
Male 28 24 (16--33) 0.0856
Female 12 NR (8-NR)
**ECOG performance status**
PS 0 35 32 (18--58) 0.0169
PS 1 4 12.5 (3-NR)
**Primary site**
Primary, bladder 21 32 (17--77) 0.3189
Others 19 23 (9--48)
**Pathology of primary site**
Pure urothelial carcinoma 33 30 (17--50) 0.707
Others 4 NR (7-NR)
**Baseline laboratory data**
Hemoglobin, \<10 g/dL 3 NR (24-NR) 0.4775
≥10 g/d+ 37 27 (16--48)
LDH, ≥200 IU/L 13 50 (8-NR) 0.5546
\<200 IU/L 27 24 (16--36)
CRP, ≥1 mg/dL 12 NR (11-NR) 0.0532
\<1 mg/dL 28 24 (16--36)
Corrected Ca, ≥10 mg/dL 3 17 (7-NR) 0.9172
\<10 mg/dL 35 27 (16--50)
eGFR, fit (≥60 mL/min./1.73 m^2^) 22 32 (17--58) 0.8792
Unfit (\<60 mL/min./1.73 m^2^) 18 23 (7-NR)
**Prior chemotherapy**
Yes 4 NR (23-NR) 0.0529
No 36 24 (16--36)
**Primary site at the initiation of chemotherapy**
Resected 27 33 (18--77) 0.0495
Not resected 13 17 (7--36)
**Metastatic site**
Lymph node, yes 24 27 (16--50) 0.8226
No 16 33 (10--77)
Lung, yes 18 24 (17--33) 0.0427
No 22 48 (17-NR)
Bone, yes 11 58 (11-NR) 0.3293
No 29 27 (16--36)
Liver, yes 5 30 (7-NR) 0.7923
No 35 27 (17--50)
Local, yes 3 10 (7-NR) 0.7371
No 37 30 (17--50)
Visceral metastasis (lung, liver, or bone), yes 25 27 (17--50) 0.4053
No 15 32 (12-NR)
**Single-organ metastasis**
Yes 20 33 (18--58) 0.2633
No 20 17 (7--50)
CI = confidence interval, NR = not reach
######
Multivariate analysis of prognostic factors after the initiation of MSC
No. of patients Hazard ratio (95% CI) p-value
------------------------------------------------ ----------------- ----------------------- ---------
ECOG performance status
PS 0 35 1 0.0601
PS 1 4 4.476 (0.930--16.88)
Primary site at the initiation of chemotherapy
Resected 27 1 0.306
Not resected 13 1.571 (0.652--3.654)
Lung metastasis
No 22 1 0.0535
Yes 18 2.2357 (0.988--5.303)
Discussion {#hyz084s9}
==========
In order to maintain the response to chemotherapy and delay disease progression, we continued the systemic chemotherapy in selected patients mainly with at least stable disease after the first-line systemic chemotherapy, when the patients agreed with the present maintenance strategy of administrating the effective agents with drug holidays. Overall, the median OS was 39 months in the MSC cohort, as compared with 14 months in the non-MSC cohort (Figure [1](#hyz084F1){ref-type="fig"}a, *P* \< 0.0001). After propensity score matching in the patients showing CR/PR/SD after first-line chemotherapy, the median OS was 37 months in the MSC cohort and 19 months in the non-MSC cohort (Figure [1](#hyz084F1){ref-type="fig"}c, *P* = 0.0573). Our observation reflected the treatment outcome of real-world clinical practice, not a clinical trial, and a well-controlled randomized study is necessary to determine the clinical benefit of the present maintenance strategy. However, our observation suggested that long-term systemic chemotherapy could be performed with a drug-free interval, and the maintenance of cytotoxic drugs could be one of the treatment options for long-term disease control. Gemcitabine plus CDDP or carboplatin was dominantly utilized in an MSC setting (*n* = 28), usually every 3 months (2-month drug holiday). Recently, we routinely replace CDDP with carboplatin to minimize the accumulation of renal toxicity when considering MSC.
A maintenance strategy is not a new concept for metastatic urothelial cancer treatment. Grivas et al. investigated the role of sunitinib maintenance in patients with advanced UC showing stable disease or a partial or complete response after 4 to 6 chemotherapy cycles. Participants were randomly assigned to sunitinib at a dose of 50 mg/day (4 weeks on and 2 weeks off) or placebo, and the primary endpoint was the 6-month progression rate. The study was prematurely closed due to poor accrual (sunitinib: *n* = 26, placebo: *n* = 28, predefined accrual goal: 42 participants per treatment arm), and maintenance sunitinib did not improve the 6-month progression rate (sunitinib: 71.7%, placebo: 64.3%) ([@hyz084C14]). Powles et al. also did not observe a clinical benefit of maintenance lapatinib (HER1 and HER2 tyrosine kinase inhibitor) in patients with HER-1 and HER-2 bladder cancer, who showed stable disease during 4 to 8 cycles of chemotherapy for advanced metastatic UC ([@hyz084C15]). In terms of chemotherapeutic agents, several previous studies suggested possible clinical activity. García-Donas et al. reported the outcomes of maintenance therapy with vinflunine ([@hyz084C16]). The 87 patients were included in their study after disease control with 4 to 6 cycles of a cisplatin and gemcitabine regimen and were randomly assigned to receive vinflunine every 3 weeks plus best supportive care, or best supportive care alone. The median progression-free survival of 6.5 months in the vinflunine group with an acceptable safety profile, which was significantly longer than the 4.2 months achieved in the best supportive care group (hazard ratio = 0.59, *P* = 0.031). Muto et al. reported their experiences of maintenance monotherapy with gemcitabine ([@hyz084C17]). A total of 33 patients underwent maintenance therapy after a mean of 2.7 courses of prior chemotherapy. Gemcitabine (1000 mg/m2) was administered on an outpatient basis every 4 weeks, and a median of 9 courses was administered. They observed that the median cancer-specific survival was 15 months after the induction of maintenance chemotherapy. Also in other malignancies, maintenance treatment has been performed using a drug different from that in the induction regimen, for example, maintenance olaparib after platinum-based chemotherapy in advanced ovarian cancer patients ([@hyz084C18]), or single agents of the induction regimen such as pemetrexed after pemetrexed plus cisplatin in advanced non-squamous non-small-cell lung cancer patients ([@hyz084C19]). Because we continued the combination regimen with the aid of drug holidays, our strategy might represent relative dose reduction. Regarding immune checkpoint inhibitors, 'Testing the PD-1 Inhibitor Pembrolizumab as Maintenance Therapy After Initial Chemotherapy in Metastatic Bladder Cancer (NCT02500121)' is ongoing.
In the present cohort, PS0 (*P* = 0.0169), the absence of lung metastasis (*P* = 0.0387), and resection of the primary site (*P* = 0.0495) were associated with long-term survival after the initiation of MSC (Table [4](#hyz084TB4){ref-type="table"}). These factors might be associated with maintaining a good health status, including the absence of local symptoms and a good respiratory function during MSC treatment, enabling the continuation of long-term systemic chemotherapy. When dividing the outcomes by the timing of maintenance initiation (first-line or second/third -line), we did not find any ignificant difference in survival between the two cohorts. Although the present cohort was very small, our observations suggest that the maintenance strategy could be utilized in a salvage regimen if at least stable disease is observed during treatment.
We recognize that our study was limited by its retrospective nature and small sample size. Assessment of the radiological response might not have been as strict as that in prospective clinical trials. We did not have data on adverse events or the quality of life during MSC treatment. We could not come to a conclusion regarding the appropriate indication for MSC chemotherapy, or its ideal duration. As described above, a future prospective study is needed to clarify the survival benefit of MSC. Nevertheless, we consider that several important findings were generated by the present study.
Conclusions {#hyz084s10}
===========
In the selected patients, long-term systemic chemotherapy could be performed with a 1--3-month drug-free interval. Our maintenance strategy with cytotoxic drugs may become one of the treatment options for long-term disease control.
Conflict of interest statement {#hyz084s11}
==============================
None declared.
| {
"pile_set_name": "PubMed Central"
} |
Introduction
============
School violence affects negatively both the learning en vironment and the role of educational institutions on the formation of values, attitudes and pro-social be haviors that allow children and adolescents to live with others based on tolerance, respect and equity ([@B15]; [@B41]). Studies in Mexico show that violence among students is a problem that affects the educational institutions of the country ([@B7]; [@B1]; [@B37]; [@B40]; [@B48]; [@B57]).
Bullying differs from other forms of aggression because of its systematic nature and its manifestation in the context of interpersonal relationships with marked asymmetries of power between the perpetrator and the victim ([@B47]; [@B62]). Among other negative effects, this problem leads to decreased academic performance, emotional disorders, and suicide attempts for victims ([@B36]; [@B38]). The attackers also are affected, as they show social adjustment difficulties during adolescence and adulthood, which are manifested in substance abuse, employment difficulties, violence in relationships and antisocial behavior ([@B18]; [@B24]; [@B39]; [@B42]; [@B56]).
This study assumes it is necessary to address bullying from an ecological perspective since it is considered to be the result of interactions in which students operates within different contexts. This implies that there are characteristics of the student, family, school and community that are risk factors for the presence of bullying ([@B23]; [@B54]).
Based on the preceding, this study investigated the relationship of variables located in the individual (challenging behavior), in the family (parental conflict) and community (community violence) with bullying in elementary school students. Although there are studies that address the relationship of these variables with bullying in the international literature ([@B5]; [@B9]; [@B26]; [@B32]; [@B34]) it should be noted that few studies address this topic in Mexico ([@B43]; [@B50]; [@B58]). Consequently, it is valuable to establish the relationship of these variables to bullying in the context of Mexican elementary schools.
Challenging behavior is manifested in a stable pattern of anger, violations of social rules, and direct or indirect confrontation with authority figures ([@B53]). The literature suggests that these behavioral problems relate positively with bullying ([@B11]; [@B49]; [@B59]) and that when these two problems overlap, aggressive behavior is more stable over time ([@B45]; [@B64]).
Parental conflict is related to parents showing less emotional availability and involvement with children ([@B33]; [@B52]; [@B63]), excessive psychological control and inconsistent discipline practices ([@B2]; [@B35]). This conflict, which has a negative effect on parenting practices, favors the presence of externalized behavior problems (aggression, bullying) and internalization (depression, anxiety) in children ([@B6]; [@B25]; [@B27]). It also damages adolescents\' and children\'s sense of family identity and safety, effects that are exacerbated when conflict is common, if it is handled in a hostile way and when it relates to or involves children ([@B13]; [@B29]).
Because interactions established within the community contribute to the internalization of expectations, values and social norms, the community exercises a major influence on the emotional development of adolescents. The influence of the community is investigated through three elements: sense of belonging, feelings of safety and exposure to vioalence ([@B65]). There is evidence that the perception of danger in the place where one lives influences the presence of aggressive behavior in young people ([@B19]; [@B30]; [@B32]).
We addressed adolescent exposure to various types of violence in the community, including both personal experiences of aggression and exposure to incidents targeting other people who are known victims in their community. Exposure to violence is related to the presence of behavioral problems such as bullying in adolescents ([@B3]; [@B10]; [@B31]; [@B54]).
A conflictive atmosphere in the community leads to violence legitimization, and, therefore, the existence of aggressors and victims becomes perceived as standard within schools ([@B9]; [@B10]; [@B46]).
This study set out to determine the effects of the presence of challenging behavior problems, parental conflict and exposure to violence in the community on the likelihood of aggressive behavior by elementary students towards their peers. It is hypothesized that the above variables significantly increase the likelihood of students behaving aggressively toward their peers.
Method
======
Participants
------------
In a non-probabilistic way, 664 students enrolled in grades 4 to 6 in 12 public elementary schools from four school zones (three per school zone) of a city in northwestern Mexico were selected. 80 (12.04%) students were identified as aggressors by the criterion that their average score was \> 2 *(Sometimes,* three or four times a month) on the measuring scale of violence among students. Of these, 45 (56.25%) were male and 35 (43.75%) female. Their average age was 13.8 years (SD = .86).
Of the remaining 584 students, who were not identified as aggressors, a subsample randomly selected 80 in order to perform analyses in groups of similar size. This group was composed of 42 (52.5%) males and 38 (47.5%) females with an average age of 13.6 (SD = 1.13).
Instruments
-----------
*Violence among students.* The self-report scale developed by [@B57] was used, in which students were questioned about how often they assaulted weaker students during the last month (e.g. *Hitting peers,* or *Insulting their classmates).* This instrument consists of six items that are answered by a Likert-type scale with five response options 0 (Never), 1 *(Almost never, once or twice a month),* 2 *(Sometimes, three or four times a month),* 3 *(Almost always, five to seven times a month)* and 4 *(Always, more than seven times a month).*
A confirmatory factor analysis demonstrated that the scale is a one-dimensional model to sustainably and empirically measure the construct already mentioned ~^*(X2*^~ = 19.73, *p* = .019; AGFI = .97; CFI = .99; RMSEA = .04, IC 90 \[.02 - .07\]). Cronbach\'s alpha was .83.
*Challenging behavior.* This instrument was developed by [@B53]. This study used the subscale that measures the presence of challenging behavior through 11 items (e.g. *I disobey the instructions and rules of adults).* Using factor analysis with Oblimin maximum likelihood and rotation the factor structure of the scale was confirmed ~^*(X2*^~ = 850. 42, *p* \< .000; KMO = .83) which was able to explain 62% of the variance of scores. Cronbach\'s Alpha, which demonstrates the reliability of the measurement scale, was .87.
The instrument utilized Likert-type scales with five response options 0 (Never), 1 *(Almost never),* 2 *(Sometimes),* 3 *(Almost always)* and 4 *(Always).* From the average score on the scale, students were classified into two groups: without challenging behaviors *(M* \< 1) and with challenging behaviors (M \> 1).
*Parental conflict.* We used the subscale of the instrument developed by [@B21], which used five items to assess childrens\' perception of the frequency and intensity of parental conflict (e.g. *I often see my parents arguing, My parents go crazy when arguing).* An exploratory factorial analysis with Oblimin maximum likelihood and rotation, modeled the one-dimensional character of the scale ~^*(X2*^~ = 506. 50, *p* \< .000; KMO = .83), which explained 60% of the variance of the scores. The reliability of the scores measured by Cronbach\'s Alpha was .90.
The instrument used a Likert-type scales with four response options: 0 *(Strongly disagree),* 1 *(Disagree),* 2 *(Disagree)* and 3 *(Strongly agree).* According to the average total scores on the scale, students were classified into a first group, which includes those who expressed a perception of low parental conflict (M \< 1) and a second group, which reflect a high level of parental conflict (M \> 1).
*Exposure to violence in the community.* An instrument based on the review of developed instruments that measured related aspects was developed for this study ([@B20]; [@B65]). Seven risk situations are presented to the student (e.g. *Fights in the neighborhood, Presence of gangs)* and then they\'re asked to select the option that best represents how often this situation occurs in their community.
It was answered by a Likert-type scale with five response options 0 (Never), 1 *(Almost never),* 2 *(Sometimes),* 3 *(Often)* and 4 *(Always).* Average scores were used to form a group of students with low exposure to violence *(M* \< 1) and another group with high exposure *(M* \> 1). The measurement reliability with Cronbach\'s Alpha was .77.
Procedure
---------
After presenting the study objective, we obtained authorization from school authorities for access. The written and informed consent of the parents of students who participated in the study was also required. Finally, we requested the voluntary cooperation of the students, guaranteeing them confidentiality
In data analysis, descriptive and inferential statistics were used, particularly logistic regression. The calculation of the regression model was conducted with support from SPSS software v. 22.
Results
=======
[Table 1](#t1){ref-type="table"} shows that the predictive variables differ significantly between groups of students with and without reports of aggression toward peers. The group that reports higher levels of aggression has a higher proportion of students with challenging behaviors, high parental conflict and greater exposure to violence in the community.
Table 1Frequencies of the predictive variables of aggression in elementary studentsNot aggressors (n = 80)Not aggressors (n = 80)Aggressors (n = 80)Aggressors (n = 80)Aggressors (n = 80)Variablen%n%X2(1)Challenging behaviorAbsence5771.25162040.93\*\*\*Presence2328.756480Parental conflictLow243067.513.55\*\*\*High56707492.5Violence in the communityLow1316.2556.259.04\*\*High6783.757593.75[^1]
The ~^*R2*^~ value of .39 shows that the predictive variables integrate a model with better predictive power than the base model. Hosmer-Lemeshow\'s test ~^*(X2*^~ = 1.178, *df* = 4, *p* = .88) was not significant, indicating that there are no differences in the current distribution and the predicted values of the dependent variable.
All Beta coefficient values were significantly positive, implying that the probability of belonging to the group of aggressors and the incidence of challenging behaviors, perception of parental conflict and exposure to violence in the community were increased (see [Table 2](#t2){ref-type="table"}).
Table 2Summary of logistic regression analysis for predicting assault toward pairsVariables*BESOR*Wald's StatisticChallenging behavior2.06.387.8328.96\*Parental conflict1.33.543.775.38\*Community violence1.68.855.363.87\*[^2]
Predictive variables allowed the proper classification of 76.4% of students in the groups of aggressors and non-aggressors, which is considered acceptable for an analysis of this type ([@B8]; [@B22]). It should be noted they present slightly better discriminative power to identify aggressors than non-aggressors (see [Table 3](#t3){ref-type="table"}).
Table 3Analysis of the classification aggressors and not aggressor\'s groupsPredicted group membershipPredicted group membershipPredicted group membershipPredicted group membershipCurrent membership groupAggressorsAggressorsNot aggressorsNot aggressorsnn%n%Aggressors806378.71821.3Not aggressors8020756025[^3]
Discussion
==========
The present study analyzes the effects on elementary students of the presence of challenging behavior problems, parental conflict and exposure to violence in the community on the likelihood of aggressive behavior towards their peers. The results support the hypothesis of the study in the sense that they suggest that both challenging behavior problems, such as parental conflict and exposure to violence in the community make up a model that predicts the presence of aggressive behavior toward peers in elementary students. These findings reaffirm the importance of addressing bullying from an ecological framework that considers the effects of variables located in various contexts where the individual grows up ([@B23]; [@B54]).
Matching reports in the literature, it was found that behavioral problems increase the likelihood of students to bully their peers ([@B11]; [@B49]; [@B60]), which suggests that bullying in some children is the manifestation of a symptom within a pattern of maladaptive behaviors ([@B12]; [@B16]). Although this requires more investigation, it suggests that these students form part of the group of socially marginalized aggressors, who experience little acceptance and praise from peers ([@B17]; [@B54]; [@B44]).
It was found that perception of parental conflict increases the likelihood of the student developing bullying behaviors. This is consistent with reports in the literature which suggest frequent and hostile parental conflict is related to the presence in the children of aggressive behavior toward peers at school ([@B6]; [@B25]; [@B27]).
Parental conflict explains the presence of aggressive behavior of children through their observation of aggressive parental models [@B4] and because of the frustration associated with this situation, which they channel to aggression towards others ([@B14]).
Lastly, the results suggest that exposure to violence in the community increases the likelihood of students behaving like bullies, which is consistent with the literature ([@B3]; [@B9]; [@B10]; [@B31]; [@B55]). Interestingly, the impact of this variable was greater than the presence of parental conflict which is a closer variable to child development. This suggests that interaction in extra-family social contexts, especially the community, has an important influence on the formation and expression of moral values, social norms and behavioral patterns ([@B28]; [@B51]; [@B61]).
Conclusions
===========
We conclude that the analysis of bullying from an ecological perspective is fruitful as it allows us to consider the impact of variables located in different contexts of individual development. From the results, we conclude that measures to prevent bullying should include: interventions, a focus on developing students\' pro-social behavior, the strengthening of family life, and a reduction of the various manifestations of violence in the community. Our finding concerning the influence of exposure to violence in the community on the presence of bullying is particularly striking because it is a subject about which there have been few studies in Mexico, despite the high levels of violence in many communities and regions throughout the country. This shows that, for the prevention of bullying, public policies should be developed to prevent violence at the societal level as well as to strengthen the economic, cultural and social capital of communities.
The present study provides knowledge that can be useful for the prevention of bullying in schools. Nonetheless, it does present limitations, including the fact that we didn\'t consider interactions between variables, which would establish their share effects on the problem of bullying. Each context variable should also be integrated to allow for a broader perspective of violence among students. The study has a cross-sectional design that doesn\'t allow us to establish causal relation among variables. We suggest future studies be undertaken with an experimental and longitudinal design.
[^1]: \*p \< .05. \*\* *p* \< .01. \*\*\* *p* \< .001.
[^2]: \* *p* \< .05.
[^3]: *Note.* Globally correctly classifies 76.4% of students
| {
"pile_set_name": "PubMed Central"
} |
INTRODUCTION
============
The coronavirus 2019 (COVID-19) pandemic has disrupted patient care across the NHS. Following the suspension of elective surgery, priority was placed in providing urgent and emergency surgery for patients with no alternative treatment.
The pandemic has caused significant morbidity and mortality to the general population \[[@ivaa143-B1]\]. Furthermore, the outcome of patients undergoing thoracic surgery and major abdominal surgery is known to be poor \[[@ivaa143-B2]\]. Although routine cardiac surgery comes with a low risk, it does require a hospital stay of ∼5--7 days including time in intensive care and on a ventilator. Developing COVID-19 infection perioperatively would likely increase the morbidity and mortality above and beyond associated with surgery. However, the outcomes of patients developing COVID-19 in cardiac surgery patients during their hospital stay are unknown.
We aim to assess the outcomes of patients undergoing cardiac surgery who have COVID-19 infection diagnosed in the early postoperative period.
PATIENTS AND METHODS
====================
All patients undergoing elective or urgent cardiac surgery at St Bartholomew's Hospital from 1 March to 27 March 2020 were included in this study. They were identified from a national cardiac surgery database. This time period was prior to routine COVID-19 screening. Patients diagnosed with COVID-19 infection via positive throat swab taken due to clinical suspicion postoperatively were reviewed. Data from electronic patient records collected contemporaneously were reviewed retrospectively and presented.
RESULTS
=======
During the study period, we performed 97 cardiac surgical procedures. Seven patients were diagnosed with COVID-19 infection in the immediate postoperative period. Mean age was 59 (21--73) years, all patients were male and the majority of operations were performed on an urgent basis with one done electively. Mean EuroSCORE II was 2.06 (0.77--7.43). For the urgent cases, the mean length of hospital stay preoperatively was 12 (4--22) days. The majority of this time was in the referring general hospital (Table [1](#ivaa143-T1){ref-type="table"}).
######
Demographics and outcomes of patients diagnosed with COVID-19 in the postoperative period
Patient Male/ female Age Preoperative length of stay Euro SCORE II Elective/urgent Operation Bypass/ cross-clamp Day of COVID diagnosis Outcome Time to outcome
--------- -------------- ----- ----------------------------- --------------- ----------------- ------------ --------------------- ------------------------ --------- -----------------
1 Male 62 18 2.5 Urgent CABG 75/96 3 Died 4
2 Male 56 1 1.03 Elective CABG 77/86 8 Home 14
3 Male 21 4 0.77 Urgent AVR 70/92 9 Home 12
4 Male 73 6 7.43 Urgent MVR + CABG 144/173 8 Died 16
5 Male 72 22 2.97 Urgent AVR + CABG 133/152 3 Died 11
6 Male 57 5 0.88 Urgent CABG 84/104 4 Home 6
7 Male 71 16 1.26 Urgent CABG 37/77 1 Home 16
8 Male 59 23 2.29 Urgent MVR 84/101 23 Home 7
9 Female 79 1 4.91 Elective MVR + TVR 73/123 37 Died 2
AVR: aortic valve replacement; CABG: coronary artery bypass grafts; COVID-19: coronavirus 2019; MVR: mitral valve replacement; TVR: tricuspid valve repair.
Procedures were coronary artery bypass grafts (CABG) *n* = 4, aortic valve replacement *n* = 1, aortic valve replacement + CABG *n* = 1 and mitral valve replacement + CABG *n* = 1. Mean bypass time was 111 min (77--173) and cross-clamp time 89 min (37--144). There were no resternotomy for bleeding, no stroke and no renal replacement therapy.
COVID-19 diagnosis was made on postoperative day 1, day 2, day 3 (×2), day 8 (×2) and day 9. Note the day 1 diagnosis was following a negative swab 2 days previously.
There were a variety of presentations at the time of diagnosis. On the day of diagnosis, 4 patients were pyrexial, 1 patient had raised white cell count, 6 patients had raised CRP, 3 patients had lymphopenia and 4 patients had CXR changes consistent with COVID-19.
Mortality was *n* = 3 (43%) whilst the other 4 patients were discharged home well. Of the 3 patients who died, time from diagnosis to death was 4, 13 and 16 days. All 3 died from respiratory failure, refractory to medical management. One patient had additional renal failure.
In addition, there were 2 patients, operated on previously but remained inpatients during the study period, with a late diagnosis. Mean age was 60 years (59--79), one was male and the other, female. Mean EuroSCORE was 3.6% (2.29--4.91). Their operations were urgent mitral valve replacement and elective mitral valve replacement plus tricuspid valve repair. Time from surgery until diagnosis was 23 and 39 days. One patient died 2 days following diagnosis from respiratory failure and systemic sepsis, the other was discharged home 7 days following diagnosis (Table [2](#ivaa143-T2){ref-type="table"}).
######
Clinical features on the day of COVID-19 diagnosis
Patient Day of COVID diagnosis Outcome Pyrexia Raised WCC Lymphopenia Raised CRP CXR changes
--------- ------------------------ --------- --------- ------------ ------------- ------------ -------------
1 3 Died 38.4 No Yes Yes Yes
2 8 Home 38 Yes No Yes No
3 9 Home 38.5 No No Yes No
4 8 Died 36.9 No Yes Yes No
5 3 Died 36.7 No Yes Yes Yes
6 4 Home 38.1 No No Yes Yes
7 1 Home 36.5 No No No Yes
8 23 Home 38.6 No Yes No No
9 37 Died ? Yes No Yes Yes
COVID-19: coronavirus 2019; CRP: c-reactive protein; CXR: chest x-ray; WCC: white cell count.
All patients discharged home were alive and well at 4-week follow-up.
DISCUSSION
==========
COVID-19 is a respiratory disease caused by the novel corona virus. It is highly contagious and causes severe respiratory failure, mainly in people with comorbid conditions. It has an asymptomatic incubation period of 5--14 days \[[@ivaa143-B1]\].
COVID-19 infection following cardiac surgery is associated with very poor outcomes and in our series of 9 patients, a mortality of 44%. During the same period, our mortality was 5.5% for patients not identified as having COVID-19 infection.
During the period in which these patients were operated on, there was no established COVID-19 screening protocol preoperatively or postoperatively. Only one patient (patient 7) had a preoperative swab, which was negative. However, a second swab on the day of the operation came back as positive the following day. For those patients who developed COVID-19 in the early postoperative period, it is likely that they contracted the virus preoperatively, either prior to admission or whilst in hospital. However, the clinical manifestations only appeared in the postoperative period. We have shown a widely variable presentation in those patients testing positive. It is unclear as regards the contribution of cardiopulmonary bypass, postoperative ventilation and obvious cardiac comorbidities to the outcomes in these patients. However, the outcomes were very poor, even for this patient population requiring urgent major cardiac surgery. Furthermore, the hospital stay for these patients is significantly longer than that expected for routine elective and urgent cardiac surgery.
The cause of death in all 4 of our patients was respiratory failure with 1 concomitant renal failure. We did not see thrombosis in any patients which has been seen in other series of COVID-19 infection.
Those patients who were diagnosed late in their hospital stay most likely contracted the virus whilst in hospital. It is impossible to determine the source, but it could be hypothesized that it originated from asymptomatic patients, staff members or visitors. Without routine screening of staff or patients at this time, such transmission is likely and the incidence of this unknown.
During the COVID-19 pandemic, despite the cancellation of all elective operations, there is a need to maintain urgent and time-critical emergency cardiac surgical services. As part of the Pan London Emergency Cardiac Surgery (PLECS) system, our institution and one other were tasked with providing this service for the 7 usual providers across London. We aim to provide this care in a COVID-19 protective environment to prevent the poor outcomes seen in this patient series. This will remain important during resumption and recovery of elective surgical activity.
We have instituted local protocols aimed at thorough screening of all patients before and on arrival in our unit to ensure that we do not admit or operate on anyone with the virus and isolating patients with unknown COVID-19 status. Our screening protocol has been published previously \[[@ivaa143-B5]\] and consists of 2 throat swabs, lymphocyte count, LDH and ferritin levels and a computed tomography chest. Furthermore, all patients have bronchoalveolar lavage for viral testing at the time of intubation in operating theatre. In line with Public Health England guidance, all surgeries are now performed with full personal protection equipment and patients are cared for in critical care with staff wearing full personal protection equipment. We minimize staff movement through clinical areas and interaction with patients. On the ward setting, staff wear face masks, apron and gloves. We have also banned visitors in order to reduce transmission.
Our observations are limited by the lack of routine testing of all patients for COVID-19 infection; therefore, the infection rate may be higher; however, poor outcomes appear to be in those patients who show signs of infection.
CONCLUSION
==========
In conclusion, the outcome of cardiac surgical patients who contracted COVID-19 infection perioperatively is extremely poor. In order to offer emergency and time-critical urgent cardiac surgery, units must implement rigorous protocols aimed at maintaining a COVID-19 protective environment to minimize additional life-threatening complications related to this virus infection.
The authors thank the Barts Consortium of Surgeons whose patients were included in this series.
**Conflict of interest:** none declared.
Reviewer information
====================
Interactive CardioVascular and Thoracic Surgery thanks Martin Andreas and the other, anonymous reviewer(s) for their contribution to the peer-review process of this article.
| {
"pile_set_name": "PubMed Central"
} |
1. Introduction {#sec1-ijerph-16-01887}
===============
Despite the substantial progress made to reduce child mortality, 5.6 million children aged less than five years died worldwide in 2016, and the majority of these deaths occurred within the first year of life \[[@B1-ijerph-16-01887]\]. Furthermore, it was estimated in 2011 that malnutrition was responsible 3.1 million of these deaths worldwide \[[@B2-ijerph-16-01887]\]. Appropriate child feeding practices include optimal breast feeding and complementary feeding, and are key determinants of child nutritional wellbeing and health \[[@B2-ijerph-16-01887],[@B3-ijerph-16-01887],[@B4-ijerph-16-01887]\]. The large bodies of literature support the notion that the initiation of breast feeding within the first hour of birth (EIBF) protects children from common childhood diseases, such as diarrhoea and acute respiratory infections (ARI) \[[@B5-ijerph-16-01887],[@B6-ijerph-16-01887],[@B7-ijerph-16-01887]\], and can significantly reduce neonatal mortality \[[@B8-ijerph-16-01887],[@B9-ijerph-16-01887]\]. Exclusive breast feeding (EBF) also decreases the incidences of diarrhoea and ARI \[[@B7-ijerph-16-01887],[@B10-ijerph-16-01887]\] and prevents growth faltering and acute malnutrition \[[@B11-ijerph-16-01887],[@B12-ijerph-16-01887]\]. Furthermore, recent systematic reviews and meta-analyses have reported that children who are breastfed for longer periods have fewer dental malocclusions, are less likely to develop childhood leukemia, and have greater intelligence than their counterparts \[[@B13-ijerph-16-01887],[@B14-ijerph-16-01887]\]. Optimal breast feeding also reduces the risks of childhood obesity, adult overweightness/obesity, and type 2 diabetes \[[@B3-ijerph-16-01887],[@B15-ijerph-16-01887]\].
The timely initiation of CF at six months similarly protects against childhood illnesses and promotes health and nutritional wellbeing \[[@B16-ijerph-16-01887]\]. It has been shown by recent studies that the initiation of CF too early is associated with undernutrition \[[@B17-ijerph-16-01887]\], overweightness in children \[[@B18-ijerph-16-01887]\], as well as obesity and overweightness in later life \[[@B19-ijerph-16-01887]\], and that the delayed introduction of CF increases the risk of celiac disease in children \[[@B20-ijerph-16-01887]\], lower child growth velocity and increases childhood infections \[[@B21-ijerph-16-01887],[@B22-ijerph-16-01887]\]. For these reasons, the World Health Organization (WHO) recommended EIBF followed by EBF for six months and then the introduction of CF at six months with breastfeeding \[[@B4-ijerph-16-01887]\].
However, despite the numerous short- and long-term advantages of optimal breast feeding, only 37% of children younger than six months of age are exclusively breastfed in low-income and middle-income countries \[[@B14-ijerph-16-01887]\]. Astonishingly, the EIBF percentage is less than 50% in low income countries, e.g., 42% in South Asia, 35% in West and Central Africa, and 45% in Sub-Saharan Africa \[[@B23-ijerph-16-01887]\]. Similarly, approximately one third of infants aged 4--5 months are already prematurely weaned onto solid foods, whereas about 20% of 10--11-month olds have never consumed solid foods. Furthermore, the timely introduction of complementary foods is immensely more problematic in developing countries, such as those situated in Latin America, the Caribbean, East Asia and the Pacific, where nearly half of all infants between 4 and 5 months of age are already consuming solid foods \[[@B24-ijerph-16-01887]\].
Many studies have attempted to identify factors associated with suboptimal breast feeding \[[@B25-ijerph-16-01887],[@B26-ijerph-16-01887],[@B27-ijerph-16-01887],[@B28-ijerph-16-01887],[@B29-ijerph-16-01887],[@B30-ijerph-16-01887]\] and the timely initiation of complementary feeding \[[@B31-ijerph-16-01887],[@B32-ijerph-16-01887],[@B33-ijerph-16-01887],[@B34-ijerph-16-01887]\]. Maternal socio-economic and demographic factors \[[@B26-ijerph-16-01887],[@B27-ijerph-16-01887],[@B35-ijerph-16-01887]\], the receipt of adequate prenatal care services, postnatal visits \[[@B25-ijerph-16-01887],[@B26-ijerph-16-01887],[@B29-ijerph-16-01887]\], breast related problems \[[@B30-ijerph-16-01887]\], and pre-lacteal feeds \[[@B35-ijerph-16-01887]\] have been well reported to be associated with EIBF and EBF. Similarly, maternal factors such as living in urban settlements, education, and the utilization of maternal care services (e.g., antenatal follow ups, institutional delivery, and postnatal check-ups) have been reported to predict the initiation of CF at the appropriate time \[[@B31-ijerph-16-01887],[@B32-ijerph-16-01887]\]. However, such studies are sparse in Nepal.
Despite having effectively implemented the Infant and Young Child Feeding (IYCF) programme, recent studies that utilized nationally representative data from the Nepal Demographic and Population Health Survey (NDHS) in 2016 demonstrated that two thirds of Nepalese mothers exclusively breastfed children for ≤5 months, initiated breastfeeding within an hour of childbirth \[[@B36-ijerph-16-01887],[@B37-ijerph-16-01887],[@B38-ijerph-16-01887]\], and that 16% of mothers did not introduce complementary foods at 6--8 months. Furthermore, breast feeding and complementary feeding practices vary dramatically across Nepal \[[@B39-ijerph-16-01887]\]. Therefore, it is essential that program managers and policy makers understand local and regional child feeding practices, as well as factors affecting these practices, before establishing customized strategies aimed at improving the health and nutritional status of younger Nepalese children. The present study was undertaken to identify maternal factors associated with infant feeding practices and to determine the effects of maternal care service utilization on infant feeding practices in rural areas of Southern Nepal.
2. Materials and Methods {#sec2-ijerph-16-01887}
========================
2.1. Study Design and Participants {#sec2dot1-ijerph-16-01887}
----------------------------------
Data from a cluster randomized controlled trial 'MATRI-SUMAN' conducted during 2015--2016 were used to identify associations between infant feeding practices, maternal factors and the utilization of maternal care services. A multistage stratified random sample of 426 pregnant women was selected from six rural village development committees (VDCs) in the Dhanusha district of Nepal. The overall response rate was 94.3%. However, of these 426 pregnant women, 379 who breastfed their child and were followed up for seven months postnatally were included in the present study. The MATRI-SUMAN trial focused on the capacity building of female community health volunteers (FCHVs) and providing information about maternal and child health care services (MCH) to pregnant women by mobile text messaging. The capacity building of FCHVs was performed by one day extensive reinforcement training to equip them with the knowledge and skills of MCH services, and mobile text messaging intervention was done through periodic mobile short messaging service (SMS) system to either pregnant woman or their family members (who could convey messages to the participant) in the intervention arm concentrating on MCH services. Details of sample selection and procedures have been found elsewhere \[[@B40-ijerph-16-01887]\].
2.2. Data Collection and Measures {#sec2dot2-ijerph-16-01887}
---------------------------------
Trial data were collected from MCH register of FCHVs and face-to-face interview by female research assistants using survey questionnaire at different time interval. The details of the follow ups and measurement can be found elsewhere \[[@B40-ijerph-16-01887]\]. We used datasets that contained information on the use of maternal care services, socio-demographic characteristics, and infant feeding practices.
2.3. Outcome Measures {#sec2dot3-ijerph-16-01887}
---------------------
The dependent variables of the study were: (i) early initiation of breast feeding within an hour of birth (EIBF), (ii) exclusive breastfeeding (EBF), and the (iii) initiation of complementary feeding (CF) at six months. EIBF was classified 'yes' if the mother initiated breast feeding within the first hour of after birth and not otherwise. EBF was recorded as 'yes' if a mother fed her baby only breast milk (excepting syrups and medicines). Similarly, the initiation of the CF at six months recorded as yes or no. These outcome variables were adapted from UNICEF and WHO recommendations and standards \[[@B4-ijerph-16-01887],[@B41-ijerph-16-01887]\] and the measurement was based on the MCH register of FCHVs and survey questionnaires.
2.4. Measurement of Exposure {#sec2dot4-ijerph-16-01887}
----------------------------
Two separate dimensions of exposure variables were assessed, that is, maternal care services received and sociodemographic characteristics. Maternal care services included antenatal visits, place of delivery, and postnatal care received. ANC visits were recorded as continuous variables and categorized as \< or ≥4 ANC visits. Similarly, the place of delivery was categorized as either the home or health facility. PNC visits were categorized as 'yes' or 'no'. The child's sex was classified as either male or female.
2.5. Socio-Demographic Variables {#sec2dot5-ijerph-16-01887}
--------------------------------
Socio-demographic variables were adapted from previous studies. Age was categorized as: (i) \<20 years, (ii) 20--34 years, or (iii) ≥35. Caste/ethnicity were classified as (i) upper caste---Brahmin, Chhetri, and Terai (Yadav, Teli, Thakur, Koiri), (ii) Adibasi/Janjati---Janjati and indigenous, and (iii) Dalit \[[@B42-ijerph-16-01887]\]. Education was recorded as years of completed education using the following three categories: (i) no education, (ii) primary---1 to 5 years of schooling, and (iii) secondary and higher ≥6 years of education \[[@B39-ijerph-16-01887],[@B43-ijerph-16-01887]\]. Occupations were categorized as: (i) business, private or government, or households work; (ii) agricultural work if in own farms, and (iii) skilled or unskilled manual work. Monthly incomes were classified by tertiles: (i) 1st tertile (income \< 14,333 Nepalese Rupees (Nrs)/month) (ii) 2nd tertile (Nrs 14,334--23,666/month) (iii) 3rd tertile (\>Nrs. 23,666/month). Families were classified as nuclear or joint, parity as primi or multi, and birth origin as Terai or Hill on the basis of the geographical regions of Nepal \[[@B44-ijerph-16-01887]\].
2.6. Statistical Analyses {#sec2dot6-ijerph-16-01887}
-------------------------
Infants feeding practices are reported as percentages of all infants. Chi-square tests (χ^2^) were performed to assess associations between independent variables with EIBF, EBF for six months, and the initiation of CF at six months. Variables found to be significant by Chi-square testing were subjected to multivariate logistic regression analysis. Independent variables with a p-value of \<0.1 were entered into the multivariate analysis. Unadjusted and adjusted odds ratios with 95% confidence intervals (CIs) are reported. The statistical analysis was conducted using SPSS ver. 21.0 (SPSS, IBM, Armonk, NY, USA).
2.7. Ethics {#sec2dot7-ijerph-16-01887}
-----------
Ethical approval for the 'MATRI-SUMAN' protocol was obtained from the Nepal Health Research Council, Nepal (approval no: 101) and the ethics committee of the Institute of Medical Sciences, Banaras Hindu University, India (approval no: ECR/526/Inst/UP/2014 Dt.31.1.14), and the District Public Health Office, Dhanusha, Nepal (Ref. 2245). Additional ethical approval was also received from the Institutional Review Board of Janaki Medical College for the data analysis. The aims and objectives of the study were explained to all study subjects, who provided written informed consent. All personal identifiers were removed before the data analysis.
3. Results {#sec3-ijerph-16-01887}
==========
3.1. Infant Feeding Practices of Mothers {#sec3dot1-ijerph-16-01887}
----------------------------------------
Infant feeding practices of among mothers in rural Terai are detailed in [Table 1](#ijerph-16-01887-t001){ref-type="table"}. For the 379 study subjects, feeding practices were as follows; EIBF 41.4%, EBF 53.0%, and 43% initiated CF at six months.
3.2. Maternal Characteristics and Utilization of Maternal Care Services {#sec3dot2-ijerph-16-01887}
-----------------------------------------------------------------------
The majority of the study subjects (69.9%) were 20--34 years old, 62.5% were from an upper caste group, 71.8% were Terai by birth, 23.5% had no education, 83.1% worked in the agricultural or service/business/household sectors, 65.7% had family incomes in the second or third tertile, slightly more than half (52.5%) were from joint families, 60.9% were multiparous, and 52.3% of the babies were female. Maternal factors such as birth origin, education, occupation, family income, and family type were significantly associated with EIBF; caste/ethnicity, education, occupation, family income, and parity were significantly associated with EBF; and birth origin, education, occupation, and child sex were significantly associated with the initiation of CF at six months ([Table 2](#ijerph-16-01887-t002){ref-type="table"}).
The utilization of maternal care services associated with infant feeding practices is summarized in [Table 3](#ijerph-16-01887-t003){ref-type="table"}. Half (50.9%) of the study subjects received MATRI-SUMAN intervention, 60.9% visited an ANC more than four times, 58.0% of childbirths were at home, and 57.8% visited a PNC. All variables associated with the utilization of maternal care services were significantly associated with the three child feeding practices.
3.3. Maternal Factors and the Utilization of Maternal Care Services Associated with Infant Feeding Practices {#sec3dot3-ijerph-16-01887}
------------------------------------------------------------------------------------------------------------
Maternal factors and the utilization of maternal care services associated with infant feeding practices with unadjusted odds ratios are provided in [Table 4](#ijerph-16-01887-t004){ref-type="table"}. A number of maternal factors and maternal care service utilization factors, that are caste/ethnicity, birth origin, education, occupation, family income, family type, MATRI-SUMAN intervention, ANC visits, place of delivery, and PNC visits, were significantly associated with EIBF; education, occupation, family income, parity, MATRI-SUMAN intervention, ANC visits, place of delivery, and PNC visits were significantly associated with EBF; and birth origin, education, occupation, family income, child sex, MATRI-SUMAN intervention, ANC visits, place of delivery, and PNC visits were significantly associated with the initiation of CF at six months by univariate analyses.
Multiple logistic regression model results with adjusted odds ratio are summarized in [Table 5](#ijerph-16-01887-t005){ref-type="table"}. Maternal education secondary and higher (aOR 2.2; 95% CI (1.2--4.2)), occupation in the service/business/household sectors (aOR 2.2; 95% CI (1.2--4.2)), receipt of MATRI-SUMAN intervention (aOR 1.7; 95% CI (1.1--2.9)), ≥4 ANC visits (aOR 3.2; 95% CI (1.2--8.0)), and delivery in a health facility (aOR 1.9; 95% CI (1.0--3.4)) had higher odds ratios for EIBF.
Mothers with a primary level of education (aOR 2.2; 95% CI (1.1--4.4)), a secondary or higher level of education (aOR 5.5; 95% CI (1.8--16.1)), working in the service/business/household sectors (aOR 2.1; 95% CI (1.1--4.1)), primipara (aOR 2.2; 95% CI (1.2--4.0)), that received MATRI-SUMAN intervention (aOR 1.8; 95% CI (1.1--3.2)), that had visited an ANC ≥ 4 times (aOR 3.1; 95% CI (1.3--7.5)) and visited a PNC (aOR 2.4; 95% CI (1.1--5.6)) had higher odds ratios of EBF. However, mothers aged 35--45 years were less likely (aOR 0.3; 95% CI (0.1--0.7)) to have performed EBF than their counterparts.
Mothers with a secondary or higher education level (aOR 2.2; 95% CI (1.2--3.9)), male baby (aOR 1.7; 95% CI (1.1--2.7)), received MATRI-SUMAN intervention (aOR 1.6; 95% CI (1.0--2.7)), and visited a PNC (aOR 2.3; 95% CI (1.0--5.1)) had higher odds ratios for initiating CF at six months.
4. Discussion {#sec4-ijerph-16-01887}
=============
In the study, 41.4%, 53%, and 43% of 379 ever-breastfed mothers initiated breast feeding within an hour of the birth (EIBF), practiced exclusive breast feeding (EBF), and initiated complementary feeding (CF) at six months, respectively, and these percentages are lower compared with those in a recent Nepalese Demographic and population health survey report 2016 \[[@B39-ijerph-16-01887]\]. However, the proportion of mothers that performed EIBF in the present study is similar to those reported in other developing countries like Brazil \[[@B45-ijerph-16-01887]\] and Ethiopia \[[@B46-ijerph-16-01887]\], but higher than that found in an Indian study \[[@B47-ijerph-16-01887]\]. Likewise, the prevalence of EBF in Nigerian \[[@B48-ijerph-16-01887]\] and Ethiopian \[[@B49-ijerph-16-01887]\] studies was reported to be 37.3% and 54.5%, respectively, values which are substantially higher than that found in a study conducted in Indian urban slums, in which only 7.8% were exclusively breast fed \[[@B50-ijerph-16-01887]\]. However, a recent Nepalese study conducted using nationally representative data also reported an EBF rate for children of ≤five months was 66.3%, which is higher than that found in the present study. These differences were probably due to different study population sizes and settings, for example, some used nationally representative datasets \[[@B36-ijerph-16-01887],[@B38-ijerph-16-01887]\], whereas our study was conducted in the rural Southern Terai. In addition, different levels of education, wealth quintiles, and existing local traditions and beliefs might have impacted results.
In the present study, only slightly more than two fifths of mothers initiated CF at 6 months, which is lower than those found in some Ethiopian studies performed in different parts of the country \[[@B32-ijerph-16-01887],[@B33-ijerph-16-01887],[@B51-ijerph-16-01887]\], which reported \>60% of mothers initiated CF at six months. Studies conducted in other developing countries like Ghana \[[@B52-ijerph-16-01887]\], Ethiopia \[[@B53-ijerph-16-01887]\], and Bangladesh \[[@B54-ijerph-16-01887]\], also presented higher percentages of mothers initiated CF at six months. Recent local Nepalese studies \[[@B55-ijerph-16-01887],[@B56-ijerph-16-01887]\] also reported 57% of mothers initiated CF at six months, and these values are also higher than those found in the present study. This wide variation in the prevalence of the timely initiation of CF may have been due to the fact that the present study was conducted in rural areas of the South-East region of Nepal, whereas the two other Nepalese studies were conducted in Western Nepal. Differences between Nepalese studies may also have been influenced by different sample sizes, literacy statuses, and methodological differences. Importantly, we dichotomized mothers based on whether they initiated CF at exactly six months, and several other studies used a range of 6--8 months \[[@B52-ijerph-16-01887],[@B54-ijerph-16-01887],[@B57-ijerph-16-01887]\], which might also have contributed to reported differences.
We found maternal education and the receipt of MATRI-SUMAN intervention increased the odds ratio of initiating EIBF and the practice of EBF and CF. Several other reports concur regarding a positive association between maternal education and EIBF, EBF and CF \[[@B28-ijerph-16-01887],[@B30-ijerph-16-01887],[@B36-ijerph-16-01887],[@B53-ijerph-16-01887],[@B58-ijerph-16-01887],[@B59-ijerph-16-01887]\], presumably because educated women are more aware about healthy and timely child feeding practices. In fact, maternal education has been reported to be associated with improved child feeding practices \[[@B60-ijerph-16-01887]\]. MATRI-SUMAN intervention on the other hand might have influenced rural women to adopt recommended child feeding practices. MATRI-SUMAN intervention has two components, namely the training of female community health volunteers and mobile text messaging directed at pregnant and post-natal women; both approaches were directed at increasing the utilization of maternal and child health care services, and thereby, improved maternal and child health. The MATRI-SUMAN intervention has been previously described in detail \[[@B40-ijerph-16-01887]\].
In addition, we observed that a maternal occupation in the service/business/household sectors resulted in higher odds ratios for EIBF and EBF than those observed in the manual labor sector. This finding agrees with some previous reports \[[@B25-ijerph-16-01887],[@B28-ijerph-16-01887],[@B61-ijerph-16-01887]\], but conflicts with others \[[@B62-ijerph-16-01887]\]. Alzaheb et al. and Adugna et al. reported working mothers were less likely to adopt EBF \[[@B25-ijerph-16-01887],[@B61-ijerph-16-01887]\], and a review paper issued by Alzaheb et al. showed working mothers adopted EIBF less frequently than non-working mothers \[[@B28-ijerph-16-01887]\]. However, in rural areas of Southern Terai, small businesses are conducted in homes or nearby, and we categorized these mothers as working at home, whereas women engaged in manual labor usually work long hours far away from their homes, and thus find breastfeeding more difficult. We also found older mothers (35--45 years old) were less likely to practice EBF. The relation between maternal age and EBF is unclear. For example, a recent study concluded that younger women were less likely to practice EBF \[[@B63-ijerph-16-01887]\], whereas others support the notion that an older maternal age predicts the use of non-exclusive breast feeding \[[@B47-ijerph-16-01887],[@B48-ijerph-16-01887]\]. In addition, a study performed in Nigeria discovered that the practice of EBF increases with a mother's age to a peak at around 32 years, and that teenage mothers and mothers older than 32 years tended not to practice EBF \[[@B64-ijerph-16-01887]\]. It is possible older women have more roles and responsibilities in addition to child care and that they are required to look after other family members, respectively in poor rural areas. On the other hand, several studies have reported primiparity as a potential risk factor for not practicing EBF \[[@B65-ijerph-16-01887],[@B66-ijerph-16-01887],[@B67-ijerph-16-01887],[@B68-ijerph-16-01887]\], whereas we found primiparous mothers were more likely to practice EBF than multiparous mothers. Reports differ regarding the association between parity and EBF. Studies conducted in China and Sweden found no association between the two \[[@B69-ijerph-16-01887],[@B70-ijerph-16-01887]\], whereas others \[[@B71-ijerph-16-01887],[@B72-ijerph-16-01887]\] reported primiparous mothers were more likely to adopt EBF, which highlights the importance of educating multiparous women. Furthermore, as has been reported previously \[[@B51-ijerph-16-01887],[@B73-ijerph-16-01887]\], we found CF feeding at six months was more common for male babies, which may reflect the long-standing traditional gender norm that discriminates against female child feeding \[[@B51-ijerph-16-01887]\].
Importantly, our study confirms that some specific components of maternal care services, such as antenatal visits, childbirth at a health facility, and postnatal visits, significantly impact child feeding practices. Mothers that visited ANCs four or more times were found to be significantly more likely to practice both EIBF and EBF, and childbirth at a health facility was significantly associated with EIBF. In addition, PNC visits appeared to increase the likelihood of EBF and the initiation of CF at six months. A number of studies have demonstrated a significant positive association between prenatal care visits and EIBF \[[@B33-ijerph-16-01887],[@B35-ijerph-16-01887],[@B46-ijerph-16-01887],[@B74-ijerph-16-01887]\] and EBF \[[@B26-ijerph-16-01887],[@B50-ijerph-16-01887]\]. Similarly, we found a higher odds ratio of EIBF among mothers that gave birth at a health facility, which concurs with reports issued in other developing countries \[[@B25-ijerph-16-01887],[@B48-ijerph-16-01887]\]. In particular, an Ethiopian and a Nepalese study identified that institutional delivery increased EIBF \[[@B32-ijerph-16-01887],[@B36-ijerph-16-01887]\], and another Nepalese study found home delivery was associated with EBF \[[@B37-ijerph-16-01887]\]. Nonetheless, a more recent study conducted in Nigeria concluded that non-exclusive breast feeding was attributable to no antenatal care visits, home delivery, and delivery assisted by a non-health professional, which is in line with our study findings \[[@B75-ijerph-16-01887]\]. These findings may have arisen because maternal exposure to breast feeding counseling by health workers during childbirth increased maternal knowledge about breastfeeding practices. Furthermore, maternal exposure to health education classes on breast feeding and an adequate knowledge of breast feeding improves breast feeding practices \[[@B76-ijerph-16-01887]\]. Thus, institutional child birth increases the likelihood of mothers adopting EIBF and EBF \[[@B29-ijerph-16-01887]\]. These observations suggest modifiable factors, such as the promotion of childbirth at health facility and education about breast feeding, which can present excellent means of promoting breast feeding practices in low income countries like Nepal. In addition, an Ethiopian study also showed that postnatal visits had a significant positive association with the timely initiation of complementary feeding \[[@B31-ijerph-16-01887]\], and in other studies, childbirth at a health institution were found to predict timely complementary feeding \[[@B32-ijerph-16-01887],[@B77-ijerph-16-01887]\]. We recommend that interventions to promote complementary feeding at recommended times in health care settings should be promptly implemented.
Our study has several valuable strengths. First, the response rate was very high. Second, we assessed the effects of a number of maternal and maternal care service factors on child feeding practices using three dependent variables, that is, EIBF, EBF, and timely initiation of CF using the WHO recommendation and indicators in the same cohort who were involved in the trial. Third, our study made use of a dataset obtained at different time intervals from the same cohort, which may provide reliable information. However, the study has some specific limitations that should be understood. First, our measurements were based on the MCH register as well as maternal recall at different time intervals of the same cohort, and thus may have been influenced by measurement and recall biases. Second, we only assessed whether mothers started complementary feeding at six months and our study could not indicate the appropriateness of the first CF in terms of dietary energy. Third, since the MATRI-SUMAN trial ended follow up visits at seven months after the birth of the baby, our methods could not assess other potentially important indicators of complementary feeding practices, such as meal frequency, dietary diversity, and minimum acceptable diet. Fourth, like many other cross-sectional studies, the present study does not address causality.
5. Conclusions {#sec5-ijerph-16-01887}
==============
In the present study, reported rates of EIBF, EBF, and CF were found to be lower than those reported in other developing countries. Maternal education and the receipt of MATRI-SUMAN intervention positively impacted EIBF, EBF, and CF, and maternal occupation in the service/business/household sector positively influenced EIBF and EBF. Primiparity was found to be protective, but an older age (35--45 years) negatively influenced EBF. In addition, it was observed that male babies were more likely to receive complementary feeding at the recommended time. The utilization of maternal care services, such as ≥4 antenatal care visits significantly and positively promoted the practice of EIBF and EBF, whereas childbirth at a health facility was significantly associated with EIBF alone. In addition, a PNC visit was found to significantly increase the probability of EBF and of initiating CF at six months. Thus, we recommend that the utilization of maternal care services, such as ANCs, the use of health institutions for child birth, and postnatal visits should be recommended to improve child feeding practices, and that maternal factors which impact child feeding practices be carefully considered when designing strategies and interventions. In addition, we suggest a community-based study be undertaken to improve the understanding of the effects of sub-optimal breast feeding and complementary feeding among Nepalese infants.
We thank all study participants for participating in the study and the Janaki Medical College Teaching Hospital (JMCTH) for logistic support provided during the MATRI-SUMAN trial and for approving the use of the data collated for further analysis.
D.A. and J.K.S. conceptualized the study and performed the statistical analysis. D.A., J.K.S. and R.K. drafted the manuscript with the help of S.-J.Y., J.-H.P. and K.L. All authors contributed to the preparation, editing, and revision of the manuscript, and all approved the final version.
The study was supported by a grant from the Janaki Medical College Teaching Hospital (JMCTH; grant no. RES-11-2072-73), but the JMCTH was not involved in the writing or in the decision to submit the manuscript for publication.
The authors have no conflict of interest to declare.
ijerph-16-01887-t001_Table 1
######
Infant feeding practice among mothers in rural Southern Nepal, 2016.
Infant Feeding Practices Yes, *n* (%) No, *n* (%) Total, *N* (%)
--------------------------------------------------------- -------------- ------------- ----------------
Initiation of breast feeding within first hour of birth 157 (41.4) 222 (58.6) 379 (100)
Exclusive breast feeding for 6 months 201 (53.0) 178 (47.0) 379 (100)
Initiation of complementary feeding at 6 months 163 (43.0) 216 (57.0) 379 (100)
ijerph-16-01887-t002_Table 2
######
Association between maternal characteristics of mothers and infant feeding practices in rural Southern Nepal, 2016.
--------------------------------------------------------------------------------------------------------------
Variables Total, *N* (%)\ Infant Feeding Practices, Yes, *n* (%)
379 (100)
--------------------- ----------------- ---------------------------------------- -------------- --------------
Age (years) *p* = 0.371 *p* = 0.545 *p* = 0.127
\<20 86 (22.7) 41 (47.7) 50 (58.1) 44 (51.2)
20--34 265 (69.9) 106 (40.0) 136 (51.3) 105 (39.6)
≥35 years 28 (7.4) 10 (35.7) 15 (53.6) 14 (50.0)
Caste/ethnicity *p* = 0.141 *p* \< 0.001 *p* = 0.296
Dalit 58 (15.3) 18 (31.0) 132 (55.7) 99 (41.8)
Adibasi/Janajati 84 (22.2) 40 (47.6) 51 (60.7) 42 (50.0)
Upper caste Group 237 (62.5) 99 (41.8) 18 (31.0) 22 (37.9)
Birth origin *p* = 0.025 *p* = 0.097 *p* = 0.038
Hill 107 (28.2) 54 (50.5) 64 (59.8) 108 (39.7)
Terai 272 (71.8) 103 (37.9) 137 (50.4) 55 (48.6)
Women education *p* \< 0.001 *p* \< 0.001 *p* \< 0.001
No education 89 (23.5) 23 (25.8) 16 (18.0) 22 (24.7)
Primary 135 (35.6) 42 (31.1) 69 (51.1) 48 (35.6)
Secondary & higher 155 (40.9) 92 (59.4) 116 (74.8) 93 (60.0)
Women occupation *p* \< 0.0001 *p* \< 0.001 *p* \< 0.001
Labor 64 (16.9) 17 (26.6) 13 (20.3) 16 (25.0)
Agricultural work 122 (32.2) 36 (29.5) 55 (45.1) 47 (38.5)
Service/business/\ 193 (50.9) 104 (53.9) 133 (68.9) 100 (51.8)
household works
Family income *p* = 0.006 *p* = 0.002 *p* = 0.125
1st tertile 130 (34.3) 40 (30.8) 53 (40.8.) 47 (36.2)
2nd tertile 121 (31.9) 53 (43.8) 68 (56.2.) 54 (44.6)
3rd tertile 128 (33.8) 64 (50.0) 80 (62.5) 62 (48.8)
Types of family *p* = 0.029 *p* = 0.260 *p* = 0.341
Nuclear 180 (47.5) 85 (47.2) 90 (50.0.) 82 (45.6)
Joint 199 (52.5) 72 (36.2) 111 (55.8) 81 (40.7)
Parity *p* = 0.430 *p* \< 0.001 *p* = 0.177
Primi 148 (39.1) 65 (43.9) 95 (64.2) 70 (47.3)
Multi 231 (60.9) 92 (39.8) 106 (45.9) 93 (40.3)
Sex of Child *p* = 0.05 *p* = 0.097 *p* = 0.004
Male 178 (47.7) 77 (43.3) 102 (57.3) 90 (50.6)
Female 195 (52.3) 79 (40.5) 95 (48.7) 70 (35.9)
--------------------------------------------------------------------------------------------------------------
ijerph-16-01887-t003_Table 3
######
Association between utilization of maternal care services and infant feeding practices in rural Southern Nepal, 2016.
----------------------------------------------------------------------------------------------------------
Variables Total, N (%)\ Infant Feeding Practices\
379 (100) (Yes, %)
-------------------------------- --------------- --------------------------- -------------- --------------
MCH intervention (MATRI-SUMAN) *p* = 0.002 *p* = 0.001 *p* = 0.002
Received 193 (50.9) 95 (49.2) 118 (61.1) 98 (50.8)
Not received 186 (49.1) 62 (33.3) 83 (44.6) 65 (34.9)
Number of ANC visit *p* \< 0.001 *p* \< 0.001 *p* \< 0.001
\<4 ANC 148 (39.1) 30 (20.3) 31 (29.9) 35 (23.6)
4 or more 231 (60.9) 127 (55.0) 170 (73.6) 128 (55.4)
Place of Delivery *p* \< 0.001 *p* \< 0.001 *p* = 0.004
Home 220 (58.0) 61 (27.7) 88 (40.0) 73 (33.2)
Health facility 159 (42.0) 96 (60.4) 113 (71.1) 90 (56.6)
PNC visit *p* \< 0.0001 *p* = 0.01 *p* \< 0.001
Yes 219 (57.8) 120 (54.8) 162 (74.0) 39 (24.4)
No 160 (42.2) 37 (23.1) 39 (24.4) 124 (56.6)
----------------------------------------------------------------------------------------------------------
MCH, maternal and child health; MATRI-SUMAN, Maternal Alliance for Technological Research Initiative on Service Utilization and Maternal Nutrition; ANC, antenatal care; PNC, postnatal care.
ijerph-16-01887-t004_Table 4
######
Maternal factors and utilization of maternal care services associated with infant feeding practice in rural Southern Nepal, 2016 by Unadjusted Odds Ratio (OR).
------------------------------------------------------------------------------------------------------------
Variables Infant Feeding Practices (OR, 95% CI)
-------------------------------- --------------------------------------- ------------------ ----------------
Age (Years)
\<20 1.00 1.00 1.00
20--34 0.7 (0.4--1.1) 0.7 (0.4--1.2) 0.6 (0.3--1.0)
35--45 0.6 (0.2--1.4) 0.8 (0.3--1.9) 0.9 (0.4--2.2)
Caste/ethnicity
Dalit 1.00 1.00 1.00
Adibasi/Janajati 1.5 (0.8--2.9) 1.2 (0.7--2.0) 1.6 (0.8--3.2)
Upper caste group 2.0 (1.1--4.0) 0.3 (0.1--0.6) 1.1 (0.6--2.1)
Birth origin
Terai 1.00 1.00 1.00
Hill 1.6 (1.1--2.6) 1.4 (0.9--2.3) 1.6 (1.0--2.5)
Women education
No education 1.00 1.00 1.00
Primary 1.2 (0.7--2.3) 4.7 (2.5--9.0) 1.6 (0.9--3.0)
Secondary & higher 4.1 (2.3--7.4) 13.5 (7.0--26.0) 4.5 (2.5--8.1)
Women occupation
Labor 1.00 1.00 1.00
Agricultural work 1.1 (0.5--2.2) 3.2 (1.5--6.5) 1.8 (0.9--3.6)
Service/business/\ 3.2 (1.7--6.0) 8.6 (4.4--17.1) 3.2 (1.7--6.0)
household works
Family Income
1st tertile 1.00 1.00 1.00
2nd tertile 1.7 (1.1--2.9) 1.8 (1.1--3.0) 1.4 (0.8--2.3)
3rd tertile 2.2 (1.3--3.7) 2.4 (1.4--3.9) 1.6 (1.0--2.7)
Types of family
Joint 1.00 1.00 1.00
Nuclear 1.5 (1.0--2.3) 1.2 (0.8--1.8) 0.8 (0.5--1.2)
Parity
Multi 1.00 1.00 1.00
Primi 1.8 (0.7--1.7) 2.1 (1.3--3.2) 1.3 (0.8--2.0)
Sex of Child
Female 1.00 1.00 1.00
Male 1.1 (0.7--1.6) 1.4 (0.9--2.1) 1.8 (1.2--2.7)
MCH Intervention (MATRI-SUMAN)
Not received 1.00 1.00 1.00
Received 1.9 (1.2--2.9) 2.0 (1.3--3.1) 1.9 (1.2--2.9)
ANC visit
\<4 ANC 1.00 1.00 1.00
4 or more 4.8 (2.9--7.2) 10 (6.4--17.2) 4.0 (2.5--6.3)
Place of delivery
Home 1.00 1.00 1.00
Health facility 3.9 (2.5--6.1) 3.6 (2.3--5.7) 2.6 (1.7--4.0)
PNC visit
No 1.00 1.00 1.00
Yes 4.0 (2.5--6.3) 8.8 (5.5--14.1) 4.0 (2.5--6.3)
------------------------------------------------------------------------------------------------------------
MCH, maternal and child health; MATRI-SUMAN, Maternal Alliance for Technological Research Initiative on Service Utilization and Maternal Nutrition; ANC, antenatal care; PNC, postnatal care.
ijerph-16-01887-t005_Table 5
######
Adjusted Odds Ratio (aOR) for infant feeding practices among rural mothers of Southern Nepal, 2016.
Variable Infant Feeding Practices (aOR, 95% CI)
----------------------------------- ---------------------------------------- ----------------- ----------------
Age (Years)
\<20 1.00 1.00 1.00
20--34 0.4 (0.1--1.2) 0.3 (0.1--1.1) 0.6 (0.2--1.7)
35--45 0.5 (0.2--1.7) 0.3 (0.1--0.7) 0.9 (0.4--1.6)
Caste/ethnicity
Dalit 1.00 1.00 1.00
Adibasi/Janajati 1.7 (0.5--5.1) 1.1 (0.4--2.7) 1.0 (0.4--2.1)
Upper caste group 0.7 (0.4--1.4) 0.6 (0.3--1.2) 1.1 (0.6--1.9)
Birth origin
Terai 1.00 1.00 1.00
Hill 1.4 (0.7--2.5) 1.1 (0.3--3.0) 1.1 (0.6--1.9)
Women education
No education 1.00 1.00 1.00
Primary 1.2 (0.4--3.3) 2.2 (1.1--4.4) 2.2 (0.8--5.6)
Secondary & higher 2.2 (1.2--4.2) 5.5 (1.8--16.1) 2.2 (1.2--3.9)
Women occupation
Labor 1.00 1.00 1.00
Agricultural work 1.3 (0.4--3.8) 1.5 (0.5--5.0) 1.1 (0.6--3.0)
Service/business/household works 2.2 (1.2--4.2) 2.1 (1.1--4.1) 1.4 (0.5--4.0)
Family Income
1st tertile 1.00 1.00 1.00
2nd tertile 1.0 (0.5--2.0) 1.2 (0.6--2.4) 1.0 (0.4--2.4)
3rd tertile 1.1 (0.6--2.1) 1.5 (0.7--3.2) 1.1 (0.5--2.0)
Types of family
Joint 1.00 1.00 1.00
Nuclear 1.3 (0.8--2.1) 1.6 (0.9--2.9) 1.0 (0.6--1.6)
Parity
Multi 1.00 1.00 1.00
Primi 1.2 (0.6--2.1) 2.2 (1.2--4.0) 1.3 (0.8--2.0)
Sex of child
Female 1.00 1.00 1.00
Male 1.1 (0.6--1.7) 1.2 (0.7--2.0) 1.7 (1.1--2.7)
MCH Intervention (MATRI-SUMAN)
Not received 1.00 1.00 1.00
Received 1.7 (1.1--2.9) 1.8 (1.1--3.2) 1.6 (1.0--2.7)
ANC visit
\<4 ANC 1.00 1.00 1.00
4 or more 3.2 (1.2--8.0) 3.1 (1.3--7.5) 1.2 (0.5--2.8)
Place of delivery
Home 1.00 1.00 1.00
Health facility 1.9 (1.0--3.4) 1.2 (0.6--2.4) 1.1 (0.6--2.0)
PNC visit
No 1.00 1.00 1.00
Yes 1.0 (0.4--2.3) 2.4 (1.1--5.6) 2.3 (1.0--5.1)
MCH, maternal and child health; MATRI-SUMAN, Maternal Alliance for Technological Research Initiative on Service Utilization and Maternal Nutrition; ANC, antenatal care; PNC, postnatal care. Variables entered for each outcome variables: age group, caste/ethnicity, birth origin, women's occupation, women's education, family income, types of family, MCH intervention received, parity, ANC visit, place of delivery, PNC, sex of child.
| {
"pile_set_name": "PubMed Central"
} |
1.. Introduction
================
The air-assisted sprayers used in fruit production must be carefully and effectively regulated to ensure that crops are treated successfully. Four main factors affect the deposition efficiency: the nozzle type, the fluid pressure, the ground speed, and the volumetric flow rate of the air. The combination of these parameters determines the applied volume rate. This factor directly influences the quality of the treatment \[[@b1-sensors-15-02399],[@b2-sensors-15-02399]\].
Considering the effect of the air volumetric flow rate, the fan setting directly affects the treatment quality \[[@b3-sensors-15-02399]\]. If the air velocities or volumes produced by the fans are too low, insufficient pesticide will reach the trees. On the other hand, if these velocities or volumes are too high, the pesticide will be blown over and through the trees. Commonly, the operating manuals provided with these sprayers do not provide information on the characteristics of the air flow generated by the machine. In this sense, the air flow generated by the sprayer can be characterized by using high-precision anemometers such as sonic anemometers (2D and 3D) which are used to measure the velocity components for different heights, sections, and distances from the sprayer \[[@b4-sensors-15-02399],[@b5-sensors-15-02399]\].
The use of experimental methods to characterize the air flow generated by a sprayer would be difficult and expensive \[[@b6-sensors-15-02399]\]. Instead, the air velocity in the vicinity of a sprayer can be simulated by applying integrated computational fluid dynamics (CFD) \[[@b7-sensors-15-02399],[@b8-sensors-15-02399]\]. CFD models are a useful tool for improving the design and development of these machines by enabling the testing of different design alternatives and reducing the number of measurements to those required to validate the models. Nevertheless, some measured data remains necessary and, to achieve this, the use of high-precision anemometers is required.
CFD models can be used to estimate the air velocity by considering the sprayer to be either stationary or moving \[[@b9-sensors-15-02399]--[@b11-sensors-15-02399]\]. The first step in analysing the air flow generated by a sprayer consists of determining the air flow patterns while the sprayer is stationary, with no forward motion of the sprayer and without considering any interaction with the crop \[[@b11-sensors-15-02399]\]. The static characterisation of the air flow pattern allows us to understand how the sprayer will perform when moving forward at a certain speed. Several studies have shown the relationship between the measurements obtained when the sprayer is either stationary or moving forward \[[@b10-sensors-15-02399],[@b11-sensors-15-02399]\], whereby the air flow pattern of a moving sprayer exhibits a decrease in the air velocity relative to the values obtained when the sprayer is stationary \[[@b10-sensors-15-02399]\]. This decrease becomes more pronounced as the forward speed of the sprayer increases.
CFD models have been successfully applied to the simulation of the operation of different sprayers used in agriculture: boom sprayers \[[@b12-sensors-15-02399]\], the rotary atomizers used on agricultural aircraft for ultra-low volume spraying operations \[[@b13-sensors-15-02399]\], recycling tunnel sprayers for orchards \[[@b14-sensors-15-02399]\], and to a greater degree, in the air-assisted sprayers used in orchards \[[@b15-sensors-15-02399],[@b16-sensors-15-02399]\].
Although 2D CFD models were used originally, 3D models are better suited to representing the air flow generated by a sprayer \[[@b17-sensors-15-02399]\]. In fact, CFD models have been improved to study further aspects affecting the efficacy of the treatment, including the interaction between the air flow and the crop or the influence of environmental parameters (wind speed, temperature) on the treatment \[[@b18-sensors-15-02399],[@b19-sensors-15-02399]\].
Most CFD models that have been developed to simulate the air flow generated by an air-assisted sprayer have assumed the use of the air velocity measured at the fan outlet as a boundary condition \[[@b6-sensors-15-02399],[@b11-sensors-15-02399],[@b20-sensors-15-02399]\]. This methodology is very useful for analysing the characteristics of the air flow generated by a sprayer considering its specific configuration: outlet width, deflector position, fan blade angle, and fan speed. However, taking the design of the sprayer into consideration, it is important to obtain simulation models capable of simulating the air flow produced by the fan at different settings, so that the air velocity at the fan outlet is not a boundary condition. This type of model would be a flexible tool for simulating different air outlet geometries for a specific air flow, thus avoiding the need to measure the air speed at the fan outlet for each design. In this case, the total air flow produced by the sprayer must be measured for each fan configuration, requiring experimental work that, in any case, the sprayer manufacturer must perform to obtain technical information on the air flow generated by the fan. This is necessary because the operational adjustment of a sprayer should be based on a knowledge of the air flow created by the fan and the geometry of the discharge outlet \[[@b10-sensors-15-02399]\].
The air flow pattern depends on the number and configuration of the sprayer fans. Currently, conventional orchard sprayers consist of a fan located behind a tank. For this kind of machine, different studies have shown the viability of using CFD to estimate the generated air velocity \[[@b6-sensors-15-02399],[@b11-sensors-15-02399],[@b15-sensors-15-02399],[@b16-sensors-15-02399],[@b18-sensors-15-02399]--[@b20-sensors-15-02399]\]. However, no CFD models have been developed for a sprayer equipped with two axial fans. This kind of machine has been developed in recent years, with several different designs appearing \[[@b7-sensors-15-02399],[@b21-sensors-15-02399]\]: two fans placed side by side behind the tank, two fans placed in line behind the tank, one fan placed in front of the tank and another behind, two fans behind the tank at different heights, and so on. In addition, the models have been based on the use of air velocities measured at the fan outlet as a boundary condition.
The objective of this work was to develop and validate, by experiment, a CFD model for predicting the air velocity distributions for an air-assisted sprayer equipped with two axial fans, assuming the total air flow instead of the outlet air velocity to be the main parameter of the CFD model and to validate the model by using 3D sonic anemometers to take actual measurements. The model was developed for different fan speeds and assuming the sprayer to be stationary.
2.. Experimental Section
========================
2.1.. Air Sprayer
-----------------
The analysed sprayer was a product of Gar Melet S. L. (Huesca, Spain). The sprayer ([Figure 1](#f1-sensors-15-02399){ref-type="fig"}) was equipped with two reversed-rotation axial fans \[[@b4-sensors-15-02399]\], one placed behind the tank and the other placed in front. These fans spin in opposite directions to ensure a uniform distribution of the chemical product to be applied. When viewed from the tractor the front fan spins anticlockwise and the rear fan clockwise. Each fan sucks air axially from the outer area of the machine and throws it radially ([Figure 2](#f2-sensors-15-02399){ref-type="fig"}). The diameter of the front fan was 800 mm and that of the rear fan was 830 mm. The CFD model was developed assuming the operation of both of these fans.
The fan blade angles can be adjusted, through a range enumerated as 1 to 5, to create different air flows. The air flows were measured according to the ISO 9898:2000 standard \[[@b22-sensors-15-02399]\], considering three fan blade positions (1.5; 3; 4.5). The air flow produced by the rear fan was measured at the fan\'s inlet, using a TESTO 0635 1041 hot-wire anemometer (Testo AG, Lenzkirch, Germany; accuracy: 0.03 m/s; range: 0 to 20 m/s). The measurements were carried out with the power take-off (PTO) turning at 540 rpm.
2.2.. Experimental Measurements
-------------------------------
The velocity of the air generated by the sprayer was measured in the absence of any wind by using a WindMaster 3D sonic anemometer (Gill Instruments, Lymington, UK) according to the methodology developed by García-Ramos *et al.* \[[@b4-sensors-15-02399]\]. The accuracy of the sonic anemometer was 1.5% (for wind speeds up to the maximum measureable value) with an air velocity range of 0 to 45 m/s, and a resolution of 0.01 m/s. The air velocity data was recorded at a frequency of 1 Hz.
Measurements were carried out with the sprayer operating but stationary, for three different settings of the rear fan blades (1.5, 3, and 4.5) and, therefore, three different air volumetric flow rates ([Table 1](#t1-sensors-15-02399){ref-type="table"}). The air velocity was measured in three sections of the sprayer: A, B, and C ([Figure 3](#f3-sensors-15-02399){ref-type="fig"}). For each section, measurements were performed on both sides of the machine at 1.5, 2.5, and 3.5 m from the centre of the sprayer, at heights of 1, 2, 3, and 4 m above the ground ([Figure 4](#f4-sensors-15-02399){ref-type="fig"}). For each measuring point and sprayer setting, the air velocity was recorded over a 60-s duration. The anemometer orientation for the measurements is shown in [Figure 3](#f3-sensors-15-02399){ref-type="fig"}.
2.3.. Model Formulation
-----------------------
A CFD simulation software, specifically, the commercial ANSYS-CFX 14.5 CFD code, was used to analyse the air flow generated by the sprayer. This study used the Spalart--Allmaras model \[[@b23-sensors-15-02399]\]. This is a single-equation model ([Equation (1)](#FD1){ref-type="disp-formula"}) that solves a modelled transport equation for a turbulent kinematic viscosity. The Spalart--Allmaras model is implemented specifically for applications involving wall-bound flows \[[@b24-sensors-15-02399]--[@b26-sensors-15-02399]\]. For this reason, this model is used for turbomachinery applications \[[@b27-sensors-15-02399]--[@b30-sensors-15-02399]\]. The Spalart--Allmaras model adds an extra condition, specifically, that turbulent kinematic viscosity at the walls is zero: $$\frac{\partial\overline{v}}{\partial t} + u_{i}\frac{\partial\overline{v}}{\partial x_{i}} = G_{v} + \frac{1}{\sigma}\left\lbrack {\nabla\left( {v + \overline{v}} \right)\ \nabla\overline{v}} \right\rbrack + c_{b2}\left| {\nabla v} \right|^{2} - Y_{v}$$ $$G_{v} = c_{b1}\left\lbrack {1 - f_{t2}} \right\rbrack\ \overset{\sim}{S}\overline{v} + \frac{c_{b1}}{k^{2}}f_{t2}\left( \frac{\overline{v}}{d} \right)^{2}$$ $$Y_{v} = c_{w1}f_{w}\left( \frac{\overline{v}}{d} \right)^{2}$$
[Equation (1)](#FD1){ref-type="disp-formula"} represents the governing equation for the Spalart--Allmaras model. In these equations, *G~v~* and *Y~v~* are the production and the destruction of turbulent viscosity, respectively ([Equations (2)](#FD2){ref-type="disp-formula"} and [(3)](#FD3){ref-type="disp-formula"}). This effect occurs in the near-wall region due to the effects of wall blocking and viscous damping.
In [Equation (1)](#FD1){ref-type="disp-formula"}, *u* represents the fluid velocity (obtained previously by solving the Navier--Stokes (N--S) equations), *v* is the molecular kinematic viscosity, and *d* ([Equations (2)](#FD2){ref-type="disp-formula"} and [(3)](#FD3){ref-type="disp-formula"}) is the distance to the closest surface, with: $$\begin{matrix}
\begin{matrix}
{\overset{\sim}{S} = S + \frac{\overline{v}}{k^{2}d^{2}}f_{v2};} & {f_{v2} = 1 - \frac{X}{1 + Xf_{v1}};} & {f_{v1} = \frac{X^{3}}{X^{3} + {C_{v1}}^{3}};} & {C_{v1} = 7.1} \\
\end{matrix} \\
\begin{matrix}
{S = \sqrt{2\Omega_{ij}\Omega_{ij}};} & {\Omega_{ij} = \frac{1}{2}\left( {\frac{\partial u_{i}}{\partial x_{j}} - \frac{\partial u_{j}}{\partial x_{i}}} \right)} \\
\end{matrix} \\
\begin{matrix}
{f_{w} = g\left\lbrack \frac{1 + {C_{w3}}^{6}}{g + {C_{w3}}^{6}} \right\rbrack^{1/6};} & {g = r + C_{w2}(r^{6} - r);} & {r = \frac{\overline{v}}{\overset{\sim}{S}k^{2}d^{2}};} & {C_{w2} = 0.3} \\
\end{matrix} \\
\begin{matrix}
{f_{t2} = C_{t3}\exp( - C_{t4}X^{2});} & {C_{t3} = 1.2;} & {C_{t4} = 0.5} \\
\end{matrix} \\
\begin{matrix}
{C_{w1} = \frac{C_{b1}}{k^{2}} + \frac{(1 + C_{b2})}{\sigma};} & {C_{b1} = 0.1355;} & {C_{b2} = 0.622;} & {K = 0.41;} & {\sigma = \frac{2}{3}} \\
\end{matrix} \\
\end{matrix}$$
The turbulent kinematic viscosity, *v~t~*, is given by: $$\begin{matrix}
{v_{t} = \overline{v}f_{v1};} & {f_{v1} = \frac{X^{3}}{X^{3} + {C_{v1}}^{3}};} & {X = \frac{\overline{v}}{v}} \\
\end{matrix}$$
Detailed information about the Spalart-Allmaras model is given in \[[@b23-sensors-15-02399]\].
2.4.. Simulation Domain and Load Cases
--------------------------------------
To develop the study, a numerical analysis was performed. The simulation domain used in the numerical analysis was 15 m long in the driving direction, 15 m wide, and 8.9 m high. These dimensions are necessary to avoid errors in the results due to the contours of the simulation domain. [Figure 5](#f5-sensors-15-02399){ref-type="fig"} shows the simulation domain.
In this simulation domain, the research group modelled an air-assisted sprayer with two fans. The air-assisted sprayer was located in the centre of the domain base, with the rotors of the fans 0.9 m above ground level, as shown in [Figure 6](#f6-sensors-15-02399){ref-type="fig"}. The diameters of the front and rear fans were 0.8 m and 0.9 m, respectively. The distance between the fans was 3 m. In addition, the air-assisted sprayer was equipped with a water tank between the two fans, which was also modelled.
The tractor was assumed to be stationary during the research carried out. Considering the air flow at the entrance of the fans, the actual air flows were measured experimentally in such way that the effect of the tractor on the air input was included in the measurement. On the other hand, the CFD numerical model considered the experimental air flow measured at the entrance of both fans as input parameter, then, the air flow sucked by the fans in the numerical analysis was not affected by the presence of the tractor because it was an input parameter. Considering the air velocity generated by each fan in the vicinity of the sprayer, the air flows generated in static way were perpendicular to the tractor-sprayer system ([Figure 2](#f2-sensors-15-02399){ref-type="fig"}). In this sense, the tractor does not affect the air velocity values at the measurement points for each section of the sprayer. For this reasons, the effects of the tractor on the results would be negligible. Consequently, the tractor was not included in the numerical model.
The meshing process was carried out in two phases. In the first phase, an optimizing mesh process was carried out in order to fulfill the convergence criteria imposed. The convergence criterion applied in the study was that residual values obtained in the numerical analysis were below 10^−4^. Once the convergence criterion was fulfilled, a mesh-independence study was carried out in order to verify the accuracy of the numerical model. In this study the number of elements has been increased up to 50% with respect to the initial number of elements. The numerical results obtained from the different meshes generated in the mesh-independence study showed similar values, with mean differences around 2.1%, but the computational cost was greatly increased by increasing the number of elements in the model. Once this process was finished, the conclusion obtained was that the numerical model obtained after the optimization phase of the mesh was correct. This numerical model was configured using hexahedral and shell elements. The hexahedral elements were used to simulate the fluid, while the shell elements were used to define the fans of the air-assisted sprayer and the sides of the simulation domain. The model consisted of 4,682,881 nodes and 948,735 elements.
The first boundary condition imposed in the simulation was to constrain the movement of the air-assisted sprayer. In addition, the air flow generated by the fans can cross the lateral sides and top of the simulation domain. In these areas a pressure outlet boundary condition has been imposed. At the bottom of the domain, the air flow cannot cross the ground.
In the developed study, three load cases were analysed. These corresponded to three possible settings of the fan blades. The manufacturer of the sprayer enumerates these positions as "1.5, 3, and 4.5".
To implement the simulation, two components of the air velocity in the fan are required for each load case. These components are the axial (c~m~) and rotational (c~t~) velocity of the air at the outlet of the fan blades ([Figure 7](#f7-sensors-15-02399){ref-type="fig"}). The values of c~m~ and c~t~ were calculated using the equations derived from [Figure 7](#f7-sensors-15-02399){ref-type="fig"}.
In modern air-assisted sprayers, the fans have two types of blades. The first of these are fixed blades, called "stators", while the other type can move, and are called "rotors". The function of the stators is to guide the air flow so as to improve the performance of the machine when the air comes into contact with the moving blades. The inclinations of the stator blades and rotor blades are designated γ and α, respectively.
The initial data for the analysis, in addition to the inclinations of the blades, includes the geometric data of the machine as well as the air flow to the fan inlet (Q, [Table 1](#t1-sensors-15-02399){ref-type="table"}) and the angular velocity of the blades (w). The values of Q and w were obtained by experiment.
The c~te~ parameter corresponds to the rotational component of the air flow caused by the stator. This is calculated from the axial component of the air flow at the fan inlet (c~m~) and the inclination of the stator blades (γ).
The c~t~ parameter corresponds to the rotational component of the air flow caused by the relative velocity of the rotor. It is calculated from the axial component of the air flow at the fan inlet (c~m~) and the inclination of the rotor blades (α).
[Equations (4)](#FD4){ref-type="disp-formula"}--[(8)](#FD8){ref-type="disp-formula"} show the relationship between the parameters shown in [Figure 6](#f6-sensors-15-02399){ref-type="fig"}. $$\overset{\rightarrow}{\text{Q}} = \text{S} \cdot \overset{\rightarrow}{\text{c}_{\text{m}}}$$ $$\overset{\rightarrow}{\text{v}_{\text{e}}} = \overset{\rightarrow}{\text{c}_{\text{te}}} + \overset{\rightarrow}{\text{c}_{\text{m}}}$$ $$\overset{\rightarrow}{\text{c}} = \overset{\rightarrow}{\text{u}} + {\overset{\rightarrow}{\text{v}}}_{\text{l}}$$ $$\overset{\rightarrow}{\text{c}} = \overset{\rightarrow}{\text{c}_{\text{m}}} + \overset{\rightarrow}{\text{c}_{\text{t}}}$$ $$\overset{\rightarrow}{\text{u}} = \text{R} \cdot \overset{\rightarrow}{\text{w}}$$
The values of the parameters used in this study are listed in [Table 1](#t1-sensors-15-02399){ref-type="table"}, where S and r are the air passage of the fan and the radius of the fan, respectively.
2.5.. Statistical Analysis
--------------------------
The development of the CFD model was analysed by calculating the coefficient of determination (R^2^) between the simulated and actual measurements. To analyze the effect of distance to the machine (horizontal and vertical) we calculated the correlation between the two measures, assuming that both fans are running, for all possible configurations.
The effects of the main variables affecting the air velocity values, as determined with the CFD model ([Table 2](#t2-sensors-15-02399){ref-type="table"}), were analysed statistically. A generalized linear model with a gamma distribution and a logarithmic link function was used to establish the principal effect of the variables on the air velocity \[[@b31-sensors-15-02399]\]. In addition, two-way interactions between the variable "analysis type" and other fixed effects were analysed. The analysis of other interactions was not considered relevant to validate the viability of the CFD model.
A total of 432 data were used in the statistical analysis, 216 obtained from the experimental measurements (considering the mean air velocity for each measurement point), and 216 from the CFD model ([Table 2](#t2-sensors-15-02399){ref-type="table"}).
To analyse the differences between the measured and simulated values, the data matrix was partitioned and the data was analysed, one on one, for each combination of variables, paying attention to the significance of the double interactions after completion of the generalized linear model. The set of experimental observations of the variable velocity of the air generated by the sprayer follows a normal distribution and homocedasticity, however, obtained with CFD model does not (Kolomogorov-Smirnov test 0.226; *p* \< 0.001). The air velocity was transformed into a logarithmic function to obtain a normalized distribution. The Student t test was used, with an appropriate value being chosen according to the homogeneity of the variances after the application of the Levene test. All statistical analyzes were performed with SAS (SAS Inst. Inc., Cary, NC, USA).
3.. Results and Discussion
==========================
3.1.. Correlation between CFD and Measured Results
--------------------------------------------------
[Figures 8](#f8-sensors-15-02399){ref-type="fig"}, [9](#f9-sensors-15-02399){ref-type="fig"} and [10](#f10-sensors-15-02399){ref-type="fig"} show the CFD results for a specific fan configuration (both front and rear fans set to fan setting 3), considering three sections of the sprayer and viewing the sprayer from behind. The simulation shows the asymmetry of the air flow generated by the fans due to the direction in which it is turning. Because of this, the use of two fans spinning in opposite directions compensates for the effect of the asymmetry, thus producing an overall air flow that is more uniform on both sides of the machine.
The air velocity values generated by the CFD model were in good agreement with those obtained by \[[@b11-sensors-15-02399]\] who showed, through a test with a stationary traditional sprayer, that the air velocities decreased with the distance and height from the sprayer, but increased with the air flow. The variation in the air velocity values was similar to that obtained by \[[@b10-sensors-15-02399]\], who concluded that, for a static test using a traditional sprayer, the air velocities at a distance of 3.5 m were half the value at 1.75 m.
The overall R and R^2^ coefficients, considering all of the data, were 0.927 and 0.859, respectively. This shows that there is a good correlation between the simulated and measured data. The results were excellent ([Table 3](#t3-sensors-15-02399){ref-type="table"}), with R^2^ more than 0.856, with an improvement in the areas closest to the sprayer.
Considering the effect of the air flow (fan settings of 1.5, 3, and 4.5) the best adjustment was obtained at the lower air flows. On the other hand, the worst adjustment was that corresponding to a fan setting of 4.5, although the correlations were satisfactory with R^2^ = 0.825.
The obtained correlation values indicate that the CFD model can be used as a useful tool for predicting the actual values of the air velocity in the vicinity of a sprayer. However, it is necessary to analyse the error level and the influence of each variable (distance, height, air flow) in the model in more detail.
3.2.. Influence of Spraying Variables on CFD Model
--------------------------------------------------
The air velocity was significantly affected by all of the variables except for the "analysis type" and the "sprayer side" ([Table 4](#t4-sensors-15-02399){ref-type="table"}). The fact that the "analysis type" had no significant effect on the air velocity, coupled with the high values of the coefficients of determination, supports the validity of the use of the CFD model to estimate the air velocity in the vicinity of the sprayer.
Considering the effect of the "sprayer side" on the CFD and measured air velocities, no significant differences were found between the different sides of the sprayer. This fact has been analysed by other researchers \[[@b10-sensors-15-02399]\], who concluded that, for most sprayers, it is possible to observe at least a small difference in the air field between the left and right sides. On the other hand, a sprayer design in which the fans turn in opposite directions produces an air flow that is symmetrically distributed.
The degree of influence of the variables in the CFD model on the air velocity, are listed in [Table 4](#t4-sensors-15-02399){ref-type="table"}. In descending order of importance, they are measurement section, height, horizontal distance, air flow, analysis type, and sprayer side.
Considering the interaction of the "analysis type" with the other variables, it was not significant for "air flow", "sprayer side" and "height". However, two significant interactions were found for the case of "analysis type" × "measurement section" and "analysis type" × "horizontal distance" ([Table 4](#t4-sensors-15-02399){ref-type="table"}).
Considering the effect of the "measurement section", the velocity values were found to be significantly higher in the centre section for the CFD model. However, no significant differences were found in the front and rear sections ([Table 5](#t5-sensors-15-02399){ref-type="table"}). [Figure 11](#f11-sensors-15-02399){ref-type="fig"} shows the 95% confidence interval. Considering the mean values for the air velocity listed in [Table 5](#t5-sensors-15-02399){ref-type="table"}, the mean absolute errors obtained were 3.79 for the rear section, 33.55% for the centre section, and 13.89% for the front section.
This fact is concordant with that shown in [Table 4](#t4-sensors-15-02399){ref-type="table"} which reflects a significant effect of the interaction of the variables "analysis type" and "measurement section" on the air velocity values. Considering the interaction between the variables, this significant effect ([Table 4](#t4-sensors-15-02399){ref-type="table"}) is explained because the CFD data were lower than experimental ones for the front section and, by the other hand, for rear and centre sections, the CFD model values were higher than the experimental ones ([Table 5](#t5-sensors-15-02399){ref-type="table"}).
For the variable "height", no significant differences were obtained except at a height of 3 m ([Table 6](#t6-sensors-15-02399){ref-type="table"} and [Figure 12](#f12-sensors-15-02399){ref-type="fig"}). The air velocities were found to fall with an increase in the measurement height. Considering that Mediterranean fruit orchards are typically between 1 and 4 m in height, the CFD model was able to predict the air velocities with a mean error of 9.92%.
The air velocity values obtained with the CFD model did not differ significantly from the measured values for horizontal distances up to 2.5 m ([Table 7](#t7-sensors-15-02399){ref-type="table"} and [Figure 13](#f13-sensors-15-02399){ref-type="fig"}) with a mean error of 5.72%. This means that, given that fruit orchards typically have planting widths of 5 m, the CFD model is in fairly good agreement with the measured values. This fact is concordant with that shown in [Table 4](#t4-sensors-15-02399){ref-type="table"} which reflects a significant effect of the interaction of the variables "analysis type" and "horizontal distance" on the air velocity values, showing that. Analysing this interaction, and considering low horizontal distances (until 1.5 m), CFD data were lower than experimental ones However this fact was the opposite for distances of 2.5 and 3.5 m where the CFD model values were higher than the experimental ones.
Upon analysing the effects of the air flow generated by the fans on the air velocity values ([Table 8](#t8-sensors-15-02399){ref-type="table"} and [Figure 14](#f14-sensors-15-02399){ref-type="fig"}), no significant differences were found. The results were similar for both the low and high air flows with no clear trends apparent.
A comparison of the measured and estimated values for the air velocity, considering a zone no more than 3 m in height and less than 2.5 m from the sprayer, which corresponds to a typical treatment area in a Mediterranean fruit orchard, revealed a mean error of 9.19%. These results were in good agreement with those of previous research. In this sense, \[[@b11-sensors-15-02399]\] incurred error of 25% for a sprayer equipped with a rear fan and operating while stationary, at a measurement distance of 1.75 m. In a similar way, \[[@b19-sensors-15-02399]\] incurred errors of less than 20% in 95% of the measurements made for three sprayers at forward speeds of 7.1 km/h, with mean values having an error of less than 11%.
3.3.. Usefulness of the CFD Model
---------------------------------
The obtained correlation values between experimental and simulated data indicate that the CFD model can be used as a useful tool for predicting the actual values of the air velocity in the vicinity of the sprayer. Besides, information of [Table 4](#t4-sensors-15-02399){ref-type="table"} demonstrates the usefulness of the CFD model since the statistical analysis carried out shows that the results obtained by numerical simulation were not significantly different from those obtained experimentally. This fact, added to the high values of the coefficient of determinations between experiment and model ([Table 3](#t3-sensors-15-02399){ref-type="table"}), supports the use of the CFD model as a valid alternative to the experimental methods. In addition, the CFD model has shown its robustness to analyze the relative importance of the variables that affect the characteristics of the airflow generated as it has been showed in [Tables 5](#t5-sensors-15-02399){ref-type="table"}, [6](#t6-sensors-15-02399){ref-type="table"}, [7](#t7-sensors-15-02399){ref-type="table"} and [8](#t8-sensors-15-02399){ref-type="table"}. The fitting of the data supplied by the CFD model with the experimental measurements also supports the use of the Spalart Allmaras turbulence model to analyse the air flow generated by the sprayer.
Considering the experimental measurements, the research shows the utility of using high precision sensors (sonic and hot-wire anemometers) to measure the components of the air velocity for different heights, sections, and distances from the sprayer. The use of this type of sensors is required to obtain accurate experimental data which will be used to validate the CFD model. This fact is in concordance with previous studies \[[@b4-sensors-15-02399],[@b5-sensors-15-02399]\].
As conclusion, one of the main advantages of the proposal numerical model is the use, for each fan, of the total air flow aspirated as the main parameter to be introduced in the model instead of the outlet air velocities used in traditional models. This methodology reduces the requirement of specific experimental measurements for each configuration of air flow in the fans, and reduces the time required to analyse the performance of a sprayer equipped with two fans. In this sense, the proposed CFD model will be an useful tool for sprayer manufacturers to improve the design phase of the machine by predicting, for different fan configurations, the characteristics of the air velocities generated by the sprayer in the vicinity of the machine which will let to analyse the influence of different designs of the machine, with different dispositions of the front and back fans.
4.. Conclusions
===============
The application of CFD models to the estimation of the air velocity distributions generated by an air-assisted sprayer equipped with two axial fans and operating while stationary, considering the total air flow rather than the outlet air velocity as the main parameter of the CFD model, was validated as being an effective method.
Considering all of the data, the air velocity values obtained with the CFD model were found to be in good agreement with the measured data. The global coefficient of determination between the CFD model and measured data was 0.859. Considering a zone of up to 3 m in height and 2.5 m from the sprayer, the mean error between the measured and estimated values was found to be 9.19%. The degree of influence of the variables used in the CFD model on the air velocity, in descending order of importance, was: the measurement section, height, horizontal distance, air flow, analysis type, and sprayer side. Significant differences arose only for specific combinations of three variables, with the values predicted by the computer simulation being significantly higher than those measured by experiment. These variables were, in descending order of importance, the measurement section (centre), height (3 m), and horizontal distance (3.5 m).
The authors express their gratitude to Gar Melet S. L. for their assistance with this research.
F. J. García-Ramos, H. Malón, and M. Vidal have contributed to all phases of the work (CFD modelling, testing, data analysis, and report writing). J. Aguirre contributed to the data analysis and report-writing. A. Boné contributed to the testing, and J. Puyuelo contributed to the CFD modelling and data analysis.
The authors declare no conflict of interest.
{#f1-sensors-15-02399}
{#f2-sensors-15-02399}
{#f3-sensors-15-02399}
{#f4-sensors-15-02399}
{#f5-sensors-15-02399}
{#f6-sensors-15-02399}
{#f7-sensors-15-02399}
{#f8-sensors-15-02399}
{#f9-sensors-15-02399}
{#f10-sensors-15-02399}
{#f11-sensors-15-02399}
{#f12-sensors-15-02399}
{#f13-sensors-15-02399}
{#f14-sensors-15-02399}
######
Input data for CFD analysis.
**Parameter** **Rear Fan Air Flow** **Front Fan Air Flow**
------------------ ----------------------- ------------------------ ------- ------- ------- -------
**c~te~ (m/s)** 13.03 17.56 19.38
**γ(°)** 60 60 60 60 60 60
**W (rad/s)** 255 255 237 198 198 184
**α (°)** 36.91 47.16 58.21 36.91 47.16 58.21
**u (m/s)** 76.5 76.5 71.1 59.3 59.3 55.2
**S (m^2^)** 0.59 0.59 0.59 0.46 0.46 0.46
**v~e~ (m/s)** 26.06 35.12 38.76
**c~m~ (m/s)** 22.57 30.42 33.57 16.62 23.51 26.76
**c~t~ (m/s)** 33.40 30.71 30.90 37.18 37.49 38.60
**Q (m^3^/s)** 13.31 17.94 19.8 7.60 10.81 12.32
**r (m)** 0.45 0.45 0.45 0.4 0.4 0.4
**w~s~ (rad/s)** 111 102 103 124 125 128
######
Variables affecting air velocity values determined with CFD model. N: number of observations.
**Variable Name** **Variable Configurations** **N (Each Level)**
----------------------------------- --------------------------------- --------------------
Analysis type Experimental; simulated (CFD) 216
Measurement section Front (A); Centre (B); Rear (C) 72
Air flow 1.5; 3; 4.5 24
Sprayer side (viewed from behind) Left; Right 12
Horizontal distance 1.5 m; 2.5 m; 3.5 m 4
Height 1 m; 2 m; 3 m; 4 m 1
######
Coefficients of determination (R^2^) according to horizontal and vertical distance from machine. Overall values for three measurement sections (A, B, C) and three fan settings (1.5, 3, 4.5).
**Height (m)** **Horizontal Distance (m)**
---------------- ----------------------------- ------- -------
**4** 0.911 0.883 0.856
**3** 0.922 0.885 0.895
**2** 0.971 0.947 0.929
######
Effect of variables on air velocity when using generalized linear model with sum of squares Type III, results contrast Omnibus: Chi square likelihood ratio = 440.292; *p* \< 0.001. Deviation obtained with this model in relation to a gamma distribution with a logarithmic link function was 137.738.
**Variable (Interactions)** **Wald Chi-Square** **df** **Significance**
----------------------------- ------------------------- -------- ------------------ ---------
**(Interception)** 1806.389 1 \<0.001
**Analysis type** 3.054 1 0.081
**Measurement section** 368.217 2 \<0.001
**Air flow** 15.722 2 \<0.001
**Sprayer side** 0.498 1 0.480
**Horizontal distance** 79.843 2 \<0.001
**Height** 137.583 3 \<0.001
**Analysis type ×** **Measurement section** 14.455 2 \<0.001
**Air flow** 0.670 2 0.715
**Sprayer side** 0.801 1 0.371
**Horizontal distance** 13.388 2 \<0.001
**Height** 6.691 3 0.082
######
Air velocity values according to analysis type and measurement section. The Student t test was used to compare the means.
**Measurement Section** **Analysis Type** **Probability**
------------------------- ------------------- ----------------- ------ -------- ---------
**Rear** 6.83 6.58 0.25 0.374 0.709
**Centre** 2.03 1.52 0.51 4.586 \<0.001
**Front** 2.85 3.31 0.45 −0.311 0.756
######
Air velocity values according to the analysis type and the height. The Student t test was used to compare the means.
**Height (m)** **Analysis Type** **Probability**
---------------- ------------------- ----------------- ------ -------- -------
1 5.64 6.24 0.60 −0.593 0.555
2 4.93 4.59 0.34 0.468 0.640
3 3.42 2.91 0.51 2.006 0.048
4 2.04 1.94 0.10 0.716 0.475
######
Air velocity values according to analysis type and horizontal distance. The Student t test was used to compare the means.
**Horizontal Distance (m)** **Analysis Type** **Probability**
----------------------------- ------------------- ----------------- ------ -------- -------
1.5 5.31 5.67 0.36 −0.425 0.672
2.5 3.71 3.53 0.18 0.966 0.336
3.5 3.01 2.56 0.45 2.062 0.041
######
Air velocity values of air flows generated by fans. The Student t test was used to compare the means.
**Air Flow** **Analysis Type** **Probability**
-------------- ------------------- ----------------- ------ ------- -------
1.5 3.37 3.33 0.04 0.637 0.525
3 4.22 4.18 0.04 0.800 0.425
4.5 4.43 4.25 0.17 0.735 0.464
[^1]: These authors contributed equally to this work.
[^2]: Academic Editor: Gonzalo Pajares Martinsanz
| {
"pile_set_name": "PubMed Central"
} |
Cell polarity is a fundamental organizing principle in metazoa that is necessary for cell division, differentiation and morphogenesis. Polarization of epithelia is implicit in the development of lumens, which are essential for glandular tissues to carry out their normal functions, but polarity is disrupted in diseases such as cancer. Understanding mechanisms of polarity will therefore improve future prospects for therapy.
The establishment of epithelial polarity involves a coordinated series of events by several protein complexes, leading to the asymmetric segregation of plasma membranes into apical and basolateral domains. Polarity is accompanied and maintained by asymmetric distribution of intracellular organelles, together with dynamic cytoskeleton and membrane trafficking^[@R1]-[@R4]^. However, the mechanisms defining the spatial orientation of polarity are not well understood.
Cell interactions with the extracellular matrix (ECM) contribute to the organisation of the basolateral surface, which creates an apical face on the opposite membrane^[@R5]-[@R7]^. Although the formation of apical domains are relatively well understood, little is known about how cells respond to an ECM in order to locate their apical membranes correctly. We hypothesised that an intracellular network regulated by cell-ECM interactions determines the orientation of epithelial polarity and therefore lumen formation.
Signals from the ECM are transmitted through integrins, which connect to the cytoskeleton, and to adaptor proteins within adhesion complexes^[@R8],\ [@R9]^. Integrins control cell shape, migration, proliferation and differentiation, and also have a role in the establishment of polarity. For example, β1 integrin ablation resulted in a loss of polarity leading to defective arterial lumen formation and asymmetric cell division in skin epithelia^[@R10]^ ^[@R11]^. In kidney epithelia, function-perturbing antibodies revealed that β1-integrins regulate basement membrane (BM) assembly, which is necessary for polarity^[@R12]^. However it is not known how BM-integrins subsequently signal inside the cell to establish the apico-basal polarity axis.
The mammary gland epithelium is organized into a branched network of polarized ducts and acini. Lactating acini are composed of a monolayer of luminal epithelial cells surrounded by a sparse network of myoepithelia, and jointly subtended by a BM. The luminal cells are polarized, with the apical secretory face adjacent to a lumen and the basolateral surface in contact with the BM. This asymmetric orientation of apico-basal membranes ensures that the cells secrete their milk products into the lumen.
We previously showed that β1-integrins determine the differentiated function of mammary glands^[@R13]^. We now demonstrate that β1-integrins also control the orientation of epithelial polarity and thereby the formation of lumens in differentiated acini. Our results reveal the mechanism by which β1-integrins regulate tissue polarity and lumen formation downstream of cell interactions with the BM^[@R12]^.
Results {#S1}
=======
β1-integrins are required for glandular lumen formation downstream of cell-BM interactions {#S2}
------------------------------------------------------------------------------------------
The role of cell-matrix interactions in lumen formation have been studied in kidney epithelia cultured in collagen I gels. Here, function-blocking anti-β1-integrin antibodies prevent the formation of polarized cysts. This is because integrin signalling via IRSp53-Rac1 induces the cells to deposit and organize a BM around the cyst periphery, which then contributes to lumen formation^[@R12],\ [@R14],\ [@R15]^. However the intracellular mechanism for lumen formation downstream of BM is unknown.
To distinguish the intracellular function of integrins from their role in BM assembly, we used primary luminal mammary epithelial cells (MECs) cultured within an exogenous BM-matrix, in which the cells form lactational acini. We generated *β1^fx/fx^;CreER™* mice, which permitted β1-integrin gene deletion in MECs using 4-hydroxy-tamoxifen (4OHT)^[@R16]^. Immunofluorescence staining showed that untreated wild type (WT) acini develop lumens with apical f-actin, lateral E-cadherin, and basolateral β1-integrins ([Fig. 1a](#F1){ref-type="fig"}). Treatment with 4OHT at the time of plating cells caused β1-integrin gene deletion (β1-KO), and the acini were unable to develop lumens ([Fig. 1a,c](#F1){ref-type="fig"}). Lumen formation in MECs from non-transgenic ICR mice was unaffected by 4OHT ([Fig. 1b,c](#F1){ref-type="fig"}). Thus, β1-integrins are required for MECs cultured on BM to form hollow acini.
To confirm the role of β1-integrins in acinar morphogenesis, we analysed mammary glands from *β1^fx/fx^;Blg-Cre* mice *in vivo* (*β1−/−*). Cre recombinase driven by both the Blg and Wap promoters is activated in mid-pregnancy, specifically in luminal epithelial cells but not in myoepithelia^[@R13]^. By employing this strategy, myoepithelial cells were still able to make and deposit a BM around the acini, allowing integrin contribution downstream of BM assembly to be analysed ([Fig. 1f](#F1){ref-type="fig"}). At lactation day 2 (L2), β1-integrin gene deletion resulted in defective acinar morphogenesis with epithelial cells filling the luminal space whilst WT acini (β1^fx/fx^, with no Cre) displayed a single central lumen ([Fig. 1d, e](#F1){ref-type="fig"}).
β1-integrins are thus required for normal mammary lumen formation, both *in vivo* and in a primary culture model downstream of a BM.
Integrin mediated lumen formation requires ILK but not Rac1 {#S3}
-----------------------------------------------------------
To determine whether Rac1 is required to establish glandular lumens, we generated *Rac1^fx/fx^;LSLYFP;CreER™* mice. In MECs from these mice, 4OHT specifically deleted Rac1 and the cells expressed YFP ([Supplementary Fig. 1a-c](#SD3){ref-type="supplementary-material"}). Unlike β1-integrin, Rac1 deletion did not prevent mammary acini from developing lumens ([Fig. 1c,g](#F1){ref-type="fig"}). We confirmed this by generating *Rac1^fx/fx^:LSLYFP:WAPiCre* mice to delete the Rac1 gene *in vivo* ([Supplementary Fig. 1d-g](#SD3){ref-type="supplementary-material"}). Lactating mammary acini were still able to form polarized lumens ([Fig. 1h,i,j](#F1){ref-type="fig"}). These data indicate that Rac1 is not required for lumen formation downstream of a BM-integrin axis and that integrins establish intracellular polarity via a distinct mechanism to the molecular pathway involved in BM assembly.
To identify proximal integrin signalling components controlling lumen formation, we analysed two focal adhesion proteins, integrin-linked kinase (ILK) and focal adhesion kinase (FAK). We reasoned that these proteins might be involved because deletion of β1-integrins in MECs resulted in displacement of ILK from the basal cell surface and dephosphorylation of FAK^Y397^ ([Supplementary Fig. 2a](#SD3){ref-type="supplementary-material"}). We generated *Ilk^fx/fx^:CreER™* mice and analysed acini after 4OHT treatment to remove the ILK gene ([Supplementary Fig. 1h-k](#SD3){ref-type="supplementary-material"}). ILK deletion resulted in 90% of acini containing filled lumens ([Fig. 1c,k](#F1){ref-type="fig"}). In contrast, FAK deletion in MECs isolated from *FAK^fx/fx^* mice did not inhibit lumen formation (not shown).
To determine whether these integrin effectors control lumen formation *in vivo*, we generated *FAK^fx/fx^;Blg-Cre* and *Ilk^fx/fx^;Blg-Cre* mice ([Supplementary Fig. 1l,m](#SD3){ref-type="supplementary-material"}). ILK deletion resulted in abnormal morphogenesis similar to the *β1−/−* phenotype, with cells filling the luminal space of acini ([Fig. 1l-n](#F1){ref-type="fig"}). In contrast, lactating FAK-null glands showed no morphogenesis defects^[@R17]^.
These data show that downstream of cell-BM interactions, ILK has a key role in linking integrins with MEC lumen formation, but FAK and Rac1 do not.
Mechanisms not involved in integrin-mediated lumen formation {#S4}
------------------------------------------------------------
Integrin signalling is instructive in controlling lumen formation and utilizes a distinct pathway to the Rac-dependent pathway for BM assembly in MDCK cells. To identify possible mechanisms, we analysed whether apoptosis of internal cells caused lumen formation^[@R18]^. In WT MECs, low levels of apoptosis were observed over a time-course of acinar morphogenesis both in the internal cells and those contacting the BM ([Fig. 2a, b](#F2){ref-type="fig"} arrows). Lumens developed in the presence of the caspase inhibitor zVAD ([Fig. 2d](#F2){ref-type="fig"}). Thus, as with some other epithelial models, apoptosis is not the mechanism for lumen formation in primary MECs^[@R19]-[@R22]^. Moreover, β1-integrin deletion did not alter apoptosis ([Fig. 2c](#F2){ref-type="fig"}).
We investigated the possibility that β1-integrin ablation disrupted tissue organization by affecting adherens junctions. Scribble and E-cadherin localized to cell-cell junctions of internal cells in β1-KO MEC acini and *β1−/−* glands ([Fig 1a](#F1){ref-type="fig"}, [Fig. 2e](#F2){ref-type="fig"}, [Supplementary Fig. 3](#SD3){ref-type="supplementary-material"},), as well as ILK-deleted glands and acini ([Fig. 1k, m](#F1){ref-type="fig"}). These findings are consistent with other cell types, e.g. E-cadherin localizes normally in *β1−/−* keratinocytes^[@R23]^.
β1-integrin-null acini are therefore not filled with cells because of defective apoptosis or altered intercellular adhesions.
β1 integrins orient epithelial polarity via ILK {#S5}
-----------------------------------------------
Lumen formation requires apical polarity which is typically controlled by the PAR, Scribble and Crumbs complexes^[@R24]^. We asked whether integrins are needed to establish apico-basal polarity. A time-course of MEC lumen development revealed that 1 day after plating, acini lacked lumens, and aPKC was unpolarized ([Fig. 3a,b,d](#F3){ref-type="fig"}). During β1-integrin engagement with BM, aPKC was displaced away from the basolateral membrane and multiple aPKC-lined lumens developed, progressively forming a single lumen. However in the absence of β1-integrins, aPKC, ZO-1 and PAR3 did not relocalize but instead were inverted at the outer membrane ([Fig. 3c,d](#F3){ref-type="fig"} [Supplementary Fig. 4a](#SD3){ref-type="supplementary-material"}, [Movies S1](#SD5){ref-type="supplementary-material"}, [S2](#SD6){ref-type="supplementary-material"}). Electron microscopy revealed a proper apical membrane containing microvilli and tight junctions bordering the BM in ®1-KO acini ([Fig. 3e](#F3){ref-type="fig"}, [Supplementary Fig. 4b](#SD3){ref-type="supplementary-material"}). Transferrin receptors were mislocalised, indicating lost basolateral polarity, and the ability of MECs to internalize transferrin-488 from the media was diminished ([Supplementary Fig. 4c,d](#SD3){ref-type="supplementary-material"}). These results suggest that apical domain assembly is not integrin-independent. Rather, ®1-integrins are required to form the epithelial basolateral surface and to separate it topologically from the apical domain.
®1-integrin deletion caused polarity inversion despite culturing acini on a BM suggesting that other laminin-binding ®-integrins were unable to compensate for β1-integrin loss. Analysis of the laminin-binding α6- and ®4-integrins showed they were juxtaposed to the BM in WT acini, but redistributed from the basal domain in ®1-null acini ([Fig 3f-h](#F3){ref-type="fig"}). In previous studies, genetic deletion of α6- and ®4-integrins had no observable effect on mammary acinar morphogenesis in vivo^[@R25]^.
We investigated whether integrin polarity signals require ILK. 90% of acini lacking ILK had inverted apical polarity and filled lumens ([Fig. 3i,j](#F3){ref-type="fig"}). Moreover, adenoviral expression of ILKEGFPf reverted polarity and lumens in 89% of ILK-KO acini ([Fig. 3k](#F3){ref-type="fig"}).
β1-integrins therefore orient apical polarity away from the cell-BM interface by establishing BM-cell interactions and by controlling intracellular signalling via ILK. This determines the location of the apical domain and thereby the position of the lumen.
β1-integrins and ILK control internal organelle polarity {#S6}
--------------------------------------------------------
Trafficking of newly synthesized proteins from the trans Golgi network (TGN) is essential for cell polarization^[@R26]^. We investigated whether reduced β1-integrin signalling alters Golgi positioning. WT acini contained sub-apical Golgi (GM130) (80% of cells, [Fig. 4a](#F4){ref-type="fig"}). However,®1-integrin deletion resulted in either redistribution of Golgi to the periphery (51.1%) or Golgi ribbon fragmentation (35.9%) ([Fig. 4c](#F4){ref-type="fig"}, left). Fragmentation of Golgi correlated with a loss of apical membrane polarity ([Fig 4d](#F4){ref-type="fig"}). *In vivo*, WT acini displayed apical Golgi (95% of cells), which were scattered or undetectable in *β1−/−* glands ([Fig. 4b,c](#F4){ref-type="fig"}, right). ILK deletion reproduced the effects of β1-integrin deletion ([Fig. 4e-g](#F4){ref-type="fig"}).
Golgi positioning at the pericentrosome is principally controlled by the microtubule (MT) cytoskeleton, a process driven by dynein and dynactin motors, and treatments with MT depolymerising reagents (Nocodazole) disperse Golgi into ministacks^[@R27]^. In polarized acini cultured from non-transgenic ICR MECs, Nocodazole caused Golgi fragmentation similar to the integrin/ILK knockouts ([Fig. 4h](#F4){ref-type="fig"}).
Although individual Golgi units are functional for membrane trafficking ^[@R27]^, Golgi scattering in β1-integrin and ILK null epithelial cells correlated with the loss of polarized trafficking. This could additionally contribute to abnormal lumen development.
β1-integrin adhesions control the spatial orientation of microtubules {#S7}
---------------------------------------------------------------------
Since ®1-integrins and ILK deletion caused Golgi to redistribute or disperse into ministacks, we reasoned that integrin adhesion might control polarity by organizing MTs. We examined microtubule orientation by analysing their plus tips. In WT acini, MTs aligned along the apico-basal polarity axis with EB1 positioned towards the basolateral membrane ([Fig. 5a,b](#F5){ref-type="fig"}, left panel). Over the time-course of acinar polarization, integrin reorganisation coincided with recruitment of EB1 towards the basolateral membrane ([Supplementary Fig. 5](#SD3){ref-type="supplementary-material"}). By contrast, in the KO MECs, MTs were not able to form a discrete apical-basal orientation and EB1 was scattered throughout the cell, suggesting the presence of short unstable MTs ([Fig. 5a,b](#F5){ref-type="fig"}, right panel). This was confirmed by immunostaining acetylated MTs which were enriched near the apical cortex in WT acini, but destabilized in the absence of ILK ([Fig. 5c](#F5){ref-type="fig"}). Moreover, WT acini showed reduced MT acetylation at the basolateral pole suggesting that they are more dynamic here and undergo temporary stabilizations only. Destabilizing MTs with Nocodazole treatment of polarized acini also disrupted polarity and lumens, but this effect was reversed when the drug was washed out ([Fig. 5d,e](#F5){ref-type="fig"}).
The results suggest that integrin adhesions recruit EB1, which then establishes apico-basal MT orientation and cell polarity. We therefore determined whether EB1 is physically associated with integrin adhesions. *In situ* Proximity Ligation Assays showed that β1-integrin adhesions formed a complex with EB1 specifically at the basal surface of polarized WT acini ([Fig. 5f](#F5){ref-type="fig"}). Notably this complex was disrupted in the absence of ILK, indicating that ILK links integrin adhesions with EB1 ([Fig. 5g](#F5){ref-type="fig"}). To determine if EB1 is required for apical lumens, we depleted it from MECs using an shRNA lentivirus ([Supplementary Fig. 6](#SD3){ref-type="supplementary-material"}). EB1 knockdown resulted in defective formation of mammary acini, with fewer than 95% forming discernable lumens ([Fig. 5h, i](#F5){ref-type="fig"}).
These findings reveal that integrin-ILK complexes anchor MT plus ends to the basolateral cell surface via EB1 thereby providing a mechanism for MT orientation. A possible outcome is that integrin anchoring of MT plus ends may temporarily stabilize MTs, thus allowing deposition of basolateral trafficking cargo at this membrane. It follows that some of this cargo may trigger the removal of apical proteins from the outer membrane, thereby re-orienting polarity.
Microtubules are necessary for the orientation of apico-basal polarity {#S8}
----------------------------------------------------------------------
To address the role of MT dynamics in polarity re-orientation, we tested the effects of MT disrupting drugs on the formation of mammary acini. In time-course studies, paclitaxel stabilization prevented MT reorientation along the apicobasal polarity axis, and simultaneously inhibited aPKC redistribution from the cell-BM interface to the luminal surface. The acini failed to form lumens ([Fig. 6a](#F6){ref-type="fig"}). Similarly, Nocodazole inhibited lumen formation ([Fig. 6b](#F6){ref-type="fig"}).
This result suggests that the outer membrane of acini remodels when it contacts a BM, and that MTs might be important to actively direct apical proteins away from the cell-BM interface. To test this hypothesis, we used MEC monolayers in order to be able to experimentally manipulate the apical cell surface, by adding BM proteins to the culture media to form a new cell-BM interface (BM-overlay).
Monolayer cultures of MECs were polarized, with ZO1 at apical tight junctions ([Fig. 6fi](#F6){ref-type="fig"}). ®1-integrins were absent from the top surface of the monolayer ([Fig 6cii](#F6){ref-type="fig"}), but BM-overlay caused integrins to locate at the top surface ([Fig 6dii](#F6){ref-type="fig"}) and ZO1 was removed ([Fig 6f iii](#F6){ref-type="fig"}). Under these conditions, exogenous Tfr488 uptake increased, indicating that the top surface of the cells was functioning as a basolateral membrane ([Fig 6c-e](#F6){ref-type="fig"}). MT stabilization prevented the BM-induced removal of ZO1 ([Fig 6fiv](#F6){ref-type="fig"}).
MTs are therefore needed to remove apical proteins away from the cell-BM interface and to orient apico-basal polarity.
β1-integrins orient epithelial polarity via endocytosis of apical components from the cell-BM interface {#S9}
-------------------------------------------------------------------------------------------------------
To investigate the mechanism for BM-matrix induced apical protein removal, we tested whether endocytosis is involved by examining the requirement of the endocytic trafficking dynamin GTPase. Dynamin inhibitors, MiTMAB and dynasore, blocked endocytosis in MECs because they inhibited Tfr488 and cholera toxin-B uptake from the media ([Supplementary Fig. 7](#SD3){ref-type="supplementary-material"}). Both inhibitors, as well as adenoviral expression of dominant-negative K44A-dynamin 1 and 2, prevented BM-induced removal of ZO1 ([Fig. 7a](#F7){ref-type="fig"} and not shown). Moreover, K44A-dynamin 1 and 2 prevented polarity establishment and lumen formation in mammary acini ([Fig. 7b](#F7){ref-type="fig"}). To confirm the role of early endosomal trafficking in polarity, we expressed dominant-negative S34NRab5a in MECs, which also prevented lumen formation ([Fig. 7c](#F7){ref-type="fig"}).
These results suggest that apical proteins are removed from the cell-BM interface by endocytosis. We investigated the involvement of ®1-integrins using the BM-overlay assay. Whilst BM removed ZO1 from the apical surface of WT, it was unable to do so after β1-integrin gene deletion ([Fig. 7d](#F7){ref-type="fig"}). To confirm the role of ®1-integrins in apical protein endocytosis, we tracked the internalization of a transmembrane tight junction protein, Claudin 7, in response to 6 h BM-overlay using surface biotinylation. An increased pool of endocytosed Claudin 7 was detected in response to BM-overlay ([Fig. 7e](#F7){ref-type="fig"}, compare lanes 4 with 6) but this was reduced after deleting β1-integrins ([Fig. 7e](#F7){ref-type="fig"}, lane 7).
Our results show that the engagement of BM proteins with ®1-integrins coordinate MTs to trigger the endocytic removal of apical proteins. This remodels the BM-cell interface into a basal domain and creates an apical domain on the opposing membrane, allowing lumen development. In the absence of ®1-integrins or MT dynamics, the apical components remain on the basolateral surface.
β1-integrins maintain apical polarity {#S10}
-------------------------------------
We also investigated whether ®1-integrins maintain the polarized orientation of apical proteins. Deleting the ®1-integrin gene in *β1^fx/fx^;CreER™* MECs after the acini had fully developed caused inversion of the apical components, actin and aPKC, to the outer membrane ([Fig. 8a](#F8){ref-type="fig"}). Maintenance of polarity is therefore an active process that is governed by ®1-integrins, and requires the continuous endocytic removal of apical components from the basolateral membrane.
Discussion {#S11}
==========
In this paper, we provide mechanistic insights into how integrins control apico-basal polarity in epithelia, leading to the formation of lumens and functional glandular tissues. BM interactions with epithelia provide a microenvironmental cue to govern the spatial orientation of cell polarity. However, the intracellular pathways by which integrin-BM interactions contribute to polarity was previously unknown.
Using genetic deletion in a primary 3D mammary culture model, we have shown that ®1-integrins function through ILK in organising MTs, which then promote the establishment of polarity at multiple levels. ®1-integrins and MTs collaborate in the endocytic removal of tight junction proteins from the outer membrane, thereby defining a basolateral surface. The integrin-ILK-MT network also positions the Golgi sub-apically in order that they can govern polarized trafficking of proteins and create a new apical face on the opposing membrane ([Fig. 8b](#F8){ref-type="fig"}).
β1-integrins are not involved in establishing an apical domain, however, active BM-®1-integrin interactions are required to maintain an apical surface adjacent to the lumen, because without ®1-integrins, luminal tight junctions disrupt and reassemble on the surface next to the BM. A possible mechanism is that apical cargo is initially targeted to the basolateral surface, internalized and then transcytosed to the apical domain. This mode of protein transport is utilized by a number of epithelial cell types, e.g. hepatocytes, intestinal epithelia and MDCK cells ^[@R4]^. Thus a loss of integrin signalling in MECs may inhibit endocytosis at the basolateral membrane, resulting in an accumulation of apical polarity proteins and thereby inversion of polarity.
A model emerges in which β1-integrins recruit ILK and capture EB1 plus tips in order to orient MTs. Consistent with this model, mass spectrometry approaches have identified integrin adhesion connections to MT plus ends, e.g. ILK interacts with α and β-tubulin, with IQGAP1 and mDia, and fibronectin-β1-integrin adhesions contain EB1 ^[@R28],\ [@R29]^ ^[@R30]^. Since ILK deletion affects MT stability, integrin-ILK adhesions may bind MT plus ends to organize and stabilise them. Indeed MT stability at the apical cortex was reduced in the absence of ILK suggesting that plus-end anchoring of MTs at the basolateral membrane may serve to redirect apical components along stabilised MTs towards the apical membrane. Similarly, in colonic epithelia, the kinesin KIF17 stabilizes MTs by interacting with EB1, which contributes to epithelial polarization^[@R31]^.
MTs serve as tracks for polarized vesicular transport by orienting their minus ends towards the apical domain and plus ends towards the basal surface ^[@R32],\ [@R33]^. One possibility is that β1-integrin-EB1 connections may deliver factors to the basolateral membrane that facilitate the endocytosis of apical components. Caveolae might have a key role here, because ®1-integrin depletion blocks caveolar endocytosis in skin fibroblasts and caveolae are involved with endocytosis of tight junctions in both gut and brain endothelia ^[@R34]-[@R36]^. Moreover in keratinocytes, MT stabilisation by ILK is essential for targeting and insertion of caveolae at the plasma membrane^[@R30]^. This suggests that depletion of ®1-integrins/ILK, or inhibiting MT dynamics, may perturb caveolar trafficking to the MEC cell surface, thereby preventing the internalisation of apical proteins.
We have shown that ®1-integrins are required for endocytosing apical proteins and that inhibiting early endosomal trafficking with DN-Rab5a blocks lumen formation. This suggests that some endocytosed apical components may be recycled to generate a new luminal membrane. Consistent with our studies, in MDCK cells, Rab11 positive recycling vesicles transport the polarity complexes, Par3-aPKC and Crumbs3-Pals1-PatJ to early lumens, but the full contribution of synthesis vs recycling is yet to be established ^[@R37],\ [@R38]^.
The work presented here identifies a mechanism by which cell-matrix interactions provide orientation cues for organising cell polarity within tissues. Future investigations with live cell imaging will characterize in more detail the roles of these integrin adhesion complexes in the morphogenesis of the mammary gland.
Online Methods {#S12}
==============
Mouse strains {#S13}
-------------
*β1^fx/fx^; Blg-Cre* mice, *Ilk^fx/fx^; Blg-Cre* mice, *FAK^fx/fx^; Blg-Cre* mice and β*^1fx/fx^CreER™* mice have been described previously ^[@R13],\ [@R17]^. The *Ilk^fx/fx^* or *Rac1^fx/fx^ :LSLRosaYFP ^[@R39]^* and CreER™ lines were crossed to produce the *ILK^fx/fx^;CreER™* or *Rac1^fx/fx^:LSLYFP;CreER™* mice. For the Rac1 *in vivo* analysis, Cre-mediated specific *Rac1* gene deletion in luminal mammary epithelial cells was achieved by crossing *Rac1^fx/fx^ :LSLRosaYFP* mice with *WapiCre^Tg/•^* mice to produce *Rac1^fx/fx^YFP :WapiCre^Tg/•^* (*Rac1*^−/−^) mice. To avoid problems in feeding of pups by mothers with potentially defective mammary glands, only the male mice of the breeding pairs carried the *Cre* transgene. *Rac1^fx/fx^:LSLYFP* mice that lacked the *Cre* gene were used as wild type (WT) controls. Genotyping was performed using DNA prepared from ear punches^[@R39]-[@R41]^. Note that the mammary specific Blg-Cre and WAPiCre promotors were used to conditionally delete the β1 integrin^fx/fx^ gene in luminal MECs but not in myoepithelia or stromal cells ^[@R13]^. BlgCre is activated in nulliparous mice between 8 and 12 weeks whereas WAPiCre is activated during pregnancy. In some experiments, non-transgenic ICR mice were used as the source of primary MECs. Mice were housed and maintained according to the University of Manchester and UK Home Office guidelines for animal research.
Primary cell culture and gene deletion {#S14}
--------------------------------------
Primary MECs were harvested from 15.5-17.5 day pregnant mice and cultured as described in ^[@R42]^. Cre-mediated deletion of β1-integrin or ILK or Rac1 in primary MEC cultures was achieved by harvesting the MECs from *β1^fx/fx^:CreER™* or *Ilk^fx/fx^;CreER™* mice, or *Rac1^fx/fx^:LSLYFP:CreER™* and treating with 100 nM 4-hydroxytamoxifen dissolved in ethanol. β1-integrin gene deletion after acini had polarized was performed by addition of 4OHT to 4-5 day old cultures.
Cells were plated onto Collagen 1 for monolayer cultures, BM-matrix (Matrigel; BD Biosciences) to form acini and cultured in growth media (Ham's F12 medium (Sigma) containing 5 μg/ml insulin, 1 μg/ml hydrocortisone (Sigma), 3 ng/ml epidermal growth factor (EGF), 10% fetal calf serum (Biowittaker), 50 U/ml Penicillin/Streptomycin, 0.25 μg/ml fungizone and 50 μg/ml gentamycin).
For BM-overlay assays, monolayer MECs were overlaid with diluted BM-matrix (1:50) in DMEM:F12 medium (containing 5 μg/ml insulin, 1 μg/ml hydrocortisone, 3 ng/ml EGF, 50 U/ml Penicillin/Streptomycin, 0.25 μg/ml fungizone and 50 μg/ml gentamycin) for 48h, or 1:25 for 1-6h ^[@R43]^.
Inhibitors: In some experiments cells were treated with 10 μM zVAD; dynamin inhibitors 40 μM MitMAB (Calbiochem), 100 μM Dynasore (Sigma); or microtubule inhibitors 200 ng/ml Nocodazole, 100 nM or 1 μM paclitaxel (Sigma).
Genomic PCR {#S15}
-----------
Genomic DNA was isolated from mammary tissue or cultured *Ilk^fx/fx^:CreER™* or *Rac1^fx/fx^:LSL- YFP:CreER™* MECs following 4OHT addition (24h) and analysed by PCR. DNA was extracted with Direct PCR lysis reagent (Viagen Botech, LA) containing 10 μg/ml proteinase K. The position of PCR was carried out as described ^[@R44]^. The PCR reaction products were 2.1 kB (Ilk flox) and 230bp (Ilk flox recombined and Ilk null), or 333bp (Rac1 flox) and 175bp (Rac1 flox recombined and Rac1 null).
Adenovirus infection {#S16}
--------------------
Primary cells were infected in suspension, and then replated onto BM-matrix for 48h to form acini ^[@R45]^. Greater than 80% infection was achieved. HA tagged K44A Dynamin 1, K44A Dynamin 2, S34NRab5a and Tetracyclin regulator (TetR) adenoviruses ^[@R46]^ ^[@R47]^. To avoid overexpression of viruses, cells were cultured in the presence of 5 ng/ml doxycycline. Ad-GFP and Ad-ILKEGFPf (wild type ILK with farnesylated GFP) have been previously described ^[@R17]^. For rescue experiments β1 integrin or ILK was depleted first in monolayer with 4OHT. Cells were trypsinized, infected in suspension with Ad-GFP or Ad-ILKEGFPf and plated onto BM-matrix as previously described ^[@R17]^. Cells were harvested 48h later for immunofluorescence.
Lentiviral ShRNA knockdown {#S17}
--------------------------
The mouse EB1 and scrambled shRNA sequences were as described ^[@R48]^. Double stranded oligonucleotides were cloned into shRNA transfer vectors pLVTHM (Tronolab). ShRNA constructs were transfected into Swiss 3T3 fibroblasts using Lipofectamine plus (invitrogen) to verify EB1 knockdown. Lentivirus production in 293T cells was as previously described ^[@R49]^. EPH4 mammary epithelial cells were cultured in DMEM;F12 containing 5% fetal calf serum, 5 μg/ml insulin and 50 U/ml Penicillin/Streptomycin. Lentivirus infection of EPH4 MECs was performed by adding lentiviral particles to 50% confluent monolayers in the presence of polybrene. Greater that 80% infection was achieved. Media was replaced after 12h and infected cells were cultured for 3-4 days. Cells were replated onto BM-matrix for a further 4 days to form acini.
Immunostaining {#S18}
--------------
Expression and distribution of proteins were visualized by indirect immunofluorescence as in ^[@R45]^. For visualising EB1, cells were fixed for 10 min in 90% methanol/ 3% (w/v) paraformaldehyde/ 5mM sodium carbonate (pH 9). F-actin was detected by incubating cells with TRITC or FITC-phalloidin, (Sigma) or Alexa 647 phalloidin (Molecular Probes), for 1 h at RT, and nuclei were stained using 4 μg/ml Hoechst 33258 (Sigma) for 5 min at RT, followed by mounting in prolong gold antifade (Molecular Probes). Acini were visualized by confocal imaging. Images were collected on a Leica TCS SP5 AOBS inverted confocal using a *63x Plan Fluotar* objective. The confocal settings were as follows, pinhole *1 airy unit*, scan speed *1000Hz unidirectional*, format *1024 × 1024*. For Z stacks, 0.2μm sections were taken and Leica software was used to determine the optimal number of Z sections.
Images, 3D rendering and movies were developed with Volocity software (Perkin-Elmer) and Image J64. Quantification of acini was scored by analyzing 100 acini for each condition. Where approximately half the cells within an acinus displayed a change in morphology, these acini were scored as positive for that change. Non-biased cell counts were performed by concealing the identity of each slide.
For some experiments cells were immunostained with primary antibodies followed by in situ Duolink ™ proximity ligation assays to detect protein interactions, performed according to the manufacturer's instructions (Olink Biosciences, Sweden).
Immunofluorescence of mammary tissue was performed on paraffin-embedded tissue or cryosections (7μm)^[@R17]^ and luminal surface was detected with wheat germ agglutinin-488 (Invitrogen) and imaged using confocal microscopy. Primary antibodies used for immunofluorescence have been described in [Supplementary table 1](#SD4){ref-type="supplementary-material"}. Alexa Flour 488-conjugated wheat germ agglutinin (Molecular Probes), secondary antibodies conjugated to Cy2, Rhodamine-RX and Cy5 (Jackson Immunoresearch).
Histological analysis {#S19}
---------------------
Mammary tissue sections were stained with haematoxylin and eosin (H&E) and imaged as in ^[@R17]^.
TEM {#S20}
---
Acini were fixed in 1.5% Glutaraldehyde (in 0.1M cacodylate buffer) for 2h, washed 3x in 0.1M sodium cacodylate buffer and embedded in a serum plug ^[@R50]^. After several washes in 0.1M sodium cacodylate buffer containing 3mM calcium chloride (Agar Scientific Ltd, Stansted, UK), the cell plug was diced into small pieces and post-fixed in 1% osmium tetroxide (Agar Scientific Ltd) in 0.05M sodium cacodylate buffer pH 7.3 for 1h at 4°C followed by a rinse in buffer. Cells were dehydrated in an ascending alcohol series, treated twice with propylene oxide (15 mins each) then left in a 1:1 mix of propylene oxide and TAAB epoxy resin (Taab Laboratories Equipment Ltd., Aldermaston, UK) for 1h at RT followed by rotating overnight at 4°C in a mixture of 1:3 propylene oxide and epoxy resin. The cell plug was given two changes of fresh resin at 45°C for 1h each before being embedded in gelatin capsules and polymerized for 72h at 60°C.
Ultrathin sections were prepared with a diamond knife, mounted on copper grids and stained with uranyl acetate and lead citrate. Images were captured on a Philips CM10 electron microscope at an accelerating voltage of 80kV, with a Deben camera.
Cell fractionation {#S21}
------------------
Acini were isolated using Matrisperse (BD Biosciences), lysed in hypotonic lysis buffer (10mM Tris pH7.5, 1.5mM magnesium chloride, 10mM sodium chloride 10μg/ml leupeptin, 10μg/ml aprotinin, 1mM sodium fluoride, 1mM sodium orthovanadate, 1mM PMSF) and cells were fractionated into cytosolic and membrane fractions ^[@R51]^. Alternate fractions were separated by SDS-PAGE and immunoblotted with indicated antibodies.
Surface biotinylation {#S22}
---------------------
Assays for surface biotinylation were as described in ^[@R52]^. In brief, confluent MEC monolayers on 100 mm dishes were labelled with 0.5mg/ml sulfo-NHS-SS-biotin (Pierce, Rockford) for 30 min on ice with rocking. Free biotin was quenched using 50 mM ammonium chloride. Cells were left untreated or overlaid with BM-matrix (1:25) and incubated at 37°C for 6 hrs to induce endocytosis of surface apical proteins. To remove biotin on the cell surface, cells were stripped three times with 100mM MESNA buffer (pH 8.6) for 20 min each at 4°C, followed by three 5min incubations with 5mg/ml iodoacetamide at 4°C to quench free SH groups. To detect efficient stripping of the surface, some cells were kept at 4°C throughout the procedure to block endocytosis and stripped as above with MESNA or treated with the same buffer without MESNA. Cells were lysed in 1x Nonidet P40 buffer (10% w/v glycerol, 50 mM Tris-HCl pH7.5, 100 mM NaCl, 1% w/v Nonidet-P40, 2 mM MgCl~2~ and fresh protease/phosphatase inhibitors) and biotinylated proteins were isolated with NeutrAvidin Agarose.
Protein analysis {#S23}
----------------
Proteins were extracted and immunoblotted as in ^[@R45]^. Equal amounts of proteins were used and equivalent loading assessed by referral to controls, such as Calnexin (Bioquote).
Clathrin and caveolar endocytic assays {#S24}
--------------------------------------
Cells were incubated with 25 μg/ml labelled transferrin (Tfr488; Invitrogen) for 30 min or 10 μg/ml cholera toxin B (CtxB-FITC; Sigma) for 15 min at 37°C to internalize fluorescent markers. Cells were washed in PBS and incubated with an acid-salt wash buffer (0.2 M HAc, 0.5 M NaCl) for 10 min at 4°C to strip surface-bound fluorescent marker, followed by PBS wash, and fixed in 4% formaldehyde for 10 min. Internalized fluorescent markers were either analysed by confocal microscopy or quantitated in black 24 well clear bottom plates using a Berthold fluorescent plate reader. P values were determined using the Student's t test. \*p \< 0.007; \*\*p \< 0.02; \*\*\*p \< 0.05.
Supplementary Material {#SM}
======================
We thank Carolyn Jones for doing the electron microscopy, Sandra Schmid and Brian Ceresa for dynamin and Rab5a adenoviruses, Cary Wu for anti-ILK monoclonal, Martin Lowe for GM130 polyclonal, and Peter March for confocal microscope training. The FLS Bioimaging Facility microscopes were purchased with grants from BBSRC, Wellcome and the University of Manchester Strategic Fund. Thanks to Pat Caswell, Tim Hardingham, Martin Humphries, Martin Lowe and Paul Lu and for critical appraisal of the manuscript. This work was supported by the Wellcome Trust \[\#081203/Z/06/Z\]. The Wellcome Trust Centre for Cell-Matrix Research is supported by core funding from the Wellcome Trust \[\#088785/Z/09/Z\].
**Author contributions** NA conceived ideas, performed experiments, analysed and interpreted the data and wrote the manuscript.
CHS conceived ideas and wrote the manuscript.
{ref-type="supplementary-material"}, [2](#SD3){ref-type="supplementary-material"}.](emss-50423-f0001){#F1}
{ref-type="supplementary-material"}, [8](#SD3){ref-type="supplementary-material"}.](emss-50423-f0002){#F2}
{ref-type="supplementary-material"}, [8](#SD3){ref-type="supplementary-material"}, [Movies S1](#SD5){ref-type="supplementary-material"} and [S2](#SD6){ref-type="supplementary-material"}](emss-50423-f0003){#F3}
{ref-type="fig"}).\
(h) ICR acini, treated with DMSO or Nocodazole (24h). Arrow indicates Golgi dispersal after MT disruption. Bar: 10μm.](emss-50423-f0004){#F4}
{ref-type="supplementary-material"}, [6](#SD3){ref-type="supplementary-material"}.](emss-50423-f0005){#F5}
{#F6}
![β1-integrins orient polarity via an endocytic mechanism\
(a) Confocal images of monolayer ICR MECs (i) untreated or pre-treated for 1 h with (ii) MitMAB or (iii) dynasore prior to 6 h BM-overlay. Blocking dynamin prevented ZO1-internalization by the BM-overlay. Note that dynamin inhibition by itself slightly reduced apical tight junctions compared to untreated controls, possibly as a result of decreased vesicle budding from the Golgi ^[@R53]^, but no further loss occurred in response to the BM-overlay. Bar: 8μm.\
(b) MECs infected with adenoviruses expressing TetR alone, TetR + K44A-dynamin1 or TetR + K44A-dynamin2, then cultured in 3D on BM-matrix. DN-dynamin prevented relocation of aPKC to the internal luminal surface and polarization of acini. Note that these acini resemble day 1 of the developmental time course (cf [Fig. 3a](#F3){ref-type="fig"}) Bar: 10μm.\
(c) MECs infected with adenoviruses expressing TetR alone, TetR + S34NRab5a and then cultured in 3D on BM-matrix. DN-Rab5a prevented relocation of aPKC to the internal luminal surface and polarization of acini. Bar: 10μm.\
(d) Left panels: confluent monolayers of WT MECs displayed apical tight junctions (ZO1) and basolateral β1-integrin. (i) Z-section (ii, iii) confocal view. Right panels: BM-overlay induced ZO1 disruption in WT but not β1-KO MECs. Confocal plan views were merged from the top and middle to show both ZO1 and nuclei. Bar: 8μm\
(e) Surface biotinylation of cells followed by 6 h BM-overlay then blotting of endocytosed Claudin 7. i) (1-5) Monolayer controls, (6,7) BM-overlay. (1,2) Efficiency of biotin stripping from the surface at 4°C: (1) unstripped, and (2) stripped monolayers. (3-7) To induce endocytosis, cells were switched to 37°C after biotin labelling: (3) no biotin, (4) monolayer, (5) monolayer +4OHT, (6) BM-overlay, (7) BM-overlay +4OHT. ii) ImageJ64 quantification of endocytosed Claudin7 immunoblots. Histogram represents mean values +/− s.e.m for error bars of n=3 experiments. Note that BM-overlay increased Claudin 7 endocytosis (cf 4,6), which was prevented by β1-integrin deletion (cf 6,7).\
See also [Supplementary Figs 7](#SD3){ref-type="supplementary-material"}, [8](#SD3){ref-type="supplementary-material"}.](emss-50423-f0007){#F7}
{#F8}
| {
"pile_set_name": "PubMed Central"
} |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.