ADAR1 polymorphisms are related to severity of liver fibrosis in HIV/HCV-coinfected patients

The adenosine deaminase acting on RNA (ADAR1) gene is an interferon-stimulated gene involved in liver injury protection. Our aim was to analyze the association of polymorphisms within this gene with the severity of liver disease in European HIV/HCV-coinfected patients. We performed a cross-sectional study in 220 patients that underwent a liver biopsy. Five SNPs in the ADAR1 gene (rs1127326, rs1127317, rs1127314, rs1127313, rs2229857) were genotyped by GoldenGate assay. The outcome variables were fibrosis stage and necroinflammatory activity grade by METAVIR-score, aspartate aminotransferase to platelet ratio index (APRI), FIB-4 index, and fibrosis progression rate (FPR). In multivariate analysis, fibrosis progression rate (FPR) (aAMRs = 0.97) decreased in a dose-dependent manner with the presence of rs2229857_T, rs1127313_G, rs1127314_G and rs1127317_G; while rs1127326_T allele had only significant associations with FIB-4 (aAMRs ≤ 0.63) and FPR (aAMRs ≤ 0.97). Moreover, carriers of rs2229857_T, rs1127314_G, rs1127317_G, and rs1127326_T alleles were protected against advanced fibrosis (F ≥ 3) (adjusted ORs (aORs) ≤ 0.44), APRI ≥ 1.5 (aORs ≤ 0.33), and FPR ≥ 0.075 (aORs ≤ 0.45). rs1127313_G carriers showed lower odds of having F ≥ 3 (aORs = 0.39), FIB4 ≥ 3.25 (aOR = 0.22) and FPR ≥ 0.075 (aORs = 0.44). In conclusion, ADAR1 polymorphisms protected against severe liver disease in HIV/HCV-coinfected patients. These results could be used to improve therapeutic decision-making in clinical practice.

ADAR1 genotypes and liver disease. The additive model of inheritance was the model that best fit to our data. With this model, we explored the genetic association between ADAR1 polymorphisms and liver disease stages (Tables 3-5).
Secondly, the association values with ordinal variables are shown in Table 4. Patients carrying the rs2229857 T, rs1127313 G, rs1127314 G, rs1127317 G, and rs1127326 T alleles were protected against the increase of fibrosis stage in the multivariate analysis [adjusted ORs (aORs) ≤ 0.64; p < 0.05]. No significant values were found for the activity grade.
Next, the association values with dichotomous variables are shown in Table 5. In the multivariate analysis, carriers of rs2229857 T, rs1127314 G, rs1127317 G, and rs1127326 T alleles were protected against advanced liver fibrosis (F ≥ 3) (aORs ≤ 0.44; p < 0.05), values of APRI ≥ 1.5 (aORs ≤ 0.33; p < 0.05), and FPR ≥ 0.075 (aORs ≤ 0.45; p < 0.05). In the same way, patients carrying the rs1127313 G allele showed lower odds of having advanced liver fibrosis (F ≥ 3) (aORs = 0.39; p = 0.002), values of FIB4 ≥ 3.25 (aOR = 0.22; p = 0.008) and FPR ≥ 0.075 (aORs = 0.44; p = 0.002). No significant values were found for the activity grade. ADAR1 haplotypes and liver disease. Linkage disequilibrium (LD) analysis (Fig. 2) showed that there was a high LD (non-random association of alleles at different loci) between ADAR1 SNPs (D' > 0.98), meaning that there is no evidence of a possible recombination between these SNPs. Moreover, the R-squared among all studied SNPs, except for rs1127326, was high (R 2 > 0.96). R 2 values among all SNPs and rs1127313 is less than 0.44, meaning that rs1127313 did not provide the same information and could not be substituted one for the other. Furthermore, we performed a Tag SNPs analysis in the region studied by using Haploview 4.2 software and we determined that rs1127313 and rs1127317 are representative SNPs in our study.
Haplotype analysis (Table 6) showed that only three major ADAR1 haplotypes (comprised of rs1127313 and rs1127317) were found: 51% AT (unfavorable alleles), 28.8% GG (favorable alleles), and 19.4% GT. The AT haplotype (unfavorable alleles) showed higher odds for having advanced liver fibrosis (F ≥ 3) (aOR = 2.63; p = 0.006), SCIENtIFIC RepoRTs | 7: 12918 | DOI: 10.1038/s41598-017-12885-4 In silico analysis. The in silico prediction analysis (Supplementary Tables 1-4) showed functional predictions in all the markers studied. rs2229857 is a missense variant located in exon 2 and produces an amino acid substitution (K/R), which does not seem to affect the protein structure according to SIFT and PolyPhen tools. Furthermore, according to the F-SNP database, rs2229857 seems to alter a putative exonic splicing enhancer (ESE). Functional analysis of rs2229857, rs1127313, rs1127314, rs1127317 and rs1127326 polymorphisms by using rSNPBase reported that each of them could be involved in transcriptional regulation. Additionally, rs1127313 and rs1127314 are 3′UTR variants and seem to be involved in the miRNA binding site of hsa-miR-4516 and hsa-miR-363-5p, respectively.

Discussion
In this study, the major findings were that ADAR1 rs2229857 T, rs1127313 G, rs1127314 G, rs1127317 G, and rs1127326 T alleles protect against the development of liver fibrosis, (fibrosis stage, but not activity grade) and the development of high levels of non-invasive indexes (APRI and FIB-4) in HIV/HCV-coinfected patients. We have detected these effects both with ADAR1 SNPs alone and ADAR1 haplotypes; similar associations were observed in both analyses. To the best of our knowledge, this is the first study showing a genetic association between ADAR1 SNPs and the severity of liver fibrosis in HIV/HCV-coinfected patients. The immune response against HCV infection initially attempts to eradicate the viral infection, but the immune system also causes hepatocyte damage and fibrosis through direct cellular toxicity and the release of inflammatory cytokines 21 . The innate response against viruses produces type I IFN (alpha, beta), which induces the expression of ISGs such as ADAR1 22 . In fact, a high level of ADAR1 protein has been detected in HCV-infected patients at weeks   23 therapy. Additionally, the inhibition of ADAR1 in vitro stimulates the production of an HCV-replicon, indicating that ADAR1 has a role in limiting HCV replication 18 . Moreover, ADAR1 SNPs have been related to HCV clearance in Caucasian American patients treated with IFN2a/r 24 , and HBV clearance in Chinese patients, both spontaneously and with interferon treatment 25 . HCV-related liver fibrosis is a progressive disease that results from an impaired immune response and subsequent loss of control of HCV replication 7 . The lack of induction of an ISG-mediated antiviral response has been associated with rapid fibrosis progression in HCV infected and HCV/HIV coinfected patients 26 . Additionally, the ADAR1 gene seems to contribute to the regulation of innate immunity, especially in hepatocytes 16 . Thus, ADAR1 may have implications in liver diseases involving excessive inflammation that may induce liver fibrosis in HCV-infected patients 18 . Therefore, any genetic variant at ADAR1 gene that impairs its protein expression or stability, could be affecting to liver fibrosis progression. In addition, recent reports have shown that the ADAR1 gene is also related to HIV infection, as ADAR1 protein seems to be incorporated into HIV virions 20 , where it is required for efficient replication in CD4+ T cells 27 .
With respect to the functionality of the ADAR1 SNPs, we have performed an in silico prediction of the putative role of each ADAR1 polymorphism on protein function or expression. The rs2229857A > G is a missense polymorphism that generates a lysine to arginine amino acid change at position 427 in the protein. This change could affect the solvent accessibility of the protein and the transcriptional regulation 28 . The in silico prediction showed that rs2229857T > C seems to alter a putative exonic splicing enhancer (ESE), which is an element that plays an important role in constitutive and alternative splicing 28 . Therefore, this variant could alter not only the structure of the protein, but also the RNA sequence and secondary structure. Furthermore, the functional analysis of rs1127313, rs1127314, rs1127317 and rs1127326 polymorphisms reported that each one could be involved in transcriptional regulation, being putative binding sites for different transcription factors.
ADAR proteins also play an important role in microRNA (miRNA) biogenesis. miRNAs are small non-coding cellular RNAs that are involved in post-transcriptional regulation of messenger RNAs (mRNAs). The mRNA is degraded or not transcribed depending on the complementarity between the miRNA and the target sequence of the mRNA. ADAR proteins act on nuclear double-stranded RNA structures that are pre-mRNA transcripts. Therefore, it is likely that they edit pri-miRNA transcripts prior to their export from the nucleus. In addition, ADAR1 is known to form a complex with Dicer 13 , an RNAse III enzyme involved in miRNA maturation.
All the studied SNPs, except for rs1127326, have been related to miRNA biogenesis 29 . Thus, the presence of polymorphisms in miRNA processing machinery genes such as ADAR1 has an unpredictable and broad effect on transcriptional regulation. In addition, the presence of polymorphisms within the target sequence of the mRNA will affect miRNA recognition and therefore, the transcriptional regulation of the mRNA will be altered. In this respect, rs1127313 and rs1127314 are in the target sequence for the binding of hsa-miR-4516 and  hsa-miR-363-5p, respectively. When the G allele for both SNPs is present, these miRNAs can bind to the ADAR1 mRNA and thus regulate its expression, while when the alternative allele (T) is present, the target is disrupted and the miRNA cannot bind to it. In our study, the presence of G alleles at both polymorphisms (rs1127313 and rs1127314), alone or in haplotype analysis, were associated with advanced liver fibrosis. On the one hand, hsa-miR-4516 seems to be involved in fibrosis reduction in muscle 30 . On the other hand, hsa-miR-363-5p has been identified as being upregulated in cirrhotic livers 31 , therefore its binding to ADAR1 could be related to a poor prognosis of liver disease. The strong association between ADAR1 SNPs and the severity of liver fibrosis has not been previously found in two previously performed genome-wide association studies (GWAS) 32,33 . This issue could be due to the fact   that a GWAS usually does not have the power to detect all significant effects, only the biggest ones. In the GWAS reported 32,33 , the patients origins are not stated, therefore the lack of detection could be due to a different genetic background of the patients enrolled. In addition, both GWAS were performed with HCV patients, while our CHC patients were coinfected with HIV. In this case, the HIV could be affecting the relationship between the ADAR1 polymorphisms and liver fibrosis development. Finally, several aspects must be considered for the correct interpretation of the results. Firstly, this report has a cross-sectional design with a relatively small number of patients, which could limit the statistical power to detect some differences between groups. Therefore, it would be necessary to perform an independent replication of this study in a prospective study with a higher sample size, including different ethnic groups. Also, data was collected retrospectively, which entails a lack of uniformity. Secondly, although we found a strong association with ADAR1 polymorphisms and liver fibrosis, we cannot assure that any of them are the causal mutation. Therefore, it cannot be discarded that the detected association is due to other SNPs in high linked disequilibrium with ADAR1 polymorphisms. Thirdly, as discussed above, the in silico analysis showed that ADAR1 polymorphisms alone could have some influence on the expression of the proteins involved, but it would be necessary to demonstrate if these ADAR1 polymorphisms modify the expression of the ADAR1 gene. Fourthly, all selected patients met a set of criteria for starting HCV treatment (e.g., no alcohol abuse, high CD4 cell counts, controlled HIV replication, and good treatment adherence), and this may have introduced a selection bias. Fifthly, this study was performed on patients with European ancestry, and it would be interesting to perform these analyses on different ethnic groups. Finally, our study only included HIV/HCV-coinfected patients and it would be interesting to know the role of the studied polymorphisms in HCV-monoinfected patients, but we did not have access to a cohort of HCV-monoinfected patients. In conclusion, genetic variants in the ADAR1 gene protect against the development of liver fibrosis in HIV/ HCV-coinfected patients. These data suggest that ADAR1 polymorphisms may play a major role in the pathogenesis of CHC in HIV/HCV-coinfected patients, and it could be used to improve therapeutic decision-making in clinical practice.

Materials and Methods
Study population. We performed a cross-sectional study in 220 HIV/HCV-coinfected naïve patients that underwent a liver biopsy between September 2000 and November 2008. The study was conducted in accordance with the Declaration of Helsinki, and patients gave their informed consent for the study. The Institutional Review Board and the Research Ethics Committee of the Instituto de Salud Carlos III approved the study.
Liver biopsies were performed on patients who were potential candidates for anti-HCV therapy and had not received previous IFN therapy (naïve for HCV-treatment). Selection criteria were: no clinical evidence of hepatic decompensation, detectable HCV RNA by polymerase chain reaction (PCR), negative hepatitis B surface antigen, availability of DNA sample, CD4 + lymphocyte count higher than 200 cells/µL, and stable combination antiretroviral therapy (cART) for at least 6 months before study entry or no need for cART according to treatment guidelines used in the study period 34,35 . Patients with active opportunistic infections, active drug addiction, and other concomitant severe diseases were excluded. All subjects included in our study were white Europeans.
Epidemiological and clinical data. Clinical and epidemiological data were obtained from medical records.
Consumption of more than 50 g of alcohol per day for at least 12 months was considered as a high intake. Body mass index (BMI) was calculated as the weight in kilograms divided by the square of the height in meters. The duration of HCV infection for patients with a history of intravenous drug use (IDU; 87.7% of patients) was estimated starting from the first year they shared needles and other injection paraphernalia, which are the most relevant risk practices for HCV transmission 36 . When the initiation of HCV infection could not be estimated with certainty, the duration was not calculated (n = 16). For 204 out of 220 patients with the duration of HCV infection, the FPR was calculated by dividing the fibrosis stage (0 to 4) by the estimated duration of HCV infection in years (fibrosis units per year) 37 .  Biochemistry panel was measured using an autoanalyzer Hitachi 912 (Boehringer Mannheim, Germany), while patients were fasting. The degree of insulin resistance was estimated for each patient using the homeostatic model assessment (HOMA) 38 : fasting plasma glucose (mmol/L) times fasting serum insulin (mU/L) divided by 22.5. Non-invasive fibrosis indexes were calculated according to the formula originally described for the APRI 39

HCV assays. HCV infection was documented in all patients by enzyme-linked immunosorbent assay and
PCR test. HCV genotype was determined by hybridization of biotin-labeled PCR products to oligonucleotide probes bound to nitrocellulose membrane strips (INNO-LiPA HCV II, Innogenetics, Ghent, Belgium). Plasma HCV-RNA viral load was measured by PCR (Cobas Amplicor HCV Monitor Test, Branchburg, NJ, USA) and real-time PCR (COBAS AmpliPrep/COBAS TaqMan HCV test), and results were reported in terms of international units per milliliter (IU/mL). The limit of detection varied between 10 IU/mL and 600 IU/mL depending on the HCV diagnostic test used at the time of biopsy. There was no patient with HCV viral load below the limit of detection. The HCV RNA level of 500,000 UI/ml was chosen as cut-off for low and high viral load (circulating HCV RNA) 41,42 . Liver biopsy. Liver biopsies were performed as outpatient service following the recommendations of the Patient Care Committee of the American Gastroenterological Association 43 , as we have previously described 44 . All samples were evaluated by the same pathologist, who was unaware of the patients' clinical or laboratory data. Liver fibrosis and necroinflammatory activity were estimated according to Metavir score as follows 45 : F0, no fibrosis; F1, mild fibrosis; F2, significant fibrosis; F3, advanced fibrosis; and F4, definite cirrhosis. The degree of necroinflammation (activity grade) was scored as follows: A0, no activity; A1, mild activity; A2, moderate activity; A3, severe activity.
Genotyping of DNA polymorphisms. The most common SNPs in the ADAR1 gene (chromosome 1) were selected using the databases of the HapMap Project (http://snp.cshl.org/cgi-perl/gbrowse/hapmap_B35/) and NCBI (dbSNP) (http://www.ncbi.nlm.nih.gov/entrez/). The selection criteria were: i) SNPs located in putative regulatory regions, or splicing control elements (SCE), or missense variants; ii) minor allelic frequency (MAF) greater than 20% in European people in order to have guarantees to obtain significant associations when working with small sample sizes 46 . With this setting, we found five SNPs in the ADAR1 gene (rs1127326, rs1127317, rs1127314 and rs1127313 located at 3′UTR; and rs2229857 in exon 2) Genomic DNA was extracted from peripheral blood with Qiagen kit (QIAamp DNA Blood Midi/Maxi; Qiagen, Hilden, Germany). DNA samples were genotyped at the Spanish National Genotyping Center (CeGen; http://www.cegen.org/) by using GoldenGate assay with VeraCode Technology (Illumina Inc., San Diego, California, USA).
We obtained the allelic and genotypic frequency for healthy subjects, from the 1000 Genomes Project website (http://www.1000genomes.org/home). This database provides a broad representation of common human genetic variation from multiple populations 47 . We selected the IBS population that included 107 individuals.
Outcome variables. HIV/HCV-coinfected patients were classified according to the severity of liver disease developed after a minimum follow-up time of 10 years with HCV infection. The outcome variables were the fibrosis stage and the necroinflammatory activity grade according to METAVIR score, and values of APRI, FIB-4 and FPR. The cut-off values to determine the severity of liver disease were 48 : a) advanced fibrosis (F ≥ 3, APRI ≥ 1.5, and FIB-4 ≥ 3.25), b) severe activity grade (A3); and c) rapid fibrosis progression rate (FPR) higher than median (FPR ≥ 0.075).

Statistical analysis.
For the description of the study population, values were shown as absolute numbers (%) and median (percentile 25; percentile 75). Frequencies of alleles and genotypes for ADAR1 SNPs in HIV/ HCV-coinfected patients were compared to healthy subjects by using Chi-squared test.
Hardy-Weinberg equilibrium (HWE) for all SNPs was assessed by a χ 2 test, considering equilibrium when p > 0.05. In addition, pair-wise linkage disequilibrium (LD) analysis was computed to detect the inter-marker relationship using the standardized D' and r 2 values by Haploview 4.2 software. The haplotype frequencies were inferred with the Expectation-maximization algorithm by using Haploview 4.2 software (https://www.broadinstitute.org/haploview/haploview). Haplotype association tests are obtained by summing the fractional likelihoods of each patient for each haplotype. The haplotype-based association test was performed by multivariate logistic regression adjusted by the most important clinical and epidemiological covariates, as cited above. These analyses were performed with Plink software (http://zzz.bwh.harvard.edu/plink/) and adjusted by the same variables as for association with individual SNPs. All p-values were two-tailed and statistical significance was defined as p < 0.05.
For the genetic association study, Generalized Linear Models (GLMs) were used to study the association of ADAR1 polymorphisms with outcome variables according to different models of inheritance: dominant, recessive and additive. We used the Akaike information criterion (AIC) value and Bayesian information criterion (BIC) to evaluate the model that best fit our data. On the one hand, a GLM with a normal distribution was used to investigate the association between ADAR1 polymorphisms and continuous outcome variables (APRI, FIB-4, and FPR). This test gives the differences between groups and the arithmetic mean ratio (AMR) in continuous outcome variables. On the other hand, a GLM with multinomial distribution (cumlogit-link) was used to investigate the association of ADAR1 polymorphisms with ordinal outcome variables (fibrosis stage and activity grade). Next, a GLM with binomial distribution (logit-link) was used to investigate the association of ADAR1 polymorphisms with categorical outcome variables (F ≥ 3, A3, APRI ≥ 1.5, FIB4 ≥ 3.25, and FPR ≥ 0.075). These tests give the differences between groups and the odds ratio (OR) for categorical outcome variables. Additionally, GLM tests were adjusted by the most important clinical and epidemiological characteristics (gender, age, alcohol intake, BMI, HOMA, nadir CD4+ T-cells, AIDS, undetectable HIV-RNA (<50 copies/ml), CD4+ T-cells, time on cART, type of cART, HCV-RNA ≥ 500,000 IU/ml, HCV genotype, IL28B rs12980275 polymorphism). All analyses were performed by using the IBM SPSS Statistics for Windows, Version 21.0 (IBM Corp, Chicago, Armonk, NY, USA).
All association analyses were corrected by multiple comparison testing by using the SDM software (http:// www.sdmproject.com/utilities/?show=FDR) that calculate the False Discovery Rate (FDR) 49 .
We also performed an in silico analysis, for evaluating the possible functional implication of each polymorphism, by using different web-tools: a) F-SNP (http://compbio.cs.queensu.ca/F-SNP/) that provides information about the functional effects of SNPs obtained from 16 bioinformatics tools and databases 28 b) Ensembl's Variant Effect Predictor (VEP) tool (http://grch37.ensembl.org/info/docs/tools/vep/index.html) 50 c) rSNPBase (http:// rsnp.psych.ac.cn/) that provides regulatory annotations on rSNPs; and d) miRDB target prediction (http://mirdb. org/miRDB/index.html) for identification of miRNA target sequences 51,52 . Meetings at which parts of the data will be presented. Some