Several previous studies suggested that HLA-Class II may be associated with susceptibility to primary biliary cirrhosis (PBC), but data from individual studies remain controversial. Therefore, a systematic review and meta-analysis is needed to comprehensively evaluate the association between HLA-Class II and PBC risk.
All published reports of an association between HLA class II and PBC risk were searched in PubMed, EMBASE (updated to 22 May 2012). ORs with 95% confidence intervals (CIs) were extracted from each included study and the meta-analysis was performed using the fixed- or random-effects model.
A total of 3,732 PBC patients and 11,031 controls from 34 studies were included in the meta-analysis. An assessment of study quality revealed that the majority of studies included (18 studies) were of high quality. The serological group DR8 was found to be a risk factor for PBC (OR = 2.82, 95%CI: 1.84–4.30). At the allelic level, HLA-DR*08 and HLA-DR*0801 were identified as risk factors for PBC (OR = 2.30, 95%CI: 1.76-3.00; OR = 3.23, 95%CI: 2.22–4.70, respectively), whereas HLA-DR*11 and HLA-DR*13 were potent protective factors (OR = 0.31, 95%CI: 0.27-0.38; OR = 0.62, 95%CI: 0.48-0.81, respectively). HLA-DQB1 and HLA-DQB1*0402 conferred a predisposition to PBC development (OR = 3.47, 95%CI: 2.35–5.13), whereas HLA-DQB1*0604 was protective against PBC (OR = 0.3, 95%CI: 0.18–0.58). No HLA-DPB1 allele was observed to be associated with PBC susceptibility (P > 0.05).
The present study revealed that HLA-Class II components are closely associated with the development of PBC.
Citation: Qin B, Wang J, Chen J, Liang Y, Yang Z, et al. (2013) Association of Human Leukocyte Antigen Class II with Susceptibility to Primary Biliary Cirrhosis: A Systematic Review and Meta-Analysis. PLoS ONE 8(11): e79580. doi:10.1371/journal.pone.0079580
Editor: Aftab A. Ansari, Emory University School of Medicine, United States of America
Received: August 27, 2013; Accepted: September 27, 2013; Published: November 12, 2013
Copyright: © 2013 Qin et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Funding: This study was supported by two grants from the China National Natural Science Foundation Council (81001333 and 81072479). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing interests: The authors have declared that no competing interests exist.
Primary biliary cirrhosis (PBC) is a chronic immune-mediated liver disease characterized by progressive intrahepatic bile-duct destruction, leading to cirrhosis and eventual liver failure . Although the etiology of PBC is unclear, it has been suggested that both genetic and environmental factors may initiate and promote the disease . Specifically, the high concordance for PBC in monozygotic twins, family clustering, and female predominance suggest that genetic factors may play an important role in the development of PBC [3-5].
Human leukocyte antigens (HLA) are encoded by genes within the major histocompatibility complex (MHC), located on chromosome 6p21. Genes located in the centromeric class II region of the HLA encode polymorphic HLA-DR, HLA-DQ, and HLA-DP, which are expressed on antigen-presenting cells and can participate in antigen processing. Although accurate mechanisms to explain how the HLA class II gene induces autoimmunity are unclear, several HLA class II alleles have been identified to confer genetic risk for or protection against many autoimmune diseases , and such associations between certain HLA class II alleles and the development of PBC have been extensively investigated.
The most commonly detected HLA genes involved in susceptibility to PBC are the Class II DRB1*08 allele family, especially DRB1*0801 and DRB*0803, which were implicated as risk factors for PBC in some studies [7-9]. However, data to suggest an association between DRB1*08 and PBC risk are controversial at this time. In addition to PBC risk conferred by HLA Class II alleles, several studies suggest that some alleles could be protective against PBC such as DRB1*11 and DRB1*13 [10,11]. These results support the concept that HLA Class II alleles may have diverse roles in PBC pathogenesis.
In addition to HLA alleles, some haplotypes, such as DRB1*0803-DQA1*0103-DQB1*0601 and DRB1*0801-DQA1*0401-DQB1*402, were also found to be associated with PBC risk in studies of patients with diverse ethnic origins [8-10]. Further evidence from recent genome-wide association studies (GWAS) also revealed a strong association between polymorphisms in the HLA II region and PBC. In GWASs, the HLA-DRB1 and HLA-DQB1 locus were reported to be significantly associated with susceptibility to PBC [12,13]. To better understand GWAS results regarding roles of HLA class II in PBC, quantitative analyses are needed to pool data and estimate associations.
Data supporting a protective role for HLA Class II alleles is of interest to investigators researching molecular mechanisms of PBC disease development . Although many studies have focused on the involvement of HLA genes in PBC, results from these investigations were inconsistent or inconclusive due to small sample sizes. In the view of existing heterogeneity among these studies, we conducted a meta-analysis to comprehensively assess the association between HLA Class II and PBC risk.
A literature search of electronic databases including PubMed, EMBASE, (updated to 22 May 2013) was conducted independently by two investigators (BDQ and JC). The search included only published studies on association between HLA-Class II (HLA-DRB1, HLA-DQB1, and HLA-DPB1) and susceptibility to PBC. The search strategies including Mesh terms and keywords were as follows: “primary biliary cirrhosis”, ”human leukocyte antigens”, “major histocompatibility complex”, and “HLA-DR/DQ/DP antigens”. No limit was placed on publication language or geographic area.
Studies included in the meta-analysis fulfilled selection criteria as follows: a) the study should be a case-control design; b) PBC patients should be diagnosed according to internationally accepted criteria; c) the study should contain sufficient published data for the evaluation of an odds ratio (OR) with a 95% confidence interval (95%CI); and d) the study should have been published in journals as full articles. Studies which had one of the following exclusion were discarded: a) the study was based on family or sibling pairs; b) the study had repeated reports on the same population or subpopulation; c) the study had no control group; d) there was insufficient published data for extraction; and e) the paper was a review or abstract.
All included studies were retrieved and the data were extracted independently in duplicate using a standard protocol by two authors (BDQ and JC). Study characteristics were extracted, including the first author, year of publication, country in which the study was launched, ethnicity of the population sample, HLA genotyping technique, sample size, and total number of cases and controls (Table 1). In the literature review, data were chiefly calculated using two methods. One method compared the difference in the number of individuals who carried HLA-DRB1, HLA-DQB1, HLA-DPB1 alleles, comparing PBC patients and controls (na/N, na is number of individuals who carried the alleles, N is the total number of PBC patients or controls). The second method used differences in the frequency of the alleles between PBC patients and controls (nb/2N, nb is the number of some alleles and N is the total number of PBC patients or controls).
|Author||Year||Country||Ethnicity||Sample size||Number of PBC patient/control||Specific technique|
|Miyamori H, et al.||1983||Japan||Asian||72||22/50||microlymphocytotoxicity|
|Bassendine MF, et al.||1985||UK||Caucasian||275||75/200||microlymphocytotoxicity|
|Johnston DE, et al.||1987||UK||Caucasian||200||71/129||microlymphocytotoxicity|
|Briggs DC, et al.||1987||UK||Caucasian||403||96/307||microlymphocytotoxicity|
|Gores GJ, et al.||1987||USA||Caucasian||285||114/171||microlymphocytotoxicity|
|Prochazka EJ, et al.||1990||USA||Caucasian||1,581||35/1,546||microlymphocytotoxicity|
|Manns MP, et al.||1991||Germany||Caucasian||201||25/176||microlymphocytotoxicity|
|Underhill J, et al.||1992||UK||Caucasian||321||159/162||PCR/RFLP|
|Morling N, et al.||1992||Denmark||Caucasian||1,227||23/1,204||microlymphocytotoxicity|
|Gregory WL, et al.||1993||UK||Caucasian||493||130/363||RFLP|
|Begovich, et al.||1993||USA||Caucasian||291||51/240||PCR-SSO|
|Seki, et al.||1993||Japan||Asian||191||41/150||microlymphocytotoxicity/PCR|
|Onishi S, et al.||1994||Japan||Asian||246||31/215||PCR-SSO|
|Zhang L, et al.||1994||UK||Caucasian||83||40/43||PCR/RFLP|
|Mehal WZ, et al.||1994||USA||Caucasian||125||64/61||RFLP|
|Underhill JA, et al.||1995||UK||Caucasian||185||82/103||PCR-SSO|
|Mella JG, et al.||1995||Germany||Caucasian||79||32/47||PCR-SSO|
|Akimoto S, et al.||1999||Japan||Asian||228||13/215||NA|
|Donaldson P, et al.||2001||UK||Caucasian||266||164/102||PCR-SSO|
|Stone J, et al.||2002||USA||Caucasian||370||154/216||PCR-SSP|
|Wassmuth R, et al.||2002||Germany||Caucasian||257||99/158||PCR-SSO|
|Invernizzi P, et al.||2003||Italy||Caucasian||670||112/558||PCR-SSO|
|Bittencourt PL, et al.||2003||Brazil||Caucasoid/African American/Amerindian||144||61/83||PCR-SSP|
|Jiang XH, et al.||2004||China||Asian||95||52/43||PCR-SSP|
|Mullarkey ME, et al.||2005||USA||Caucasian||453||72/381||PCR-SSO|
|Chen C, et al.||2005||China||Asian||132||72/60||PCR-SSP|
|Liu HY, et al.||2006||China||Asian||496||65/431||PCR-SSP|
|Donaldson PT, et al.||2006||UK/Italy||Caucasian||823||492/331||PCR-SSO|
|Zhao J, et al.||2006||China||Asian||107||40/67||PCR-SSP|
|Invernizzi P, et al.||2008||Italy||Caucasian||2,656||664/1,992||PCR-SSO|
|Vázquez-Elizondo G, et al.||2009||Mexico||Mexican/Spanish||390||9/381||PCR-SSP|
|Nakamura M, et al.||2010||Japan||Asian||592||334/258||PCR-SBT|
|Chong VH, et al.||2010||Brunei Darussalam||Asian||74||9/65||PCR-SSO|
|Umemura T, et al.||2012||Japan||Asian||752||229/523||PCR-SSO|
Using these methods, data were pooled and a meta-analysis was performed. Any disagreement in the data was resolved by discussion and consensus (The data would be extracted independently by other authors using the same standard protocol).
Assessment of study quality
Due to the lack of standardized quality criteria for meta-analyses of single nucleotide polymorphism studies, we chose the Newcastle-Ottawa scale (NOS) to assess the quality of these non-random studies. According to the NOS, the criteria for evaluation include selection of cases and controls, comparability of cases and controls, and ascertainment of exposure . The scoring system provided a summary numeric score of quality that ranged from 0 star to 9 stars. The studies were graded into the three categories: high (>=7 stars), medium (4-6 stars) and low (<=3 stars) quality .
The final effect ORs and 95%CIs were calculated in a random-effect or fixed-effect model to evaluate the strength of the association between HLA Class II and PBC risk. Heterogeneity of effects among studies was estimated using the means of χ2-based Q test and I2 test. For the Q test, a P value less than 0.1 was considered to be representative of significant heterogeneity, and an I2 statistic represented the percentage of total variation contributed by a between-study variation that ranged from 0 to 100% . If no significant heterogeneity was observed, a fixed-effect model was used to pool the data. Otherwise, a random-effect model was used and meta-regression analysis was used to identify the potential source of heterogeneity if sufficient studies were included. Funnel plots, Egger’s test, and Begg’s test were used to determine publication bias. A sensitivity analysis was performed to test data stability, and a cumulative meta-analysis was used to determine whether the strength of a relationship was stable through the repeated performance of meta-analysis whenever a new study became available for inclusion. The number or frequency of HLA Class II was zero in some included studies, so these zero total event studies were included to provide the most generalizable estimated ORs . Studies with no events in both groups were excluded. All analyses were conducted in STATA 11.0 software (Stata Corp, College Station, TX). P values less than 0.05 were considered significant.
Studies included in the meta-analysis
From the initial published work search, a total of 580 non-overlapping articles were identified and screened from the previously described electronic databases. A total of 468 articles were excluded based on screening of abstracts or titles. After retrieving the full-text articles, 78 articles were excluded based on the exclusion criteria, leaving 34 relevant studies for the meta-analysis [7-11,19-47]. (Figure 1)
Figure 1. Flowchart of the present meta-analysis.doi:10.1371/journal.pone.0079580.g001
Characteristics of included studies
The thirty-four studies were published between 1983 and 2012 and were comprised of 31 English-language papers and 3 Chinese-language papers, with data for 3,732 PBC patients and 11,031 controls. All 34 studies fulfilled the inclusion criteria and an OR with a 95%CI could be obtained from each study. Study information and characteristics are depicted in Table 1.
Assessment of study quality
According to the scoring system given, the quality of each study was assessed and 7 studies scored 8 stars, 11 scored 7 stars, 7 scored 6 stars, 7 scored 5 stars and 2 scored 4 stars. (Table S1).
Serological HLA-DR and PBC
A total of 13 studies contained data on serological HLA-DR from 5,400 subjects (788 cases of PBC and 4,612 controls). The meta-analyses for associations between each HLA-DR and PBC susceptibility revealed that only 6 of 16 associations were statistically significant. The serological groups HLA-DR2, DR5, DR12, DR16, and DR52a were potent protective factors against PBC, whereas DR8 was found to be a risk factor for PBC. The pooled ORs and 95%CIs were 2.82 (1.84-4.30), 0.70 (0.53-0.91), 0.55 (0.39-0.78), 0.40 (0.19-0.86), 0.21 (0.05-0.97), and 0.60 (0.38-0.95), respectively, for DR8, DR2, DR5, DR12, DR16, and DR52a. (Figure S1)
There was no heterogeneity among the studies with respect to the association between DR2, DR5, DR12, DR16, and DR52a and the risk of PBC, but a high degree of heterogeneity was found to exist between DR8 and PBC risk (I2= 54.8%, P = 0.011). A subgroup meta-analysis by ethnicity indicated that a significant association was discovered in both Caucasian and Asian groups, but the Caucasian group was significantly heterogeneous. Subsequently, the cumulative meta-analysis revealed that the estimated OR and 95%CI were stable in order of year of publication and the convergence of evidence across numerous studies confirmed the strength of HLA-DR8 as a risk factor for PBC. A statistically significant association was detected as far back as 1990  (Figure 2).
Figure 2. Cumulative meta-analysis of 12 studies of HLA-DR8 and susceptibility to PBC by the Mantel-Haenszel method with the random-effect model.doi:10.1371/journal.pone.0079580.g002
HLA-DQB1 alleles and PBC
A total of 14 meta-analyses of studies of HLA-DQB1 alleles and PBC susceptibility were conducted. HLA-DQB1 (*02, *04, *0401, *0402 and *0601) were found to be risk factors for PBC, and the pooled ORs and 95%CI were 1.40 (1.00–1.97), 2.24 (1.46–3.46), 1.41 (1.07–1.85), 3.47 (2.35–5.13), and 1.99 (1.57–2.53), respectively (Figure S2). In contrast, HLA-DQB1 (*0301, *06, *0602 and *0604) were protective factors, and the pooled ORs and 95%CIs were 0.62 (0.48–0.79) and 0.48 (0.35–0.68), 0.58 (0.39–0.85), 0.61 (0.42–0.87), and 0.53 (0.37–0.75), 0.3 (0.18–0.58), respectively (Figure S3). No significant heterogeneity was observed in the meta-analysis for these alleles.
HLA-DPB1 alleles and PBC
A total of 13 meta-analyses have been conducted, but no statistically significant association between HLA-DPB1 alleles and PBC risk was observed (data not shown).
HLA-DRB1 alleles and PBC
Compared with HLA-DQB1 and HLA-DPB1 alleles, studies describing HLA-DRB1 alleles and PBC were more abundant. HLA-DRB1 alleles were also most extensively studied in the context of a relationship between HLA-Class II and PBC. HLA-DRB1 (*01, *03, *0405, *07, *08, *0801 and *0803) were found to predisposition individuals to PBC, and the pooled results were 1.26 (1.05–1.51), 1.38 (1.16–1.65), 1.43 (1.16–1.76), 2.03 (1.29–3.21) and 1.47 (1.27–1.70), 2.30 (1.76–3.00) and 2.48 (1.60–3.84), 3.23 (2.22–4.70), and 2.64 (1.55–4.51), 3.00 (1.89–4.76), respectively (Figure 3; Table 2). Conversely, HLA-DRB1 (*11, *13, *1101 and *1501) were observed to confer resistance to PBC, and the pooled results were 0.45 (0.33–0.62) and 0.31 (0.27–0.38), 0.41 (0.24–0.68), 0.82 (0.68–0.99), and 0.45 (0.31–0.65), respectively (Figure 4; Table 2). There was no significant heterogeneity in meta-analyses for most HLA-DRB1 alleles, but high heterogeneity existed in meta-analyses for HLA-DRB1*08, *0803 and *13. The subgroup meta-analysis also demonstrated that HLA-DRB1*08 is a risk factor for PBC in Caucasian and Asian populations, but significant heterogeneity was observed in subgroup analyses. The subgroup meta-analysis for HLA-DR*0803 after stratification by ethnicity indicated that HLA-DR*0803 was a risk factor in Asian groups, not Caucasian groups. Also, the subgroup meta-analysis by ethnicity for HLA-DR*13 indicated that HLD-DR*13 was protective against PBC in Caucasian groups with a smaller heterogeneity.
Figure 3. Meta-analysis of the studies of HLA-DRB1*08, *0801, *0803 and PBC risk.
a: differences in the number of individuals who carried the HLA-DRB1 allele, comparing PBC patients and controls. b: differences in the frequency of the HLA-DRB1 allele between PBC patients and controls.doi:10.1371/journal.pone.0079580.g003
|HLA Class II||studies included (N)||OR (95%CI)||Q test||I2 test||Begg' test||Egger' test||Association|
|HLA-DR2||7||0.70 (0.53–0.91)||0.372||7.4%||0.553||0.230||Protective factor|
|HLA-DR5||8||0.55 (0.39–0.78)||0.268||20.3%||0.902||0.610||Protective factor|
|HLA-DR8||12||2.82 (1.84–4.30)||0.011||54.8%||1||0.901||Risk factor|
|HLA-DR12||3||0.40 (0.19–0.86)||0.583||0||0.296||0.207||Protective factor|
|HLA-DR16||3||0.21 (0.05–0.97)||0.900||0||1||0.124||Protective factor|
|HLA-DR52a||3||0.60 (0.38–0.95)||0.459||0||1||0.613||Protective factor|
|HLA-DQB1*02 a||3||1.40 (1.00–1.97)||0.204||37.2%||1||0.710||Risk factor|
|HLA-DQB1*0301 a||3||0.62 (0.48–0.79)||0.468||0||1||0.997||Protective factor|
|HLA-DQB1*0301 b||4||0.48 (0.35–0.68)||0.245||27.8%||1||0.558||Protective factor|
|HLA-DQB1*04 a||4||2.24 (1.46–3.46)||0.692||0||0.308||0.247||Risk factor|
|HLA-DQB1*0401 b||3||1.41 (1.07–1.85)||0.607||0||0.296||0.459||Risk factor|
|HLA-DQB1*0402 a||4||3.47 (2.35–5.13)||0.560||0||0.308||0.150||Risk factor|
|HLA-DQB1*06 a||3||0.58 (0.39–0.85)||0.433||0||1||0.836||Protective factor|
|HLA-DQB1*0601 b||4||1.99 (1.57–2.53)||0.130||46.9%||1||0.778||Risk factor|
|HLA-DQB1*0602 a||3||0.61 (0.42–0.87)||0.180||41.8%||0.296||0.229||Protective factor|
|HLA-DQB1*0602 b||4||0.53 (0.37–0.75)||0.330||12.6%||1||0.589||Protective factor|
|HLA-DQB1*0604||3||0.3 (0.18–0.58)||0.667||0||0.296||0.328||Protective factor|
|HLA-DRB1*01 b||6||1.26 (1.05–1.51)||0.134||40.7%||0.452||0.128||Risk factor|
|HLA-DRB1*03 b||5||1.38 (1.16–1.65)||0.920||0||0.221||0.065||Risk factor|
|HLA-DRB1*0405 b||4||1.43 (1.16–1.76)||0.749||0||0.734||0.328||Risk factor|
|HLA-DRB1*07 a||3||2.03 (1.29–3.21)||0.415||0||1||0.874||Risk factor|
|HLA-DRB1*07 b||5||1.47 (1.27–1.70)||0.836||0||0.806||0.083||Risk factor|
|HLA-DRB1*08 a||8||2.30 (1.76–3.00)||0.508||0||0.035||0.185||Risk factor|
|HLA-DRB1*08 b||6||2.48 (1.60–3.84)||0.035||58.4%||1||0.714||Risk factor|
|HLA-DRB1*0801 a||4||3.23 (2.22–4.70)||0.617||0||0.308||0.272||Risk factor|
|HLA-DRB1*0801 b||4||2.64 (1.55–4.51)||0.695||0||0.734||0.982||Risk factor|
|HLA-DRB1*0803 b||6||3.00 (1.89–4.76)||0.043||56.3%||0.707||0.786||Risk factor|
|HLA-DRB1*11 a||5||0.45 (0.33–0.62)||0.729||0||0.462||0.265||Protective factor|
|HLA-DRB1*11 b||5||0.31 (0.27–0.38)||0.327||13.6%||0.806||0.190||Protective factor|
|HLA-DRB1*1101 b||4||0.41 (0.24–0.68)||0.336||11.4%||0.734||0.965||Protective factor|
|HLA-DRB1*13 b||5||0.82 (0.68–0.99)||0.369||6.6%||0.462||0.225||Protective factor|
|HLA-DRB1*1501 b||3||0.45 (0.31–0.65)||0.186||40.6%||1||0.175||Protective factor|
Figure 4. Meta-analysis of the studies of HLA-DRB1*11, *1101 and *13 and PBC risk.doi:10.1371/journal.pone.0079580.g004
HLA haplotype and PBC
Of the 34 studies included, only 6 studies included discussions regarding the association between HLA haplotypes and PBC susceptibility [7,10,20,29,37,46]. Due to less overlap among HLA haplotypes detected in these 6 studies, a meta-analysis could not be conducted.
Sensitively analysis and assessment of bias
To assess the influence of each individual study on the overall OR, a sensitivity analysis was conducted by repeating the meta-analysis sequentially excluding one study at a time. Except for HLA-DRB1*13, the estimated pooled ORs were constant and the overall results were relatively stable in the present meta-analysis. The results indicate that the frequency of the HLA-DRB1*13 allele in PBC patients is significantly higher than in controls, but this relationship was not observed in individuals who carried the HLA-DRB1*13 allele between the PBC group and control group (OR = 0.668, 95%CI: 0.428–1.043; I2 = 59.0%, P = 0.045). However, after excluding Bittencourt’s study , the difference in the number of individuals who carried the HLA-DRB1*13 allele, comparing PBC patients and controls, achieved significance (OR = 0.553, 95%CI: 0.419–0.729), and this occurred with a lower heterogeneity (I2 = 13.8%, P = 0.323), supporting the hypothesis that inclusion of this particular study may lead to bias.
As shown in Table 2, significant heterogeneity (P < 0.1) existed in the three meta-analyses including DR8 and HLA-DRB1 (*08, *0803). To investigate this finding, a meta-regression analyses for ethnicity, year of publication, and sample size was performed, and to avoid ecological bias and to limit type I errors, we perform no other exploratory regression analyses, especially for patient-level factors such as sex and age . The meta-regression analysis showed that the major source of the heterogeneity listed in the meta-analyses for HLA-DR8 and HLA-DRB1*08 could not be attributed to ethnicity, year of publication, or sample size (data not shown). However, the results obtained from meta-regression analysis for HLA-DRB1*0803 indicated that the year of publication may contribute to findings of heterogeneity, not ethnicity or sample size. After excluding two studies conducted before 2000 [7,9], significant heterogeneity disappeared (I2 = 0%, P = 0.677). The meta-analyses of HLA-DRB1*08 included a study in which PBC patients were a “mixed population”, which may explain heterogeneity. However, the heterogeneity in the meta-analysis of HLA-DRB1*08 did not decrease after excluding this study (I2 = 63.8%, P = 0.026) .
The funnel plot for associations between HLA Class II and PBC susceptibility was drawn (data not shown) and publication bias was also measured by formal testing for funnel plot asymmetry using Egger’s test and Begg’s test. As expected, no significant publication bias was detected (Table 2).
To our knowledge, this is the first study to systemically review and meta-analyze all eligible published data to assess an association between HLA Class II and PBC risk. We discovered that common genetic variants of HLA-Class II were significantly associated with the development of PBC. A total of 26 HLA Class II alleles or antigens were found to be associated significantly with PBC as indicated by our meta-analysis of worldwide published data (Table 2). Also, 13 of those variants were found to be significant risk factors for PBC including DR8, HLA-DQB1 (*02, *04, *0401, *0402 and *0601), and HLA-DRB1 (*01, *03, *07, *08, *0801, *0803 and *0405). Protective factors against PBC were found to include HLA-DR2, DR5, DR12, DR16, DR52a, HLA-DQB1 (*06, *0301, *0602 and *0604) and HLA-DRB1 (*11, *13, *1501, and *1101). No consistent conclusion regarding the association of HLA haplotypes and PBC risk could be drawn due to the small number of included studies. However, our findings provide important insight into the role of HLA class II in the immune-pathogenesis of PBC.
A total of 34 studies were included in the present study (1983 to 2012), and 18 of these were determined to be of high quality; 16 were of medium quality; and no study was considered to be of low quality. Therefore, the meta-analysis presented here should reliably clarify which HLA-Class II genes may be responsible for PBC.
Our data confirmed that HLA-DR*08, *0801, and *0803 were potential risk factors for PBC (corresponding serological type DR8). Previous studies indicated that HLA-DR8 was significantly associated with PBC, with ORs ranging from 2.4 to 3.3 based on the population examined . In contrast, several reports have failed to confirm this association [30,33]. Nevertheless, evidence for this positive association was powerful in the present investigation, with the same direction of effect detected (OR = 2.82, 95%CI: 1.84–4.30). DRB1*08 was also observed to have a strong genetic association with PBC among many reports for Caucasian or Asian populations [8,10,38], but other studies failed to show a association between DRB1*08 and PBC [36,41,44].
Our work supports the hypothesis that both the number of individuals carrying the HLA-DRB1*08 allele and the frequency of the HLA-DRB1*08 allele in PBC groups were significantly greater than in control groups. Moreover, the results showed that HLA-DRB1*0801 and *0803 alleles conferred a predisposition to PBC. HLA-DRB1*08, *0801 and *0803 and HLA-DRB1*01, *03, *0405 and *07 were also identified to be risk-conferring alleles, but with relatively weak effects. Although the specific role of HLA-Class II risk alleles in PBC remains unclear, HLA molecules encoded by risk alleles such as DR8, might preferentially bind and present auto-antigens to T cells and trigger an autoimmune response [50,51]. Further studies are required to elucidate mechanisms by which HLA-class II genetic polymorphisms and pathogenesis cooperate in disease states such as PBC.
Strong protective associations between HLA-DRB1*11, *1101, *13 and *1501 alleles and PBC were observed, and, although these findings have been replicated by several other studies, the data from these reports were inconsistent [39-41]. We therefore combined quantitative evidence from these studies to offer direct support for the conclusion that HLA-DRB1*11, *1101, *13 and *1501 alleles had strong protective effects against PBC, which more clearly integrates the relationship between HLA-class II and PBC . Previous studies support the finding that HLA-DRB1*11 and *13 alleles were potent protective factors against some viruses [52-54]; thus, these HLA-Class II alleles may influence resistance to several infectious agents, and the lack of such alleles may lead to the onset of PBC through molecular mimicry of infectious agents. Several studies suggest that infection may be critical to the development of PBC [4,55,56], and our data also support a potential role for infections in PBC etiology.
Data derived from the GWAS of Canadian and Japanese patients indicated that HLA-DQB1 had the strongest association with PBC. Here, we found 8 HLA-DQB1 alleles that were associated with PBC and the association of the HLA-DQB1*0402 allele with PBC (OR 3.47, 95%CI 2.35–5.13) was strongest among them. All conclusions in 4 studies regarding the association between HLA-DQB1*0402 and PBC risk were consistent; this allele is a risk factor for PBC [10,20,31,38]. In contrast, the HLA-DQB1*03, *06, *0602, and *0604 alleles conferred protection against PBC. However, the meta-analyses indicated that no HLA-DPB1 alleles were significantly associated with PBC. Most previous studies have also provided strong evidence for no association between HLA-DPB1 and PBC. Until now, only two studies have suggested that HLA-DPB1*0301 and *0501 are associated with PBC, but these were based on relatively small sample populations [27,34]. A recent study from a large Italian cohort indicates that HLA-DPB1*0301 was a predisposing risk allele, a finding that is consistent with the previous study of a small German cohort . However, as indicated in previous studies, our data support the finding that most HLA associations with PBC could be attributed to specific associations with HLA-DRB1 and HLA-DQB1 alleles rather than HLA-DPB1.
As mentioned before, the GWAS provided suggestive evidence for a strong association of PBC with HLA region polymorphisms. The strength of the association between HLA-class alleles and the risk of PBC in our current study was similar to data from other large-scale studies such as GWAS. In Juran’s study, 7 HLA class II alleles were reported to achieve genome-wide significance including HLA-DQA1*0501, HLA-DQB1 (*0301, *0302 and *0402) and HLA-DRB1 (*0801, *1101 and *1501) . With the exception of HLA-DQB1*0302 and HLA-DQA1*0501, six other HLA-class II alleles were included in the present study. Data show that roles of these HLA-class II alleles in PBC development were quite consistent between Juran’s study and our study which suggested that HLA-DQB1*0402 and HLA-DRB1*0801 contribute to the PBC susceptibility and that conversely, HLA-DQB1*0301, HLA-DRB1*1101, and HLA-DRB1*1501 were protective alleles against PBC. Our estimated ORs were also quite similar to Juran’s results obtained from three independent large-scale cohorts. Collectively, this information indirectly reflects the accuracy and reliability of the present meta-analysis to study the role of HLA class II in PBC.
Studies to associate specific HLA haplotypes and susceptibility to PBC are scarce and those which exist indicate that several HLA haplotypes were significantly increased or reduced in PBC patients, suggesting that they confer susceptibility or protection to PBC, respectively. For example, the haplotype HLA-DRB1*0801-DQB1*0402 was identified as a risk factor for PBC in Caucasian subjects [7,34], and genes of the haplotype were found to be in linkage disequilibrium in the Caucasian population . Furthermore, the frequency of the HLA-DRB1*0801-DQA1*0401-DQB1*0402 haplotype was reported to be increased in patients who had progressed to late stages of PBC, but not in those with early stage disease, suggesting that this haplotype may be a specific marker for the overall disease course . We could not perform a meta-analysis to ascertain the association of HLA haplotype and PBC risk due to too few relevant original articles. Still, understanding and identifying these haplotype effects may improve research into or clinical treatment of PBC so future studies should be conducted to determine whether specific HLA haplotypes are associated with PBC susceptibility.
Our meta-analysis had some limitations. First, we included a wide variety of articles to determine the role of HLA class II in PBC development, so specific differences within studies may be a potential source of bias. Second, all available studies were published data; unpublished data were not identified. This fact alone suggests that publication bias cannot be absolutely excluded even though no significant publication bias was observed by funnel plot analysis, Egger’ test, and Begg’s test in the meta-analyses. Next, it is also impossible to completely exclude the influence of confounding factors inherent in these included studies, although subgroup analyses by ethnicity were performed. Other confounders such as age, sex, country, specific technique could not be excluded, and this may explain our findings. Although we performed 32 meta-analyses, 11 of these only included 3 eligible studies. We could not conduct sufficient subgroup meta-analyses after stratification by relevant characteristics in each analysis. Finally, the data regarding PBC and other HLA-Class II alleles were extremely sparse and inadequate, limiting our ability to draw conclusions from the meta-analysis.
In summary, our investigations suggest that distinct HLA class II genetic variants conferred both a predisposition and a resistance to PBC. HLA-DQB1 (*02, *04, *0401, *0402 and *0601) and HLA-DRB1 (*01, *03, *0405, *07, *08, *0801, and *0803) were identified as risk factors for PBC, whereas HLA-DQB1 (*0301, *06, *0602 and *0604), and HLA-DRB1 (*11, *1101, *13 and *1501) were potent protective factors. Also, DR8 was identified to be a predisposing factor. These results expand the repertoire of HLA-Class II genes with potential roles in PBC pathogenesis, however follow-up biological studies are needed to confirm these associations.
Meta-analysis of the studies of HLA-DR serological antigens and PBC risk.
Meta-analysis of the studies of HLA-DQ risk alleles and PBC.
Meta-analysis of the studies of HLA-DQ protective alleles and PBC.
Methodological quality of included studies according to the NEWCASTLE-OTTAWA Quality Assessment Scale.
Conceived and designed the experiments: BDQ RQZ. Performed the experiments: BDQ JQW JC. Analyzed the data: JQW JC. Contributed reagents/materials/analysis tools: YL ZXY. Wrote the manuscript: BDQ ZXY RQZ.
- 1. Sood S, Gow PJ, Christie JM, Angus PW (2004) Epidemiology of primary biliary cirrhosis in Victoria, Australia: high prevalence in migrant populations. Gastroenterology 127: 470-475. doi:10.1053/j.gastro.2004.04.064. PubMed: 15300579.
- 2. Selmi C, Mayo MJ, Bach N, Ishibashi H, Invernizzi P et al. (2004) Primary biliary cirrhosis in monozygotic and dizygotic twins: genetics, epigenetics, and environment. Gastroenterology 127: 485-492. doi:10.1053/j.gastro.2004.05.005. PubMed: 15300581.
- 3. Selmi C, Invernizzi P, Zuin M, Podda M, Gershwin ME (2005) Genetics and geoepidemiology of primary biliary cirrhosis: following the footprints to disease etiology. Semin Liver Dis 25: 265-280. doi:10.1055/s-2005-916319. PubMed: 16143943.
- 4. Smyk D, Cholongitas E, Kriese S, Rigopoulou EI, Bogdanos DP (2011) Primary biliary cirrhosis: family stories. J Autoimmune Dis, 2011: 189585. PubMed: 21687641.
- 5. Mitchell MM, Lleo A, Zammataro L, Mayo MJ, Invernizzi P et al. (2011) Epigenetic investigation of variably X chromosome inactivated genes in monozygotic female twins discordant for primary biliary cirrhosis. Epigenetics 6: 95-102. doi:10.4161/epi.6.1.13405. PubMed: 20864813.
- 6. Cho JH, Gregersen PK (2011) Genomics and the multifactorial nature of human autoimmune disease. N Engl J Med 365: 1612-1623. doi:10.1056/NEJMra1100030. PubMed: 22029983.
- 7. Begovich AB, Klitz W, Moonsamy PV, Van de Water J, Peltz G et al. (1994) Genes within the HLA class II region confer both predisposition and resistance to primary biliary cirrhosis. Tissue Antigens 43: 71-77. doi:10.1111/j.1399-0039.1994.tb02303.x. PubMed: 8016844.
- 8. Mullarkey ME, Stevens AM, McDonnell WM, Loubière LS, Brackensick JA et al. (2005) Human leukocyte antigen class II alleles in Caucasian women with primary biliary cirrhosis. Tissue Antigens 65: 199-205. doi:10.1111/j.1399-0039.2005.00351.x. PubMed: 15713222.
- 9. Onishi S, Sakamaki T, Maeda T, Iwamura S, Tomita A et al. (1994) DNA typing of HLA class II genes; DRB1*0803 increases the susceptibility of Japanese to primary biliary cirrhosis. J Hepatol 21: 1053-1060. doi:10.1016/S0168-8278(05)80617-8. PubMed: 7699227.
- 10. Donaldson PT, Baragiotta A, Heneghan MA, Floreani A, Venturi C et al. (2006) HLA class II alleles, genotypes, haplotypes, and amino acids in primary biliary cirrhosis: a large-scale study. Hepatology 44: 667-674. doi:10.1002/hep.21316. PubMed: 16941709.
- 11. Invernizzi P, Selmi C, Poli F, Frison S, Floreani A et al. (2008) Human leukocyte antigen polymorphisms in Italian primary biliary cirrhosis: a multicenter study of 664 patients and 1992 healthy controls. Hepatology 48: 1906-1912. doi:10.1002/hep.22567. PubMed: 19003916.
- 12. Nakamura M, Nishida N, Kawashima M, Aiba Y, Tanaka A et al. (2012) Genome-wide association study identifies TNFSF15 and POU2AF1 as susceptibility loci for primary biliary cirrhosis in the Japanese population. Am J Hum Genet 91: 721-728. doi:10.1016/j.ajhg.2012.08.010. PubMed: 23000144.
- 13. Hirschfield GM, Liu X, Xu C, Lu Y, Xie G et al. (2009) Primary biliary cirrhosis associated with HLA, IL12A, and IL12RB2 variants. N Engl J Med 360: 2544-2555. doi:10.1056/NEJMoa0810440. PubMed: 19458352.
- 14. Invernizzi P (2011) Human leukocyte antigen in primary biliary cirrhosis: an old story now reviving. Hepatology 54: 714-723. doi:10.1002/hep.24414. PubMed: 21563204.
- 15. Oremus M, Oremus C, Hall GB, McKinnon MC (2012) Inter-rater and test-retest reliability of quality assessments by novice student raters using the Jadad and Newcastle-Ottawa Scales. BMJ Open 2: ([MedlinePgn:]) PubMed: 22855629.
- 16. Liang Y, Yang Z, Zhong R (2012) Primary biliary cirrhosis and cancer risk: a systematic review and meta-analysis. Hepatology 56: 1409-1417. doi:10.1002/hep.25788. PubMed: 22504852.
- 17. Higgins J, Thompson SG (2002) Quantifying heterogeneity in a meta-analysis. Statist Med 21: 1539-1558. doi:10.1002/sim.1186.
- 18. Friedrich JO, Adhikari NK, Beyene J (2007) Inclusion of zero total event trials in meta-analyses maintains analytic consistency and incorporates all available data. BMC Med Res Methodol 7: 5. doi:10.1186/1471-2288-7-5. PubMed: 17244367.
- 19. Morling N, Dalhoff K, Fugger L, Georgsen J, Jakobsen B et al. (1992) DNA polymorphism of HLA class II genes in primary biliary cirrhosis. Immunogenetics 35: 112-116. PubMed: 1735557.
- 20. Underhill J, Donaldson P, Bray G, Doherty D, Portmann B et al. (1992) Susceptibility to primary biliary cirrhosis is associated with the HLA-DR8-DQB1*0402 haplotype. Hepatology 16: 1404-1408. doi:10.1002/hep.1840160616. PubMed: 1359995.
- 21. Briggs DC, Donaldson PT, Hayes P, Welsh KI, Williams R et al. (1987) A major histocompatibility complex class III allotype (C4B 2) associated with primary biliary cirrhosis (PBC). Tissue Antigens 29: 141-145. PubMed: 3496684.
- 22. Johnston DE, Kaplan MM, Miller KB, Connors CM, Milford EL (1987) Histocompatibility antigens in primary biliary cirrhosis. Am J Gastroenterol 82: 1127-1129. PubMed: 3479012.
- 23. Manns MP, Bremm A, Schneider PM, Notghi A, Gerken G et al. (1991) HLA DRw8 and complement C4 deficiency as risk factors in primary biliary cirrhosis. Gastroenterology 101: 1367-1373. PubMed: 1936808.
- 24. Prochazka EJ, Terasaki PI, Park MS, Goldstein LI, Busuttil RW (1990) Association of primary sclerosing cholangitis with HLA-DRw52a. N Engl J Med 322: 1842-1844. doi:10.1056/NEJM199006283222603. PubMed: 2348837.
- 25. Bassendine MF, Dewar PJ, James OF (1985) HLA-DR antigens in primary biliary cirrhosis: lack of association. Gut 26: 625-628. doi:10.1136/gut.26.6.625. PubMed: 3859460.
- 26. Gores GJ, Moore SB, Fisher LD, Powell FC, Dickson ER (1987) Primary biliary cirrhosis: associations with class II major histocompatibility complex antigens. Hepatology 7: 889-892. doi:10.1002/hep.1840070516. PubMed: 3498674.
- 27. Mella JG, Roschmann E, Maier KP, Volk BA (1995) Association of primary biliary cirrhosis with the allele HLA-DPB1*0301 in a German population. Hepatology 21: 398-402. doi:10.1016/0270-9139(95)90099-3. PubMed: 7843712.
- 28. Miyamori H, Kato Y, Kobayashi K, Hattori N (1983) HLA antigens in Japanese patients with primary biliary cirrhosis and autoimmune hepatitis. Digestion 26: 213-217. doi:10.1159/000198892. PubMed: 6409699.
- 29. Underhill JA, Donaldson PT, Doherty DG, Manabe K, Williams R (1995) HLA DPB polymorphism in primary sclerosing cholangitis and primary biliary cirrhosis. Hepatology 21: 959-962. doi:10.1002/hep.1840210411. PubMed: 7705806.
- 30. Zhang L, Weetman AP, Bassendine M, Oliveira DB (1994) Major histocompatibility complex class-II alleles in primary biliary cirrhosis. Scand J Immunol 39: 104-106. doi:10.1111/j.1365-3083.1994.tb03346.x. PubMed: 7904771.
- 31. Donaldson P, Agarwal K, Craggs A, Craig W, James O et al. (2001) HLA and interleukin 1 gene polymorphisms in primary biliary cirrhosis: associations with disease progression and disease susceptibility. Gut 48: 397-402. doi:10.1136/gut.48.3.397. PubMed: 11171832.
- 32. Gregory WL, Mehal W, Dunn AN, Cavanagh G, Chapman R et al. (1993) Primary biliary cirrhosis: contribution of HLA class II allele DR8. Q J Med 86: 393-399. PubMed: 7909617.
- 33. Mehal WZ, Gregory WL, Lo YM, Cross SJ, Fleming KA et al. (1994) Defining the immunogenetic susceptibility to primary biliary cirrhosis. Hepatology 20: 1213-1219. doi:10.1002/hep.1840200517. PubMed: 7927254.
- 34. Seki T, Kiyosawa K, Ota M, Furuta S, Fukushima H et al. (1993) Association of primary biliary cirrhosis with human leukocyte antigen DPB1*0501 in Japanese patients. Hepatology 18: 73-78. doi:10.1016/0270-9139(93)91821-9. PubMed: 8100798.
- 35. Akimoto S, Abe M, Ishikawa O, Takagi H, Mori M (2001) HLA-DRB1 and DQB1 genes in anticentromere antibody positive patients with SSc and primary biliary cirrhosis. Ann Rheum Dis 60: 639-640. doi:10.1136/ard.60.6.639. PubMed: 11386258.
- 36. Invernizzi P, Battezzati PM, Crosignani A, Perego F, Poli F et al. (2003) Peculiar HLA polymorphisms in Italian patients with primary biliary cirrhosis. J Hepatol 38: 401-406. doi:10.1016/S0168-8278(02)00440-3. PubMed: 12663229.
- 37. Stone J, Wade JA, Cauch-Dudek K, Ng C, Lindor KD et al. (2002) Human leukocyte antigen Class II associations in serum antimitochondrial antibodies (AMA)-positive and AMA-negative primary biliary cirrhosis. J Hepatol 36: 8-13. doi:10.1016/S0168-8278(01)00237-9. PubMed: 11804658.
- 38. Wassmuth R, Depner F, Danielsson A, Hultcrantz R, Lööf L et al. (2002) HLA class II markers and clinical heterogeneity in Swedish patients with primary biliary cirrhosis. Tissue Antigens 59: 381-387. doi:10.1034/j.1399-0039.2002.590504.x. PubMed: 12144621.
- 39. Bittencourt PL, Palácios SA, Farias AQ, Abrantes-Lemos CP, Cançado EL et al. (2003) Analysis of major histocompatibility complex and CTLA-4 alleles in Brazilian patients with primary biliary cirrhosis. J Gastroenterol Hepatol 18: 1061-1066. doi:10.1046/j.1440-1746.2003.03091.x. PubMed: 12911663.
- 40. Jiang XH, Zong RQ, Fang XY, An F, Hu Y et al. (2004) [Relationship between alleles of HLA-DRB and HLA-DQB1 and Chinese patients with primary biliary cirrhosis]. Zhonghua Gan Zang Bing Za Zhi 12: 436. PubMed: 15268814.
- 41. Liu HY, Deng AM, Zhou Y, Yao DK, Xu DX et al. (2006) Analysis of HLA alleles polymorphism in Chinese patients with primary biliary cirrhosis. Hepatobiliary Pancreat Dis Int 5: 129-132. PubMed: 16481299.
- 42. Vázquez-Elizondo G, Ponciano-Rodríguez G, Uribe M, Méndez-Sánchez N (2009) Human leukocyte antigens among primary biliary cirrhosis patients born in Mexico. Ann Hepatol 8: 32-37. PubMed: 19221531.
- 43. Chen S, Zhu GJ, Deng GH, Zhu Y (2005) Clinical pathology and HLA-DRB1 allele associations of primary biliary cirrhosis and autoimmune cholangitis in Han Chinese. World Chin J Digestology 13: 2450-2454.
- 44. Chong VH, Telisinghe PU, Jalihal A (2010) Primary biliary cirrhosis in Brunei Darussalam. Hepatobiliary Pancreat Dis Int 9: 622-628. PubMed: 21134832.
- 45. Nakamura M, Yasunami M, Kondo H, Horie H, Aiba Y et al. (2010) Analysis of HLA-DRB1 polymorphisms in Japanese patients with primary biliary cirrhosis (PBC): The HLA-DRB1polymorphism determines the relative risk of antinuclear antibodies for disease progression in PBC. Hepatol Res 40: 494-504. doi:10.1111/j.1872-034X.2010.00631.x. PubMed: 20374297.
- 46. Umemura T, Joshita S, Ichijo T, Yoshizawa K, Katsuyama Y et al. (2012) Human leukocyte antigen class II molecules confer both susceptibility and progression in Japanese patients with primary biliary cirrhosis. Hepatology 55: 506-511. doi:10.1002/hep.24705. PubMed: 21953406.
- 47. Zhao J, Su HB, Zhu B, Chi SP, Wu YC et al. (2006) Relationship between HLA-DRB alleles and primary biliary cirrhosis in patients living in northern China. Acad J Second Mil Med Univ 27: 1286-1289.
- 48. Higgins JP, Thompson SG (2004) Controlling the risk of spurious findings from meta-regression. Stat Med 23: 1663-1682. doi:10.1002/sim.1752. PubMed: 15160401.
- 49. Mells GF, Floyd JA, Morley KI, Cordell HJ, Franklin CS et al. (2011) Genome-wide association study identifies 12 new susceptibility loci for primary biliary cirrhosis. Nat Genet 43: 329-332. doi:10.1038/ng.789. PubMed: 21399635.
- 50. Carson DA (1992) Genetic factors in the etiology and pathogenesis of autoimmunity. FASEB J 6: 2800-2805. PubMed: 1634042.
- 51. Thorsby E, Lie BA (2005) HLA associated genetic predisposition to autoimmune diseases: Genes involved and possible mechanisms. Transpl Immunol 14: 175-182. doi:10.1016/j.trim.2005.03.021. PubMed: 15982560.
- 52. Alric L, Fort M, Izopet J, Vinel JP, Charlet JP et al. (1997) Genes of the major histocompatibility complex class II influence the outcome of hepatitis C virus infection. Gastroenterology 113: 1675-1681. doi:10.1053/gast.1997.v113.pm9352872. PubMed: 9352872.
- 53. Hendel H, Caillat-Zucman S, Lebuanec H, Carrington M, O'Brien S et al. (1999) New class I and II HLA alleles strongly associated with opposite patterns of progression to AIDS. J Immunol 162: 6942-6946. PubMed: 10352317.
- 54. Malhotra U, Holte S, Dutta S, Berrey MM, Delpit E et al. (2001) Role for HLA class II molecules in HIV-1 suppression and cellular immunity following antiretroviral treatment. J Clin Invest 107: 505-517. doi:10.1172/JCI11275. PubMed: 11181650.
- 55. Padgett KA, Selmi C, Kenny TP, Leung PS, Balkwill DL et al. (2005) Phylogenetic and immunological definition of four lipoylated proteins from Novosphingobium aromaticivorans, implications for primary biliary cirrhosis. J Autoimmun 24: 209-219. doi:10.1016/j.jaut.2005.01.012. PubMed: 15848043.
- 56. Smyk DS, Bogdanos DP, Kriese S, Billinis C, Burroughs AK et al. (2012) Urinary tract infection as a risk factor for autoimmune liver disease: from bench to bedside. Clin Res Hepatol. Gastroenterologist 36: 110-121. doi: 10.1016/j.clinre.2011.07.013
- 57. Invernizzi P, Ransom M, Raychaudhuri S, Kosoy R, Lleo A et al. (2012) Classical HLA-DRB1 and DPB1 alleles account for HLA associations with primary biliary cirrhosis. Genes Immun 13: 461-468. doi:10.1038/gene.2012.17. PubMed: 22573116.
- 58. Juran BD, Hirschfield GM, Invernizzi P, Atkinson EJ, Li Y et al. (2012) Immunochip analyses identify a novel risk locus for primary biliary cirrhosis at 13q14, multiple independent associations at four established risk loci and epistasis between 1p31 and 7q32 risk variants. Hum Mol Genet 21: 5209-5221. doi:10.1093/hmg/dds359. PubMed: 22936693.