Introduction

Infection with HBV is a major public health issue throughout the world, especially in Asian countries (Schweitzer et al. 2015). Despite the effective use of the potent HBV vaccine and antiviral treatments, HBV infection maintains a high prevalence worldwide, with an estimated 240 million people affected chronically in 2015 (World Health Organization 2015; http://www.who.int/mediacentre/factsheets/fs204/en/), and is the world’s tenth leading cause of death (Trepo et al. 2014). Approximately 60 % of hepatocellular carcinoma (HCC) is associated with chronic hepatitis B (CHB), and the infection is responsible for many cases of liver cirrhosis (LC) (Lai et al. 2003; Perz et al. 2006).

Infection with HBV and its progression to LC and HCC are influenced by viral, environmental, and host factors (Thursz 2001). Genome-wide association studies (GWAS) have been recently used increasingly to identify genetic variants in HBV-related liver diseases, which revealed several novel common variants located in ubiquitin-conjugating enzyme E2L3 (Hu et al. 2013), solute carrier family 10 member 1 (Peng et al. 2015), glutamate receptor ionotropic kainate 1 (Li et al. 2012), kinesin family member 1B (Li et al. 2012), and signal transducer and activator of transcription 4 (STAT4) (Jiang et al. 2013). However, many of these newly identified susceptibility gene findings have not been replicated in independent studies.

STAT4 is located on chromosome 2q32.3 and encodes a crucial transcription factor (Zhong et al. 1994). Several cytokines, such as interleukin (IL)-2 (Wang et al. 1999), IL-12 (Bacon et al. 1995) and interferon (IFN)-α/β (Nguyen et al. 2002), are reported to activate STAT4 for transmitting signals to its downstream target genes. In natural killer (NK) and T cells, STAT4 is phosphorylated predominately by the binding of IL-12 and its receptor. Then the phosphorylated STAT4 stimulates NK and T cells to produce IFN-γ, an important cytokine for antiviral replication and antitumor activity (Wang et al. 2015). Moreover, it has been reported that STAT4 plays critical roles in other viral infections, such as respiratory syncytial virus (Dulek et al. 2014), herpes simplex virus (HSV) 2 (Svensson et al. 2012), and lymphocytic choriomeningitis virus (Holz et al. 1999). A recent GWAS found that SNP rs7574865 in STAT4 was associated with HBV infection and HBV-related HCC (Jiang et al. 2013), but this association has not been replicated yet in any other Chinese Han population.

To further investigate the association of STAT4 polymorphisms with HBV-related liver diseases, we performed an association study of single and multiple SNPs within STAT4 in a large Chinese Han sample.

Materials and methods

Subject recruitment and demographic characteristics

This case–control study included 3033 Chinese Han individuals comprising 1610 with chronic HBV infection (CHBVI) and 1423 uninfected control subjects (Table 1). A detailed description of this sample has been reported previously (Tao et al. 2015).

Table 1 Clinical demographics of subjects in each subgroup
Full size table

Based on the laboratory testing results and the “Guideline of Prevention and Treatment for Chronic Hepatitis B” defined by the Chinese Medical Association (Chinese Society of and Chinese Society of Infectious Diseases 2011), 1610 CHBVI patients could be identified as asymptomatic carriers (AsC; n = 388), CHB (n = 853), LC (n = 296), or HCC (n = 73). The diagnostic criteria for these HBV-related liver diseases were as follows: (1) AsC: seropositive for either HBsAg or HBV-DNA and persistently normal ALT; (2) CHB: seropositive for either HBsAg or HBV-DNA and recurrently abnormal or continuously elevated ALT for >6 months; (3) LC: presence of ascites, hypersplenism, elevated ALT, and appropriate imaging findings; (4) HCC: abnormal serum alpha-fetoprotein and imaging findings. The 1423 control subjects were classified as healthy (n = 749) or HBV clearance (n = 674). The definition of healthy subjects and clearance was: (1) healthy: negative for HBsAg, HBcAb, and HBsAb; and (2) clearance: negative for HBsAg but positive for HBcAb and HBsAb.

Subjects were excluded if they were: (1) co-infected with any other hepatitis virus; (2) suffered from any other type of cancer or autoimmune disease; (3) positive for human immunodeficiency virus; (4) effectively immunized against HBV; (5) not Chinese Han ethnicity; or (6) pregnant.

DNA extraction, SNP selection and genotyping, and quality control

Genomic DNA was extracted from 3 ml of peripheral blood of each subject using the Qiagen DNA purification kit. Selection of 13 SNPs in STAT4 was based on the HapMap database (Release #27) with a minor allele frequency (MAF) >0.1 in the Chinese population (Fig. 1; Table S1). Of the 13 selected variants, SNPs rs3821236, rs1517352, rs10168266, rs11889341, rs4274624, rs7574865, rs8179673, rs11685878, rs7572482, and rs897200 have been reported to be associated with autoimmune or viral disease, including systemic lupus erythematosus (SLE) (Kawasaki et al. 2008), inflammatory bowel disease (Jostins et al. 2012), primary biliary cirrhosis (PBC) (Joshita et al. 2014), autoimmune Addison’s disease (Mitchell et al. 2014), Behcet’s disease (Hou et al. 2012), HBV-related HCC (Jiang et al. 2013), or HSV-2 infection (Svensson et al. 2012).

Fig. 1
figure 1

Location of selective SNPs in STAT4 with a total length of 122 kb. Exons or UTRs for STAT4 are shown as black bars and introns as horizontal black lines. This figure is shown in its original scale

Full size image

All SNPs included in the study were genotyped in a 384-well microplate format using the TaqMan SNP Genotyping Assay (Applied Biosystems, Foster City, CA, USA). The genotyping rate for each SNP was required to be >95 %, and the P value for the Hardy–Weinberg equilibrium was set to >0.05.

Literature search for meta-analyses on the association of rs7574865 with HBV infection and clearance

Our meta-analysis was performed according to the criteria suggested in the PRISMA guidelines (Moher et al. 2009). We searched PubMed and Web of Science for all eligible studies that could be included in this study. The key words used in the search strategy were: “rs7574865” or “STAT4” or “signal transducer and activator of transcription 4” and “hepatitis B” or “hepatitis B virus” or “HBV”. The identified reports were checked individually for reference to studies that were not found in the abovementioned electronic databases. When different samples of the same population were included in a study, each one was treated as an independent one in this report. For each study, the following data were extracted: author, publication year, ethnicity, study design, genotyping method, sample size, mean age, and sex ratio.

The inclusion criteria for the meta-analysis were that: (1) the study was designed as a case–control study; (2) the study provided the number of CHBVI patients, healthy individuals, and subjects with HBV clearance, and the allele frequency of rs7574865; (3) the study presented sufficient information for calculating the odds ratio (OR) and the 95 % confidence interval (CI).

Statistical analysis

The population admixture of subjects was assessed by principal component (PC) analysis using 291 ancestry informative markers (AIMs) selected from the Illumina arrays. For individual SNP-based association analysis, the association of STAT4 polymorphisms with HBV-related liver diseases was performed under an additive genetic model using PLINK (v. 1.07) (Purcell et al. 2007). The ORs and 95 % CIs were calculated by logistic regression with age, sex, and the first seven PCs as covariates. After Bonferroni correction for multiple tests, the appropriate significance of individual SNP was set to P = 3.8 × 10−3 (0.05/13). Pairwise linkage disequilibrium (LD) was determined by the D′ value, and haplotype blocks were generated by the Haploview software (v. 4.2) (Barrett et al. 2005). The association of each haplotype with HBV-related liver diseases was analyzed using R package “haplo.stats” (http://cran.r-project.org/web/packages/haplo.stats; Yang and Li 2014). The P value was given by the score test under the same genetic model and covariates as used in the individual SNP-based association analysis (Schaid et al. 2002). The significance of each haplotype was adjusted by the number of major haplotypes (frequency >5 %) included in the analysis. For the SNPs within the significant haplotypes, variant effects on the regulation of transcription were analyzed using HaploReg (v. 2) (http://www.broadinstitute.org/mammals/haploreg/haploreg_v2.php). In this database, evolutionarily conserved regions are predicted by genome evolutionary rate profiling (GERP) (Ward and Kellis 2012), the enhancer histone mark prediction was based on the evidence of local H3K4Me1 modification, and variant effects on regulatory motifs were evaluated according to TRANSFAC, JASPAR and protein-binding microarray experiments (Ward and Kellis 2012).

Meta-analysis was performed to assess the association of rs7574865 with HBV infection and clearance using Comprehensive Meta-Analysis software (v. 2.0; Biostat Inc., Englewood, NJ, USA, as we described previously (Ma et al. 2015a, b)). Heterogeneity among the studies was assessed by Cochran’s Q and I 2 statistic tests (Higgins et al. 2003). If the P value of the Q test is <0.1, indicating heterogeneity among the studies, the random-effects model is more applicable. Otherwise, use of the fixed-effects model is encouraged. The I 2 test is used to analyze the effect of heterogeneity and the degree of inconsistency among the studies. If Cochran’s Q P value is >0.1 and the I 2 value is ≤50 %, we consider that there exists no strong evidence for heterogeneity (Higgins et al. 2003; Taylor et al. 2015). In this work, both the random-effects and fixed-effects models were used in the meta-analysis. The Z statistical test was used to calculate the significance of pooled ORs. A funnel plot was applied to assess potential publication bias. Sensitivity analysis was performed to detect the stability of results. The significance level was set to P < 0.05 (two-sided) for meta-analysis.

Results

Association analysis of STAT4 polymorphisms with HBV infection and clearance

To identify the association of STAT4 polymorphisms with HBV infection and clearance, we compared individual SNP-based association analysis of 13 SNPs in 1610 CHBVI patients vs. 749 healthy subjects and 1610 CHBVI patients vs. 674 clearance subjects under an additive genetic model after adjusting for age, sex, and the first seven PCs (Table 2).

Table 2 Association of SNPs in STAT4 with HBV infection and clearance under additive genetic model
Full size table

We found that SNP rs7574865 was significantly associated with HBV infection risk (OR 1.15; 95 % CI 1.00, 1.31; P = 0.046), providing an independent replication of a previous report (Jiang et al. 2013). Furthermore, this SNP was negatively correlated with HBV clearance (OR 1.17; 95 % CI 1.02, 1.33; P = 0.028). That is, rs7574865 plays a role in HBV susceptibility, as well as a retarding role in HBV clearance.

An additional four SNPs showed a nominally significant correlation with HBV infection and clearance (Table 2). When comparing CHBVI with the healthy group, we obtained the OR and its corresponding P value for SNPs rs10168266 (OR 1.19; P = 0.013), rs11889341 (OR 1.18; P = 0.018), rs4274624 (OR 1.13; P = 0.056), and rs8179673 (OR 1.14; P = 0.047). As for the HBV clearance, the ORs and their corresponding P values were as follows: rs10168266 OR 1.15; P = 0.043; rs11889341 OR 1.15; P = 0.041; rs4274624 OR 1.15; P = 0.050; and rs8179673 OR 1.15; P = 0.040. Thus, all of the SNPs showed a risk effect on HBV infection and clearance.

Association analysis of STAT4 polymorphisms with disease progression

To reveal the relation of STAT4 variants to the progression of HBV-related liver diseases, we performed the following three comparisons: (1) AsC vs. CHB; (2) CHB vs. LC; and (3) LC vs. HCC. We found no significant association of the STAT4 variants with disease progression after adjusting for age, sex, and the first seven PCs (detailed results not shown).

Haplotype analysis of STAT4 polymorphisms with HBV infection and clearance

Haplotype-based association analysis was performed for all SNPs in STAT4 regardless of whether they were significant or non-significant according to the individual association analysis. One haplotype, CTCTT, formed by rs8179673, rs7574865, rs4274624, rs11889341, and rs10168266, showed a significant association with HBV infection and clearance (Table 3; Fig. 2). When comparing the CHBVI with the healthy group, the haplotype CTCTT was significantly associated with HBV infection (Hap Score = −2.290; P = 0.022). Similarly, the haplotype CTCTT was significantly associated with HBV clearance when comparing the CHBVI with the clearance group (Hap Score = −2.370; P = 0.018). Together, these results indicate that the haplotype formed by the minor alleles of the abovementioned SNPs is protective against HBV susceptibility and beneficial for HBV clearance.

Table 3 Haplotype analysis of STAT4 polymorphisms with HBV infection and clearance
Full size table
Fig. 2
figure 2

LD maps and haplotype blocks of STAT4 generated by Haploview software (v. 4.2). The LD map was determined by the D′ value (D′ >0.9), which was drawn from the genotypic data of the CHBVI and healthy subjects (a) and the genotypic data of the CHBVI and clearance subjects (b). LD linkage disequilibrium, CHBVI chronic HBV infection (includes AsC, CHB, LC, and HCC), AsC asymptomatic carrier, CHB chronic hepatitis B, LC liver cirrhosis, HCC hepatocellular carcinoma

Full size image

Meta-analysis of association of rs7574865 with HBV infection and clearance

To determine the effects of rs7574865 on HBV infection and clearance with a large sample, we performed meta-analysis for those reported studies along with the current one. Based on this search strategy, we included 10 reported studies and one unpublished data set totaling 8944 CHBVI patients, 8451 healthy individuals, and 2081 subjects with HBV clearance (Table S2). Under the fixed-effects model, rs7574865 was significantly associated with HBV infection (pooled OR 1.14; 95 % CI 1.07, 1.21; pooled P = 3.8 × 10−5; Fig. 3a) and clearance (OR 1.20; 95 %CI 1.07, 1.35; P = 0.002; Fig. 4a). Similarly, we performed a meta-analysis under the random-effects model, which again revealed rs7574865 to be significantly associated with HBV infection (pooled OR 1.14; 95 % CI 1.07, 1.21; pooled P = 3.8 × 10−5; Fig. 3b) and clearance (pooled OR 1.25; 95 % CI 1.01, 1.55; pooled P = 0.044; Fig. 4b). Considering that we found no significant evidence for the presence of heterogeneity among the studies for the meta-analysis of HBV infection (P = 0.82; I 2 = 0) and clearance (P = 0.19; I 2 = 34 %), we believed the results from either model are acceptable, and both models provide further evidence for the significant association of rs7572865 with HBV infection and clearance. Next, we performed sensitivity analysis by sequentially omitting individual studies under the fixed-effects model and found that the P value remained significant for all cases (data not shown).

Fig. 3
figure 3

Forest plots of meta-analysis results for the association of rs7574865 with HBV infection in the CHBVI and healthy groups under the fixed-effects model (a) and the random-effects model (b). The Z and P values of each study are shown by rows. The central vertical solid line represents ORs that equal 1 under the null hypothesis. The ORs and 95 % CIs are represented by the square and horizontal bars, respectively. The pooled OR is indicated by the diamond

Full size image
Fig. 4
figure 4

Forest plots of meta-analysis results for the association of rs7574865 with HBV clearance in the CHBVI and clearance groups under the fixed-effects model (a) and the random-effects model (b). The Z and P values of each study are shown by rows. The central vertical solid line represents ORs that equal 1 under the null hypothesis. The ORs and 95 % CIs are represented by the square and horizontal bars, respectively. The pooled OR is indicated by the diamond symbol

Full size image

Function prediction analysis of SNPs in STAT4

We applied the HaploReg tool (Ward and Kellis 2012) to predict the function of the five SNPs in the significant haplotype CTCTT (Table S3). Of the five, SNP rs7574865 was predicted to be conserved based on the GERP; SNP rs11889341 was associated with histone modifications in B cells from peripheral blood that were suggestive of enhancer activity. In addition, four SNPs (rs10168266, rs11889341, rs7574865, and rs8179673) were predicted to affect specific regulatory motifs such as Foxp1, Foxo3, HNF1, and HNF4. These predictions indicated that the SNPs have important effects on transcription by influencing the binding of transcription factors or regulatory motifs.

Discussion

A recent GWAS study identified STAT4 as a susceptibility gene for HBV infection and HBV-related HCC in the Chinese Han population (Jiang et al. 2013). However, Liao et al. failed to validate such an association using 1312 Chinese Han individuals (Liao et al. 2014). To determine whether this gene indeed plays any role in HBV-related liver diseases, we carried out this study with a much larger sample.

In individual SNP-based association analysis under the additive genetic model, SNP rs7574865 in STAT4 not only increased the probability of HBV infection (OR = 1.15; P = 0.046) but also reduced the probability of viral clearance (OR = 1.17; P = 0.028). The meta-analysis of the data on this SNP, incorporating all the data from 8944 CHBVI patients, 8451 healthy and 2081 HBV clearance individuals, also indicated SNP rs7574865 to be a risk factor for HBV infection and an inhibitor of clearance. During the preparation of this manuscript, Lu et al. (2015) reported that rs7574865 is related to HBV clearance with 596 subjects, whose result is consistent with ours. It has also been reported that rs7574865/G was significantly associated with lower mRNA values for STAT4 (Jiang et al. 2013). Because STAT4 was able to bind to the first intron of interferon-gamma, higher STAT4 expression was expected to enhance the activation of IFN-γ, a soluble cytokine with important effects on modulating antiviral activity and immunoregulatory functions (Morinobu et al. 2002), leading to the increased ability for HBV clearance and resistance to viral infection. Also, we found SNPs rs10168266, rs11889341, rs4274624, and rs8179673 to be nominally associated with HBV infection and viral clearance. It has been reported that SNPs rs10168266, rs11889341, and rs8179673 are significantly correlated with PBC and SLE susceptibility (Joshita et al. 2014; Kawasaki et al. 2008); and SNP rs4274624 is related to susceptibility to autoimmune Addison’s disease (Mitchell et al. 2014). These results suggest that STAT4 plays dual roles in autoimmune disease and viral infection.

We also investigated the roles of STAT4 polymorphisms in each stage of liver disease progression. None of the SNPs showed a significant association, which indicates that STAT4 plays different roles at different stages of HBV infection. STAT4 protects against HBV infection and increases HBV clearance, but it may play no significant role in disease progression. However, in a recent comparison of 712 LC patients and 2601 HBV-positive controls, Jiang et al. (2015) found that rs7574865 was significantly associated with disease progression, which is inconsistent with our findings. This discrepancy might be caused by a relatively small sample in a few HBV-related liver disease categories, and a large sample is required for future study.

Further, we performed haplotype-based association analysis and found a high protective CTCTT haplotype formed by rs8179673-rs7574865-rs4274624-rs11889341-rs10168266, which showed significant activity against HBV infection (Hap Score = −2.290; P = 0.022) and increased the probability of viral clearance (Hap Score = −2.370; P = 0.018). This suggested that these five SNPs jointly contribute to HBV infection and clearance, although they were not significantly associated with HBV-related liver diseases after Bonferroni correction based on individual SNP-based association analysis.

We then evaluated the potential biological functions of these five SNPs located in noncoding regions using the HaploReg tool (Ward and Kellis 2012). It predicted that SNP rs11889341 influences the enhancer activity in B cells on the basis of its location with a histone modification. Additionally, regulatory motif prediction suggested that SNPs rs10168266, rs11889341, rs7574865, and rs8179673 affect the affinity for binding of transcription factors. Many transcription factors are associated with viral infection and liver diseases. For rs11889341, the alternate allele increases affinity for Foxo3, which regulates CD8 T cell responses during chronic viral infection (Sullivan et al. 2012). For rs7574865, the alternate allele is predicted to have a lower affinity for Foxp1, which is reported to be associated with the risk of LC progression to HBV-related HCC (Li et al. 2015). We suspect that rs7574865 influences oncogenesis by changing the affinity for Foxp1. For rs8179673, its alternate allele is predicted to have lower binding of transcription factors HNF1 and HNF4 to STAT4. HNF1α repression can lead to activation of JAK/STAT pathways and facilitate the progression of chronic liver diseases (Qian et al. 2015), and HNF4 also plays an important role in HBV replication (Shin et al. 2014). However, it remains to be determined how these five SNPs contribute to HBV infection and clearance by affecting STAT4 transcriptional regulation.

In conclusion, this study confirmed the association of STAT4 polymorphisms with HBV susceptibility and viral clearance in a much larger sample than previously studied. In addition, haplotype-based association analysis revealed a highly protective CTCTT haplotype, which is also significantly associated with HBV infection and clearance. Finally, functional prediction analysis using the HaploReg tool indicated that these five SNPs likely play critical roles in the regulation of transcription of the STAT4 gene. However, further functional study is required to validate these predictions.