Polymorphisms in MICA, but not in DEPDC5, HCP5 or PNPLA3, are associated with chronic hepatitis C-related hepatocellular carcinoma

Recently, the MICA rs2596542 and DEPDC5 rs1012068 variants in Japanese individuals as well as the HCP5 rs2244546 and PNPLA3 rs738409 variants in European individuals have been found associated with hepatocellular carcinoma (HCC). The present study determined which single nucleotide polymorphism (SNP) is the most predictive for developing hepatitis C virus (HCV)-related HCC in a Japanese cohort. Of the 4 SNPs analysed, only the MICA genotypes were significantly associated with development of HCC (p = 0.0185). The major (MA), hetero (HE), and minor (MI) genotypes occurred in 40%, 41%, and 19% of HCC patients and in 43%, 47%, and 10% of non-HCC patients, respectively. Interestingly, the MICA genotype was significantly correlated with MICA mRNA and soluble protein levels. In patients older than 70 years, the MI genotype was significantly associated with HCC development. In addition, the MI genotype was related to HCC development when the platelet count range was 10–15 × 104/μL, corresponding with the fibrosis stage; but not when the range was less than 10, indicating advanced fibrosis; or greater than 15 × 104/μL, as mild fibrosis. Thus, polymorphisms in MICA, but not in DEPDC5, HCP5 or PNPLA3, are associated with HCC development in Japanese patients with chronic HCV infection.

between the HCC (n = 142) and non-HCC groups (n = 575) were observed in age, sex, aspartate transaminase (AST), alanine transaminase (ALT), platelet count, percentage of prothrombin time (PT%), albumin, and AFP (p < 0.0001) but not in HCV viral load or the IL28B or ITPA SNP. The mean age of the patients with HCC was significantly greater than that of the patients without HCC (67 vs. 59 years old). Seventy percent of patients with HCC were male, a value significantly greater than the 43% observed among patients without HCC. AST and ALT levels were significantly higher in the HCC group than in the non-HCC group (69 ± 41 and 67 ± 43 IU/L vs. 49 ± 34 and 55 ± 49 IU/L, respectively). The platelet counts, PT%, and albumin levels in patients with HCC were significantly lower than those in patients without HCC (12.3 ± 6.1 × 10 4 /μL, 88.0 ± 17.5% and 3.8 ± 0.4 g/dL vs. 16.7 ± 6.0 × 10 4 /μL, 99.1 ± 16.2% and 4.2 ± 1.8 g/dL, respectively). Moreover, the mean levels of the tumour markers AFP and PIVKA-II were 630 ng/mL and 2,098 mAU/mL, respectively, in the HCC group, which were significantly higher than those in the non-HCC group (11 ng/mL and 23 mAU/mL, respectively).

Association between the risk allele of SNPs in 4 genes and the development of HCC in patients with CHC.
To investigate the association between the MICA SNP and HCC development, we genotyped 717 samples by using TaqMan SNP genotyping assays. Major (MA), hetero (HE), and minor (MI) genotypes were present in 57 (40%), 58 (41%), and 27 (19%) patients with HCC, respectively, and in 246 (43%), 269 (47%), and 60 (10%) patients without HCC, respectively, indicating a significant association between the MICA genotype and HCC development in patients with CHC (p = 0.0185, Fig. 1). The MI allele frequencies (MAFs) in the HCC and non-HCC groups were 0.394 and 0.338, respectively. In addition to the association of all 717 patients, a significant association also was observed in sub-cohorts of 541 chronic (p = 0.0057, Fig. 2a) and 176 cirrhosis patients (p = 0.0453, Fig. 2b).
In addition to the MICA SNP, we also analysed three other reported HCV-related HCC SNPs. For the DEPDC5 SNP, we found MA, HE, and MI genotypes in 97, 40, and 5 patients with HCC, respectively, and in 412, 151, and 12 patients without HCC, respectively (p = 0.5161); the MAFs were 0.18 and 0.15 for the HCC and non-HCC groups, respectively. For the HCP5 SNP, we observed the MA, HE, and MI genotypes in 105, 34, and 3 patients with HCC, respectively, and in 455, 108, and 10 patients without HCC, respectively (p = 0.3668). Finally, the MA, HE, and MI PNPLA3 genotypes were observed in 39, 58, and 33 patients with HCC, respectively, and in 151, 262, and 129 patients without HCC, respectively (p = 0.7466, Fig. 1).
Independent factors related to HCC development. MICA SNP and variables with p values < 0.0001 in the univariate analysis (Table 1) were subjected to logistic regression analysis. These variables including age, sex, albumin, prothrombin time, AFP, AST, ALT, PIVKA-II concentration, platelets, and genotype of the MICA SNP were categorized and used to analyse associations with binary outcomes (HCC or non-HCC). Logistic regression analysis indicated that age (older than 65 years old), male sex, albumin ≤ 4 g/dL, prothrombin time ≤70%, AFP concentration ≥ 20 ng/mL, PIVKA-II concentration ≥ 40 mAU/mL, and minor genotype of the MICA SNP were independent factors that were significantly associated with HCC development ( Table 2).
The MICA SNP was correlated with MICA mRNA and soluble protein levels. Because the MI genotype of the MICA SNP was associated with a high risk of HCC development, we assessed whether rs2596542 was correlated with MICA expression in patients with HCV-related HCC. We examined the transcription level of MICA using paired HCC and adjacent non-tumour liver tissues from 21 individuals with HCV. As shown in Fig. 3a, real-time quantitative PCR assays revealed a significant decreased mRNA expression of MICA MI genotype in tumour tissues compared to MA genotype (p = 0.048). Furthermore, sMICA levels were measured in 36   serum samples (n = 6 of each genotype from patients with or without HCC) by using ELISAs. The results showed that in the HCC samples, median protein levels of sMICA from patients with the MA, HE, and MI genotypes were 80, 50, and 0 pg/mL, respectively (Fig. 3b). Although these protein levels tended to be higher in MA and HE, and lower in MI genotypes, they were not significantly different (p = 0.1284 by Kruskal-Wallis test). However, interestingly, in the non-HCC group, the sMICA levels were significantly correlated with its genotypes (p = 0.0498 by Kruskal-Wallis test, Fig. 3). These results suggest that MICA mRNA and protein expression likely correlate with the MICA genotype.
The MICA MI genotype is related to HCC development in patients older than 70 years. We determined whether the risk allele rs2596542 was related to HCC development when patients were stratified by age. By comparing patients with and without HCC across 3 groups of age (younger than 65 years old, 65-70 years old, and above 70 years old), we found that the MI genotype was significantly associated with HCC development in the subset of patients with CHC above 70 years old (p = 0.004, Fig. 4). Moreover, we analysed the ratio of patients with HCC to those without HCC with respect to the MICA genotypes in 5-year age ranges: younger than 55 years old, 55-59 years old, 60-64 years old, 65-69 years old, 70-74 years old, and above 74 years old. This ratio varied among MI allele carriers from 0.06 to 2.0, which was greater than the range of 0.06 to 0.46 observed among patients with the non-MI allele (Fig. 5).
The MICA MI genotype is associated with the development of HCC in patients with platelet counts in the range of 10-15 × 10 4 /μL. We took advantage of knowing that platelet count reflects stage of liver fibrosis 11 in which a decrease in platelet count in accordance with severity of fibrosis in CHC patients 12,13 . In this study, we determined the platelet count range at which MI rs2596542 was related to the development of HCC. We found that the MICA MI genotype was associated with HCC development when the platelet count was 10-15 × 10 4 /μL, corresponding with the fibrosis stage; but not when the range was less than 10, indicating advanced fibrosis; or greater than 15 × 10 4 /μL, as mild fibrosis (Fig. 6).

Discussion
Both host genetic and environmental factors affect the progression of liver disease over multiple stages. With respect to host genetic factors, variations in the IL28B gene affect the progression of CHC due to HCV exposure 14,15 , and variations in the ITPA gene are associated with the outcomes of interferon and ribavirin combination therapy 16 . The genetic factors that play a role in late-stage liver disease remain controversial, although several studies have shown that variants in MICA, DEPDC5, HCP5 and PNPLA3 affect HCC development [6][7][8]17 . The SNPs listed above were associated with HCV-related HCC in four separate cohorts. This study is the first to evaluate the association between these SNPs and HCV-related HCC in a unique cohort from Japan. We found a strong association between MICA SNP and HCC development, particularly in older patients or patients with fibrosis.
In our cohort, the MI allele A of rs2596542 in MICA was determined to be a likely risk factor for the development of HCC. This result is consistent with findings from a previous study in Japan, which has reported similar MAFs of 0.398 and 0.333 in HCC and non-HCC groups, respectively 7 . In contrast, Lange et al., in the Swiss Hepatitis C Cohort Study (SCCS), have found that the MI allele exerts a protective effect against HCC development 8 . The MI allele A of rs2596542 might be a risk allele in Japanese patients but a protective allele in European patients with HCC. In the SCCS, a lower MAF of 0.241 in patients with HCC and a higher MAF of 0.361 in patients without HCC supported this hypothesis. In addition, the proportion of patients with HCC in our cohort was 20% (142/717), a value much greater than the 3% (64/1924) observed in the SCCS. The association of MICA genotype with HCC development was also confirmed in both CHC and LC background patients of our cohort (Fig. 2). Importantly, our findings were validated using another cohort of 638 patients with CHC, which showed the same results with the first cohort (Supplementary Table S1 and Supplementary Fig. S1).
We did not find any associations between the 3 other SNPs tested (i.e., DEPDC5 rs1012068, HCP5 rs2244546 and PNPLA3 rs738409) and the development of HCC in patients with CHC. Although the DEPDC5 variant was associated with progression to HCC in a previous Japanese cohort 6 , our data did not show this association. The disparity between these results may be due to differences in study design. Miki studied patients over 55 years old, whereas our study enrolled patients regardless of age. This difference resulted in a higher proportion of patients with HCC in Miki's study (30%) compared with the 20% in our cohort. One additional explanation is  8 . Furthermore, the well-known alcoholic and nonalcoholic fatty liver disease-associated variant in PNPLA3 rs738409 9,19,20 has recently been shown to affect the development of HCC in European individuals with CHC 9,10,17 . In contrast, Japanese ethnicity may have resulted in no significant association of rs738409 with HCC development in our cohort. In Japan, a study showed that the PNPLA3 minor genotype is associated with the age at onset of HCC 21 , and another study with a small number of total patients as well as HCC patients has indicated that the PNPLA3 SNP is indirectly associated with HCC development via serum AFP level 22 . The greater proportion of HCC patients in our cohort might explain the apparent discrepancy. Moreover, Trepo 23 has noted that the results obtained from European patients with HCV-related HCC remain controversial because 2 other studies have not found a positive relationship between the PNPLA3 risk allele and HCC development in HCV-infected patients 24,25 . Furthermore, a secondary analysis of the data from the American HALT-C trial has not identified a significant association between rs738409 and HCC development in patients with HCV 26 . Additionally, MICA mRNA and protein levels were significantly correlated with the MICA variants, thus further supporting the association between the MICA MI genotype and the risk of HCC development. In normal tissues, MICA expression is relatively low but is elevated in tumour tissues 27 . Here we also found the MICA mRNA level were higher in liver tumour tissues compared to non-tumour area, down-regulated in MICA HE genotype, and lowest in MI genotype (Fig. 3a). As a natural killer (NK) group 2D ligand, membrane-bound MICA triggers the immune system, thereby resulting in the elimination of target tumour cells via NK and CD8+T cells 28,29 . In our study, which was consistent with the data from another Japanese cohort 7 , patients with HCC carrying the MI allele A of rs2596542 expressed low levels of sMICA protein, which is probably excreted through shedding of membrane-bound MICA protein, thus eliminating NK and CD8 + cell activity in viral-infected cells. Consequently, risk allele carriers are more likely to develop HCC from CHC. This same sMICA trend was found in patients with CHC, whose sMICA levels were 0 pg/mL in individuals with the MI genotype but gradually increased in those with the HE and MA genotypes. These findings were similar to those of Kumar, who has reported median sMICA levels of 0, 69, and 65 pg/mL for the MI, HE and MA genotypes, respectively, and levels of 0, 65 and 78 pg/ mL for patients without HCV, with CHC, and with HCC, respectively 7 . These results suggest that sMICA levels are elevated during the chronic stage of disease, and this condition has the potential to develop into HCC.
Hepatic fibrosis is a major risk factor for HCC in patients with CHC; however, age is another important factor because HCC can develop in older patients with CHC and mild fibrosis. To determine the age threshold at which the MI genotype of MICA is most related to HCC development, we compared the MI genotype in patients with and without HCC with respect to age. Our results suggest that older patients (>70 years old) with HCC exhibit an increased carriage of MICA MI. Other studies have used age cutoffs of 60 or 65 years old to analyse the  association between age and HCC development 6,7 . Moreover, our data suggest that MICA is a protective factor against HCC development in older patients because the MICA non-MI genotype predominated in older patients without HCC compared with those with HCC. Notably, the ratio of patients with HCC to those without HCC with the rs2596542 MI allele was higher than that in patients without the MI allele across all 5-year age ranges. In addition, with increasing age, the slope of the MI allele carriers was steeper than that of the non-MI allele carriers. Developing HCC was significantly associated with MICA and DEPDC5 SNPs in two separate studies; Kumar et al 7 . did not mention the ratio of HCC to non-HCC in their study of the MICA SNP, whereas this trend corroborates findings of Miki's study of the DEPDC5 SNP 6 . These results confirmed that HCC develops more often in older patients and that the risk allele is also associated with the development of HCC in older patients.
There is a reverse correlation between peripheral platelet count and liver fibrosis stage, in which the more platelet count the less liver fibrosis stage 11 . In our cohort, the prevalence of MICA genotypes in patients with HCC differed from that in patients without HCC with a platelet count of ≤10 × 10 4 /µL. Furthermore, the difference became significant in patients with a platelet count in the range of 10-15 × 10 4 /µL. No differences were found in patients with platelet counts greater than 15 × 10 4 /µL. Since the platelet count in the range of 10-15 × 10 4 /µL corresponds with the fibrosis stage 13,30 , our data suggest that the MI MICA genotype may associate with HCC in patients with fibrosis.
As with previous studies of these 4 SNPs, the present study has limitations. First, these studies were retrospective, cross-sectional analyses. With a longer follow-up period, more patients with continuous HCV infection would be likely to develop HCC, and patients' genetic backgrounds are not the only predictor of HCC development. Also, there was no information about past treatment for CHC in the present study. In addition, sMICA levels were examined in only 36 patients. However, the results were consistent with MICA mRNA level in HCC and non-tumour tissues.

Conclusion
The MICA rs2596542 genotype was correlated with MICA mRNA and protein levels, and the MI genotype of MICA was associated with an increased risk of HCC development in patients with HCV infection. This finding was particularly true for patients older than 70 years, even those with fibrosis. Our data suggest that MICA plays an important role in the development of HCC in patients with CHC.

Patients and Methods
Patients. This study was a cross-sectional analysis. A total of 717 patients were recruited at Osaka City University Hospital between December 2004 and December 2013. All patients had either a viral load of >10 5 IU/ mL according to the COBAS AMPLICOR HCV Monitor test, version 2.0 (Roche Diagnostics, Branchburg, NJ, USA), or a viral load of >5 log copies/mL as determined by the COBAS TaqMan HCV test (Roche Diagnostics). HCC was diagnosed at the conclusion of the data collection in December 2013. All patients provided written informed consent, and all methods were carried out in accordance with the ethical guidelines of the 1975 Declaration of Helsinki. All experimental protocols were approved by the ethical committee of Osaka City University, Graduate School of Medicine (approval No. 1646).
The exclusion criteria included a history or evidence of a serious chronic or poorly controlled medical or psychiatric condition and infection with human immunodeficiency virus or hepatitis B virus. Patients with autoimmune liver diseases, primary biliary cirrhosis, and heavy alcoholic habits were also excluded. Figure 6. MICA SNP MI genotype and platelet counts. The vertical axis shows the percentage of each genotype in HCC or non-HCC patients, and the data table shows the number of independent samples tested in each group of platelet counts. Plt ≤ 10, platelet counts equal or less than 10 × 10 4 /μL blood; 10 < Plt < 15, platelet counts in the range of 10-15 × 10 4 /μL blood; Plt ≥ 15, platelet counts equal or more than 15 × 10 4 /μL blood.
Measuring soluble MICA (sMICA) protein levels. We randomly chose a group of 6 patients with each MICA genotype from either the HCC or non-HCC group for further analysis of soluble MICA levels. The characteristics of these 36 patients were present in Supplementary Table S2. Serum was collected after the blood was allowed to clot. The clot was removed by centrifugation, and sMICA level in the resulting supernatants from 36 samples were quantified using a RayBio Human MICA ELISA Kit as described in the manufacturer's instructions (RayBiotech, Norcross, GA, USA). We assessed whether the soluble MICA levels decreased from the major to minor genotype in either group of HCC or non-HCC. We used the Kruskal-Wallis test for analysis.

Statistical analyses.
All data analyses were conducted using the JMP program, version 9.0 (SAS Institute, Cary, NC, USA). Individual between-group characteristics were evaluated using Wilcoxon's two-sample test for continuous variables or Fisher's exact test for categorical variables. Hardy-Weinberg equilibrium (HWE) was tested to assess the quality of the SNP data. To control for confounders, variables exhibiting p values < 0.0001 in univariate analysis were subjected to logistic regression analysis 31 . A p value of < 0.05 was considered to be statistically significant.