Interplay of PNPLA3 and HSD17B13 Variants in Modulating the Risk of Hepatocellular Carcinoma among Hepatitis C Patients

A single-nucleotide polymorphism causing a C to G change in the PNPLA3 gene (rs738409) is associated with disease severity and development of hepatocellular carcinoma (HCC) in nonalcoholic fatty liver disease; the insertion variant rs72613567:TA of the 17β-hydroxysteroid dehydrogenase type 13 (HSD17B13) mitigates this detrimental effect. Our aim was to evaluate if the same holds true in chronic hepatitis C virus infection (HCV). With a case control retrospective study design, we selected 110 patients who developed HCC on a background of HCV infection, matching each patient for sex and age (±30 months) to three HCV-infected, non-HCC patients. All participants underwent genotyping for PNPLA3 and HSD17B13 gene variants. Both univariate and multivariate analyses of risk factors for advanced disease and HCC were performed. Carriage of PNPLA3 G∗ allele was associated with a trend of progressively more severe liver disease, from mild fibrosis to significant fibrosis, cirrhosis, and HCC (p = 0.007). When the HSD17B13:TA status of these patients was taken into account, the abovementioned trend was strengthened among HSD17B13 major allele homozygotes and completely blunted among carriers of the minor allele (p = 0.0003 and 0.953, respectively). In a conditional logistic regression model including diabetes and AST to platelet ratio index among predictor variables, the unfavourable genetic profile characterized by the coexistence of the PNPLA3 minor allele and HSD17B13 major allele (vs. all other possible combinations) was an independent risk factor for HCC (OR = 2.00, 95% CI: 1.23-3.26) together with a history of alcohol abuse. In conclusion, carriage of the combination PNPLA3 minor allele and HSD17B13 major allele may represent a risk factor for HCC among HCV-infected patients. The interplay between the two genes may explain some of the controversy on this topic and may be exploited to stratify HCC risk in hepatitis C.


Introduction
The natural history of chronic liver diseases is highly variable, ranging from minimal histological damage to cirrhosis and hepatocellular carcinoma (HCC) [1,2]. Some of this variability is genetically determined. The carriage of the G allele in the single-nucleotide polymorphism (SNP) I148M (rs738409) of the adiponutrin gene (PNPLA3) on a background of metabolic, alcoholic, and viral disease is one of the best examples of a gene variant affecting the natural history of chronic liver diseases [3]. In fact, carriers of the G allele at the abovementioned locus have a reduction in the enzymatic activity of adiponutrin, resulting in higher intracellular triglyceride levels [3], which might then increase their risk to develop advanced fibrosis, cirrhosis, and HCC. As far as hepatitis C virus-(HCV-) infected patients are concerned, however, the association between PNPLA3 variants and HCC has been debated, suggesting that, in this condition, other factors may mask the alleged effect on hepatocarcinogenesis [4,5].
Recently, a protein-truncating variant responsible for a loss-of-function mutation of in the HSD17B13 gene (rs72613567) has been associated with a decrease in aminotransferase levels and a reduction of liver injury in the setting of a fatty liver, mitigating the risk conferred by the carriage of PNPLA3 I148M variant [6,7]. HSD17B13 codes for a retinol dehydrogenase, 17β-hydroxysteroid dehydrogenase type 13, which is highly expressed in the liver. Although the physiological role of this protein has not been fully elucidated yet, it appears to be involved in lipid biosynthesis, as well as in the modulation of sex hormone, bile acid, and fatty acid redox processes [8,9]. Importantly, the truncated variant has been shown to be protective against HCC development among patients with alcoholic liver disease [10].
Since alcoholic and nonalcoholic fatty livers have an impact on the natural history of HCV infection [11], we wondered whether carriage of HSD17B13:TA, by modulating the detrimental effects of carrying an unfavorable PNPLA3 genotype, may protect against cirrhosis and HCC development in patients with long-standing HCV infection.

Study Population.
We designed a case-control study, including N = 440 patients attending the liver clinic of an academic hospital in northern Italy, affected by longstanding chronic HCV infection at different stages. The study has been conducted in strict accordance with the principles of the Declaration of Helsinki; all the patients gave an informed consent to their participation to the study, which was approved by the local ethical committee (Comitato Etico Interaziendale di Novara, no. 176/18).
Because timing of infection is hard to determine, we chose to include only patients aged 60 years or older, thus reducing the probability of a duration of infection shorter than 20 years. Therefore, we applied the following inclusion criteria: age ≥ 60 years and laboratory evidence of chronic HCV infection.
The only exclusion criterion was the inability to give informed consent to participation in the study.
To assess the role of rs72613567:TA in liver disease progression, we considered N = 110 patients with HCV-related HCC of any-BCLC (Barcelona Clinic Liver Cancer) stage [12]. For each of them, we selected three sex-and age-(±30 months) matched controls among patients affected by chronic HCV infection without HCC development, referred to the liver clinic for the diagnosis and treatment of their condition.
We collected demographic, clinical, and biochemical data simultaneously to blood sampling.

HCC Diagnosis.
The diagnosis of HCC has been made by imaging (computed-tomography scan or magnetic resonance imaging) [13] and was confirmed by a liver biopsy when clinically appropriate, according to EASL Clinical Practice Guidelines [12]. All the HCC patients included liver cancer superimposed to cirrhosis.
2.4. Genetic Studies. Genomic DNA was extracted from whole blood or buffy coat, using a commercial kit (Invitrogen, Carlsbad, CA, USA), according to the manufacturer's instructions. DNA was then amplified by polymerase chain reaction (PCR). The PCR primer sequences used for PNPLA3 amplification were as follows: forward: 5 ′ -CCTGCAGGCAG GAGATGTGT-3 ′ ; reverse: 5 ′ -GCCCTGCTCACTTGGA GAAA-3 ′ . The PCR primer sequences used for HSD17B13 amplification were as follows: forward: 5 ′ -GTCTGAGGC ATGAGAATTGCT-3 ′ ; reverse: 5 ′ -GGCCTGTATTGGAG ACAGATG-3′. To define the genotype of the two target genes, we performed a restriction fragment length PCR. NLA-III and TRU1I restriction enzymes (Life Technologies, Thermo Fisher Scientific; Carlsbad, California, US) were used to digest PNPLA3 and HSD17B13, respectively. All samples were amplified twice; when discordant, they were run the third time. The lab technicians performing the genetic studies were blinded about the case/control state of the subjects enrolled.

Statistical
Analysis. Statistical analysis of data was carried out with the software package Stata, version 13.1 (StataCorp LP, College Station, Texas, US). The measures of centrality and dispersion of data chosen were medians and interquartile ranges. Continuous variables were compared between groups by the Mann-Whitney test; the nonparametric test chosen to identify a trend across ordered groups was that developed by Cuzick. Exact Fischer's test and Pearson's chi-square test were used, as appropriate, to explore the associations of categorical variables. Moreover, to better understand the relationship between severity of liver disease (in four ordered categories: mild, moderate, or advanced fibrosis and HCC) and variants of the two genes of interest, we run two ordered logistic models (one with and one without an interaction term), followed by a likelihood ratio test to evaluate whether these two models were statistically different.
The χ 2 G test "Goodness of Fit" was employed to verify whether the proportions of the two polymorphisms were distributed in patients in accordance with the Hardy-Weinberg equation. Finally, a conditional logistic regression analysis was performed to test the independency of the association between the genes of interest and HCC development.

Gastroenterology Research and Practice
The level of significance chosen for all statistical tests was 0.05 (two-tailed). Table 1 presents the main demographic, anthropometric, and clinical characteristics of the study population. Cases and controls were well-matched with regard to age and sex and were of similar body mass index. Diabetics and individuals with a history of excess alcohol consumption (as defined by the Italian drinking guidelines for the general population) [17] were overrepresented among cases. Also, AST levels were slightly, but significantly, higher among cases. Lastly, HCC patients showed a significantly higher liver fibrosis with either of the two proxy measures (LS, APRI) we used. The characteristics of controls when categorized in three subgroups based on METAVIR stage are shown in Supplementary Table 1. 3.2. Genotype and Allele Frequencies for rs738409 and rs72613567. Table 2 shows genotype and allele frequencies of the rs738409 (PNPLA3) SNP and the insertion variant rs72613567 (HSD17B13) in the study population. The proportion of patients who carried the rs738409 G allele increased moving from mild to moderate fibrosis, to cirrhosis, and to HCC. This trend was far stronger among rs72613567 major allele homozygotes and completely blunted among carriers of the rs72613567 insertion variant allele ( Figure 1). Moreover, in the first of two ordered logistic models having severity of liver disease as an outcome variable and variants of the genes of interest (without an interaction term) as predictor variables, the likelihood ratio chi-square test was 7.62 (p = 0:022). In the second model, also having severity of liver disease as an outcome variable but including as predictor variables variants of the genes of interest plus an interaction term between them, the likelihood ratio chisquare test was 13.9 (p = 0:011). When compared with a likelihood ratio test, the two models were statistically different (p = 0:012), confirming the importance of the interaction between the two genes.

Characteristics of the Study Population.
Carriers of the rs738409 G allele who also carried the major allele for rs72613567, the combination shown to increase the risk of developing fibrosis in the presence of a fatty liver, were N = 109/440 (25%). The proportion of patients with such "high-risk" combination was significantly higher than that of patients with HCC compared to age-and sex-matched controls without HCC (38/110, 35% vs. 71/330, 22%, p = 0:007).

Logistic Regression Analysis Models for HCC.
At conditional logistic regression analysis, being a carrier of the dummy variable "high-risk genetic combination" (1 = carriage of both rs738409 minor allele and rs72613567 major allele; 0 = all other combinations between the two genes of interest) and having a history of alcohol abuse were predictors of HCC independent of diabetes and AST to platelet ratio index. Table 3 summarizes the results of such logistic model (N of observations = 440). Running two further conditional logistic regression models (i.e., substituting the dummy variable "high-risk genetic combination" either with each variant of the genes of interest, or with an interaction

Gastroenterology Research and Practice
term) demonstrated only a nonsignificant trend (p = 0:061) for the stratum of patients corresponding to carriers of the "high-risk combination", but the likelihood ratio test did not detect any significant difference between the two models (p = 0:355).

Discussion
The main finding of the present study is the demonstration that specific variants of HSD17B13 and PNPLA3 genes interplay in the determination of genetic predisposition to chronic liver disease progression and liver cancer in HCV-infected subjects. These data will be discussed at the light of the current literature on this topic.
Chronic HCV infection remains a leading cause of chronic liver diseases (CLD) worldwide [18], despite the high rates of sustained viral response obtained with direct antiviral agents [19]. The burden of long-standing HCV-infected patients who have developed significant liver damage constitutes an important target for surveillance, also after having achieved viral clearance [20]. Indeed, in cross-sectional and case-control studies, HCV infection is associated with a 15to 20-fold risk of developing HCC compared with HCVnegative subjects [21].
Liver carcinogenesis is the result of a multistep process, in which the main risk factor is the presence of chronic liver damage leading to persistent inflammation [2]. However, in the last years, genome-wide association studies have identified some genetic variants potentially involved in liver disease progression and, eventually, in HCC development [3,10,22]. Among them, the best validated relationship is that occurring between the I148M variant of adiponutrin-coding gene and CLD of different etiologies [3]. Specifically looking to HCV infection, PNPLA3 I148M variant has been related to an  Gastroenterology Research and Practice increased risk of liver fibrosis progression [23]. Although the role of PNPLA3 mutation in fibrogenesis is widely acknowledged, the underlying mechanisms are poorly understood.
What is known is that rs738409 C>G variant causes the accumulation of the mutated protein on the surface of lipid droplets, impairing the mobilization of triglycerides [24,25]. Furthermore, in humans, this missense mutation also leads to a deranged fat efflux from the liver [26] and to impaired hydrolysis of retinyl esters in hepatic stellate cells [27]. As a result, lipid turnover is deeply altered, and this may favor persistence of inflammatory trigger(s). The evidence on an association between PNPLA3 mutation and HCC is less striking. In fact, while the rs738409 mutation has been convincingly associated to liver carcinogenesis in alcoholic liver disease [28,29] and in nonalcoholic steatohepatitis [30,31], this association in viral hepatitis has been the subject of controversy [32,33]. While some authors reported an increased risk of HCC development in HCV patients carrying the G allele [34][35][36][37][38], others failed to confirm these findings [4,[39][40][41][42]. The reason for this discrepancy may belong to the different ethnicities of the study populations; in fact, the largest part of the studies failing to demonstrate the association of this SNP with HCC has been conducted on Asian populations. In a very recent meta-analysis, the association between HCC and rs738409 has been confirmed in Caucasians [5]. This is in line with other reports, according to which carriage of the PNPLA3 mutation is a risk factor for liver disease progression and HCC in Caucasians but not in Asians [23,43]. Taking these findings together, it is reasonable to postulate that carriage of the G allele in the rs738409 polymorphism contributes to both fibrogenesis and liver carcinogenesis, but other genetic contributors might modulate its effect, thus explaining the differences observed among ethnic groups.
In 2017, Abul-Husn et al. firstly reported that the loss of function variant HSD17B13:TA mitigates the risk of liver disease progression conferred by carriage of PNPLA3 I148M among NAFLD patients [6]. Further recent data suggest a potential protective role of HSD17B13:TA with regard to CLD progression in HCV infection [44], although its interplay with PNPLA3 polymorphism has not been investigated in this context. Our data are consistent with both studies. In fact, in a population of chronic HCV-infected patients, the prevalence of rs738409 C>G variant of adiponutrin gene increases for progressive degrees of liver fibrosis, being lowest for mild fibrosis and highest in patients affected by cirrhosis. Even more strikingly, this trend is abolished by the concomitant presence of HSD17B13:TA allele and strengthened in HSD17B13 major allele homozygotes, with a significant effect of the interaction between the two genes. This is in line with the hypothesis that the loss of function of HSD17B13 might be protective against the increased risk of liver fibrosis progression attributable to adiponutrin mutation. The mechanisms of the interplay between these two proteins are still unknown, though they are probably related to lipid turnover. In fact, HSD17B13 encodes a protein localized on the surface of lipid droplets, which is upregulated in the livers of patients and mice with NAFLD. Its overexpression is associated to an increase in the number and size of lipid droplets. Consistent with this interpretation, the hepatic overexpression of HSD17B13 in murine models significantly increases lipogenesis and triglyceride content in the liver, leading to a fatty liver phenotype [45]. The HSD17B13:TA variant results in a truncated protein with reduced enzymatic activity [46].
The second question that we tried to address was whether carrying the allele HSD17B13:TA is also protective against HCC development in HCV patients. In a recent study by Yang et al., the presence of HSD17B13:TA was protective against HCC in alcoholic liver disease, but not in HCV-infected subjects; however, the authors did not analyze the contribution of PNPLA3 to the overall genetic risk [10]. According to our data, the protective effect of HSD17B13:TA is extended to HCC. In fact, when HCC and non-HCC patients are compared, the strict association between the G allele of PNPLA3 and the diagnosis of HCC is lost in the group of patients carrying at least one HSD17B13 minor allele.
In a conditional logistic regression model, we confirmed the independent contribution given by the combination of 5 Gastroenterology Research and Practice the two gene variants to the risk of developing HCC along the course of HCV infection. The coexistence of PNPLA3 G allele and HSD17B13 major allele homozygosity is an unfavourable, high-risk genetic profile for HCC, which is independent from alcohol abuse, degree of fibrosis, and diagnosis of diabetes mellitus. Thus, the association with hepatocarcinogenesis may not to be entirely accounted for by the consequences of accelerated fibrogenesis that-not unexpectedly-do play a major role. A clinical implication of these findings relates to the potential application of genotyping to screening strategies. Risk stratification for HCC development may take advantage, in the future, from gene-based prediction models, which might find a place in clinical practice.
A further interesting finding emerging from our conditional regression analysis is the independent contribution of alcohol abuse to liver cancer in HCV-infected subjects. There is ample evidence that alcohol synergistically contributes to the detrimental viral insult on liver disease progression [47,48]. HCV-infected patients are already discouraged from alcohol consumption: whether this needs to be reinforced in the presence of an unfavourable genetic profile, even after resolution of HCV infection, remains to be ascertained.
The retrospective, cross-sectional design of the present paper does not allow us to infer any causal relationship. Moreover, the fact that the association between the two genes of interest and HCC did not reach statistical significance when all their combinations were tested together suggests that our findings await confirmation in larger studies on prospectively recruited patients. However, to our knowledge, this is the first study designed to test the interplay between PNPLA3 and HSD17B13 on a homogeneous cohort of longstanding HCV-infected patients, being controlled for two major factors influencing the course of HCV infection, i.e., age and sex.
In conclusion, HSD17B13:TA mitigates the increased risk of liver fibrosis progression and HCC in HCV-infected patients with PNPLA3 I148M variant; further studies are required to better elucidate the underlying pathogenic mechanisms and to evaluate the potential clinical applications of this observation.