The PNPLA3 rs738409 148M/M Genotype Is a Risk Factor for Liver Cancer in Alcoholic Cirrhosis but Shows No or Weak Association in Hepatitis C Cirrhosis

Background An isoleucine>methionine mutation at position 148 in the PNPLA3 gene (p.I148M, rs738409) has recently been identified as a susceptibility factor for liver damage in steatohepatitis. Here, we studied whether the PNPLA3 rs738409 polymorphism also affects predisposition to hepatocellular carcinoma (HCC). Methods We compared distributions of PNPLA3 genotypes in 80 and 81 Caucasian patients with alcoholic and hepatitis C virus (HCV)-associated HCC to 80 and 81 age- and sex-matched patients with alcohol-related and HCV-related cirrhosis without HCC, respectively. PNPLA3 genotypes in 190 healthy individuals from the same population served as reference. Potential confounders obesity, diabetes, HCV genotype and HBV co-infection were controlled by univariate and multivariate logistic regression with forward variable selection. Results PNPLA3 genotypes were in Hardy-Weinberg equilibrium for all study groups. The frequency of the 148M allele was significantly (p<0.001) increased in alcoholic cirrhosis with (53.7%) and without HCC (36.2%) but was not different between healthy controls (22.9%) and patients with cirrhosis (25.3%; p = 0.545) and HCC (30.2%; p = 0.071) due to hepatitis C. HCC risk was highest in 148M/M homozygous patients with alcoholic liver disease (odds ratio (OR) 16.8 versus healthy controls; 95% confidence interval (CI) 6.68–42.43, p<0.001). Finally, multivariate regression confirmed 148M/M homozygosity (OR 2.8; 95%-CI: 1.24–6.42; p = 0.013) as HCC risk factor in alcoholic cirrhosis. In HCV-related cirrhosis only HCV genotype 1 was confirmed as a HCC risk factor (OR 4.2; 95%-CI: 1.50–11.52; p = 0.006). Conclusion The PNPLA3 148M variant is a prominent risk factor for HCC in patients with alcoholic cirrhosis, while its effects are negligible in patients with cirrhosis due to HCV. This polymorphism provides an useful tool to identify individuals with particularly high HCC risk in patients with alcoholic liver disease that should be taken into account in future HCC prevention studies.


Introduction
Hepatocellular carcinoma (HCC) is a leading cause of cancerrelated death worldwide [1,2]. HCC is mainly attributed to chronic viral hepatitis B and C in developing countries [3], whereas in Europe and North America approximately 45% of HCCs are caused by increased alcohol consumption [4].
Great interest has come from a genome-wide association study that identified a single-nucleotide polymorphism (rs738409C/G) in the PNPLA3 gene on chromosome 22, encoding an isoleuci-neRmethionine substitution (p.I148M) of patatin-like phospholipase A3, also termed adiponutrin as risk factor of steatohepatitis and liver cirrhosis in alcoholic and non-alcoholic fatty liver disease [5,6]. Adiponutrin is a transmembrane protein expressed in human adipose tissue and liver. Although its precise in vivo function is still unknown, current data indicate a pivotal role in lipid homeostasis. Adiponutrin expression is down-regulated during fasting and is induced during high calorie intake because gene activity is up-regulated in response to glucose, insulin, and thyroid hormones. Thus, adiponutrin is an important regulator of hepatic lipid metabolism during nutritional excess [6,7]. However, this genetic variant was not correlated to body mass index, visceral or subcutaneous fat content, insulin sensitivity or peripheral blood lipid levels [8,9]. Furthermore, it correlated inversely with carotid artery intima thickness, suggesting that the I148M polymorphism selectively affects fat deposition in the liver but is not linked to a general metabolic disorder [6,10,11]. However, the 148M variant was correlated with raised serum alanine aminotransferase levels [12,13,14] elevated hepatic fat content and increased rates of fibrosis in alcoholic and non-alcoholic fatty liver disease [15,16]. Here, we studied if the adiponutrin 148M allele had any effects on the risk of HCC development among cirrhotic patients with alcoholic liver disease as compared to chronic viral hepatitis C, an alternative strong risk factor for HCC.

Ethics Statement
The reported studies were approved by the Institutional Review Boards of the Bonn and Berlin University Ethics Committees. Written informed consent was obtained from the patients prior to sample collection. Samples were coded and data stored anonymously.

Study groups
We recruited 161 patients with hepatocellular carcinoma (HCC) at the Bonn and Berlin University Departments of Gastroenterology between 2005 and 2009. In 80 patients HCC was related to alcoholic cirrhosis and in 81 patients to chronic hepatitis C. These HCC patients were compared to 80 and 81 sex and age (63 years)-matched patients with alcoholic cirrhosis and HCVassociated cirrhosis, who did not have liver cancer. Cirrhotic patients without liver cancer had at least one year of follow-up to guard against the possibility of occult malignancy. 190 healthy volunteers from the same background population served as a reference. Patients were considered to have alcoholic cirrhosis if their history indicated average alcohol consumption to exceed 300 g ethanol per week. Patients with mixed HCV infection and increased alcohol consumption (.300 g/week) had been excluded from this study. Determination of rs738409 p.I148M alleles Genomic DNA was extracted from 200 mL EDTA-blood using the QIAamp Blood Mini Kit (Qiagen, Hilden, Germany) according to the manufacturer's protocol. Determination of the PNPLA3 rs738409 polymorphism was performed by LightCycler real time PCR (Roche, Mannheim, Germany) using a commercial LightSNiP (SimpleProbe) assay from TIB-MolBiol (Berlin, Germany) according to the manufacturer's recommendations.

Statistical analysis
Genotype frequencies were determined and tested for consistency with the Hardy-Weinberg equilibrium using an exact test. Allele and genotype frequencies were compared between cases and controls by Pearson's goodness-of-fit chi 2 test and Armitage's trend test, respectively (http://ihg2.helmholtz-muenchen.de/cgi-bin/ hw/hwa1.pl). Differences between groups were analyzed by t-test and Mann-Whitney-U test as appropriate.
To take into account further potential confounding risk factors of liver cancer the effects of HCV-genotype, HBV infection, diabetes and obesity were assessed by univariate comparisons (ANOVA and chi 2 -statistics) followed by multivariate logistic regression with forward variable selection. This analysis was performed separately in patients with alcohol-and HCV-related liver disease. Parameters with effects at p,0.1 were entered into the multivariate analysis with P,0.05 for inclusion and p.0.1 for exclusion as selection criterion for parameters in the final statistical models.
Statistical analysis was performed with SPSS 18.0 (SPSS, Munich, Germany). Data are reported as mean 6 standard deviation, unless stated otherwise.

Study population
Obesity and diabetes occurred more frequently in alcoholic cirrhosis without HCC than in alcoholic patients with HCC (table 1; p,0.05). Significantly higher GGT values where observed in the two groups with alcoholic liver disease than in patients with HCV infection (p,0.05), while ALT levels were significantly higher in the patients with hepatitis C (p,0.01 each). In patients with HCV-associated liver cirrhosis distribution of Child-Pugh classes was equivalent between patients with and without HCC. In contrast, Child Pugh classes B and C prevailed in the control group with alcohol-induced cirrhosis without HCC (p,0.001).

PNPLA3 (rs738409) gene polymorphism and liver cancer
The distributions of PNPLA3 p.I148M alleles were consistent with the Hardy Weinberg equilibrium in all study groups (table 2). Of note, the distribution of PNPLA3 I148M alleles was similar in healthy controls and patients with HCV-associated cirrhosis and HCV-related HCC giving rise to 22.9%, 25.3% (p = 0.545) and 30.2% (p = 0.071) frequencies of the 148M allele, respectively. In contrast, the prevalence of patients carrying a PNPLA3 148M variant was significantly increased in the two groups with alcoholic liver disease and reached allele frequencies of 53.7% (p,0.001 versus healthy controls) and 36.2% (p = 0.033 versus healthy controls) in patients with and without hepatocellular carcinoma, respectively (figure 1).
The distributions of PNPLA3 p.I148M genotypes are summarized in table 2: Patients with alcohol-related HCC comprised significantly more 148I/M heterozygous and 148M/M homozy-gous individuals than the healthy controls and the two groups of patients with alcoholic cirrhosis without liver cancer. The frequency of the PNPLA3 148M allele was increased both in alcohol-induced cirrhosis (OR = 1.92; 95%-CI: 1.28-2.86; p,0.002) and alcohol-associated liver cancer (OR = 4.37; 95%-CI: 2.97-6.43; p,0.001) as compared to healthy controls. In patients with alcoholic cirrhosis the risk of HCC was particularly linked to the homozygous 148M/M genotype (OR = 16.84; 95%-CI: 6.68-42.43; p,0.001). When we compared the risk of the PNPLA3 I148M polymorphism between alcoholic patients with and without HCC, differences in the allele frequency (OR = 2.28; 95%-CI: 1.47-3.55; p,0.001) and the frequency of the homozygous 148M/M genotype (OR = 4.33; 95%-CI: 1.68-11.16; p,0.002) still indicated a significant effect of the 148M allele on progression of alcoholic liver cirrhosis towards liver cancer. We could not find any significant associations between the PNPLA3 I148M polymorphism and serum levels of liver enzymes or markers of the metabolic syndrome (obesity, diabetes mellitus or serum lipids) (data not shown).
To exclude the possibility that the observed differences in the distribution of Child-Pugh classes might have affected our analysis, we stratified patients into the two classes A/B versus C. Then, we matched our alcoholic patients for sex, age and the dichotomized Child -Pugh class. After matching 42 pairs remained available for a supplementary analysis: In these matched pairs of alcoholic patients with and without hepatocellular carcinoma the PNPLA3 148M allele frequencies were 53.6% vs. 31.0% (OR = 2.57; 95%-CI: 1.37-4.84; p = 0.003), respectively. Likewise, frequencies of the homozygous 148M/M genotype were 31.0% vs 9.5% in matched patients with and without HCC (OR = 6.50; 95%-CI: 1.68-25.16; p = 0.004).
Finally we checked if the PNPLA3 p.I148M polymorphism remained an independent HCC risk factor when other known risk factors of HCC such as obesity, diabetes, and HBV infection were also taken into account. Including all patients with alcohol-and HCV-induced liver disease, respectively, we calculated separate Cox regression models to identify HCC risk factors in patients with alcoholic cirrhosis (table 3) and HCV-associated cirrhosis (table 4). Homozygous PNPLA3 148M/M genotype (OR 2.83; 95%-CI: 1.24-6.42; p = 0.013) was confirmed as a risk factor of HCC in alcoholic cirrhosis. Further, HCV genotype 1 (OR 4.16; 95%-CI: 1.50-11.52; p = 0.006) was confirmed as a risk factor of HCC in HCV-associated cirrhosis.

Discussion
Chronic hepatitis C and alcohol consumption are the leading causes for HCC in European populations. Here, we performed a cross-sectional analysis to study the potential role of the PNPLA3 p.I148M variant in liver cancer associated with alcoholic liver disease as compared to patients with HCV-induced HCC. We found a steadily increasing prevalence of the 148M allele in patients with alcoholic cirrhosis (36.2%) and alcohol-associated HCC (53.7%) as compared to healthy controls (22.9%), whereas differences in the prevalence of this genetic variant between healthy controls and HCV-associated cirrhosis (25.3%) and HCVrelated HCC (30.2%) were minor and not statistically significant. These findings are in line with previous observations in alcoholic liver disease and strengthens the concept that the PNPLA3 148M variant is associated with more severe liver damage and cirrhosis in alcoholic liver disease [17,18]. Here, we add the novel observation that this genetic variant is also a strong genetic risk factor for alcohol-associated HCC, which contributes to an approximately 4-fold increased risk in homozygous carriers of the 148M allele even in the presence of established cirrhosis. This idea is further emphasized by our multivariate regression model, which confirmed PNPLA3 148M homozygosity as independent risk factors for HCC among our patients with alcoholic liver disease. On the other hand, the PNPLA3 148M variant had only minor effects on the HCC risk in cirrhosis associated with chronic hepatitis C suggesting that this genetic variant is not a tumour gene per se but only acts in combination with substantial alcohol exposure and hepatic lipid accumulation.
Two recent studies from Italy reported strong effects of the PNPLA3 148M variant also on the risk of HCC in chronic hepatitis C [19,20], while a third Italian study suggested that this allele was predominantly associated with metabolic cirrhosis [21]. In line with this latter report, we found only little effect of the  (table 4), which failed to reach statistical significance with our sample size. Power calculation indicated the need of .670 patients per study group in order to detect statistical significance (p,0.05) for this putative association. On the other hand the observed discrepancies between the various studies may reflect inadvertent dietary differences between the study populations or probably the fact, that patients with combined HCV infection and alcohol intake (.300 g/week) had not sufficiently been taken into account. This explanation would be in line with the concept that hepatic lipid accumulation in chronic hepatitis C genotype 1 is linked to the presence of other additional metabolic factors such as diabetes and obesity [22]. Importantly, the effect attributable to the PNPLA3 p.I138M polymorphism in nutritional cirrhosis was not correlated to age, gender, body mass index or diabetes in line with previous reports [8].
Finally, there was a marked imbalance in liver disease severity between patients with and without HCC among alcohol-induced cirrhosis. This imbalance possibly reflects the fact that patients with early alcohol-induced cirrhosis (Child Pugh class A) only infrequently present at hospital services, if HCC is absent. Thus, patients in the alcohol-related control group without HCC had significantly more advanced cirrhosis than patients with alcoholinduced cirrhosis and HCC. To check if this finding had biased our analysis, we also analyzed the subgroup of patients, who could be matched for Child-Pugh classes in addition to sex and age. In line with our hypothesis this supplementary analysis confirmed a significant association between the PNPLA3 148M allele -and in particular the PNPLA3 148M/M genotype -and HCC (OR = 6.50; 95%-CI: 1.68-25.16; p = 0.004).
The physiological role of the PNPLA3 protein is still unclear. The PNPLA3 gene codes for an enzyme that exhibits both triglyceride hydrolase and transacylation activity in vitro [23]. The methionine substitution at position 148 disrupts triglyceride hydrolase activity [24], but PNPLA3 ablation in two different mouse strains had not resulted in hepatic lipid accumulation and liver damage under a variety of different diets [25]. On the other hand over-expression of PNPLA3 148M in mice increased hepatic lipid content favouring a gain-of-function mutation [24]. In addition, a variety of other intracellular interactions are also  discussed as potential alternative pathogenic mechanisms [26]. In the context of non-alcoholic steatohepatitis the PNPLA3 risk allele promotes hepatic lipid accumulation and severity of fibrosis [6,8,10,27,28]. Likewise, regular alcohol consumption causes liver damage via steatohepatitis. There is also evidence that liver biopsies from carriers of a PNPLA3 148M allele, in particular 148M homozygous subjects, exhibit greater hepatic inflammation and liver damage than patients with the wild type gene at the same extent of lipid deposition [27]. Thus, it may be increased hepatic inflammatory activity and fibrosis in response to lipid accumulation, which is associated with the PNPLA3 148M variant that enhances the risk for HCC in alcoholic cirrhosis. Whatever the underlying mechanisms, the link between the PNPLA3 polymorphism and HCC in alcoholic liver disease appears strong enough to bear practical consequences: the presence of a PNPLA3 148M allele, and especially PNPLA3 148M homozygosity, may provide an easy tool to identify individuals at particular risk for HCC at all stages of alcoholic liver damage. Thus, future HCC prevention and treatment strategies should also take into account the combined effect of the PNPLA3 p.I148M polymorphism and alcohol in patients with chronic liver disease.