Genetic Variation in the PNPLA3 Gene and Hepatocellular Carcinoma in USA: Risk and Prognosis Prediction

Abstract

Nonalcoholic fatty liver disease (NAFLD) is an emerging epidemic with high prevalence in Western countries. Genome-wide association studies had reported that a variation in the patatin-like phospholipase domain containing 3 (PNPLA3) gene is associated with high susceptibility to NAFLD. However, the relationship between this variation and hepatocellular carcinoma (HCC) has not been well established. We investigated the impact of PNPLA3 genetic variation (rs738409: C>G) on HCC risk and prognosis in the United States by conducting a case–control study that included 257 newly diagnosed and pathologically confirmed Caucasian patients with HCC (cases) and 494 healthy controls. Multivariate logistics and Cox regression models were used to control for the confounding effects of HCC risk and prognostic factors. We observed higher risk of HCC for subjects with a homozygous GG genotype than for those with CC or CG genotypes, the adjusted odds ratio (OR) was 3.21 (95% confidence interval [CI], 1.68–6.41). We observed risk modification among individuals with diabetes mellitus (OR = 19.11; 95% CI, 5.13–71.20). The PNPLA3 GG genotype was significantly associated with underlying cirrhosis in HCC patients (OR = 2.48; 95% CI, 1.05–5.87). Moreover, GG allele represents an independent risk factor for death. The adjusted hazard ratio of the GG genotype was 2.11 (95% CI, 1.26–3.52) compared with CC and CG genotypes. PNPLA3 genetic variation (rs738409: C>G) may determine individual susceptibility to HCC development and poor prognosis. Further experimental investigations are necessary for thorough assessment of the hepatocarcinogenic role of PNPLA3.

INTRODUCTION

Despite the multi-etiologic nature of hepatocellular carcinoma (HCC), it is difficult to pinpoint the specific underlying factors that have contributed to the increased incidence of HCC in the United States, especially in young people [1]. Concurrent with the increase in HCC incidence, the prevalence of fatty liver or hepatic steatosis has increased and is currently considered a growing global health problem. An estimated 20–40% of the adult population in Western countries have fatty liver [2]. Interestingly, each of the major risk factors of HCC such as hepatitis C virus (HCV), hepatitis B virus (HBV), alcohol consumption, prior history of obesity, and diabetes mellitus has been significantly associated with a wide spectrum of hepatobiliary diseases [3–5].

Nonalcoholic fatty liver disease (NAFLD) is a general term for multiple liver disorders. The initial condition is simple hepatocyte fatty infiltration (steatosis) defined by deposition of lipid droplets (triglycerides) in liver cell [6]. Steatosis is often reversible condition but it can progress to a second disorder named nonalcoholic steatohepatitis (NASH) with underlying hepatic inflammation and or collagen deposition (fibrosis). Approximately 10–29% of patients with NASH may progress to nonreversible end stage liver disorder (cirrhosis) within 10 yrs [7] and will be at high risk for HCC development [8]. The Serum level of alanine aminotransferase (ALT) is elevated in patients with liver injury and NAFLD than in individuals with normal liver [9].

Various genetic and environmental mediators are involved in the progression of liver diseases [10] and may explain the observed variations in individuals’ susceptibility for HCC development among high risk populations.

Recently, a genome-wide association study reported that a variation in the patatin-like phospholipase domain containing 3 (PNPLA3) gene is associated with high susceptibility to NAFLD [11]. The I148M allele (rs738409: C>G) of the PNPLA3 gene represents a cytosine to guanine substitution, resulting in an isoleucine to methionine switch at codon 148. Individuals who were homozygous for the G allele had a higher hepatic triglyceride level [11] and an elevated serum level of ALT [12]. Moreover, PNPLA3 (rs738409: C>G) may influence fibrosis severity in patients with fatty liver [13–16]. Thus, we hypothesized that PNPLA3 genetic variation (rs738409: C>G) may determine individual susceptibility to HCC and may correlate with underlying cirrhosis and poor prognosis of American men and women with HCC.

SUBJECTS AND METHODS

Population and Study Design

Case patients were prospectively recruited from the population of patients with newly diagnosed HCC who were evaluated and treated at the M. D. Anderson Cancer Center. The inclusion criteria were as follows: pathologically confirmed diagnosis of HCC and U.S. residency. The exclusion criteria were the presence of other types of primary liver cancer and concurrent or past history of cancer at another organ site.

From January 2000 through December 2009, 724 patients with suspected HCC were identified, 553 of whom were eligible for this study. We enrolled 446 eligible patients with HCC (participation rate = 80.7%); 107 were not recruited due to patients refusal, patients sickness, or inadequate time to complete the interview; 75 were Caucasians (70%). Among the total of 446 eligible HCC patients 313 cases were Caucasians (70.2%); 257 (82.1%) had sufficient and high-quality DNA samples for genotyping and were included in this investigation.

The healthy control subjects were genetically unrelated family members (spouses or in-laws) of patients at MD Anderson who had cancers other than liver, gastrointestinal, lung, or head and neck cancer. The eligibility criteria for controls were the same as for cases, except for having a cancer diagnosis. Control subjects were recruited from the Institution’s Central Diagnostic Radiology Clinics, where all cancer patients and their companions are sent for the initial cancer diagnosis or post-treatment follow-up examination. Research data coordinator of the study followed a detailed protocol to identify candidates of control subjects according to the eligibility criteria of control selection.

A short self-administered structured questionnaire was used to determine the perception of the healthy individuals to participate in medical research and to identify potential controls from the healthy individuals on the basis of the eligibility criteria. The questionnaire requested information about age, sex, race, educational level, place of residency, personal history of cancer, relationship to the cancer patient, and the patient’s type of cancer. We estimated that 83.6% of the eligible candidates for control selection agreed to participate in the study and did not differ significantly from eligible candidates who refused to participate.

Using the same short self-administered structured questionnaire, we sought to confirm the control subjects’ reasons for coming to the hospital with cancer patients and whether these reasons could have been related to the risk factors of HCC. We found that the underlying causes for the controls’ companionship were care and altruism. Moreover, all spouses of patients with other cancers who served as control subjects reported that they would have chosen to be referred to M.D. Anderson if they had been diagnosed with cancer during the same time period because they tended to share the same family physician, had the same health insurance coverage, and lived in the same geographic location. All of the above-mentioned results indicated that the patients and controls had the same catchments, which further supported the idea that the control subjects were representative of the M.D. Anderson population from which HCC patients were selected. A total of 494 Caucasian healthy control subjects who were frequency-matched to the HCC cases by gender and age (±5 yrs) were recruited and included in this investigation.

HCC cases and controls were recruited during the same time frame and were personally interviewed for 25–30 min. To control for the confounding impact of the major risk factors of HCC, the interviewers used a structured and validated questionnaire to collect information of demographic variables and HCC risk factors including cigarette smoking, alcohol consumption, chronic medical conditions, and family history of cancers.

We recently revised the study questionnaire to collect information about BMI at different ages (≤20, 20, 30, 40, and 50 yrs) prior to HCC development. Total of 155 Caucasian HCC cases and 307 Caucasian control subjects had been interviewed for obesity history using the revised questionnaire. Detailed methods of cases and controls ascertainment were previously described in Refs. [17–21].

Several clinical variables were considered in determining survival outcome and were retrieved from HCC patients’ medical records: alpha-fetoprotein levels (ng/ml) at the time of diagnosis; tumor size (≤50% vs. >50%) of the total liver size; tumor nodularity (uninodular vs. multinodular); radiological and pathological evidence of cirrhosis, fibrosis, and steatosis; presence or absence of portal thrombosis; presence or absence of macrovascular or microvascular invasion; extrahepatic metastasis; lymph node involvement; tumor differentiation (well differentiated, moderately differentiated, poorly differentiated); Child-Pugh classification (A–C); Barcelona Clinic Liver Cancer (BCLC) stage (A–D); Eastern Cooperative Oncology Group (ECOG) performance status; and the type of first treatment initiated at time of diagnosis [no therapy, chemotherapy with or without others modalities, surgical resection/transplantation with or without other modalities, and transcatheter arterial chemoembolization (TAC)].

Other modalities included percutaneous ethanol injection, radiofrequency thermal ablation, or immune therapy.

The current investigation is part of an ongoing hospital-based case–control study, which was approved by the Institutional Review Board (IRB) at The University of Texas MD Anderson Cancer Center. Written informed consent for participation was obtained from each study participant. Under IRB approval we retrieved the medical records of the eligible and non-recruited HCC patients. There were no significant differences in demographic, epidemiologic, and clinical factors, among the eligible and the included Caucasian patients with DNA samples (N = 257) as compared to eligible Caucasian patients without DNA samples (n = 56) or as compared to Caucasian patients who were eligible but not enrolled in the study (n = 75).

Laboratory Methods

Blood samples were collected in heparinized vacutainers (BD Biosciences, Franklin Lakes, NJ). DNA was extracted from mononuclear cells by using a FlexiGene DNA kit (Cat# 51206, Qiagen, Valencia, CA) and the Maxwell 16 automated system with Cell DNA Purification Kit (Cat # AS1020, Promega, Madison, WI). DNA concentration was determined by Nano-Drop ND1000 and diluted DNA (10 ng/μl) stored at 4°C for immediate use. Separated plasma samples were tested for HBV and HCV markers. HCV antibodies (anti-HCV), hepatitis B surface antigen (HBsAg), and antibodies to hepatitis B core antigen (anti-HBc) were detected by use of a third-generation enzyme-linked immunosorbent assay (ELISA; Abbott Laboratories, North Chicago, IL). Positive results prompted repeated confirmatory ELISA.

Sequenom polymerase chain reaction (PCR) and single nucleotide polymorphism (SNP) primers were designed with use of Sequenom’s Assay Design software, limiting the number of targets to 20 per multiplex reaction. Multiplex PCR reactions were amplified by using Qiagen Hotstar Taq polymerase; the PCR products were cleaned by using EXO-SAP (Sequenom); and the primer extension reactions were run by using iPLEX Gold (Sequenom). The samples were spotted onto Spectrochip II matrix chips and then run in the MassARRAY. All samples were run in duplicate. Sequenom Typer 3 software and in-house software were used to interpret the mass spectra. Moreover, Genotyping confirmation in all cases and controls was performed by using TaqMan allelic discrimination assays (Applied Biosystems, Foster City, CA).

Sample Size Consideration and Statistical Methods

We estimated the sample size based on probability of α = 0.05 and β = 0.1 and assuming that the prevalence of the risk allele (G) in the control group was 23% [11], and estimated odds ratio (OR) was 2. Approximately 2 to 1 control–case ratio was chosen to increase the power of the study. According to the above parameters the estimated sample size had enough power to assess the risk of the PNPLA3 genetic variation (rs738409: C>G) on HCC development.

Univariate analyses were conducted with χ² or Fisher exact tests for categorical variables and the Kruskal–Wallis test for continuous variables. All clinical, genetic, and epidemiological data were merged and analyzed with use of STATA software (STATA Corp, College Station, TX). For controls, deviations of the genotype frequencies from those expected under Hardy–Weinberg equilibrium were assessed by χ² tests (1 df), and genotype frequencies in cases and controls were compared by using χ² tests (2 df).

We performed multivariable logistic regression to estimate the adjusted odds ratio (AOR) and 95% confidence intervals (CI) using the maximum likelihood estimate. All AORs were adjusted for age, gender, educational level, cigarette smoking, alcohol consumption, diabetes mellitus, family history of cancer, and HBV/HCV infection. The final model was determined on the basis of backward stepwise selection procedure. In addition, we used multivariable logistic regression models to investigate possible interactions on an additive scale of PNPLA3 genotypes with HCV/HBV infection, diabetes mellitus, and alcohol consumption.

Overall survival (OS) was defined as the time between diagnosis and death (as a result of all causes) or end of follow-up (censored observations). Median survival was estimated by using the Kaplan–Meier product-limit method, and significant differences between the survival times were determined by using the log-rank test. To identify independent prognostic factors for OS, hazard ratios (HR) and 95% CIs were calculated by using Cox proportional hazard models with a backward stepwise selection procedure, considering the clinical covariates of HCC without multicollinearity between the clinical features of HCC.

RESULTS

Participants’ Characteristics

Table 1 shows the demographic characteristics of the 257 Caucasian cases with HCC and the 494 Caucasian healthy controls. Gender distribution was comparable between HCC cases and controls. Overall, cases were slightly older than controls but frequency matched by ±5 yrs; the mean age ± standard error was 62.39 ± 0.78 yrs for HCC cases and 60.35 ± 0.51 yrs for controls. Higher education (≥college degree) was more frequent among controls (46.4%) than among HCC cases (33.1%; P = 0.001). Cases and controls had a similar distribution of geographical region (U.S. state of residency) with the majority of study participants (70.7%) from Texas and adjacent states (i.e., Louisiana, Arkansas, New Mexico, and Oklahoma). The observed HCC risk factors including chronic infection with HCV/HBV, alcohol consumption, cigarette smoking, prior history of diabetes mellitus, and first-degree family history of liver cancer were consistent with those of our previous reports addressing the potential association between environmental risk factors and HCC [17–21].

Table 1.

Demographic and Epidemiological Characteristics of the Study Population

Demographic and risk factors	HCC patients		Controls		P-value
	N = 257	%	N = 494	%
Sex					0.5
Male	176	68.5	336	68
Female	81	31.5	158	32
Age (yrs old)					0.039
≤40	13	5.1	20	4
41–50	28	10.9	84	17
51–59	67	26.1	149	30.2
≥60	149	58	241	48.8
Mean age (±SE)	62.39 ±0.78		60.35 ± 0.51		0.02
Education level					<0.001
<College degree	172	66.9	265	53.6
≥College degree	85	33.1	229	46.4
State of residency					0.1
TX, LA, AK, NM, OKa	174	67.7	357	72.3
Other states	83	32.3	137	27.7
Hepatitis virus infection					<0.001
HCV−, HBsAg−, Anti-Hbc−	157	61.1	474	96
HCV+, HBsAg−, Anti-Hbc+	21	8.2	2	0.4
HCV+, HBsAg−, Anti-Hbc−	54	21	2	0.4
HCV−, HBsAg+, Anti-Hbc+	9	3.5	2	0.4
HCV−, HBsAg−, Anti-Hbc+	13	5.1	14	2.8
HCV+, HBsAg+, Anti-Hbc+	3	1.2	0	0
Alcohol consumption					<0.001
No	79	30.7	194	39.3
≤60 ml ethanol/d	119	46.3	267	54
>60 ml ethanol/d	59	23	33	6.7
Cigarette smoking					<0.001
No	88	34.2	244	49.4
≤20 pack yrs	49	19.1	109	22.1
>20 pack yrs	120	46.7	141	28.5
Prior history of diabetes					<.001
No	180	70	444	89.9
≤1 yr	9	3.5	9	1.8
>1 yr	68	26.5	41	8.3
First degree history of liver cancer					0.010
No	249	96.9	491	99.4
Yes	8	3.1	3	0.6

^aStates of Texas, Louisiana, Arkansas, New Mexico, and Oklahoma.

PNPLA3 Genotypes and HCC

The distribution of PNPLA3 genotypes (CC, GC, and GG) was significantly different between HCC cases (50.6%, 36.6%, and 12.8%) and controls (55.5%, 39.9%, and 4.7%), respectively, P < 0.001 (Table 2). The control genotype distribution was consistent with the Hardy–Weinberg equilibrium in all (P = 0.1), in men (P = 0.2), and in women (P = 0.3). Although we found deviations from the expected proportions of homozygote and heterozygote classes in HCC case population (P = 0.02), this supports the association between HCC and G allele of PNPLA3 gene.

Table 2.

PNPLA3 Genotypes and the Independent Risk of HCC Development

PNPLA3 genotypes (rs738409)	Total populationa cases/controls		Menb cases/controls		Womenb cases/controls
	257/494	AOR (95% CI)	176/336	AOR (95% CI)	81/158	AOR (95% CI)
CC	130/274	1 (reference)	85/186	1 (reference)	45/88	1 (reference)
GC	94/197	1.05 (0.71–1.55)	67/134	0.76 (.45–1.26)	27/63	1.30 (0.68–2.48)
GG	33/23	3.28 (1.68–6.42)	24/16	2.87 (1.19–6.91)	9/7	4.44 (1.38–14.33)
CC	130/274	1 (reference)	85/186	1 (reference)	45/88	1 (reference)
GG GC	127/220	1.28 (0.89–1.84)	91/150	1.58 (0.98–2.57)	36/70	1 (0.55–1.81)
CC GC	224/471	1 (reference)	152/320	1 (reference)	72/151	1 (reference)
GG	33/23	3.21 (1.68–6.41)	24/16	3.34 (1.44–7.72)	9/7	3.77 (1.26–11.33)

^aOdds ratios is adjusted for confounding of age, sex, educational level, HCV, HBV, alcohol drinking, cigarette smoking, diabetes mellitus, and family history of cancer.

^bOdds ratios is adjusted for confounding of age, educational level, HCV, HBV, alcohol drinking, cigarette smoking, diabetes mellitus, and family history of cancer.

Multivariable logistic regression analyses indicated that having the homozygous GG variant of PNPLA3 afforded a threefold increased risk for HCC compared with participants with the homozygous or heterozygous C allele (CC/CG), irrespective of the potential confounding effects of HCC risk factors. A significant risk association was observed for both men and women (Table 2).

An interaction of the GG genotype and diabetes (risk modification) was observed on the additive scale. The AOR (95% CI) was 19.11 (5.13–71.20) for diabetic patients with GG genotype as compared to non-diabetic individuals who are homozygous or heterozygous for the C allele, P for the interaction = .04 (Table 3) No risk modification was observed by hepatitis virus infection or by alcohol consumption.

Table 3.

Risk Modification of PNPLA3 Genotypes by HCC Established Risk Factors

Risk variables		Cases/controls	AOR (95% CI)	P-value
HCV/HBV	Genotypes	257/494	Model (1)a
Negative	CC/GC	133/452	1 (reference)
Positive	CC/GC	91/19	17.64 (9.71–29.82)	<0.001
Negative	GG	24/22	3.41 (1.75–6.63)	<0.001
Positive	GG	9/1	23.03 (2.71–195.82)	0.004
Risk variables		Cases/controls	AOR (95% CI)
Alcohol	Genotypes	257/494	Model (2)b	P-value
No	CC/GC	67/185	1 (reference)
Yes	CC/GC	157/286	1.60 (1.04–2.46)	0.03
No	GG	12/9	5.18 (1.91–14.05)	0.001
Yes	GG	21/14	3.52 (1.49–8.32)	0.004
Risk variables		Cases/controls	AOR (95% CI)
Diabetes	Genotypes	257/494	Model (3)c	P-value
No	CC/GC	162/424	1 (reference)
Yes	CC/GC	62/47	3.93 (2.40–6.46)	<0.001
No	GG	18/20	2.65 (1.24–5.69)	0.01
Yes	GG	15/3	19.11 (5.13–71.20)	<0.001

^aAdjusted odds ratio for age, sex, education, diabetes, alcohol drinking, cigarette smoking, and family history of cancer.

^bAdjusted odds ratio for age, sex, education, diabetes, HCV, HBV, cigarette smoking, and family history of cancer.

^cAdjusted odds ratio for age, sex, education, HCV, HBV, alcohol drinking, cigarette smoking, and family history of cancer.

Association Between PNPLA3 GG Genotype and Underlying Cirrhosis in HCC Patients

At the time of HCC diagnosis, a total of 156 HCC cases (60.7%) had pathological (n = 98) and/or radiological (n = 58) evidence of cirrhosis (n = 145), as well as fibrosis with or without steatosis (n = 11). High prevalence of underlying cirrhosis was observed in HCC cases with the GG PNPLA3 genetic variation (72.7%) compared with that in HCC cases with the CC (53.1%) or CG genotype (67%; P = 0.03). The final logistic stepwise regression model for underlying cirrhosis indicated that the GG genotype of PNPLA3 was an independent and significant predictor for cirrhosis development before HCC diagnosis. The estimated AOR (95% CI) was 2.48 (1.05–5.87; Table 4. Other significant predictors including: HCV, HBV, and diabetes mellitus. The estimated AORs were 5.77 (2.93–11.36); 5.26 (1.07–25.77); and 3.87 (1.76–8.55) for HBV, HCV, and diabetes, respectively (Table 4).

Table 4.

Predictors of Underlying Cirrhosis in Patients With HCC

Risk factors	No cirrhosis, N = 101 (%)	Cirrhosis, N = 156 (%)	AOR (95% CI)	P-value
Chronic HCV infection (anti-HCV+)
No	87 (86.1)	92 (59)	1 (reference)
Yes	14 (13.9)	64 (41)	5.77 (2.93–11.36)	0.000
Chronic HBV infection (HBsAg+)
No	99 (98)	146 (93.6)	1 (reference)
Yes	2 (2)	10 (6.4)	5.26 (1.07–25.77)	0.041
Prior history of diabetes mellitus
Non-diabetic or diabetics for ≤5 yrs	91 (90.1)	123 (78.9)	1 (reference)
>5 yrs with diabetes	10 (9.9)	33 (21.1)	3.87 (1.76–8.55)	0.001
PNPLA3 genotypinga
CC/CG	92 (91.1)	132 (84.6)	1 (reference)
GG	9 (8.9)	24 (15.4)	2.48 (1.05–5.87)a	0.039

^aAdjusted odds ratio of PNPLA3 GG genotype after controlling for age, sex, HCV, HBV, diabetes mellitus, cigarette smoking, and heavy alcohol consumption using stepwise backward logistic regression.

Association Between PNPLA3 GG Genotype and Overall Survival

The survival duration for HCC cases with the GG genotype was shorter than that of HCC cases with the CC or CG genotypes. The estimated median survival was 16.8 months (95% CI, 9.9–23.7 months) for HCC cases with the GG genotype and 25.9 months (95% CI, 21.5–30.3 months) for HCC cases with the CC/CG genotypes (P = 0.1).

In an analysis restricted to HCC patients with advanced stages, poor median survival was observed in patients with the GG genotypes [5.17 months (95% CI, 1.66–8.68)] compared with that in patients with the CC and GC genotypes [13.53 months (95% CI, 10.73–16.34); P = .005].

Cox proportional hazard models, which included predictors of prognosis, revealed that the GG allele was an independent risk factor for mortality (Table 5). Compared with homozygous CC allele or heterozygous CG allele carriers, the adjusted HR for OS for cases with the GG genotype was more than two times higher (HR = 2.04; 95% CI, 1.22–3.41; P = 0.007). In addition, male gender and poor tumor differentiation were associated with a higher risk of death after adjustment for BCLC staging, ECOC performance status, and treatment exposure. Treated patients experienced better survival than non-treated patients, with a range of 54–89% improvement in OS by different types of treatment (Table 5).

Table 5.

Prognostic Factors of Overall HCC Survival

Prognostic variables	Variable label	N	Multivariable HR (95% CI)		P-value
Sex	Female	81	1 (reference)
	Male	176		1.54 (1.04–2.28)	0.03
HCV	Negative	179	1 (reference)
	Positive	78		1.34 (0.95–1.90)	0.099
Tumor differentiationa	Well	88	1 (reference)
	Moderate	84		1.50 (1.04–2.69)	0.033
	Poor	46		1.77 (1.17–2.69)	0.007
BCLC staging system	0, A, B	83	1 (reference)
	C, D	174		2.43 (1.19–4.98)	0.004
ECOC	0, 1	217	1 (reference)
	2	27		2.56 (1.52–4.35)	<0.0001
	3	13		4.77 (2.39–11.54)	<0.0001
Treatment exposure	No treatment	31	1 (reference)
	Chemotherapy	88		0.48 (0.29–0.82)	0.007
	Chemotherapy and others	35		0.40(0.21–0.76)	0.005
	Surgical resection/liver transplant	32		0.21 (0.09–0.48)	<0.0001
	Surgery and others	41		0.26 (0.13–0.49)	<0.000
	TACc	30		0.37 (0.18–0.77)	0.008
PNPLA3 genotypes (rs7384409)b	CC/CG	224	1 (reference)
	GG	33		2.11 (1.26–3.52)b	0.004

^aMissing information of pathological tumor differentiation for 39 HCC patients.

^bHazard ratio of PNPLA3 GG genotype is adjusted for age, sex, HCV, HBV, diabetes mellitus, alcohol consumption, tumor differentiation, prior treatment, baseline ECOC performance status, baseline BCLC staging system, and presence of cirrhosis using stepwise backward Cox regression.

^cTranscatheter arterial chemoembolization.

DISCUSSION

Despite the overwhelming impact of environmental factors on HCC development, this large case–control study revealed an independent effect of GG PNPLA3 genetic variation (rs738409: C>G) on HCC development in American men and women.

The current analysis was performed in American Caucasians, in whom the prevalence of the risk allele (G) was 24.6%, similar to that presented by GWAS (23%) [11].

Two recent genomic-wide association studies reported non-synonymous sequence variation in the PNPLA3 gene in association with NAFLD and with ALT concentration [11,12]. Results from both studies have been replicated and indicated a significant risk of the GG genotype on the development of NAFLD and NASH [13–16,22–24]. Moreover, the GG genotype influenced the severity of fibrosis and enhanced the risk of cirrhosis development in patients with NAFLD. This may explain our observation of the association between GG genotype of PNPLA3 and underlying cirrhosis in HCC patients. We estimated that 60.7% of our HCC patients had underlying cirrhosis, which was supported by population [25] and clinical studies in USA [26–28].

Our results is supported by two Italian studies indicating that patients with cirrhosis complicated by HCC were more likely to be GG homozygous than were the remaining cirrhotic patients [14,24].

PNPLA3 mRNA is expressed in the liver and the adipose tissue [29], encodes 481- amino acids protein of unclear physiological and biological function. Nevertheless, PNPLA3 exhibits strong activity in lipolysis and triglyceride hydrolysis in vitro [30]. Moreover, expression of PNPLA3-I148M, but not the wild type PNPLA3, promotes triglyceride accumulation by impairing the triglyceride hydrolysis in cultured hepatocyte and in mice liver [31].

The carcinogenic role of PNPLA3 in HCC development has not been examined. However, the genetic susceptibility of PNPLA3 for HCC development may be considered as a natural extension for the previously reported association between NAFLD and PNPLA3 genetic variation (rs738409: C>G). Similarly, fatty liver disease is a common condition in patients with other HCC risk factors including HCV infection [3], diabetes mellitus, obesity [4], and alcohol drinking [5]. Both alcoholic and non-alcoholic fatty liver diseases have a similar pathological feature [32,33].

Once steatosis has developed in genetically susceptible patients, lipid peroxidation, inflammation, and generated free oxygen radicals may play a central role in NASH development, during which the initiation phase of the HCC mechanism takes place. The balance between apoptotic and anti-apoptotic factors and the disturbance in the growth factors may facilitate oval cell proliferation and the promotion phase of hepatocarcinogenesis [34].

Hyperinsulinemia and insulin resistance is another pathway involved in hepatocarcinogenesis in patients with high susceptibility for fatty liver diseases, particularly in patients with obesity and diabetes-induced NAFLD [35,36]. This may explain our observation of synergistic interaction between the GG genotype of PNPLA3 (rs738409: C>G) and diabetes mellitus on the risk of HCC development.

PNPLA3 expression is positively correlated with body mass index (BMI) and carbohydrate intake [29,37]. Because obesity is a risk factor of type 2 diabetes and NAFLD, one may argue that obesity may confound the independent effect of GG genotype of PNPLA3 (rs738409) or its interaction with diabetes mellitus. However, the statistical adjustment for the confounding effect of obesity in subset of the study population with available obesity information (155 cases and 307 controls) or the restricted analysis to those without diabetes did not meaningfully change the observed significant association between the PNPLA3 GG genotype (rs738409) and HCC risk or poor prognosis.

Since the effect of the PNPLA3 genetic variation (rs738409: C>G) on HCC risk is possibly mediated through the effect of NAFLD, fatty liver is considered an intermediate factor and should not be treated as a confounding factor [38]. In fact, adjusting for the effect of the intermediate factors may bias the results toward the null. Nevertheless, our estimated ORs were adjusted for the major etiological factors of HCC as surrogate variables of NAFLD due to the presence of fatty liver in conjunction with these etiological factors of HCC. Large follow-up studies of patients with fatty liver may be warranted to understand how PNPLA3 GG genotype (rs738409) modulates the risk of NASH to develop HCC with and without underlying cirrhosis.

The association between the PNPLA3 (rs738409) polymorphism and HCC extended to the poor survival outcome of HCC patients with the GG genotype of PNPLA3 (rs738409) compared with patients with the CC or CG genotypes. The prognostic role of the GG genotype in HCC survival was independent of the major survival predictive factors. Although there is no specific explanation for such observation, it is possible that patients with the GG genotype of PNPLA3 (rs738409) have high susceptibly for lymph node involvement, vascular invasion, poor tumor differentiation, or other clinical parameters associated with HCC prognosis. Given the complexity of the natural history of HCC, we hypothesized that manifestations of underlying cirrhosis including ascites, bleeding, and low platelet count are a huge burden to HCC clinical management, often forcing oncologists to reduce the optimal dose of aggressive chemotherapeutic agents or prohibit surgical resection. This may ultimately affects patients’ response to the therapy and limits patients’ options for effective treatment, leading to diminished survival.

Most HCC patients received several types of systemic therapy, correlation between PNPLA3 genotypes and response to each type of treatment will be very difficult to assess independently from others due to overlapping between used regimens.

It has been reported that the highest frequency of the G allele was observed among Hispanics [11]. In this case–control study, only a small number of Hispanic patients with HCC were recruited, which precluded us from determining the impact of PNPLA3 GG genotype (rs738409) on HCC development and HCC prognosis among Hispanics. Further studies are warranted to identify common genetic predisposition factors that may contribute to susceptibility to HCC development in minorities, in whom the incidence of HCC and the prevalence of metabolic syndrome are higher than in Caucasians. Should our findings confirmed, it may help to develop novel strategies in identifying high-risk individuals for HCC especially among diabetic patients and in future personalized cancer treatment.

Abbreviations

HCC

hepatocellular carcinoma

HCV

hepatitis C virus

HBV

hepatitis B virus

NAFLD

nonalcoholic fatty liver disease

NASH

nonalcoholic steatohepatitis

OS

overall survival

AOR

adjusted odds ratio