Interaction between PNPLA3 I148M Variant and Age at Infection in Determining Fibrosis Progression in Chronic Hepatitis C

Background and Aims The PNPLA3 I148M sequence variant favors hepatic lipid accumulation and confers susceptibility to hepatic fibrosis and hepatocellular carcinoma. The aim of this study was to estimate the effect size of homozygosity for the PNPLA3 I148M variant (148M/M) on the fibrosis progression rate (FPR) and the interaction with age at infection in chronic hepatitis C (CHC). Methods FPR was estimated in a prospective cohort of 247 CHC patients without alcohol intake and diabetes, with careful estimation of age at infection and determination of fibrosis stage by Ishak score. Results Older age at infection was the strongest determinant of FPR (p<0.0001). PNPLA3 148M/M was associated with faster FPR in individuals infected at older age (above the median, 21 years; −0.64±0.2, n = 8 vs. −0.95±0.3, n = 166 log10 FPR respectively; p = 0.001; confirmed for lower age thresholds, p<0.05), but not in those infected at younger age (p = ns). The negative impact of PNPLA3 148M/M on fibrosis progression was more marked in subjects at risk of altered hepatic lipid metabolism (those with grade 2–3 steatosis, genotype 3, and overweight; p<0.05). At multivariate analysis, PNPLA3 148M/M was associated with FPR (incremental effect 0.08±0.03 log10 fibrosis unit per year; p = 0.022), independently of several confounders, and there was a significant interaction between 148M/M and older age at infection (p = 0.025). The association between 148M/M and FPR remained significant even after adjustment for steatosis severity (p = 0.032). Conclusions We observed an interaction between homozygosity for the PNPLA3 I148M variant and age at infection in determining fibrosis progression in CHC patients.


Introduction
Hepatitis C virus (HCV) infection affects roughly 130-170 million people worldwide. Individuals with chronic HCV-related hepatitis (CHC) are at high risk of liver fibrosis, an in about 20% of cases CHC progresses to cirrhosis. Therefore, HCV is a leading cause of end-stage liver disease and hepatocellular carcinoma [1]. Older age at infection, genotype 3, coexistent liver diseases, male gender, alcohol abuse, obesity and diabetes, and steatosis are risk factors for disease evolution [2][3][4][5].
However, even after taking into account all known viral, host related, and environmental factors, the rate of progression to advanced fibrosis is highly variable among individuals, thereby suggesting a role for inherited variants. Indeed, single-nucleotide polymorphisms at different genetic loci have been associated with fibrosis progression in candidate gene studies [6,7]. The rs738409 C.G polymorphism of Patatin-like phospholipase domain-containing 3 (PNPLA3), encoding for the I148M protein sequence variant, is the major common genetic determinant of hepatic fat content. The I148M variant influences the metabolism of triglycerides, lipid droplets remodeling, and lipoprotein secretion in hepatocytes [8][9][10][11][12][13]. PNPLA3 I148M is the major risk factor for both alcoholic and nonalcoholic steatohepatitis, showing an age-modulated effect on liver damage susceptibility and fibrogenesis [14][15][16][17][18][19]. Recently, it has been reported that homozygosity for the 148M risk allele (henceforth PNPLA3 148M/M) also influences the risk of advanced fibrosis and hepatocellular carcinoma in CHC [20][21][22][23][24][25]. However, scant data are derived from prospective studies [23,25], and the interaction with age at infection has not been evaluated so far.
Therefore, the aim of this study was to estimate the effect size of PNPLA3 148M/M on fibrosis progression rate (FPR) and the interaction with age at infection in a unique validated cohort of 247 CHC subjects without significant alcohol intake and diabetes, characterized by a careful estimation of age at infection and with histological determination of liver damage [3].

Patients
We analyzed a previous published cohort of 247 CHC patients followed-up at the Migliavacca Liver Disease Center at the Fondazione IRCCS Ca' Granda Policlinico of Milan [3]. Briefly, inclusion criteria comprised a precise estimate of the date of infection based on the date of the first reported risk factor, a diagnostic liver biopsy before any antiviral therapy performed at least 4 years after the date of infection, and no history of past or current alcohol abuse. Patients with viral co-infections or other liver diseases were excluded, as well as patients with either type 1 or type 2 diabetes. Disease duration was calculated considering the time elapsed from the date of infection to the time of liver biopsy. The study protocol conformed to the Declaration of Helsinki and was approved by the Institutional Review Board of the Fondazione IRCCS Ca' Granda. Each patient signed a written informed consent.
The baseline features of the study cohort stratified by PNPLA3 148M/M status are presented in table 1. The study database has been uploaded as supplementary material (Data S1).

Histological evaluation
Liver histology was evaluated by a single expert pathologist. Mean biopsy length was 26.3 mm. The severity of hepatic inflammation was evaluated by the Ishak score [26], entailing a maximum of 18 points for grading (G) and 6 points for liver staging (S). All specimens were also evaluated for steatosis severity (grade 0: absent or ,5%; grade 1: 5-33%; grade 2: 34-66%; grade 3: .66% of hepatocytes affected) [27]. Steatosis assessment is missing for 14 patients due to the inability to recollect the original histological slides re-evaluated for fibrosis stage. To analyze the effect of steatosis severity in the whole study cohort, additional analyses were conducted estimating the severity of steatosis in these 14 subjects by ultrasonography [28].

Genetic analysis
Genomic DNA was obtained from peripheral whole blood as previously described [3]. The PNPLA3 rs738409 C.G single nucleotide polymorphism, encoding for the I148M protein sequence variant, was genotyped by a 5 9nuclease Taqman assay (LifeTechnologies, Carlsbad, CA, USA) at the Metabolic Liver Diseases lab at the Fondazione Ca' Granda IRCCS. Success rate was 100% and perfect concordance between duplicates and internal controls was observed.

Statistical analysis
For descriptive statistics, variables were shown as mean 6 SD and frequencies, and compared by Student's t-test or chi-square test according to data distribution. P values were considered significant when ,0.05 (two-tailed). FPR was calculated by taking the ratio between the staging value of Ishak score and the disease duration (years), and it was treated as a continuous variable. Mean disease duration was 25610 years. Log 10 transformation of FPR was employed to obtain linearity. FPR estimation assumes that no significant liver fibrosis was present at the time of infection, and that liver fibrosis progression is constant during time [2]. A first generalized linear model was formulated and fitted to the data, specifying the presence of PNPLA3 148M/M at risk genotype [21][22][23], and the covariates as explanatory variables. We considered as covariated the independent predictors of FPR in a previously reported analysis of the same cohort [3]: age at infection (encoded as . or # the median value, 21 years), gender, and HCV genotype. The product term age at infection x PNPLA3 148M/M was fitted into the model to test the interaction between PNPLA3 genotype and age at infection on FPR. To account for other potential confounders, a second model was tested, with further adjustement for the route of HCV transmission (intra-venous drug use: IVDU or other), body-mass (BMI , or $25 Kg/m 2 ), and age at liver biopsy. Finally, to check whether the relationship between PNPLA3 genotype and FPR was independent of steatosis, a final model further adjusted for the histological severity of steatosis (grade 2-3 vs. 0-1) was evaluated. Models were checked through the regression diagnostic plots to verify normality, linearity of the data, and constant variance. The effect of the explanatory variables was considered significant if p, 0.05 (two-tailed). Analyses were performed by the SPSS 21.0 (IBM, Burbank, NY, USA) statistical analysis software.

Study cohort
Detailed characterization of the study cohort has previously been reported [3]. Briefly, 52% of patients were males, the majority infected by genotype 1 (52%), during blood transfusion (75%). Median age at infection was 21 years and disease duration 25 years. At the time of histological evaluation, 16% of patients had cirrhosis, whereas severe steatosis (grade 2-3) was observed in 20% of cases. The frequency distribution of the PNPLA3 I148M variant did not violate Hardy-Weinberg equilibrium (p.0.8), and was in line with that expected unselected subjects: 120 patients were PNPLA3 148I/I (48%), 106 148I/M (43%), and 21 had the unfavorable 148M/M genotype (9%).
Clinical features of patients stratified by PNPLA3 148M/M status are presented in table 1. Demographic and anthropometric features, viral genotype distribution, age and route of infection, and duration of follow-up did not differ according to PNPLA3 148M/M status (p = ns). There was no statistically significant difference in the prevalence of grade 2-3 steatosis according to PNPLA3 148M/M status (p = ns), although a nonsignificant trend was observed in the whole cohort including subjects with ultrasonographic estimation of steatosis severity (7/21, 33% vs. 44/226, 19% in 148M/M vs. individuals carrying other PNPLA3 genotypes, p = 0.15).

Effect of age at infection and PNPLA3 148M/M on fibrosis progression
Older age at infection was strongly associated with FPR (log 10 FPR 21.160.3 vs. 20.960.3 in subjects younger or older than 21 years, median value, respectively; p,0.0001). The FPR stratified by age at infection (. or # median value, 21 years) and PNPLA3 148M/M and is presented in figure 1 and in table 2. PNPLA3 148M/M was associated with faster FPR in older (p = 0.001), but not in younger patients (p = ns). The positive association between PNPLA3 148M/M and FPR in older patients was confirmed in sensitivity analyses using different age cut-offs (table 2)

Independent predictors of fibrosis progression
At the multivariate generalized linear model considering known determinants of FPR (age at infection, gender, and viral genotype: shown in model 1, table 4) PNPLA3 148M/M was an independent predictor of FPR (p = 0.022). The model estimated an incremental progression of 0.0860.03 log 10 unit of fibrosis per year conferred by 148M/M status. Furthermore, there was a significant interaction between PNPLA3 148M/M and older age at infection in determining FPR (p = 0.025; estimate +0.0760.03). In exploratory analyses, there was no significant interaction between PNPLA3 148M/M and other determinants of FPR.
To test the robustness of the association between PNPLA3 genotype and FPR, we extended the model with further adjustment for additional potential confounders: age at biopsy, presence of overweight, and route of infection (model 2;

Discussion
In this paper, we showed that in a prospective cohort with careful estimation of the date of infection and absence of major confounders the PNPLA3 I148M sequence variant influences fibrosis progression in CHC.
Homozygosity for the I148M variant, which in line with the prevalence in the Italian population [29] was detected in about one in ten subjects, was associated an average incremental progression of fibrosis stage of as much as circa 0.8 units of Ishak score per year. The link between 148M/M status and FPR was independent of known moderators of fibrogenesis in CHC [2,3,30], and remained robust even after extensive adjustment for other potential confounders.
Our study is the first to show that the effect of PNPLA3 I148M on fibrosis progression in CHC is modulated by the age at infection. The negative impact of 148M/M on FPR in individuals infected at older age was confirmed in sensitivity analyses with lower age cut-offs, and was prominent in individuals infected in adulthood. Previous studies suggested that PNPLA3 I148M modulates liver damage progression in an age-dependent fashion. However, despite older age at onset was associated with faster progression of liver damage in alcoholic liver disease, the effect size of 148M/M was larger in younger patients [19]. Similarly, the reported effect size of the PNPLA3 I148M variant on liver damage in nonalcoholic fatty liver disease was larger in pediatric case series [14][15][16].
The reason underlying the opposite direction of modulation by age of the effect of PNPLA3 I148M on fibrosis progression in CHC is presently unknown. It could be speculated that the pattern of activation of the immune system towards HCV is different depending on the age at infection. According to this hypothesis, HCV infection after the developmental age would lead to immune-mediated hepatic damage, exerting a permissive role on PNPLA3 I148M-related fibrogenesis [31]. Alternatively, older age may synergize with PNPLA3 I148M in determining hepatic steatosis, which would favor viral replication during the acute infection [32], thus setting the stage for faster disease progression.
Indeed, evidence is accumulating that PNPLA3 I148M synergizes with host features and environmental triggers in determining liver damage associated with altered hepatic lipid metabolism [33]. In line with these data, we found that the effect of PNPLA3 148M/M on fibrogenesis was more evident in subjects with grade 2-3 steatosis, and in those with other conditions favoring hepatocellular lipid accumulation, such as infection with genotype 3 and overweight.
Of note, we reported for the first time an effect of 148M/M on fibrosis progression in genotype 3 patients, even if previous studies did not detect an association between PNPLA3 and steatosis in this subgroup [34,35]. However, since this finding was based on FPR assessment in only two 148M/M patients infected with HCV G3, it will require confirmation in future studies. Most importantly, the effect of 148M/M on fibrogenesis was independent of the histological severity of steatosis. These data are in line with results of cross-sectional studies in nonalcoholic fatty liver disease [33], and suggest that PNPLA3 I148M may promote fibrogenesis by directly altering hepatic stellate cells lipid metabolism and trans-activation [36]. Therefore, given the independent association of the PNPLA3 I148M variant with fibrosis progression and hepatocellular carcinoma risk [24], PNPLA3 I148M genotyping may help refining treatment prioritization to novel therapeutic regimens based on direct antiviral agents [37], and patients follow-up.
Limitations of this study include the assumption of the absence of liver fibrosis at the time of infection, which is likely given the absence of cofactors of liver damage, the young age at infection and low risk profile. Furthermore, the model tested assumes the consistency and linearity of FPR during time, and of fibrosis progression across Ishak score stages, which are approximations of a more complex reality [38]. Notwithstanding, the present results are corroborated by the fact that our approach was able to validate the association of FPR with the major risk factors for fibrosis progression in CHC [3], and by independent findings linking PNPLA3 148M/M with fibrogenesis in CHC [25,33].

Conclusions
In conclusion, homozygosity for the PNPLA3 I148M variant is associated with FPR in CHC. There is a significant interaction between age at infection and PNPLA3 genotype in determining fibrosis progression, but the association between PNPLA3 and  Table 4. Independent determinants of (log 10 ) FPR at multivariate generalized linear models in 247 CHC patients. FPR is independent of the histological severity of steatosis. These findings further establish the PNPLA3 I148M variant as a moderator of CHC natural history, and have possible relevance for treatment prioritization and follow-up of CHC patients.