Germline Variants Influence Chronic Liver Disease Progression through Distinct Pathways

Abstract

Cirrhosis and hepatocellular carcinoma (HCC) are long-term complications of chronic liver disease (CLD). In this large multi-ancestry genome-wide association study of all-cause cirrhosis (35,481 cases, 2.36M controls) and HCC (6,680 cases, 1.76M controls), we identified 27 loci associated with cirrhosis (10 novel) and 11 with HCC (three novel). Three novel cirrhosis loci were replicated in independent cohorts (e.g. FGF21, RPTOR, and IFNL3/4). Fifteen cirrhosis loci exhibited differential effects on cirrhosis risk via underlying etiologies, and six HCC loci influenced HCC risk indirectly via cirrhosis. In a gene-burden analysis of rare variants from whole-genome sequencing data in the VA Million Veteran Program (n=102,677), we identified GSTA5 as a novel cirrhosis-associated gene, while APOB and ATP9B were associated with and replicated for HCC. A high genetic risk score for cirrhosis was associated with a nearly doubled risk of CLD progressing to cirrhosis (HR=1.94, P=2×10^-68) and of cirrhosis progressing to HCC (HR=1.65, P=7×10^-08). Finally, among individuals with chronic hepatitis C who underwent antiviral therapy, cirrhosis risk was modified by variants in PNPLA3, IFNL3/4, and CD81 following pegylated interferon-α therapy, and by APOE lead variant following direct-acting antiviral therapy. These findings provide new insights into the complex genetic architecture of CLD progression with potential clinical and therapeutic implications.

Figure 1:. Overview of study design and analyses.

Nine complementary analyses were performed in the current study, namely: a common variant GWAS, coding variant analysis, gene-burden analysis, immunogenetic discovery of HLA alleles, causal mediation analysis of underlying etiologies, multi-state survival analysis of genetic risk scores, integrated clinical and genetic risk prediction, drug target Mendelian randomization, and genotype-specific risk of disease progression under antiviral therapy.

Figure 2:. Genome-wide association results for cirrhosis and hepatocellular carcinoma.

Two Manhattan plots display genome-wide association statistics for cirrhosis (top) and hepatocellular carcinoma (HCC; bottom). Each point represents a single nucleotide polymorphism (SNP), plotted by genomic position on the x-axis and −log_₁₀(P-value) on the y-axis. The HCC panel is inverted to facilitate visual comparison. Multi-ancestry association signals are shown, with ancestry-specific genome-wide significant associations (P < 5×10⁻⁸) overlaid for completeness. These include PDK1 and PPARGC1A (African American) and CXXC4 (European) for cirrhosis, and CD81 (African American) for HCC. Select genome-wide significant loci are annotated with the gene in each region most likely to be functionally relevant, with blue font indicating novel loci. The gray dashed lines indicate the genome-wide significance threshold. This figure highlights both shared and ancestry-specific genetic architecture for cirrhosis and HCC.

Figure 3:. Mediation of genetic effects on cirrhosis and hepatocellular carcinoma.

(a) Mediation analysis estimating the proportion of each SNP’s effect on cirrhosis that is mediated by specific liver disease etiologies. The top panel shows the percentage of the SNP effect mediated through indirect pathways, while the bottom panel displays the residual direct genetic effect on cirrhosis not explained by mediation. Percentage mediation is color-coded by etiology, namely alcohol use disorder, chronic viral hepatitis, metabolic dysfunction with elevated liver enzymes (Met-ALT), hemochromatosis, and alpha-1 antitrypsin deficiency (A1AT deficiency). Mediation results from African American analyses (CD81, PDK1, and HLA) are denoted with a star. (b) Mediation analysis of cirrhosis as a mediator between SNPs and HCC. The top panel shows the percentage of each SNP’s effect on HCC mediated through cirrhosis, while the bottom panel shows the direct effect on HCC independent of cirrhosis and other causes of liver disease. Namely, all models were additionally adjusted for covariates alcohol use, viral hepatitis, Met-ALT, hemochromatosis, alpha-1 antitrypsin deficiency, and other causes of CLD. Analyses were performed in European Americans.

Figure 4:. Probability of being in the cirrhosis state over time by etiology and genetic risk.

Seven panels, stratified by combinations of etiologies (e.g. alcohol-related liver disease, chronic viral hepatitis, and metabolic dysfunction with elevated liver enzymes) show the probability of being in the cirrhosis state over a 20-year period following first CLD diagnosis. Probabilities were estimated using multi-state models with cirrhosis, HCC, and death as competing outcomes. Each panel compares individuals in the top 5th percentile (5^th perc) versus the bottom 95th percentile (95^th perc) of the genetic risk score. The x-axis shows time since first diagnosis of CLD in years, and the y-axis the probability of being in the state of cirrhosis at the respective timepoint. Across all etiologies, higher genetic risk was associated with an increased probability of cirrhosis. Among patients with both alcohol-related liver disease and viral hepatitis, a high genetic risk was further linked to earlier onset of cirrhosis.

Figure 5:. Cirrhosis-free survival by genotype and anti-viral treatment in chronic HCV infection.

Eight panels display 10-year cirrhosis-free survival in treatment-naïve, non-cirrhotic individuals with chronic HCV infection, separated by ancestry and anti-viral treatment: direct-acting antivirals (A, C, E, G) and pegylated interferon alpha with ribavirin therapy (B, D, F, H). Kaplan–Meier curves compare difference in cirrhosis-free survival stratified by genotypes of IFNL3, PNPLA3, and APOE in European Americans, and CD81 in African Americans. The y-axis shows cirrhosis-free survival probability, and the x-axis time since treatment initiation in years.