HCV Leaves Gene Expression Scar Preceding Liver Cancer Onset

Liver biopsies taken during active hepatitis C infection carried gene expression differences that separated the patients who would go on to develop hepatocellular carcinoma (HCC) from those who would not, a median 6.1 years before any cancer was diagnosed. The differences persisted in liver tissue sampled years later, after the virus itself had been cleared.¹

What Cure Did Not Eliminate

Clearing hepatitis C sharply reduces liver cancer risk without erasing it. Across 23 studies covering 29,395 patients, HCC occurred at 1.54 per 100 person-years after sustained virologic response against 7.80 per 100 person-years in nonresponders.² Cirrhosis was the factor that most separated the patients who remained at risk, with postcure rates of 0.85 per 100 person-years without cirrhosis and 2.47 with it.

That residual risk is what keeps cured patients in surveillance programs, and fibrosis stage remains the dominant clinical predictor of who belongs there. Whether the virus contributes anything beyond the inflammation and regeneration it causes has been argued mostly from cell culture, where hepatitis C virus (HCV) induces endoplasmic reticulum and oxidative stress, interferes with DNA damage repair, and inhibits apoptosis. Those experiments used protein overexpression or high-multiplicity infection of transformed hepatoma cells, which is a considerable distance from a human liver.¹

The study, published in JHEP Reports, drew on a biobank of more than 5000 annotated liver biopsies collected at University Hospital Basel since 1996, identifying 16 patients with active chronic hepatitis C at biopsy who later developed histologically confirmed HCC. Each was matched 1:2 to controls who did not, using propensity scores built on METAVIR fibrosis stage, the standard histologic scale, plus age, sex, race and ethnicity, follow-up duration, Child-Pugh and MELD scores, viral load, genotype, transaminases, treatment and cure status, HIV coinfection, alcohol history, and diabetes.

Every case received antiviral therapy after the baseline biopsy and was cured, with 12 cases negative for HCV at the time of cancer diagnosis. Controls were followed for a median of 10.2 years, and all but 1 achieved a sustained virologic response. Matching on fibrosis stage and inflammatory activity was what allowed the comparison to ask about the virus rather than about the liver damage it produced.

MYC and Protein Stress, Detectable Years Early

Principal component analysis separated the groups after adjustment for sex, with a significant difference along the first component. Differential expression analysis identified 1347 genes at a false discovery rate below 0.05, and among the 56 that differed more than 2-fold, 24 had been linked to HCC in earlier literature.

Two pathways carried the strongest enrichment, both at an adjusted P value of .035: MYC target genes, which ranked first, and the unfolded protein response, the cellular stress program triggered when misfolded proteins accumulate. Allograft rejection and oxidative phosphorylation genes were also reported as enriched, although at adjusted values of .164 and .200. Immune cell deconvolution found no difference in immune infiltration between the groups, and the oxidative phosphorylation genes were all nuclear encoded with no difference in mitochondrial DNA per cell, which points the signal toward the hepatocytes themselves rather than the cells around them.

Two findings cut against a straightforward reading. Immunohistochemistry for c-MYC protein and Ki-67 found no significant difference between groups, so the MYC transcriptional program did not show up at the protein level. Sequencing found a median 9 somatic mutations in cases against 4 in controls, a difference that did not reach statistical significance, although mutation burden correlated strongly and inversely with time to cancer diagnosis.

Building a usable marker required separating changes caused by the virus from changes that reverted once it was gone. Comparison against 15 normal HCV-negative livers identified genes altered by chronic infection, 986 of which overlapped the baseline case-control set. Paired comparison with non–tumor tissue collected years later at cancer diagnosis removed 371 that had normalized after cure, leaving 615 persistently dysregulated genes.

Elastic net regression across 500 bootstrap iterations ranked those genes, and repeated cross-validated logistic regression narrowed them to a 19-gene signature. The investigators were explicit that the elastic net step served only for dimensionality reduction, since out-of-bag performance estimates never stabilized on a dataset this size.

No independent validation cohort exists. What the authors performed instead was a qualitative check against 2 external biopsy sets, 1 low risk and 1 high risk, without cross-cohort batch correction and without any claim of formal validation. They also cautioned that calling these changes virus-induced describes where they were observed rather than what is unique to hepatitis C, since the same pathways may activate in other chronic liver diseases.

Two practical obstacles sit between the signature and clinical use: patients cured of hepatitis C rarely undergo liver biopsy, and the analysis ran on fresh-frozen rather than the formalin-fixed tissue a pathology laboratory archives. Any eventual assay would also have to beat the clinical variables already in use, which is a bar the authors named themselves.

References

1. Boldanova T, Trulsson F, Ebrahimi F, et al. Hepatitis C virus can induce gene expression changes associated with hepatocarcinogenesis. JHEP Rep. 2026;8(8):101897. doi:10.1016/j.jhepr.2026.101897

2. Lv GJ, Ji D, Yu L, et al. Risk of hepatocellular carcinoma occurrence after antiviral therapy for patients with chronic hepatitis C infection: a systematic review and meta-analysis. Hepatol Int. 2024;18(5):1459-1471. doi:10.1007/s12072-024-10700-7