DNA Repair Protein RPA Emerges as Poor-Prognosis Marker in Liver Cancer

A new pan-cancer analysis, published in the Egyptian Liver Journal, reports that genes encoding Replication Protein A (RPA) are overexpressed in liver hepatocellular carcinoma (LIHC) and that high expression of these genes is associated with worse survival outcomes in patients with the disease. The study, conducted entirely with publicly available datasets, positions the RPA family as a candidate prognostic and therapeutic biomarker that now awaits experimental validation.

RPA is not an obscure molecular player. It is the primary single-stranded DNA-binding protein in human cells, a heterotrimer built from three subunits, RPA1, RPA2, and RPA3, with molecular weights of 70, 34, and 14 kilodaltons respectively. When DNA is damaged or a replication fork stalls, stretches of single-stranded DNA accumulate, and RPA binds these vulnerable regions with exceptionally high affinity. In doing so, it protects the exposed strands from nuclease digestion, destabilizes secondary structures that would halt DNA processing, recruits additional repair factors, and facilitates checkpoint activation. A fourth member of the family, RPA4, is a 30-kilodalton homolog of RPA2 that does not drive DNA replication but can bind DNA repair sites and reduce levels of DNA damage markers. Because of these central functions, disruption of RPA activity can lead to the accumulation of mutations, chromosomal instability, and oncogenic progression.

Although earlier pan-cancer studies had examined individual RPA genes one at a time, the authors of the new work set out to compare all four genes side by side across tumor types. Using the TCGAplot R package together with web tools including GEPIA, TIMER3, the Human Protein Atlas, cBioPortal, UALCAN, and GSCALite, the team mined transcriptomic, clinical, and proteomic data from the TCGA, CPTAC, GTEx, GDSC, and CTRP resources. Their pipeline covered differential expression, genetic mutation mapping, survival prediction, functional annotation, tumor microenvironment profiling, and drug sensitivity analysis across 33 cancer types, with all statistical analyses performed in R and corrected for multiple testing using the Benjamini-Hochberg false discovery rate.

The expression results were striking. RPA1, RPA2, and RPA3 were significantly overexpressed in 10, 12, and 13 cancer types respectively, with the strongest and most consistent overexpression seen in kidney cancers (KICH, KIRC, and KIRP), liver cancer (LIHC), and brain tumors (LGG). At the protein level, assessed through CPTAC proteomic data, RPA1, RPA2, and RPA3 were again overexpressed in the majority of cancers examined, consistent with their mRNA patterns, including in low-grade glioma and liver hepatocellular carcinoma. When the researchers narrowed their focus to LIHC and validated the findings against independent datasets, all four RPA genes were significantly overexpressed in tumor tissue compared with control tissue.

The mutational landscape told a complementary story. RPA1 exhibited alteration frequencies exceeding 2 percent in 13 cancer types, the highest of the family, while alteration frequencies for RPA2, RPA3, and RPA4 stayed below 2 percent in most cancers. Detailed mutation counting revealed 62 missense mutations in RPA1 along with 4 truncating, 12 splice, and 3 fusion mutations and no in-frame mutations, whereas RPA4 carried 76 missense and 6 truncating mutations with no in-frame, splice, or fusion mutations detected. Alpha-missense variant prediction flagged the highest number of potentially pathogenic variants in RPA1 (3,278) and RPA2 (1,553). Importantly, mutations in RPA1 and RPA2 were associated with worse survival, with hazard ratios of 1.04 and 1.1 respectively, while mutations in RPA3 and RPA4 were associated with better outcomes, with hazard ratios of 0.49 and 0.94.

Survival analysis formed the core of the paper’s clinical claim. Univariate and multivariate Cox proportional hazards regression across 33 cancer types showed that elevated expression of RPA1, RPA2, and RPA3 was associated with worse overall survival in low-grade glioma, while RPA2, RPA3, and RPA4 were associated with worse survival in liver hepatocellular carcinoma. In the dedicated LIHC validation, higher expression of the RPA genes was again linked to worse survival outcomes, and independent analysis found hazard ratios greater than one associated with higher expression. Taken together, the authors argue, these results make the RPA genes strong candidates as negative prognostic biomarkers in both glioma and liver cancer.

Why would a DNA repair protein track with aggressive disease? The tumor microenvironment data offer one possible explanation. Correlation analyses revealed significant positive associations between RPA gene expression and multiple immunomodulators, including immune-stimulatory molecules, immune inhibitors, chemokines, and chemokine receptors, in LIHC and LGG. Immune infiltration analysis using the ESTIMATE algorithm showed that in low-grade glioma, RPA1, RPA2, and RPA3 expression correlated positively with immune, stromal, and ESTIMATE scores, and with increased infiltration of pro-tumor immune cells such as M0 and M2 macrophages and neutrophils. Single-cell RNA sequencing of five LIHC datasets added a further layer: RPA genes showed highly heterogeneous expression across liver tumor cell populations, with RPA1, RPA2, and RPA3 most highly expressed in proliferating T cells, and RPA1 and RPA2 also elevated in regulatory T cells, an immunosuppressive population. The picture that emerges is a tumor microenvironment that is immune-inflamed yet immunosuppressed, which the authors suggest explains why high RPA expression coexists with poor prognosis.

Immunotherapy-response indicators pointed in a less optimistic direction. Tumor mutational burden and microsatellite instability are widely used to predict response to immune checkpoint blockade, because elevated levels generally signal higher neoantigen production and greater immunogenicity. In LGG and LIHC, however, the correlation between RPA overexpression and TMB or MSI was very weakly positive, and in some cases negative, suggesting low immunogenicity in these tumors. This stands in contrast to other cancer types, where elevated TMB or MSI was associated with RPA expression in cancers such as UCEC, ACC, COAD, and UVM. For liver cancer and glioma, the finding reinforces the interpretation of a suppressed anti-tumor immune response despite an inflamed microenvironment.

Drug sensitivity analysis offered the therapeutic counterpoint. Using the GSCA platform, which merges mRNA expression data with the half-maximal inhibitory concentrations of 265 small molecules across 860 cell lines from the GDSC and CTRP resources, the researchers found that higher RPA gene expression correlated with increased sensitivity to several anticancer drugs, including docetaxel, doxorubicin, and teniposide. The result gains plausibility from prior experimental work cited in the paper: inhibition of the RPA-RAD52 protein complex by mitoxantrone and doxorubicin has been shown to selectively kill BRCA1-mutated breast cancer cells and BRCA1/2-mutated ovarian cancer cells, and RPA knockdown has been reported to kill glioblastoma cancer stem-like cells. Functional enrichment analysis in the current study also confirmed that RPA1, RPA2, and RPA3 are aligned with DNA replication, mismatch repair, base excision repair, nucleotide excision repair, spliceosome, and cell cycle pathways, while no significant enrichment was found for RPA4.

The authors are candid about the limits of their work. All findings derive from publicly available datasets that may carry inherent biases, and the study lacks in vitro, in vivo, and clinical validation; the team calls for functional studies, including xenograft mouse models, to elucidate the mechanistic role of RPA in the tumor microenvironment. Prior experimental work does lend some support to their computational signals, including reports that transient knockdown of RPA3 inhibited proliferation, colony formation, and invasion in HepG2 liver cancer cells and that stable knockdown suppressed tumor growth in a xenograft model. For now, however, the claim that RPA genes are candidate prognostic and therapeutic biomarkers in liver hepatocellular carcinoma remains a data-driven hypothesis, one that the authors say warrants further validation through in vitro and in vivo experimental studies before it can inform patient care.

Subject of Research: Pan-cancer computational analysis of the RPA DNA-repair gene family in liver hepatocellular carcinoma

Article Title: Replication protein A in cancer: a pan-cancer analysis targeting on liver hepatocellular carcinoma

Article References: Bhushan, R., Dhurve, A., Gupta, P., Gupta, S. K., & Yadav, P. (2026). Replication protein A in cancer: a pan-cancer analysis targeting on liver hepatocellular carcinoma. Egyptian Liver Journal, 16(1), Article 67. https://doi.org/10.1186/s43066-026-00555-y

Image Credits: AI Generated

DOI: 10.1186/s43066-026-00555-y

Keywords: RPA, replication protein A, pan-cancer analysis, liver hepatocellular carcinoma, DNA repair, TCGA, tumor microenvironment, prognosis, drug sensitivity, bioinformatics, glioma, biomarkers