Divergence of the CYB5R3-mARC1 Redox Axis Across the Human MASLD-to-Hepatocellular-Carcinoma Continuum: An Exploratory Analysis of Liver-Biopsy and Tumor Cohorts.

CYB5R3 (NADH-cytochrome b5 reductase 3) and mARC1, the product of MTARC1, draw on the same cytochrome-b5 electron relay, so they are usually treated as two ends of one redox axis. Whether they behave as a unit across the continuum from metabolic dysfunction-associated steatotic liver disease (MASLD) to hepatocellular carcinoma (HCC) has not been established. In TCGA-LIHC, higher CYB5R3 expression was associated with shorter overall survival in a continuous multivariable Cox model adjusted for age, sex, and stage (n = 339, 114 deaths; HR 1.23 per SD, 95% CI 1.01-1.48, p = 0.035), although it added essentially nothing to a clinical model containing age, sex, and stage (change in Harrell C-index 0.005, 95% CI -0.015 to +0.037). The direction was reproduced in the only independent cohort examined, GSE14520 (n = 242, 96 deaths; unadjusted HR 1.42 per SD, 95% CI 1.14-1.76, p = 0.0015), in which higher MTARC1 was associated with longer survival (unadjusted HR 0.73 per SD, 95% CI 0.60-0.89, p = 0.0017); both single-gene estimates fall below conventional significance after adjustment for age, sex, and TNM stage (HR 1.24, 95% CI 0.99-1.56, p = 0.058 and HR 0.89, 95% CI 0.72-1.10, p = 0.29; n = 225, 86 deaths), although CYB5R3 remains nominally significant in an adjusted model containing both genes (HR 1.27, 95% CI 1.01-1.59, p = 0.039), so that cohort replicates direction rather than independent prognostic value. Across the primary tumors of the analysis set, no correlation between the two transcripts was detectable (Spearman rs = -0.03, 95% CI -0.13 to +0.08, p = 0.62; n = 339), so the axis is not demonstrably a unit at the expression level. In three human liver-biopsy cohorts spanning the MASLD histological spectrum (393 biopsies with fibrosis staging), CYB5R3 gave estimates that differed in sign between cohorts and were uninformative when pooled (fibrosis stage, random-effects rs = +0.01, 95% CI -0.24 to +0.26, q = 0.95, I2 = 83%). MTARC1 behaved differently, falling with fibrosis stage in all three cohorts (pooled rs = -0.22, 95% CI -0.32 to -0.13, q = 3 × 10-5, I2 = 0%), as did PARP16 (-0.17, I2 = 0%), while SCD rose in all three (+0.13, I2 = 0%). This is consistent with the inconsistency being specific to CYB5R3 rather than a property of the cohorts, although the fibrogenic positive controls were themselves heterogeneous (I2 = 53-62%), I2 is estimated from only three cohorts, and the difference between the CYB5R3 and MTARC1 coefficients was not formally tested. These observations are associative and hypothesis-generating. They provide no support for the widely assumed corollary that hepatic CYB5R3 becomes deficient as steatotic liver disease advances, so a therapeutic strategy predicated on that corollary currently rests on mouse gain-of-function data alone, without corroboration from human tissue. What they do support is narrower: within a single physically coupled electron-transfer axis, the two members behave differently before malignancy-MTARC1 tracks fibrosis stage reproducibly while CYB5R3 shows no reproducible relationship-and carry opposite prognostic directions after malignancy, significant for MTARC1 only before covariate adjustment.