To elucidate the impact of enforced RMP expression on canonical redox and immune checkpoint pathways in hepatoma, we established stable RMP-overexpressing (OE) cell lines in murine Hepa1–6 and human Hep3B cells via lentiviral transduction followed by antibiotic selection. The robust overexpression of epitope-tagged RMP was initially confirmed through immunoblotting, compared to negative-control (NC) lines generated concurrently. (Figure 2A) Densitometric analysis, normalized to GAPDH and then to the NC mean, consistently demonstrated a significant increase in RMP protein levels across three independent experiments for each cell line, thereby validating the intended perturbation (Figure 2B). In the same lysates, we examined NRF2 and PD-L1, two critical effectors associated with oxidative stress adaptation and adaptive immune resistance, respectively. In Hepa1–6 cell lines, NRF2 protein levels were elevated relative to NC, and PD-L1 levels were also increased, resulting in a profile characterized by elevated RMP, NRF2, and PD-L1 at steady state. This pattern was consistently reproducible across replicates and achieved statistical significance, as denoted by the asterisk scheme in the plots (Figures 2C, D). The data suggest that the elevation of RMP is adequate to stabilize or enhance NRF2 and to increase PD-L1 levels in hepatoma cells, thereby establishing a cell-intrinsic RMP-NRF2-PD-L1 axis at the protein level. Importantly, flow cytometry further confirmed that RMP overexpression increased the fraction of PD−L1–positive cells at the cell surface, supporting the immunologically relevant, membrane-associated PD−L1 elevation downstream of the RMP–NRF2 program (Supplementary Figures S1A, B). To investigate the dependence of PD-L1 regulation on NRF2 within the RMP-OE context more thoroughly, Hepa1–6 RMP-OE cells were treated with the NRF2-suppressive agent, brusatol. Immunoblot analysis demonstrated a reduction in NRF2 protein levels, which was associated with a corresponding decrease in PD-L1 expression compared to the control group (Supplementary Figures S2A, B). In addition, exposure of Hepa1−6 RMP−OE cells to H₂O₂ as an oxidative challenge further increased NRF2 protein levels and was accompanied by a parallel upregulation of PD−L1 (Supplementary Figures S3A, B), supporting a redox-responsive association between NRF2 activation and PD−L1 expression in the RMP-high context.

Notably, a comparable trend was observed in human Hep3B cells, wherein the overexpression of RMP was associated with elevated levels of NRF2 and PD-L1 proteins in comparison to the negative control (Supplementary Figures S4A, B). In alignment with the observed coupling at the protein level, an analysis of the TCGA-LIHC cohort further corroborates a positive association between URI1 (RMP) and NFE2L2 (NRF2), as well as PD-L1 (CD274). (Supplementary Figures S5A, B) This association occurs with minor alterations in baseline immune infiltration, highlighting the clinical significance of the RMP–NRF2–PD-L1 state and providing a rationale for our subsequent functional evaluation of RMP-driven tumor cell fitness.

Building on the immunoblot evidence supporting an RMP–NRF2–PD-L1 axis, we proceeded to quantitatively assess the impact of RMP overexpression on colony-forming potential and two-dimensional migratory behavior. In the crystal-violet colony formation assay, Hepa1–6 and Hep3B overexpression (OE) lines were seeded at low density and allowed to form colonies over a 10-day period under identical culture conditions. Following fixation and staining, colonies were enumerated and normalized to the negative control (NC) group for each cell line. Across three independent experiments, OE cells exhibited a significant increase in colony numbers. Qualitative inspection of representative plates indicated a trend towards larger colony sizes, particularly in Hepa1-6 (Figures 3A–D).

We evaluated the migratory capacity using a scratch-wound assay on confluent Hepa1–6 monolayers. Immediately following the creation of a linear wound, images were captured at the 0-hour mark, and wound closure was quantified after 24 hours under standard serum conditions. RMP-OE monolayers demonstrated accelerated wound closure compared to NC, as indicated by a higher percentage of gap reduction over the 24-hour period (Figures 3E, F). Although wound-healing assays can be affected by cell proliferation, the relatively short duration of the experiment and the use of matched conditions help to mitigate this confounding factor. Furthermore, the concurrent increase in clonogenicity suggests that RMP likely influences both proliferation-associated and motility-associated characteristics. Future investigations employing transwell migration or invasion assays and live-cell tracking could more precisely delineate these components. However, within the context of our study, the combined enhancements in clonogenicity and scratch-wound closure support a role for RMP in promoting cellular fitness.

Together, these results suggest a coordinated rewiring in which RMP overexpression simultaneously pushes hepatoma cells toward a redox−competent, immune−checkpoint−high state and toward a growth−and−motility−enhanced phenotype. The biological plausibility of this coupling is supported by prior observations that RMP, also known as URI1, can engage KEAP1 via acidic motifs to limit NRF2 degradation, thereby favoring NRF2−dependent transcriptional programs; in turn, NRF2 activity has been associated with PD−L1 regulation in multiple tumor contexts (31, 34, 35).

Subsequently, we investigated the impact of RMP overexpression on tumor behavior in vivo utilizing a subcutaneous Hepa1−6 model in immunocompetent mice. An equal number of overexpression (OE) and negative control (NC) cells were implanted into the flanks of female C57BL/6 mice (n=5 per group), and tumor growth was monitored longitudinally until predetermined endpoints were reached. (Figure 4A) Compared to the NC group, tumors in the OE group demonstrated accelerated growth, as evidenced by increased serial volumes and higher endpoint tumor weights (Figures 4B, D, E). Representative gross images and hematoxylin and eosin (H&E) stained sections displayed typical hepatoma morphology without qualitative architectural differences beyond size, indicating that RMP overexpression does not significantly alter the histological phenotype in this context (Figure 4C). Body weight trajectories did not significantly differ between the groups (Supplementary Figure S6), and H&E surveys of organs revealed no overt histopathological changes in the heart, liver, spleen, lung, or kidney under baseline conditions (Figures 4F, G), suggesting that the model did not induce systemic toxicity.

Immunohistochemistry provided a molecular framework for understanding the growth advantage observed in OE tumors, which exhibited enhanced staining for RMP, thereby validating the genetic perturbation (Figure 5A). Similarly, NRF2 staining was elevated (Figure 5B), consistent with a redox-adaptive program. PD−L1 immunoreactivity was higher in OE tumors than in NC tumors (Figure 5C), consistent with the increased PD−L1 protein levels observed in vitro (Figure 2D) and supported by the elevated cell−surface PD−L1 detected by flow cytometry (Supplementary Figures S1A, B). The proliferative index, as measured by Ki-67, was elevated in OE tumors (Figure 5D), and heme oxygenase-1 (HO-1), a canonical target of NRF2 that mediates oxidative stress responses, was also increased (Figure 5E). The collective increase in RMP, NRF2, Ki−67, and HO−1, together with higher PD−L1 immunoreactivity, characterizes a tissue state defined by proliferation and oxidative−stress adaptation. This integrated phenotype connects our cellular findings with tumor behavior: the NRF2/HO−1 shift is consistent with enhanced redox adaptation that may support tumor growth, while higher PD−L1 immunoreactivity suggests potential engagement of the PD−1/PD−L1 axis, which we test functionally by PD−1 blockade in the next section. The tissue data indicate that OE tumors display a redox−adapted and proliferative state, and the increased checkpoint−related signals provide a rationale to evaluate PD−1 blockade responsiveness in the subsequent experiments. This dual characteristic underpins the rationale for exploring PD-1 blockade therapy in the subsequent section. The a priori expectation is that pathway engagement, a necessary condition for therapeutic response, is present. However, it is anticipated that additional resistance mechanisms associated with NRF2 and HO-1 may limit the extent of therapeutic benefit.

To determine whether the composite RMP, NRF2, and PD−L1 state influences the magnitude of response to PD−1 blockade, we treated tumor−bearing mice with an anti−PD−1 antibody administered every other day for six doses at 3 mg/kg, beginning once tumors reached 50–100 mm³ (n=5 per genotype). (Figure 6A) Throughout the treatment process, the mice exhibited no significant reduction in body weight, and the structural integrity of their major organs remained intact (Supplementary Figures S7A, B). Both the NC and OE cohorts demonstrated tumor regression during therapy, as evidenced by decreases in tumor volume and reductions in endpoint tumor weight compared to their respective untreated controls (see Figures 6B–E in relation to the growth profiles in Figures 4D, E). Our findings also suggest that the absolute reduction in tumor size may be more pronounced in the OE group, even though this group presented with larger tumors at baseline. This observation partially indicates that the RMP-NRF2-PD-L1 axis plays a significant role in the response to PD-1 blockade in HCC and may be associated with the immune environment, and further analysis is warranted. However, when normalizing the treatment effect within each genotype using tumor−weight inhibition, the inhibition rates were 64.34% in NC tumors and 63.30% in OE tumors, supporting a comparable—but not enhanced—proportional response in the RMP/NRF2−high context. This finding aligns with the hypothesis that the NRF2-skewed background limits the efficacy of anti-PD-1 monotherapy. This relationship is clearly reflected in the endpoint weights, where OE tumors under anti-PD-1 treatment (OE+PD-1) remain heavier than NC tumors under anti-PD-1 treatment (NC+PD-1) (Figure 6C), despite both cohorts showing a response to therapy.

At the conclusion of PD-1 therapy, we conducted a comprehensive profiling of tumor tissues to evaluate whether RMP overexpression continues to influence the tumor microenvironment under checkpoint blockade. Immunohistochemical analysis showed that the staining intensities of RMP and NRF2, as well as PD−L1 immunoreactivity, remained higher in OE+PD−1 tumors than in NC+PD−1 tumors (Figures 7A–C). These patterns indicate that the oxidative−stress and checkpoint−related signals associated with RMP remain comparatively higher under PD−1 therapy. Additionally, Ki-67 levels were higher in OE plus PD-1 tumors (Figure 7D), indicating a sustained proliferative drive, while heme oxygenase-1 levels also remained elevated (Figure 7E), suggesting that NRF2-dependent redox adaptation persists despite therapy. These findings are compatible with a growth−permissive, redox−adapted state that may contribute to a constrained response amplitude under PD−1 blockade.

Dual-color immunofluorescence provided valuable insights into the immune context. The signals for CD3 and CD8, which are indicative of overall T-cell presence and cytotoxic T-cell infiltration, were more pronounced in the OE+PD-1 group compared to the NC+PD-1 group. Conversely, the signals for CD4 and CD25, which serve as proxies for regulatory T-cell characteristics, were generally lower in the OE+PD-1 group. Quantitative data are presented as relative fluorescence intensities across five selected regions per sample (Figures 7F–I). To achieve a more precise cell-based assessment, we further quantified the densities of CD3⁺CD8⁺ and CD4⁺CD25⁺ double-positive cells. These measurements mirrored the directional changes identified through RFI analysis, as shown in Supplementary Figures S8A, B. Collectively, these post−therapy tissue data complement the macroscopic response profile and suggest that while PD−1 blockade elicits cytotoxic T−cell infiltration, the axis linking RMP, NRF2, and heme oxygenase−1 remains active and PD−L1 remains elevated, maintaining inhibitory pressure that likely explains the lower percent inhibition in OE compared with NC.