All procedures involving animals were approved by the Institutional Animal Care and Use Committee (IACUC) of the National Translational Science Center for Molecular Medicine, Fourth Military Medical University (Approval No. 2023-NTSCMM-ID005). All mice were bred under specific pathogen-free (SPF) conditions. At the end of the study, blood samples were collected under deep anesthesia, and mice were euthanized by an intraperitoneal injection of pentobarbital.

The detailed methods are in the supplementary material.

The hydrodynamic tail-vein injection (HTVi) system has been used to establish a novel mouse model of liver cancer by injecting Sleeping Beauty (SB) plasmids and oncogenic plasmids. To more clearly demonstrate the progress and changes of murine HTVi HCC model, we conducted preliminary experiments to determine the tumor development status at each time point. At week 8, the liver showed small white tumor foci (maximum tumor diameter ≤2 mm). At week 9, the maximum tumors grew to ≤5 mm. At week 10, tumors continued to grow but remained smaller than 10 mm in diameter. However, from 11–13 weeks, numerous large nodules emerged on the liver (maximum tumor diameter ≥10 mm), while mice developed ascites and showed cachexia (Fig. 1A, Supplementary Fig. 1A–1E). H&E staining revealed the presence of necrosis within these large tumors during weeks 11 to 13 (Supplementary Fig. 1F). Therefore, we designate the 8–9 week period as earlystage HCC and the 11-13 weeks as late-stage HCC, characterized by significant increases in maximum tumor diameter and number, the emergence of larger nodules with necrotic areas, and the performance of mice (Fig. 1B). Immunohistochemical staining showed high expression of CD147 and glypian-3 (GPC3) within the tumor lesions of murine and human HCC, which are widely accepted as biomarkers for HCC (Supplementary Fig. 2A) [9,10]. Notably, early HCC lesions showed significant enrichment of cancer-associated fibroblasts (CAFs), endothelial cells, and T cells (Supplementary Fig. 2B, 2C). These immunostaining patterns aligned with observations in tumor tissues of earlystage HCC patients (Supplementary Fig. 2D). Multiplex immunofluorescence (mIHC) staining confirmed that CD8 expression within early-stage HCC tissue was markedly elevated compared to the adjacent non-tumor tissue or normal liver tissue (Fig. 1C, 1D and Supplementary Fig. 2E). Tumor-infiltrating CD8⁺ T cells from 2-week subcutaneously grown Hepa1-6 tumors had higher expression of exhaustion markers including PD-1, TIM-3 and TIGIT, and an increased frequency of PD-1⁺ TIM-3⁺ terminal exhausted T (Tex) cells compared with T cells from spleen (Fig. 1E and Supplementary Fig. 3A, 3B). Consistently, CD8⁺ T cells from 9-week tumors of HTVi HCC model showed increased levels of exhaustion markers and an increased frequency of PD-1⁺ TIM-3⁺ Tex cells (Fig. 1F and Supplementary Fig. 3C, D). These results indicate that the exhaustion of tumorinfiltrating CD8⁺ T cells show up even at the early stage of tumorigenesis.

To elucidate the impact of the TME on the functional status of CD8⁺ T cells during the progression of HCC, we performed single-cell RNA sequencing (scRNA-seq) analysis on 3 samples of early-stage and 3 samples of late-stage tumors from the HTVi HCC model (Fig. 2A left, Supplementary Fig. 4A, 4B). T cells were significantly more enriched in early-stage tumors than in late-stage tumors, whereas late-stage tumors exhibited a higher proportion of HCC cells (Fig. 2A right and Supplementary Fig. 4C). The proportion of exhausted CD8⁺ T cells increased with disease progression (Fig. 2B, Supplementary Fig. 4D). Trajectory analysis performed using monocle3 uncovered a sequential developmental trajectory that began with CD8⁺ Tn cell cells and progressed over KIR⁺TXK⁺NK-like CD8⁺ T cells into CD8⁺ Tex cells (Fig. 2C). Density analysis shows that more KIR⁺TXK⁺NK-like CD8⁺ T cells were becoming exhausted as the tumor progressed (Fig. 2D). In addition, compared with early-stage tumors, CD8⁺ Tex cells in latestage tumors showed significant enrichment for PD-L1 expression and PD-1 checkpoint pathway (Fig. 2E, left). Like NK cells, effector memory KIR+ CD8⁺ T cells are cytotoxic and can produce IFNγ [11]. CD8⁺ Tex cells from late-stage tumors highly expressed multiple exhaustion markers and certain genes related with T cell exhaustion, including Pdcd1, Tox, Tox2, Cd38, Entpd1, Batf, and Rbpj (Fig. 2E, right). The exhaustion of CD8⁺ T cells is orchestrated by a sophisticated network of exhaustion-inducing transcription factors (TFs). Therefore, we evaluated the specifically expressed TFs and regulon activity in CD8⁺ Tex cells infiltrating early- and late-stage HCC tumors by SCENIC analysis. We found that early-stage CD8⁺ Tex cells showed higher activity levels of TFs associated with effector differentiation (Rel, Nfkb1, Jund), stemness (Foxo1 and Runx1) and survival (Runx2). Late-stage CD8⁺ Tex cells showed higher activity levels of TFs associated with exhaustion, including Fli1 and Irf2 (Fig. 2F). Unsupervised analysis further classified exhausted CD8⁺ T cells into 6 sub-clusters (Supplementary Fig. 5A, 5B). Clusters 0, 2, and 4 were enriched in early-stage tumors, while clusters 1, 3, and 5 were enriched in late-stage tumors (Supplementary Fig. 5C). To reveal the dynamics of transcriptional profiles of CD8⁺ Tex cells, we applied trajectory analysis. Partition-based graph abstraction (PAGA) revealed that cluster 2 emerged the earliest while cluster 4 appeared the latest (Supplementary Fig. 5D, left). Unsupervised monocle3 pseudotime analysis confirmed that the transition tendency originated from cluster 2 enriched in early-stage tumors (Supplementary Fig. 5D, right). Cluster 2 expressed high levels of the memoryrelated transcripts Tcf7 and Il7r, which are reported as essential markers for progenitor exhausted T cells (Supplementary Fig. 5B). These findings highlight the transition tendency of exhausted T cells during HTVi tumor progression. We further analyzed the scRNA-seq data of human HCC samples at early stage and the adjacent liver tissues (Supplementary Fig. 6A, 6B) [12]. We found 3 cluster subtypes, which were defined as CD4+, CD8⁺ and Treg cells (Supplementary Fig. 6C). Consistent with the results in murine HCC models, the infiltration of CD8⁺ T cells in tumor samples was found to be significantly increased compared with adjacent liver tissues (Supplementary Fig. 6D). Genes differentially expressed in tumor-infiltrating CD8⁺ T cells also showed enrichment of PD-L1 expression and PD-1 checkpoint pathway (Supplementary Fig. 6E, 6F). In addition, CD8⁺ T cells infiltrated in HCC tissue expressed higher levels of inhibitory molecules (PDCD1, HAVCR2, TIGIT, CTLA4) and exhibited a reduced capacity to secrete effector cytokines such as IFNγ (Supplementary Fig. 6G). In summary, these results reveal a correlation between HCC progression and the molecular signature associated with CD8⁺ T cells exhaustion, which contributes to HCC progression.

To investigate how TME influences the metabolism and the effector functions of CD8⁺ T cells, we co-cultured CD8⁺ T cells with tumor mass obtained from the murine HTVi model. Notably, the presence of tumor mass induced the expression of immune exhaustion markers, including TIM-3, CTLA4, and TIGIT (Fig. 3A). As tumor progressed, the number of tumor-infiltrating CD8⁺ T cells decreased (Supplementary Fig. 6H). Genes upregulated in late-stage CD8⁺ Tex cells compared with early-stage CD8⁺ Tex cells showed enrichment of several processes related to ferroptosis, including unsaturated fatty acid metabolic process, iron ion binding and reactive oxygen species (ROS) metabolic process (Fig. 3B and Supplementary Fig. 6I). Consistently, Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analysis of differentially expressed genes (DEGs) between CD8⁺ T cells in human HCC tumors and paracancerous tissues revealed enrichment of ferroptosis signaling pathway, suggesting a potential link between the functional loss of CD8⁺ T cells and ferroptosis in the progression of HCC (Supplementary Fig. 6F). Meanwhile, diaminobenzidine-enhanced Prussian blue iron staining revealed the accumulation of iron in murine and human HCC tumors at early stage (Fig. 3C). Moreover, tumor-infiltrating CD8⁺ T cells in early-stage murine HCC tumors had higher level of ferrous iron compared with splenic CD8⁺ T cells, indicating the import of iron from TME and the intracellular labile ferrous accumulation (Fig. 3D). In addition to elevated ferrous levels, the production of 4-hydroxynonenal (4HNE), a putative biomarker of lipid peroxidation, was enhanced in CD8⁺ T cells of early-stage murine and human HCC tumors compared with CD8⁺ T cells of liver tissues (Fig. 3E, 3F). These data confirm the accumulation of lipid peroxides and iron overload in tumor-infiltrating CD8⁺ T cells in HCC tumors, which occur at the early-stage of tumor development and are simultaneously enhanced during the progression of HCC.

To determine the effects of iron overload in CD8⁺ T cells, we treated human and murine CD8⁺ T cells with ferric ammonium citrate (FAC). CD8⁺ T cells isolated from healthy donors’ peripheral blood mononuclear cells were activated and then exposed to FAC. Compared to control cells, FACtreated CD8⁺ T cells exhibited significantly higher intracellular labile Fe²⁺ level (Supplementary Fig. 7A) and increased lipid ROS (Supplementary Fig. 7B). Notably, lipid ROS accumulation induced by FAC was inhibited by ferrostatin-1, a lipid peroxide-trapping agent and ferroptosis inhibitor (Supplementary Fig. 7B). While the percentage of cell death in CD8⁺ T cells viability remained unchanged at 15 and 18 hours without FAC treatment, FAC-treated cells showed a significant increase of cell death (Supplementary Fig. 7C). Additionally, FAC-treated cells exhibited increased expression of immunosuppressive molecules PD-1 and TIM-3 (Supplementary Fig. 7D, 7E) and a significant decrease in the production of key cytotoxic cytokines, including TNFα, Perforin, and Granzyme B (Supplementary Fig. 7F–7H). Similar effects were observed in CD8⁺ T cells isolated from the spleens of mice (Fig. 4A–4D). Iron chelation with DFO effectively reduced intracellular iron levels of splenic CD8⁺ T cells induced by FAC, FeSO₄ or FeCl₃, as evidenced by decreased of lipid peroxidation. The addition of DFO also significantly mitigated the decrease IFNγ and the increase of T cell exhaustion marker PD-1 observed under iron overload conditions, highlighting the role of iron in driving CD8⁺ T cell dysfunction (Supplementary Fig. 8). Then we performed RNA-seq to detect if T cell metabolism and signaling pathways were influenced in CD8⁺ T cells with FAC stimulation. The differential expression analysis of genes between FAC-stimulated CD8⁺ T cells and control cells revealed that 703 genes were upregulated and 1,641 genes were downregulated (Supplementary Fig. 8A). FAC-stimulated CD8⁺ T cells showed decreased expression of memory control gene Cxcr4, stem-like gene Runx1 and Jun, a TF associated with productive T cell activation [13-15]. They also showed increased expression of terminal exhaustion genes, including PDCD1, CTLA4, Tox, and LAG3, TFs implicated in fostering T cell exhaustion, including Nr4a1 and Batf (Supplementary Fig. 8B) [16-18]. KEGG analysis showed that DEGs were enriched in pathways including cytokinecytokine receptor interaction, p38 MAPK signaling, and PD-L1 expression and PD-1 checkpoint pathway (Supplementary Fig. 8C). The Gene Ontology analysis revealed that they were involved in critical biological process terms associated with ferroptosis, such as response to oxidative stress, unsaturated fatty acid metabolic process and iron ion transport (Supplementary Fig. 8D). GSEA also confirmed the enrichment of PD-L1 expression and PD-1 checkpoint pathway in FAC-stimulated CD8⁺ T cells (Supplementary Fig. 8E). To explore the in vivo effects of iron overload, we fed mice with standard or high-iron diet (HID) for 16 weeks, starting at 4 weeks of age. Subcutaneous tumor model was then established, and tumor and spleen tissues were analyzed by flow cytometry 2 weeks later (Supplementary Fig. 10A). The results revealed that tumor-infiltrating CD8⁺ T cells from tumors in HID mice had increased intracellular labile Fe²⁺ levels and significantly elevated levels of TIGIT and PD-1 compared to CD8⁺ T cells from the standard diet group, with no significant change in TIM-3 expression (Fig. 4E, 4F). Similarly, CD8⁺ T cells in the spleens of HID mice showed increased iron and elevated expression of TIGIT and PD-1 relative to the standard diet group (Supplementary Fig. 10B–10D). In summary, these findings suggest that iron overload impairs CD8⁺ T cells effector function, potentially diminishing their ability to mount an effective anti-tumor immune response. Furthermore, we cultured CD8⁺ T cells under cystine-deprived and iron-enriched conditions to investigate the role of cysteine metabolism and iron accumulation in the context of CD36^-mediated T cell dysfunction. Our results showed that cystine deprivation and iron accumulation exerted a synergistic effect on the expression of T cell exhaustion markers, including PD-1 and TIM-3, accompanied by increased lipid peroxidation (Supplementary Fig. 11). These findings are consistent with the previous study which demonstrated that cystine deprivation triggers CD36^-mediated ferroptosis and T cell exhaustion [6].

To investigate the mechanisms contributing to the functional loss of CD8⁺ T cells in HCC TME enriched with iron, we stimulated CD8⁺ T cells with FAC and examined the expression of molecules related to iron homeostasis and lipid metabolism. After 12 hours of FAC exposure, CD8⁺ T cells exhibited significantly elevated expression of TfR1, while the expression of divalent metal transporter 1 (DMT1) was decreased (Supplementary Fig. 12A). Although the expres-sion of acyl-CoA synthetase long-chain family 4 (ACSL4) was increased under short-term stimulation with FAC, the effect was diminished under long-term stimulation. Meanwhile, the expression of CD36 was maintained at a high level (Supplementary Fig. 12A). The depletion of CD36 (CD36^-/-) significantly inhibited the increase of intracellular labile Fe²⁺ level and the accumulation of lipid peroxides triggered by FAC treatment (Supplementary Fig. 12B, Fig. 5A left and median). Compared with WT CD8⁺ T cells, culture of CD36^-/- CD8⁺ T cells with addition of FAC showed notably downregulated PD-1 and TIM-3 expression (Fig. 5A, right). In line with these results, FAC exposure induced elevated expression of TfR1, FTH1 and SLC40A1 in CD8⁺ T cells, and the depletion of CD36 significantly inhibited the upregulation (Fig. 5B). Overexpression of CD36 in CD36^-/- CD8⁺ T cells significantly elevated intracellular ferrous iron levels and lipid peroxidation, as well as the expression of PD-1 and TfR1 (Fig. 5C, 5D). These results confirmed the role of CD36 in mediating the dysregulation of iron homeostasis and lipid peroxidation in CD8⁺ T cells, probably by upregulating TfR1 expression. Since the expression of TfR1 can be regulated by the IRE/IRP regulatory system, we examined the effects of FAC stimulation on IRP1 and IRP2 expression under WT and CD36 knockout conditions. We found that IRP1 levels remained unchanged, while IRP2 expression was significantly increased in FAC-stimulated WT CD8⁺ T cells compared to controls, and was markedly reduced after CD36 knockout (Supplementary Fig. 12C). Therefore, the IRE/IRP system is likely involved, but its specific role requires further investigation. Among the DEGs between FAC-stimulated CD8⁺ T and control cells that were enriched in MAPK signaling pathway, MAPK11, one of the four p38 MAPKs, was upregulated (Supplementary Fig. 12D). It has been found that oxLDL uptake via CD36 in HCC CAFs leads to p38 phosphorylation and subsequent C/EBPβ activation [19]. Additionally, utilizing ENCODE chromatin immunoprecipitation sequencing (ChIP-seq) data, we identified TFRC (encoding TfR1) as a target gene for CEBPB (GSM1519638) and CEBPD (GSM1010777). Based on the above evidence, we focused on the p38-CEBPs axis to validate its regulatory effect on the expression of TfR1. We found that the simulation with FAC in WT CD8⁺ T cells significantly elevated the protein levels of p-p38, CEBPB and CEBPD, which were enhanced by oxLDL (Supplementary Fig. 12E, 12F). These effects were greatly mitigated by CD36 knockout or the CD36 inhibitor SSO (Supplementary Fig. 12E, 12F). Meanwhile, overexpression of CD36 led to a significant increase in p-p38, CEBPB and CEBPD in CD36^-/- CD8⁺ T cells (Supplementary Fig. 12E). To investigate the role of p38 in TfR1 regulation, we treated cells with a specific p38 inhibitor SB203580. Our data showed that inhibiting p38 significantly mitigated the elevation of p-p38, CEBPB, CEBPD and TfR1, as well as exhaustion markers induced by FAC and oxLDL in WT CD8⁺ T cells, indicating that p38 activation is essential for TfR1 upregulation and CD8⁺ T cells dysfunction (Supplementary Fig. 12G–12J). Immunofluorescence staining revealed that either FAC or oxLDL alone could result in a noticeable nuclear translocation of both CEBPB and CEBPD in WT CD8⁺ T cells, which was mitigated by CD36 knockout (Supplementary Fig. 13A, 13B). Furthermore, knockdown of CEBPB resulted in a significant decrease of TfR1 expression at both the protein and mRNA levels. However, while CEBPD knockdown had minimal impact, CEBPB knockdown significantly reduced TfR1 and intracellular ferrous iron levels (Supplementary Fig. 13C–13J). This suggests that CEBPB, rather than CEBPD, is the key TF mediating TfR1 upregulation. As shown in Fig. 5E left, transfection of the CEBPB plasmid into Jurkat T cells markedly increased the luciferase activity of the TFRC promoter region. The binding of CEBPB to the TFRC promoter was promoted by either oxLDL or FAC, and the combination of oxLDL and FAC caused enhanced luciferase activity. Further, suppressing CD36 by SSO reduced TFRC promoter activity (Fig. 5E, right). ChIP analysis revealed that CEBPB binds to the TfR1 promoter region, indicating direct transcriptional regulation (Fig. 5F). These findings suggest that CD36 plays a critical role in regulating TfR1 expression and iron metabolism through the p38-CEBPB axis in the presence of high-iron microenvironment. mIHC revealed high expression of CD36 on tumor-infiltrating CD8⁺ T cells in early-stage HCC tumors of the HTVi model (8 weeks post injection) (Supplementary Fig. 14A). Compared to WT mice, CD36^-/- mice exhibited slower subcutaneous tumor growth (Supplementary Fig. 14B, 14C), lower intracellular labile ferrous levels in tumor-infiltrating CD8⁺ T cells (Supplementary Fig. 14D), reduced accumulation of iron in tumors (Supplementary Fig. 14E), reduced serum levels of the lipid peroxidation product MDA (Supplementary Fig. 14F), and decreased expression of immunosuppressive molecules, including PD-1, TIM-3, and TIGIT (Supplementary Fig. 14G). These findings indicate that CD36 deficiency rescues the CD8⁺ T cells effector functions in part through reducing the iron accumulation.

In the early and late stage scRNA-seq data, GSEA showed that processes associated with long-chain fatty acid metabolism, unsaturated fatty acid metabolism, and positive regulation of lipid metabolism were also enriched in tumorinfiltrating CD36+ CD8⁺ T cells, while negative regulation of oxidative stress-induced cell death process was enriched in CD36^-/- CD8⁺ T cells (Supplementary Fig. 14H). Tumor-infiltrating CD36+ CD8⁺ T cells also had higher activation of signaling pathways involved in PD-L1 expression and PD-1 checkpoint pathway and ROS, revealed by GSEA (Supplementary Fig. 14I). Altogether, these results confirm that CD36 not only facilitates the uptake of oxLDL, but also mediates the induction of iron accumulation through the ox-LDL-p38-CEBPB-TfR1 axis in CD8⁺ T cells in the iron-enriched TME of early-stage HCC, contributing to lipid peroxidation and subsequent CD8⁺ T cells dysfunction.

The accumulation of ferrous iron in cells can lead to lipid peroxidation and ferroptosis [20]. To investigate whether the ferroptosis-resistant pathways are activated in tumor-infiltrating CD8⁺ T cells with intracellular labile ferrous accumulation, we examined the DEGs between PDCD1 (PD-1)+ HAVCR2 (TIM-3)low and PDCD1+ HAVCR2^high tumor-infiltrating CD8⁺ T cells in scRNA-seq of human early HCC [12]. PDCD1+ HAVCR2^high CD8⁺ T cells expressed lower levels of NRF2 and higher levels of KEAP1 than PDCD1+ HAVCR2^low cells (Supplementary Fig. 15A). In scRNA-seq of murine HTVi HCC model, KIR⁺TXK⁺ NK^-like CD8⁺ T cells from early-stage tumors express high levels of Nfe2l2 (encoding NRF2), while CD8⁺ Tex cells from late-stage tumors had lower expression of NRF2 than that from early-stage tumors (Fig. 6A). Furthermore, Nfe2l2^low intratumoral CD8⁺ T cells exhibited enrichment of processes associated with higher levels of lipid peroxidation and ferroptosis compared with Nfe2l2^high CD8⁺ T cells, including mitochondrial respiratory chain, fatty acid metabolic process and lipid oxidation (Fig. 6B). Nfe2l2^low cells also displayed highly enriched gene set related to ROS production (Fig. 6C).

Given these observations, we focused on the KEAP1/ NRF2 signaling pathway, which is critical for the antioxidant stress response [21]. With FAC stimulation, the expression of NRF2 and its downstream effector FSP1 significantly increased. However, the expression of other NRF2 downstream targets associated with ferroptosis defense, including NQO1, HO-1 and GPX4 declined in response to 12 hours of FAC treatment (Fig. 6D). NRF2 activity is generally inhibited through its interaction with KEAP1, which facilitates NRF2 degradation. Under conditions of oxidative stress, this interaction is disrupted, allowing NRF2 to translocate to the nucleus and activate its target genes [22,23]. Immunofluorescence analysis indicated that FAC exposure led to slightly increased nuclear localization of NRF2 in CD8⁺ T cells (Supplementary Fig. 15B). The combined stimuli of FAC and NRF2 inhibitor ML385 suppressed IFNγ production (Supplementary Fig. 15C). This result indicates that NRF2 activation is essential for the maintenance of antitumor capabilities of CD8⁺ T cells. It has been reported that ratios of NRF2 over KEAP1 point to activation of NRF2 signaling pathway and increased antioxidant capability [24]. We next investigated the effects of FAC and oxLDL on NRF2/KEAP1 ratio by treating CD8⁺ T cells with increasing concentrations of oxLDL, with or without the presence of FAC. Although oxLDL at 40 μg/mL with the addition of FAC slightly increased the NRF2/KEAP1 ratio compared with FAC-treatment, higher doses (80 and 160 μg/mL) of oxLDL reduced the ratio, likely due to the upregulation of KEAP1 (Fig. 6E). Additionally, the expression of TfR1 was induced with increasing concentrations of oxLDL, which imports iron and thus further promotes the risk of ferroptosis. Under the combined effect of these factors, the antioxidant capability of NRF2 may not be sufficient to halt the lipid peroxidation caused by the combination of FAC and oxLDL in tumor-infiltrating CD8⁺ T cells (Fig. 6E). We then stimulated CD8⁺ T cells with Ki696, a selective inhibitor of KEAP1/ NRF2 interaction. Although NRF2 expression did not exhibit significant changes, Ki696 treatment under high iron conditions resulted in a pronounced increase in NRF2 nuclear translocation (Fig. 6F and Supplementary Fig. 15D). Given that the ferroptosis-defending capability of NRF2 in CD8⁺ T cells was insufficient to counteract the oxidative damage caused by the environment enriched with iron and oxLDL, we assumed that prevention of lipid peroxidation via NRF2 activation may rescue effector functions of T cells. To test this hypothesis, we infected CD8⁺ T cells with lentivirus overexpressing a constitutively active form of NRF2 (caNRF2), which lacks the first 89 amino acids necessary for binding KEAP1 (Supplementary Fig. 16A). CD8⁺ T cells overexpressing caNRF2 demonstrated robust nuclear localization following FAC stimulation (Fig. 7A and Supplementary Fig. 16B). Furthermore, in these cells, TfR1 expression was significantly reduced, independent of the CD36^-CEBPB signaling (Supplementary Fig. 16C). Notably, CD8⁺ T cells overexpressing caNRF2 exhibited lower levels of lipid peroxidation (Fig. 7B), reduced expression of PD-1, and increased levels of key effector cytokines such as IFNγ and TNFα (Fig. 7C). These findings indicate that transfection of active NRF2 effectively reduces lipid peroxidation and mitigates T cells exhaustion.

Tumor-bearing CD8-cre mice were treated with rAAVCMV-DIO-flag-caNRF2-WPRE-hGH to enable specific overexpression of caNRF2 in tumor-infiltrating CD8⁺ T cells (Supplementary Fig. 16D). Compared with control virus, caNRF2 overexpressing in CD8⁺ T cells resulted in a deceleration in tumor growth, marked decreased lipid peroxidation and increased number of tumor-infiltrating CD8⁺ T cells (Fig. 7D–7F). Additionally, both serum MDA levels and serum iron levels were reduced in mice treated with caNRF2 AAV (Supplementary Fig. 16E, 16F).

After infecting OT-1 CD8⁺ T cells with caNRF2-lentivirus, we adoptively transferred these cells into C57BL/6 mice bearing Hepa1-6-OVA tumors (Supplementary Fig. 17A). CD8⁺ T cells overexpressing caNRF2 exhibited significantly enhanced antitumor activity compared to those transferred with a control virus, resulting in marked tumor growth inhibition (Supplementary Fig. 17B) and decreased serum iron levels (Supplementary Fig. 17C). Collectively, these data indicate that both strategies to activate NRF2 in CD8⁺ T cells could effectively rescue the anti-tumor function of tumor-infiltrating CD8⁺ T cells, eventually resulting in elimination of tumor cells.

After subcutaneous implantation of Hepa1-6 cells in WT and CD36^-/- mice, intraperitoneal injections of IgG or PD-1 antibody were administered, respectively. Compared with the WT mice, CD36^-/- mice or PD-1 antibodies alone, the combination of CD36 knockout and PD-1 antibodies significantly delayed the growth of subcutaneous tumors (Supplementary Fig. 18A, 18B), reduced the levels of lipid peroxides and iron in the serum (Supplementary Fig. 18C, 18D). Particularly, lipid peroxidation and iron accumulation were reduced in tumor-infiltrating CD8⁺ T cells (Supplementary Fig. 18E, 18F), with decreased expression of inhibitory receptors TIM-3, PD-1, TIGIT, and CTLA4 and increased expression of TNFα (Supplementary Fig. 18G– 18K), indicating enhanced antitumor function of CD8⁺ T cells. These results demonstrate that the blockade of CD36 significantly enhances the efficacy of PD-1 therapy. This finding also suggests a potential novel strategy for cancer immunotherapy, potentially offering improved outcomes for patients with tumors.

We utilized HCC tissue microarrays to conduct mIHC to detect the expression of TfR1, CD36, and CD8. The tissue microarray consists of 57 cases of T₁N₀M₀ and 29 cases of T₂N₀M₀ human HCC tissues (5 cases without stage information). We then assessed CD36 and TfR1 in tumor-infiltrating CD8⁺ T cells. We found that in HCC tissues, the proportion of CD8 and CD36 double-positive cells and the proportion of CD8 and TfR1 double-positive cells among CD8⁺ T cells were both significantly higher than those in adjacent non-cancerous tissues (Fig. 8A, 8B). The expression levels of TfR1 in tumor-infiltrating CD36+CD8⁺ T cells were significantly higher than that in CD36^-CD8⁺ T cells, validating the correlation between CD36 and TfR1 within patient-derived tumor environments (Fig. 8C, left). Given that T₁N₀M₀ generally represents early-stage HCC according to the American Joint Committee on Cancer (AJCC) tumor/ node/metastasis (TNM) classification system, we specifically analyzed the relationship between CD36 or TfR1 expression in T₁N₀M₀ HCC tissues and found the same result (Fig. 8C, right). After assessing CD36 and TfR1 in tumor-infiltrating CD8⁺ T cells in HCC tissue microarrays, we investigated the correlation between CD36 and TfR1 levels and overall survival. We categorized the CD8⁺ T cells into different groups according to the positivity percentages of CD36 or TfR1 and analyzed the prognosis of patients. Compared to the CD36^high group, the prognosis of the CD- 36low group was significantly better (Fig. 8D). Additionally, CD36 and TfR1 expression levels didn’t show strong correlations with TNM staging (Supplementary Fig. 19). These findings suggest that the expression of CD36 in CD8⁺ T cells infiltrating HCC tumors may serve as a prognostic indicator in predicting patient outcomes and highlight potential therapeutic targets for improving prognosis in HCC.