Host Metadherin Coordinates Hepatic Lipid metabolism and CD8⁺ T-Cell Immunity to Promote Tumor Progression

SUMMARY

Cancer progression is systemically influenced by distant organ dysfunction induced by primary tumors, yet how long-distance tumor-organ crosstalk regulates antitumor immunity remains unclear. Here, we identify host Metadherin (MTDH) as a critical regulator of tumor-induced immunosuppression and metabolic reprogramming via tumor-liver interactions. Using Mtdh knockout mouse models, we show that concurrent MTDH loss in hepatocytes and CD8⁺ T cells enhances effector T cell function and suppresses tumor growth and metastasis. Mechanistically, tumor-derived extracellular vesicles (EVPs) activate Kupffer cells to secrete TNFα and TGF-β, which suppress hepatic PPARα-mediated lipid oxidation via NF-κB signaling. MTDH loss restores hepatic lipid catabolism, reduces systemic lipid levels, and promotes mitochondrial metabolic reprogramming in CD8⁺ T cells under lipid-reduced conditions, thereby boosting anti-tumor immunity. Genetic or pharmacological targeting of MTDH synergizes with anti-PD-1 therapy. These findings establish host MTDH as a key mediator of tumor-liver crosstalk through metabolic and immune interactions, driving systemic cancer progression.

In brief

Tang et al. identify host MTDH as a key mediator linking tumor-liver crosstalk to antitumor immunity. Targeting MTDH in hepatocytes and T cells enhances CD8⁺ T cell function and improves anti-PD-1 therapy, revealing a host-directed strategy to overcome cancer immunosuppression.

Graphical Abstract

INTRODUCTION

Cancer is increasingly recognized as a systemic disease in which tumor progression is governed by intrinsic genetic alterations and interactions with the host environment¹. Systemic factors, including high-fat diet (HFD)^2–6 and aging^7,8, influence tumor progression by altering host immunity and whole-body metabolism. Beyond the local tumor microenvironment, tumors affect distal organs through secreted factors and vesicles, altering host metabolism, immunity, and therapeutic outcomes^9–13. Thus, tumors exert metastasis-independent effects on distant organs, highlighting the importance of long-range tumor-organ interactions in disease progression.

The liver is a central hub for metabolic and immune regulation¹⁴. Tumor-derived EVPs reprogram hepatic function by activating Kupffer cells to secrete TNFα, inducing steatosis and impairing drug metabolism¹². Extrahepatic tumors also promote hepatic myeloid infiltration, suppress hepatocyte-specific HNF4α, and disrupt lipid metabolism¹¹. Additionally, hepatocytes metabolize fructose into lysophosphatidylcholines (LPCs), which tumor cells take up to support growth¹⁵. These bidirectional tumor-liver interactions underscore the role of systemic metabolic reprogramming in cancer. However, whether tumor-liver crosstalk affects antitumor immunity remains unclear.

CD8⁺ T cells are central to tumor surveillance, but lipid-rich tumor microenvironments often compromise their function^16–19. Lipid overload impairs mitochondrial function and oxidative phosphorylation (OXPHOS), increases reactive oxygen species (ROS), and promotes ferroptosis^20–22. In pancreatic tumors and HFD-fed models, CD8⁺ T cells accumulate long-chain fatty acids and upregulate CD36, driving lipid peroxidation and dysfunction^23,24. Single-cell RNA sequencing (scRNA-seq) also reveals enriched lipid metabolism and exhaustion signatures in tumor-infiltrating CD8⁺ T cells from obese mice³. Paradoxically, HFD-induced lipid stress can enhance PD-1 blockade efficacy²⁵. Thus, lipid metabolism has complex effects on T cell function, but whether tumor-driven hepatic lipid changes alter systemic CD8⁺ T cell immunity remains unknown.

MTDH (also known as AEG-1 or LYRIC) was identified as a metastasis-promoting gene amplified in poor-prognosis breast cancer^26,27. It was later recognized as the third most frequently altered gene in human breast cancer, after PIK3CA and TP53²⁸, and is amplified or overexpressed across many solid tumors^26,29,30. Its tumor-promoting role has been validated using whole-body Mtdh knockout (KO) in genetically modified mouse models of breast²⁷, liver^31,32, prostate³³, lung, and colorectal cancers³⁴. In breast cancer, MTDH supports tumor-initiating cell survival under oncogenic or stress conditions and promotes progression and metastasis through interaction with staphylococcal nuclease domain-containing 1 (SND1)^27,35,36. Disruption of the MTDH-SND1 complex with C26-A6 suppresses breast tumor progression and improves therapeutic efficacy^37,38. MTDH also regulates hepatic lipid metabolism by suppressing PPARα-mediated lipid oxidation and enhancing NF-κB-driven inflammation, thereby promoting metabolic dysfunction-associated steatohepatitis^39–42. Despite its broad expression, MTDH loss causes no significant defects in embryonic development or postnatal growth^27,39, supporting safe therapeutic targeting of MTDH. However, the role of host MTDH in immune and metabolic regulation during systemic tumor progression remains unknown.

Here, we identify host MTDH as a central regulator of tumor-liver-immune crosstalk. MTDH integrates hepatic lipid metabolism with CD8⁺ T cell immunity to promote tumor growth, and targeting it offers a metabolic strategy to overcome immunotherapy resistance.

RESULTS

Host MTDH promotes tumor progression and lung metastasis

To assess host MTDH expression, we analyzed the Human Protein Atlas⁴³ and two independent human breast cancer scRNA-seq datasets^44,45 and found that MTDH is widely expressed in normal tissues, tumor-associated immune and stromal cells (Figures S1A-S1C). Similarly, Mtdh is broadly expressed in normal mouse cells⁴⁶ and tumor-infiltrating immune cells from mouse mammary tumors (Figures S1D-S1G).

To investigate the role of host MTDH in tumor progression, we used our previously established whole-body Mtdh KO mice²⁷ as recipients for multiple syngeneic tumor models. Injection of Py8119 or EO771 mammary tumor cells into wild-type (WT, +/+), heterozygous (HET, +/-), or KO (-/-) mice showed significantly slower tumor growth in Mtdh KO mice (Figures 1A–1B and S1H-S1I). Similarly, host MTDH loss markedly suppressed tumor growth in the PyMT-FVB model²⁷ (Figures 1C and S1J) and 4T1 mammary tumor model (Figure 1D). Moreover, in both colorectal cancer (MC38) and Lewis lung carcinoma (LLC1) models, Mtdh KO mice exhibited reduced tumor growth (Figures 1E–1F and S1K) and prolonged survival compared with WT mice (Figure S1L). Together, these findings across multiple tumor models and genetic backgrounds demonstrate that host MTDH promotes tumor progression.

Figure 1. Host MTDH promotes tumor progression and lung metastasis.

(A-F) Primary growth of Py8119 (A), EO771 (B), PyMT-FVB (C), 4T1 (D), MC38 (E) and LLC1 (F) tumors. n=7, 6, and 6 mice in (A); n=6, 5, and 6 mice in (B); n=5 mice in (C). n=8 mice in (D); n=10 mice in (E, F).

(G-H) Representative lung images and quantification of metastatic nodules from Py8119 (G) and 4T1 MFP tumors (H). Arrows denote metastatic nodules. Scale bar, 1mm. n=7, 6, and 6 in (G); n=8 in (H).

(I) Survival of mice after tail vein injection with Py8119 cells. n=11,14 and 10.

(J) Representative lung images and quantification of metastatic nodules from MC38 tail-vein metastasis. Scale bar, 1mm. n=6, 7 and 8.

(K) Primary growth of Py8119 tumors. n=6,4 and 5 mice.

Data are shown as mean ± SEM. Two-way ANOVA (A-F, K), one-way ANOVA (G, J), two-tailed unpaired Student’s t test (H) and log-rank (Mantel-Cox) test (I).

Given that tumor-intrinsic MTDH promotes breast cancer lung metastasis^26,37, we examined whether host MTDH also regulates this process. In both Py8119 and 4T1 tumors, Mtdh KO mice developed fewer spontaneous lung metastases than WT or HET mice (Figures 1G–1H). To exclude the effect of reduced primary tumor burden, we performed tail-vein injection of Py8119 cells to establish an experimental lung metastasis model. Weekly bioluminescent imaging (BLI) showed minimal lung metastatic burden in Mtdh KO mice (Figure S1M), which was accompanied by prolonged survival (Figure 1I). Host MTDH loss also reduced lung metastasis in the MC38 model (Figure 1J). Together, these findings demonstrate that host MTDH promotes lung metastasis across tumor models.

To determine whether the MTDH-SND1 interaction is required for the tumor-promoting function of host MTDH, we generated MTDH-mutant mice (Figure S1N) carrying W391A/W398A mutations that disrupt MTDH-SND1 binding^27,35. Although the MTDH-SND1 interaction is essential for the tumor-promoting function of MTDH in tumor cells^27,37,38, disrupting this interaction in host tissues had no effect on tumor growth (Figure 1K). These results suggest that host MTDH promotes tumor progression through an SND1-independent mechanism, underscoring a distinct role for MTDH in host tissues compared with its tumor-intrinsic function.

Host MTDH suppresses CD8⁺ T cell-mediated antitumor immunity

To explore how host MTDH promotes tumor progression and lung metastasis, we first investigated its impact on the immune system. scRNA-seq of CD45⁺ immune cells isolated from Py8119 tumors in WT and Mtdh KO mice identified eight clusters, corresponding to T cells, B cells, M1/M2 macrophages, dendritic cells, neutrophils, natural killer (NK) cells, and tumor cells (Figures S2A-S2B). The most prominent changes in Mtdh KO mice were increased T cells, B cells and M1 macrophages, together with reduced M2 macrophages (Figures 2A–2B). To identify which immune population mediates host MTDH-dependent tumor promotion, we utilized Rag1 KO mice, which lack functional T cells and B cells⁴⁷. Rag1 KO largely abrogated the tumor suppression observed in Mtdh KO mice in both Py8119 (Figures 2C–2D) and MC38 models (Figures S2C-S2D). Moreover, MTDH loss no longer altered the M1/M2 macrophage ratio in Rag1 KO mice (Figure S2E). These results indicate that adaptive immunity is essential for the tumor-promoting function of host MTDH.

Figure 2. Host MTDH suppresses CD8⁺ T cell-mediated antitumor immunity.

(A-B) UMAP of tumor-infiltrating CD45⁺ cells (A) and quantification of major cell subsets in WT and Mtdh KO mice (B).

(C-D) Primary Py8119 tumor growth (C) and final tumor mass (D). n=8,9 and 6 mice in (C); n=10,12 and 10 tumors in (D).

(E-F) Primary Py8119 tumor growth (E) and final tumor mass (F) after IgG or anti-CD8 treatment. WT, n=4 mice; KO, n=5 mice.

(G) Survival in the Py8119 lung metastasis model. n=9 mice.

(H-I) UMAP (H) and quantitation (I) of intratumoral T cell subsets from WT or Mtdh KO tumors.

(J) Top enriched pathways in tumor-infiltrating CD8⁺ T cells from Mtdh KO versus WT mice, with NES and p values shown.

(K) Frequency of CD8⁺ T cells, IFNγ and granzyme B MFI, and PD1⁺ frequency among CD8⁺ T cells from Py8119 tumors. n=7–10.

Data are shown as mean ± SEM. Two-way ANOVA (C, E), one-way ANOVA (D, F), log-rank (Mantel-Cox) test (G), and two-tailed unpaired Student’s t test (K).

We next depleted CD8⁺ T cells, CD4⁺ T cells, B cells, and macrophages using depleting antibodies (Figure S2F). In the Py8119 model, CD8⁺ T cell depletion largely eliminated the tumor growth difference between Mtdh KO and WT mice (Figures 2E–2F), whereas CD4⁺ T cell depletion had only a minor effect and macrophage or B cell depletion had no effect (Figures S2G-S2I). This was further validated in the MC38 model, where CD8⁺ T cell depletion abolished the tumor growth difference between WT and Mtdh KO mice (Figures S2J-S2K). Moreover, CD8⁺ T cell depletion also markedly reduced the survival benefit of MTDH deficiency in the experimental lung metastasis model (Figure 2G). Collectively, these data establish CD8⁺ T cells as the primary mediator of the host MTDH-mediated tumor promotion.

To explore how host MTDH regulates CD8⁺ T cell-mediated antitumor immunity, we conducted subclustering analysis of tumor-infiltrating T cells. Tumors from Mtdh KO mice showed a substantial increase in memory (cluster 2) and effector (cluster 4) CD8⁺ T cell populations (Figures 2H–2I and S3A-S3B). Bulk RNA sequencing of sorted tumor-infiltrating CD8⁺ T cells further revealed enrichment of antitumor immune pathways, particularly interferon signaling, in Mtdh KO CD8⁺ T cells (Figure 2J), together with increased expression of cytotoxicity markers, including Prf1, Ifng, Gzmb, and Gzmd (Figure S3C). Flow cytometry further showed increased infiltration of total T cells, CD4⁺ T cells and CD8⁺ T cells in Mtdh KO mice compared with WT mice (Figures 2K and S3D). CD8⁺ T cells from Mtdh KO mice exhibited enhanced cytotoxicity, with increased IFNγ and granzyme B production (Figures 2K and S3E), whereas IFNγ production by CD4⁺ T cells was unchanged (Figure S3D). Reduced PD-1 expression in both CD8⁺ and CD4⁺ T cells from Mtdh KO mice indicated reduced T cell exhaustion (Figures 2K and S3D). Mtdh KO mice also displayed increased systemic CD8⁺ T cell function, with elevated splenic IFNγ and granzyme B expression (Figure S3F). Consistently, both spontaneous (Figure S3G) and experimental (Figure S3H) lung metastases from Mtdh KO mice contained more infiltrating CD8⁺ T cells and fewer exhausted T cells. Histological analysis further revealed increased CD8⁺ T cell infiltration, reduced tumor cell apoptosis and proliferation in tumors from Mtdh KO mice (Figures S3I-S3K). In summary, host MTDH promotes tumor progression primarily by suppressing CD8⁺ T cell effector function.

MTDH deficiency in CD8⁺ T cells is necessary but insufficient for tumor suppression

Given the critical role of host MTDH in suppressing CD8⁺ T cell-mediated antitumor immunity, we hypothesized that MTDH expression in CD8⁺ T cells restrains their effector function and promotes tumor progression. Notably, MTDH loss did not significantly affect T cell development, maturation, or peripheral homeostasis in tumor-free mice (Figures S4A-S4F), suggesting that it is dispensable for baseline T cell homeostasis. Surprisingly, T cell-specific Mtdh deletion did not affect tumor growth in both Py8119 and MC38 models (Figures 3A and S4G-S4H), suggesting that MTDH expression in other immune cells may regulate CD8⁺ T cell-mediated antitumor immunity. To test this possibility, we crossed Mtdh-flox mice with Vav1-Cre mice to delete Mtdh throughout the hematopoietic system. Similar to the T cell-specific KO, loss of MTDH in all hematopoietic cells did not affect Py8119 or MC38 tumor growth (Figures 3B and S4I-S4J). These results indicate that MTDH deficiency in either T cells or the hematopoietic compartment alone is insufficient to suppress tumor progression, implicating a contribution from non-hematopoietic cells.

Figure 3. MTDH deficiency in CD8⁺ T cells is necessary but insufficient for tumor suppression.

(A-B) Primary growth of Py8119 tumors. n=9 and 8 mice in (A); n=6 mice in (B).

(C-E) Schematic of the bone marrow transplantation model (C). Primary growth (D) and final tumor volumes (E) of Py8119 tumors. n=8,8,7 and 6 mice.

(F) Flow cytometry of tumor-infiltrating CD8⁺ T cells and PD1⁺Tim3⁺ subsets in tumors from (E). n=8,9,8 and 7 mice.

(G-H) UMAP (G) and quantification (H) of intratumoral T cell subsets in bone marrow-transplanted mice, normalized to total CD45⁺ cells.

(I-J) Schematic of the CD8⁺ T cells transfer model and tumor volumes at week 3. n=6,6,8 and 8 tumors.

Data are shown as mean ± SEM. Two-way ANOVA (A, B, D), and one-way ANOVA (E, F, J).

To determine whether MTDH expression in non-immune cells influences tumor progression, we performed bone marrow transplantation to separate the roles of MTDH in immune and non-immune compartments (Figure 3C). Only Mtdh KO recipients transplanted with Mtdh KO bone marrow (KK mice) showed slower tumor growth in both Py8119 and MC38 tumor models than the other groups (Figures 3D–3E and S4K-S4L), indicating that MTDH loss in both immune and non-immune cells is required for tumor suppression. Consistently, only KK tumors showed increased CD8⁺ T cell infiltration and reduced T cell exhaustion (Figure 3F). scRNA-seq of Py8119 tumors from the four bone marrow chimera groups further revealed changes in T cells and macrophages only in KK mice (Figures S4M-S4N), resembling those observed in whole-body Mtdh KO mice (Figures 2A–2B). T cell subclustering further showed enrichment of CD8⁺ T cells, particularly effector CD8⁺ T cells, in KK tumors (Figures 3G–3H).

To directly assess the requirement for MTDH deficiency in CD8⁺ T cells, we performed adoptive CD8⁺ T cell transfer into Rag1 KO mice (Figure 3I). Transfer of Mtdh KO CD8⁺ T cells into Mtdh/Rag1 double-KO recipients (KK-R mice) significantly reduced tumor burden (Figure 3J). Tumor-infiltrating CD8⁺ T cells from KK-R mice also exhibited the greatest cytotoxicity and the lowest exhaustion among all groups (Figure S4O). Together, these findings demonstrate that MTDH deficiency in CD8⁺ T cells is necessary but insufficient for tumor suppression, which requires coordinated MTDH loss in both CD8⁺ T cells and non-hematopoietic host cells.

Hepatic MTDH reprograms liver lipid metabolism through PPARα activation during tumor progression

Mtdh KO mice exhibit distinct metabolic traits, including a leaner body composition and resistance to HFD-induced weight gain³⁹. Given that MTDH expression in non-immune cells is required for tumor progression, we hypothesized that MTDH expression in non-immune cells regulates systemic metabolism, thereby influencing CD8⁺ T cell-mediated antitumor immunity. Metabolomic analysis revealed only modest metabolic differences between tumor-free WT and Mtdh KO mice (Figure S5A). In contrast, tumors and serum from Py8119 tumor-bearing Mtdh KO mice showed broad lipid reductions, including lower circulating lipid levels compared with WT mice (Figures 4A–4B and S5B-S5E). Bone marrow chimera experiments further demonstrated that these lipid changes were determined primarily by the recipient genotype, indicating that tumor-derived factors reprogram systemic lipid metabolism through MTDH-dependent effects in non-immune cells (Figures 4C–4D and S5F).

Figure 4. Hepatic MTDH reprograms liver lipid metabolism through PPARα activation during tumor progression.

(A-B) Heatmap of altered serum lipids (A) and PCA of serum lipidomic profiles (B) from Py8119 tumor-bearing mice. Decreased lipids in Mtdh KO mice were summarized by lipid category. n=6 and 7.

(C-D) Heatmap of altered serum lipids (C) and PCA of serum lipidomic profiles (D) from Py8119 tumor-bearing bone marrow-transplanted mice. n=6,7,7, and 6.

(E) Representative BODIPY staining and quantification of hepatic lipid accumulation in Py8119 tumor-bearing mice. Red points indicate samples shown in representative images. n=6. Scale bar, 20 μm.

(F) GSEA showing enrichment of the ‘PPARα activates gene expression’ pathway in livers from Py8119 tumor-bearing Mtdh KO versus WT mice.

(G) Primary growth of MC38 tumors. n=10,7,6 and 6.

(H-I) Schematic of the bone marrow transplantation model and primary growth of Py8119 tumors. n=7,6,7 and 7 mice.

Data are shown as mean ± SEM. Two-tailed unpaired Student’s t test (E), and two-way ANOVA (G, I).

Given the liver’s central role in systemic lipid metabolism, we examined whether MTDH loss impacts hepatic function. Serum AST and ALT activities were comparable between WT and Mtdh KO mice under both normal and tumor-bearing conditions, indicating no overt liver damage (Figure S5G). However, tumor-bearing Mtdh KO mice exhibited reduced hepatic lipid droplet accumulation (Figure 4E), suggesting that MTDH loss suppresses tumor-induced fatty liver formation¹². Bulk RNA-seq of liver tissue from tumor-bearing mice (without liver metastases) revealed upregulation of lipid metabolic pathways, including PPARα signaling, a key regulator of lipid oxidation⁴⁸, in tumor-bearing Mtdh KO mice (Figures 4F and S5H-S5I), whereas these changes were absent in tumor-free mice (Figures S5J-S5K). These findings suggest that enhanced PPARα-driven lipid oxidation contributes to reduced systemic lipid levels in tumor-bearing Mtdh KO mice (Figure 4A).

To determine whether PPARα activation mediates tumor suppression, we generated Mtdh/Ppara double-KO mice. PPARα deletion largely abolished tumor suppression and reduced IFNγ-producing CD8⁺ T cells observed in Mtdh KO mice (Figures 4G and S6A), indicating that PPARα activation contributes to the antitumor effects of MTDH deficiency.

To assess the role of hepatic MTDH in tumor progression, we generated hepatocyte-specific Mtdh KO mice (Mtdh^Alb-cKO). Although Mtdh^Alb-cKO mice showed reductions of some serum lipids (Figure S6B), tumor growth was unchanged (Figure S6C), indicating that hepatic MTDH loss alone is insufficient for tumor suppression. However, transplantation of Mtdh KO bone marrow into Mtdh^Alb-cKO recipient mice (K-HK mice) significantly suppressed tumor growth compared with the other groups (Figures 4H–4I), accompanied by further reductions in serum lipids (Figure S6D) and enhanced IFNγ and TNFα production by tumor-infiltrating CD8⁺ T cells (Figure S6E). These findings indicate that coordinated MTDH loss in hepatocytes and immune cells is required for tumor suppression.

We also observed reduced GSH/GSSG ratios and cysteine levels in tumors from Mtdh KO mice (Figures S6F-S6G), suggesting enhanced ferroptosis⁴⁹. Consistently, tumor cells from Mtdh KO mice exhibited increased lipid ROS and cellular ROS (Figures S6H-S6I). Because CD8⁺ T cell-derived IFNγ cooperates with lipid metabolism in the tumor microenvironment to promote ACSL4-dependent tumor ferroptosis through STAT1-IRF1 signaling⁵⁰, we examined this pathway and found increased expression of Acsl4, Stat1, and Irf1 in tumors from Mtdh KO mice (Figure S6J). Moreover, Mtdh KO hosts suppressed Acsl4-WT but not Acsl4-KO tumor growth (Figures S6K-S6L), demonstrating that ACSL4 is required for tumor suppression induced by host MTDH deficiency.

Collectively, our findings suggest that host MTDH in the liver (primarily hepatocytes) and immune cells (primarily CD8⁺ T cells) cooperatively promotes tumor progression by regulating systemic lipid metabolism, antitumor immunity, and tumor ferroptosis.

Tumor-derived EVPs reprogram hepatic lipid metabolism through MTDH-mediated NF-κB/PPARα signaling

Building on the established finding that tumor-derived EVPs induce Kupffer cell secretion of TNFα to promote fatty liver formation¹², we asked whether hepatic MTDH modulates this pathway to influence systemic lipid metabolism. Consistent with this hypothesis, conditioned media from Py8119 EVP-treated Kupffer cells induced less lipid droplet accumulation but greater nuclear PPARα levels in Mtdh KO hepatocytes than in WT hepatocytes (Figures 5A–5B and S7A), accompanied by enrichment of PPAR signaling and lipid metabolism pathways (Figures 5C and S7B). These differences were absent without conditioned media (Figures S7C-S7D), indicating that hepatic MTDH functions downstream of tumor EVP-activated Kupffer cell signaling.

Figure 5. Tumor-derived EVPs reprogram hepatic lipid metabolism through MTDH-mediated NF-κB/PPARα signaling.

(A-B) Representative flow cytometry plots and quantification of hepatocyte BODIPY (A) and nuclear PPARα staining (B). n=3.

(C) GSEA showing enrichment of the ‘PPAR signaling pathway’ in treated Mtdh KO versus WT hepatocytes.

(D) GSEA of enriched pathways in livers from Py8119 tumor-bearing Mtdh KO versus WT mice.

(E-H) Representative flow cytometry plots and quantification of hepatocyte BODIPY (E-F) and nuclear PPARα staining (G-H) after treatment with TNFα or TGF-β. n=3.

(I) ELISA of TNFα and TGF-β in conditioned media from HCI-002 PDX-EVP-treated Kupffer cells. PBS as control. n=4.

(J) Representative flow cytometry plots and quantification of BODIPY staining in hepatocytes treated with conditioned media from HCI-002 PDX-EVP-treated Kupffer cell (CM). PBS as control. n=3.

(K) Schematic of in vivo Kupffer depletion.

(L) Representative flow cytometry plots and quantification of hepatocyte BODIPY staining under indicated conditions. n=9.

(M) Model of tumor-EVP-mediated regulation of hepatic lipid metabolism.

(N) Schematic of ASO treatment in mice.

(O) Immunoblot analysis of indicated proteins in primary hepatocytes from ASO-treated male and female mice.

(P) Representative flow cytometry plots and quantification of BODIPY staining in hepatocytes from Mtdh ASO- or Control (Ctrl) ASO-treated mice after treatment with conditioned media from Py8119-EVPs-treated Kupffer cells (CM). PBS as control. n=3.

Data are shown as mean ± SEM. Two-tailed unpaired Student’s t test (A-B, E-I), and one-way ANOVA (J, L, P).

GSEA further showed that TNFα/NF-κB and TGF-β signaling were downregulated in livers from tumor-bearing Mtdh KO mice but not in tumor-free mice (Figures 5D and S7E). Because TNFα is a major driver of tumor-induced fatty liver¹² and TGF-β also regulates hepatic lipid metabolism⁵¹, we examined their effects on hepatocytes (Figure S7A). Both TNFα and TGF-β induced less lipid droplet accumulation and greater nuclear PPARα levels in Mtdh KO than WT hepatocytes (Figures 5E–5H). Consistently, lipid metabolism pathways, including the PPAR signaling pathway, were positively enriched in cytokine-treated Mtdh KO hepatocytes (Figures S7F-S7I). Together, these results identify TGF-β, alongside TNFα, as upstream mediators of MTDH-dependent hepatic lipid metabolism.

Consistent with previous studies showing that Kupffer cells take up tumor-derived EVPs and produce TNFα and TGF-β^12,52, injected Py8119-derived EVPs accumulated predominantly in Kupffer cells (Figures S7J-S7L). Blocking tumor EVP secretion by Rab27a knockdown markedly reduced hepatic lipid droplet accumulation in vivo (Figures S7M-S7N). EVPs from multiple mouse and human tumor sources, including cultured mouse tumor cells, human malignant cells, tumor-bearing mouse serum, tumor explants, and patient-derived xenograft (PDX) tumor⁵³, consistently increased TNFα and TGF-β production in Kupffer cells (Figures 5I and S8A-S8F). In addition, MTDH deficiency in Kupffer cells did not affect cytokine induction (Figure S8G). Consistent with a previous report that palmitic acid in tumor-EVPs induces TNFα secretion from Kupffer cells and promotes fatty liver formation¹², we confirmed that palmitic acid indeed stimulated TNFα and TGF-β production (Figure S8H), and tumor-bearing mice exhibited elevated circulating TNFα and TGF-β levels (Figure S8I). Moreover, conditioned media from PDX-EVP-treated Kupffer cells induced robust lipid droplet accumulation in WT but less in Mtdh KO hepatocytes (Figures 5J and S8J).

To determine whether Kupffer cells mediate hepatic lipid accumulation in vivo, we depleted Kupffer cells using clodronate liposomes (Figure 5K). Kupffer cell depletion (Figures S8K-S8L) significantly reduced hepatic lipid droplet accumulation in WT tumor-bearing mice but had no additional effect in Mtdh KO mice which already have low lipid droplet levels (Figure 5L). Similarly, in vivo neutralization of TNFα and TGF-β (Figure S8M) reduced hepatic lipid droplets in WT but not further in Mtdh KO mice (Figure S8N). Collectively, these findings demonstrate that tumor-derived EVPs promote lipid droplet accumulation through Kupffer cell-derived TNFα and TGF-β, whereas hepatic MTDH deficiency disrupts this pathway.

Both TNFα and TGF-β activate NF-κB signaling in hepatocytes^54,55, whereas MTDH deficiency attenuates TNFα-induced NF-κB activation⁴². Because NF-κB activation suppresses hepatic PPARα expression and activity⁵⁶, we examined whether tumor-derived EVPs promote hepatic lipid accumulation through a Kupffer cell–TNFα/TGF-β–NF-κB/PPARα axis (Figure 5M). Consistent with this model, TNFα, TGF-β, or conditioned media from tumor EVP-treated Kupffer cells induced weaker NF-κB activation in Mtdh KO than WT hepatocytes (Figures S8O-S8S). Moreover, pharmacological inhibition of NF-κB with BAY 11–7082 or activation of PPARα with GW7647 abolished the difference in lipid droplet accumulation between WT and Mtdh KO hepatocytes after TNFα or TGF-β stimulation (Figures S8T-S8U). To selectively target hepatic MTDH, we generated a GalNAc-conjugated⁵⁷ Mtdh antisense oligonucleotide (ASO)³⁴, which efficiently reduced MTDH expression in hepatocytes (Figures 5N–5O). GalNAc-Mtdh ASO also reduced lipid droplet accumulation in hepatocytes exposed to conditioned media from tumor EVP-treated Kupffer cells (Figure 5P).

Together, these findings extend the previously established mechanism in which tumor-derived EVPs activate Kupffer cells to induce hepatic steatosis by identifying hepatic MTDH as the key signaling node that couples Kupffer cell-derived TNFα/TGF-β to NF-κB/PPARα signaling. Loss of hepatic MTDH weakens NF-κB activation, sustains PPARα activity, lowers systemic lipid (Figure 5M), and cooperates with MTDH-deficient CD8⁺ T cells to enhance antitumor immunity.

Loss of MTDH enhances mitochondrial activity and metabolic fitness in tumor-infiltrating CD8⁺ T cells

Lipid metabolism is crucial for regulating CD8⁺ T cell-mediated antitumor immunity^19,58. Given that host MTDH influences tumor progression through CD8⁺ T cells and hepatic lipid metabolism, we hypothesized that MTDH deficiency enhances CD8⁺ T cell function through metabolic reprogramming. Although individual CD8⁺ T cell subsets had few cells, scRNA-seq pathway analysis consistently revealed altered metabolic activity across multiple CD8⁺ T cell subsets in Mtdh KO tumors (Figure S9A). Seahorse analysis showed that Mtdh KO tumor-infiltrating CD8⁺ T cells exhibited increased basal and maximal oxygen consumption rates (OCR), indicating enhanced mitochondrial OXPHOS (Figures 6A–6B and S9B). Consistently, these cells displayed increased mitochondrial mass (Figure 6C). SCENITH⁵⁹ analysis further showed increased mitochondrial activity, with higher basal metabolic activity and mitochondrial dependence in Mtdh KO CD8⁺ T cells (Figures 6D–6F and S9C). Mtdh KO CD8⁺ T cells also showed increased glucose uptake and extracellular acidification rate (ECAR) (Figures S9D-S9G), indicating enhanced glycolysis. Similar metabolic changes were observed in tumor-infiltrating CD4⁺ T cells, but not tumor-associated macrophages (Figures S9H-S9K). In contrast, pre-activated splenic CD8⁺ T cells from tumor-free mice showed no differences in OCR or ECAR between Mtdh KO and HET mice (Figures S9L-S9M), indicating that these metabolic changes are restricted to the tumor microenvironment.

Figure 6. Loss of MTDH enhances mitochondrial metabolic fitness in tumor-infiltrating CD8⁺ T cells.

(A-B) Seahorse analysis of tumor-infiltrating CD8⁺ T cells showing OCR over time (A) and basal and maximal OCR (B). O, oligomycin; F, FCCP; R, Rotenone/antimycin A. n=6. Representative of two independent experiments.

(C) Flow cytometry analysis of MitoFM deep red staining in tumor-infiltrating CD8⁺ T cells. n=6.

(D-F) SCENITH analysis of puromycin MFI in tumor-infiltrating CD8⁺ T cells under the indicated conditions (D), with basal translation (E) and mitochondrial dependence (F) calculated from puromycin MFI. Ct, control; DG, 2-Deoxy-Glucose; O, Oligomycin; DGO, 2-Deoxy-Glucose plus Oligomycin. n=7.

(G) Volcano plot of OXPHOS-related differentially expressed proteins in Mtdh KO versus Het tumor-infiltrating CD8⁺ T cells. Red and blue indicate upregulated and downregulated proteins in Mtdh KO cells, respectively. n=3.

(H-N) Metabolic profiling of tumor-infiltrating CD8⁺ T cells from bone marrow transplanted mice, including puromycin MFI (H), basal translation (I), mitochondrial dependence (J), MitoFM deep red staining (K), CD36 expression (L), lipid peroxidation (M), and cell death (N). n=5 in (H-J); n=6 in (K-N).

(O) Flow cytometry analysis of p-STAT3 in tumor-infiltrating CD8⁺ T cells. n=7.

(P) GSEA showing enrichment of pSTAT3-binding peaks within the union set of ATAC-seq peaks from WT and Mtdh KO tumor-infiltrating CD8⁺ T cells.

(Q) Relative tumor cell survival after co-culture with tumor-infiltrating or splenic CD8⁺ T cells from tumor bearing mice, with or without STAT3 inhibitor Stattic. n=3–6.

Data are shown as mean ± SEM. Two-tailed unpaired Student’s t test (B, C, E, F, O) and one-way ANOVA (I-N, Q).

To further explore how MTDH affects mitochondrial activity in CD8⁺ T cells, we performed proteomic analysis of tumor-infiltrating CD8⁺ T cells and splenic CD8⁺ T cells from tumor-free mice. Mtdh KO tumor-infiltrating CD8⁺ T cells exhibited increased expression of OXPHOS- and mitochondrial function-related proteins (Figures 6G and S9N), consistent with enrichment of OXPHOS gene signatures in effector CD8⁺ T cells (Figure S9O). Tfam, a key regulator of mitochondrial OXPHOS⁶⁰, and mitochondrial fusion protein Opa1⁶¹ were upregulated, whereas the mitochondrial fission regulator Drp1⁶² was downregulated in Mtdh KO tumor-infiltrating CD8⁺ T cells (Figure S9P-S9Q). In contrast, these mitochondrial changes were absent in splenic CD8⁺ T cells from tumor-free mice (Figure S9R), indicating that they are specific to the tumor microenvironment. Collectively, these data suggest that host MTDH deficiency enhances mitochondrial activity and metabolic fitness in tumor-infiltrating CD8⁺ T cells to support antitumor immunity.

Because effective antitumor immunity requires MTDH deficiency in both hepatocytes and CD8⁺ T cells, we next dissected their respective contributions to the metabolic fitness of tumor-infiltrating CD8⁺ T cells. Analysis of bone marrow chimeras showed that CD8⁺ T cells from the KK group exhibited the highest basal metabolic activity, mitochondrial dependence, and mitochondrial mass (Figures 6H–6K), indicating maximal mitochondrial fitness.

Given that excess lipids impair CD8⁺ T cell function by promoting lipid uptake, peroxidation, and ferroptosis^20,22, we next examined lipid metabolism in tumor-infiltrating CD8⁺ T cells. Proteomic analysis revealed reduced ferroptosis- and ROS-associated proteins in Mtdh KO CD8⁺ T cells (Figure S9S). Consistently, CD36 expression, lipid peroxidation, and cell death were lowest in CD8⁺ T cells from the KK group (Figures 6L–6N). Similar reductions were observed in Mtdh KO tumor-infiltrating CD8⁺ T cells compared with controls (Figures S9T-S9W), but not in splenic CD8⁺ T cells from tumor-free mice (Figures S9X-S9Y), suggesting that these protective effects are specific to the tumor microenvironment. In addition, tumor MHC I expression was highest in the KK group (Figures S9Z-S9AA).

To determine whether altered systemic lipid availability or T cell lipid uptake contributes to the enhanced antitumor effects of MTDH deficiency, we first increased systemic lipid availability by short-term HFD feeding (Figure S10A). HFD significantly attenuated the tumor suppression conferred by whole-body MTDH loss (Figure S10B). In addition, activated Mtdh KO CD8⁺ T cells cultured in lipid-depleted medium for 12 hours exhibited higher basal and maximal OCR than WT controls, whereas no differences were observed under lipid-replete conditions (Figures S10C-S10D). Because CD36 was downregulated in Mtdh KO tumor-infiltrating CD8⁺ T cells (Figures 6L and S9T-S9U), we overexpressed CD36 to determine whether enhanced lipid uptake contributes to tumor suppression (Figures S10E-S10F). However, CD36 overexpression did not affect tumor growth, CD8⁺ T cell infiltration, or effector cytokine production in either WT or Mtdh KO mice (Figures S10G-S10J). Collectively, these data suggest that a low-lipid systemic environment, rather than altered CD36-mediated lipid uptake, is required for the enhanced antitumor activity of MTDH-deficient CD8⁺ T cells.

To further explore the molecular basis of this metabolic reprogramming, we performed pathway enrichment analysis and identified enhanced STAT3 signaling in MTDH-deficient CD8⁺ T cells, including increased STAT3 pathway signatures (Figures S9S and S10K), elevated STAT3 transcript and protein expression (Figure S10L), and increased STAT3 phosphorylation (Figure 6O). Moreover, STAT3 phosphorylation was highest in the KK group (Figure S10M), suggesting that maximal STAT3 activation is associated with MTDH deficiency in both hepatocytes and CD8⁺ T cells. These findings are consistent with previous studies showing that STAT3 can promote OXPHOS, mitochondrial fitness and antitumor immunity^21,63, and suggest that STAT3 may contribute to the metabolic adaptation of MTDH-deficient CD8⁺ T cells.

Because STAT3 regulates chromatin accessibility in tumor-infiltrating CD8⁺ T cells⁶³, we performed ATAC-seq and integrated the data with bulk RNA-seq. MTDH deficiency increased chromatin accessibility at 4,437 open chromatin regions (OCRs) and decreased accessibility at 2,734 OCRs, predominantly in intergenic and intronic regions (Figure S10N). Integration with published STAT3 ChIP-seq datasets⁶³ further revealed increased chromatin accessibility at phosphorylated STAT3-binding sites in Mtdh KO CD8⁺ T cells (Figures 6P and S10O-S10P). Furthermore, motif enrichment analysis of gained OCRs showed enrichment for AP-1, ETS, and RUNX motifs (Figure S10Q), consistent with STAT3-associated chromatin remodeling⁶³. Finally, pharmacological inhibition of STAT3 with Stattic²¹ significantly reduced Mtdh KO CD8⁺ T cell cytotoxicity (Figures 6Q and S10R), demonstrating that enhanced STAT3 activity is required for their improved effector function. Together, these findings support a model in which STAT3 activity contributes to the enhanced metabolic adaptation and effector function of MTDH-deficient CD8⁺ T cells; however, the molecular link between MTDH and STAT3 requires further investigation.

Collectively, these findings suggest that host MTDH deficiency enhances CD8⁺ T cell effector function through systemic lipid remodeling and mitochondrial metabolic reprogramming. This effect requires coordinated MTDH deficiency in both hepatocytes and CD8⁺ T cells to maximize antitumor immunity.

MTDH deficiency synergizes with anti-PD-1 therapy to enhance antitumor immunity

Given that host MTDH deficiency enhances CD8⁺ T cell-mediated antitumor immunity, we used four tumor models to test whether targeting host MTDH could improve PD-1 blockade efficacy (Figure 7A). Because metastatic breast cancer responds poorly to immune checkpoint blockade⁶⁴, we first evaluated the metastatic Py8119 mammary tumors. Consistent with our previous findings³⁷, anti-PD-1 alone had minimal effects in WT mice (Figures 7B and S11A). In contrast, anti-PD-1 further suppressed tumor growth in Mtdh KO mice (Figures 7B and S11A), prolonged tumor-free survival (Figure 7C), and increased intratumoral CD8⁺ and CD4⁺ T cell infiltration (Figure S11B). We next evaluated the Py8119 lung metastasis model, which models advanced breast cancer resistant to immunotherapy⁶⁴. The combination of Mtdh KO and anti-PD-1 significantly prolonged survival, with three of ten mice remaining tumor-free for more than five months (Figure 7D), demonstrating durable therapeutic benefit.

Figure 7. MTDH deficiency synergizes with anti-PD-1 therapy to enhance antitumor immunity.

(A) Schematic of the anti-PD-1 treatment experiments.

(B-C) Primary growth of Py8119 tumors (B) and tumor-free survival from two independent experiments (C). n=6,5,5 and 6 mice in (B); n=12,12,10, and 12 mice in (C).

(D) Survival in the Py8119 lung metastasis model. n=10.

(E-F) Progression-free survival of MC38 tumors (E) and pie charts showing treatment responses in each group (F). n=10.

(G) Schematic of ASO and T cell transfer treatment in the MC38-Ova model.

(H-J) Individual tumor growth curves (H), final tumor volumes (I), and maximal tumor reduction ratio (J) after combined treatment. Numbers in (H) indicate the fraction of large and small tumors based on the cutoff line. In (J), the maximal tumor reduction ratio was calculated as the minimum tumor volume after combined treatment divided by the pretreatment tumor volume and is shown for individual tumors (left) and by group (right). Control mice received PBS. n=10, 22, and 22 tumors.

Data are shown as mean ± SEM. Two-way ANOVA (B), log-rank (Mantel-Cox) test (C, D, E), and two-tailed unpaired Student’s t test (I, J).

We further examined this combination in the immunotherapy-sensitive MC38 model⁶⁵. Mtdh KO alone produced antitumor effects comparable to those of anti-PD-1 alone, as reflected by tumor growth (Figures S11C-S11D), progression-free survival (Figure 7E), and objective response rates (Figure 7F). Importantly, combining Mtdh KO with anti-PD-1 induced complete tumor regression in all mice (Figures S11C-S11D). All cured mice rejected tumor rechallenge (Figure S11E), indicating durable immunological memory. Similarly, in the MC38 lung metastasis model, the combination nearly eradicated lung metastases, with seven of nine mice showing no detectable lung metastatic nodules (Figure S11F).

To further validate that coordinated Mtdh deficiency in hepatocytes and CD8⁺ T cells is required for tumor suppression, we combined hepatocyte-targeted Mtdh ASO with adoptive transfer of Mtdh KO OT-I cells in the setting of anti-PD-1 therapy (Figures 7G and S11G). Using the antigen-specific MC38-Ova model, the dual Mtdh-targeting group (GalNAc-Mtdh ASO + Mtdh KO OT-I cells) exhibited enhanced anti-PD-1 efficacy compared with the control combination group (Control ASO + WT OT-I cells), as reflected by delayed tumor growth (Figure 7H), reduced tumor size (Figure 7I), greater maximum tumor regression (Figure 7J), and increased effector cytokine production by tumor-infiltrating CD8⁺ T cells (Figure S11H). Similar therapeutic benefits were observed in the Py8119-Ova mammary tumor model (Figures S11I–S11K).

Together, these results demonstrate that genetic or pharmacological targeting of host MTDH enhances the efficacy of anti-PD-1 immunotherapy across multiple tumor models by promoting T cell–mediated antitumor immunity, supporting host MTDH as a potential immunotherapy target.

DISCUSSION

Tumors influence distant organs not only through metastasis but also via secreted factors that disrupt physiological functions. EVPs are key mediators of tumor-host communication and can reprogram local and distant tissues¹³. Distant tumor-derived EVPs induce TNFα secretion from Kupffer cells, promoting hepatic steatosis, inflammation, and impaired drug metabolism¹². Although TNFα blockade reduced fatty liver and restored hepatic drug metabolism, it did not suppress primary tumor growth¹². Thus, whether this EVP–Kupffer cell–fatty liver axis affects tumor progression or antitumor immunity remained unclear.

Our study provides a mechanistic and functional extension of this established axis into CD8⁺ T cell-mediated anti-tumor immunity. We identify hepatic MTDH as a key coupling node through which tumor EVP-educated Kupffer cells reprogram systemic lipid metabolism and, consequently, tumor-infiltrating CD8⁺ T cells. Hepatic MTDH deficiency reduced lipid accumulation but did not suppress tumor growth, consistent with the inability of TNFα blockade alone to control tumors¹². Likewise, T cell-specific MTDH loss was insufficient. In contrast, coordinated MTDH deficiency in hepatocytes and CD8⁺ T cells markedly impair tumor progression. These findings establish a dual-compartment requirement in which hepatic MTDH controls the lipid environment, while T cell-intrinsic MTDH determines CD8⁺ T cell adaptation to that environment.

Mechanistically, hepatic MTDH loss limits tumor EVP-induced systemic lipid changes, thereby creating conditions that favor the metabolic fitness of Mtdh-deficient tumor-infiltrating CD8⁺ T cells. These cells display enhanced mitochondrial activity and sustain effector function within the tumor microenvironment. Thus, MTDH links tumor-induced hepatic lipid reprogramming to CD8⁺ T cell dysfunction rather than acting solely as a hepatic metabolic effector.

Tumor-derived EVPs induce Kupffer cells to produce both TNFα and TGF-β^12,52. TNFα promotes fatty liver by targeting hepatocytes¹², whereas TGF-β induces fibronectin production in hepatic stellate cells and supports liver pre-metastatic niche formation⁵². We show that TGF-β, together with TNFα, also suppresses lipid oxidation in hepatocytes. Hepatic MTDH integrates signals from both cytokines to promote lipid accumulation. MTDH loss reduced TNFα- and TGF-β-induced lipid accumulation, whereas Kupffer cell depletion or combined cytokine neutralization eliminated the lipid difference between WT and Mtdh KO hepatocytes. EVPs derived from human breast cancer PDXs similarly induced TNFα and TGF-β production by Kupffer cells and caused less lipid accumulation in Mtdh KO than WT hepatocytes, highlighting the clinical relevance of our findings.

MTDH may therefore represent an attractive immunotherapy target because it promotes tumor progression through distinct mechanisms in tumor and host compartments. In tumor cells, MTDH interacts with SND1 to support stress survival, tumor growth, and metastasis, while reducing MHCI-dependent antigen presentation through degradation of Tap1/2 mRNAs^27,36–38. Disrupting the MTDH/SND1 complex suppresses tumor progression and sensitizes tumors to chemotherapy and immunotherapy^37,38. In contrast, host MTDH acts through SND1-independent mechanisms. In hepatocytes, MTDH promotes TNFα- and TGF-β-mediated NF-κB activation, suppresses PPARα activity, and drives lipid accumulation. In CD8⁺ T cells, MTDH limits mitochondrial activity and metabolic fitness, partly by suppressing STAT3 activity. Together, these compartment-specific functions establish a coordinated mechanism through which MTDH creates a lipid-rich systemic environment and impairs CD8⁺ T cell effector function.

Why has MTDH, a gene that is non-essential for development and homeostasis, yet broadly promotes tumor progression across diverse cancer types, been retained throughout evolution? This paradox becomes even more striking considering that MTDH functions in both tumor cells and host stromal compartments to support disease progression, yet Mtdh knockout mice display no overt ill effect and even show increased survival under laboratory conditions³⁹. The answer may lie in MTDH’s ancestral role in metabolic adaptation. Lipid storage is a fundamental and evolutionarily conserved mechanism that enables animals to endure cycles of food abundance and scarcity⁶⁶. MTDH, identified as a key regulator of lipid metabolism^39,41, is highly conserved across vertebrates but absent in invertebrates⁶⁷, suggesting its emergence coincided with increasing metabolic complexity. This conservation implies that MTDH plays a beneficial physiological role, likely by facilitating energy storage and metabolic reprogramming under stress conditions. In environments where food scarcity was a constant evolutionary pressure, the ability to store and mobilize lipids efficiently would have conferred a selective advantage. However, in today’s nutrient-rich settings, the same function may become maladaptive, contributing to obesity, fatty liver disease, and, as our findings show, systemic tumor progression. Tumors exploit hepatic MTDH to reprogram lipid metabolism in support of their own growth, revealing a critical metabolic vulnerability. Mtdh KO mice resist HFD-induced MASH^39,41 and exhibit enhanced survival in tumor-bearing models, highlighting how loss of this ancient metabolic gene can be protective in modern pathological contexts. Together, these findings reveal how an evolutionarily advantageous gene may become a liability in the context of cancer and suggest that therapeutically targeting MTDH could exploit this vulnerability to improve outcomes in cancer treatment.

In summary, our study identifies host MTDH as a critical link in tumor-liver-immune crosstalk, revealing a mechanism by which cancer progression is regulated systemically. By coordinating systemic lipid metabolism and CD8⁺ T cell function, MTDH represents a therapeutic target for enhancing immunotherapy and developing metabolism-based cancer treatments. Combining hepatocyte-targeted Mtdh ASO with adoptive transfer of Mtdh KO T cells significantly enhanced anti-PD-1 efficacy, providing a path toward clinical translation.

Limitations of the study

Our study focused on how tumor-induced hepatic lipid metabolism reprogramming promotes tumor progression. However, fatty liver can also facilitate liver metastasis within the hepatic microenvironment⁶⁸. Future studies should determine whether hepatic MTDH promotes liver metastasis through lipid dysregulation, immune modulation, or both. We identify hepatocytes and CD8⁺ T cells as the major mediators of host MTDH function. However, hepatocyte-specific MTDH loss does not fully recapitulate the effects of MTDH loss in non-hematopoietic cells, suggesting that other metabolic tissues, such as adipose tissue, may also contribute. Likewise, CD8⁺ T cell depletion or Rag1 deficiency does not fully restore tumor growth in Mtdh KO mice, and scRNA-seq revealed changes in additional immune populations, indicating that other immune cells may also participate. Our findings indicate that a low-lipid systemic environment, rather than reduced CD36-mediated lipid uptake, is required for the enhanced antitumor activity of Mtdh-deficient CD8⁺ T cells. However, because HFD only partially restores tumor growth, additional metabolic mechanisms are likely involved. Although we identified STAT3-dependent chromatin remodeling and metabolic reprogramming as key downstream events, how MTDH regulates STAT3 activity warrants further investigation. Lastly, hepatic ASO-mediated MTDH inhibition combined with anti-PD-1 showed lower efficacy than genetic Mtdh deletion. Further optimization of ASO delivery and treatment regimens will be important to improve therapeutic efficacy and clinical translation.

RESOURCE AVAILABILITY

Lead contact

Further information and requests for resources and reagents can be directed to, and will be fulfilled by the lead contact, Yibin Kang (ykang@princeton.edu).

Materials availability

All unique/stable reagents generated in this study are available from the lead contact with a completed materials transfer agreement.

Data and code availability

• Raw data and uncropped blots are provided in Data S1.

• The raw RNA-seq, scRNA-seq, and ATAC-seq data have been deposited in the Gene Expression Omnibus (GEO) under accession number GSE276107.

• The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE⁶⁹ partner repository under dataset identifier PXD064796.

• The metabolomics data have been deposited to Metabolomics Workbench⁷⁰ under study identifiers ST004024 and ST004025.

• This paper does not report original code.

• Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

EXPERIMENTAL MODEL AND STUDY PARTICIPANT DETAILS

Mice

Animals were housed in specific pathogen-free facilities at Princeton University. All animal experiments were conducted in compliance with the Institutional Animal Care and Use Committee (IACUC) of Princeton University. All mice were regularly maintained on standard chow diet (PicoLab Rodent Diet, #5053). Mtdh KO mice with different backcrossed derivatives (C57BL/6J, BALB/c and FVB) and Mtdh-flox mice (C57BL/6J) were generated in our laboratory and were described previously^27,34,37. CD4-Cre (#022071), Albumin-Cre (#003574), VAV1-Cre (#035670), OT-1 (#003831), PPARα KO (#008154), RAG1 KO (#034159) and NSG (#005557) mice were purchased from The Jackson Laboratory. Mice were crossed between different genotypes according to the experimental purpose.

Cell lines

Py8119 (CRL-3278) cells were obtained from Dr. Weizhou Zhang’s laboratory. E0771 (CRL-3461) cells were obtained from Dr. Xiang Zhang’s laboratory. PyMT-FVB²⁷ and SUM159-M1a⁸⁸ cells were generated in our previous study. Acsl4-WT and Acsl4-KO MC38 cells were obtained from Dr. Weiping Zou’s laboratory. MC38 cells were purchased from Kerafast. LLC1 (CRL-1642) and 4T1 (CRL-2539) cells were purchased from the American Type Culture Collection (ATCC). Platinum-E cells (NC0066908) were purchased from Fisher Scientific. Py8119-Ova and MC38-Ova cells were generated in our lab and was described before³⁷. PyMT-FVB, MCF7, 4T1, HEK293T and Platinum-E cells were cultured in DMEM containing 10% FBS, 2 mM glutamine, 100 U penicillin and 0.1 mg/ml streptomycin. Py8119 and Py8119-Ova cells were cultured in DMEM/F12 (1:1) containing 10% FBS, 20 ng/ml epidermal growth factor, 5 μg/ml insulin, 2 μg/ml hydrocortisone,100 U penicillin and 0.1 mg/ml streptomycin. MC38 and MC38-Ova cells were cultured with DMEM media supplemented with 10% FBS, 2 mM glutamine, 0.1 mM nonessential amino acids, 1 mM sodium pyruvate, 10 mM Hepes, 50 mg/mL gentamycin sulfate, 100 U penicillin and 0.1 mg/ml streptomycin. EO771, Acsl4-WT and Acsl4-KO MC38 cells were cultured in RPMI-1640 containing 10% FBS, 10 mM Hepes, 100 U penicillin and 0.1 mg/ml streptomycin. SUM159-M1a cells were cultured with F12 medium supplemented with 10% FBS, 20 ng/ml epidermal growth factor, 10 μg/ml insulin, 2 μg/ml hydrocortisone,100 U penicillin and 0.1 mg/ml streptomycin. MCF10A and MCF10AT-cA1a were cultured with DMEM/F12 (1:1) containing 5% horse serum, 20 ng/ml epidermal growth factor, 10 μg/ml insulin, 1 μg/ml hydrocortisone,100 U penicillin and 0.1 mg/ml streptomycin. All cells were regularly checked for Mycoplasma by PCR.

METHOD DETAILS

Mouse CD36 retroviral vector construction

Total RNA was extracted from mouse bone marrow-derived macrophages and reverse transcribed into cDNA using the SuperScript^™ IV First-Strand Synthesis System (Invitrogen, 18091050). Mouse Cd36 cDNA was amplified by PCR using gene-specific primers containing EcoRI and XhoI restriction sites in the forward and reverse primers, respectively. The amplified Cd36 cDNA fragment was subsequently cloned into the retroviral expression vector pMSCV-IRES-GFP II (Addgene, #52107) using EcoRI and XhoI restriction sites. Successful cloning was confirmed by Sanger sequencing.

Forward primer: GTCAGAATTCacgaggagaatgggctgtg

Reverse primer: CTCACTCGaggctcatccactacttattttcc

Retrovirus and lentivirus preparation

For virus production, Platinum-E cells for retrovirus or HEK293T cells for lentivirus were seeded one day prior to transfection. Transfection was performed using Lipofectamine^™ 3000 Transfection Reagent (Thermo Fisher, L3000015) according to the manufacturer’s instructions. For retrovirus production, 2 μg of pMSCV-IRES-GFP II vector and 1 μg pCL-Eco packaging plasmid (Addgene, #12371) were transfected per well of a 6-well plate. For shRNA lentivirus production, 1 μg shRNA vector, 2 μg psPAX2 (Addgene, #12260) and 0.8 μg pMD2.G (Addgene, #12259) were transfected per well of a 6-well plate. Viral supernatants were collected 48 h and 72 h after transfection. The virus was concentrated using Lenti-X^™ Concentrator (Takara, 631231) according to the manufacturer’s protocol. Concentrated virus was resuspended in PBS, aliquoted, and stored at -80°C until use.

Retrovirus transduction of primary CD8⁺ T cells

Primary splenic CD8⁺ T cells from WT OT-1 or Mtdh KO OT-1 mice were isolated using a CD8a⁺ T Cell Isolation Kit (Miltenyi Biotec, #130–104-075). 500 μl of CD8⁺ T cells in the presence of 10 ng/ml IL-2 (PeproTech, 212–12) were seeded at a concentration of 1×10⁶ cells/ml in 48-well plates pre-coated with anti-CD3 and anti-CD28 antibodies (BioXcell, clones 145–2C11 and PV-1). After 24 h of activation, retrovirus was added together with 8 μg/ml polybrene, followed by centrifugation at 350g for 90 min at 32°C. Five hours after infection, the medium was replaced with fresh medium containing 10 ng/ml IL-2, and the cells were transferred into 12-well plates. Forty-eight hours after infection, cells were collected for flow cytometric validation, and GFP-positive cells were sorted for in vivo adoptive transfer experiments.

Generation of Rab27a knockdown MC38 cell lines

MC38 cells were infected with lentivirus expressing Rab27a shRNA (Sigma-Aldrich, TRCN0000100578) or control shRNA in the presence of 8 μg/ml polybrene. Forty-eight hours after infection, cells were subjected to puromycin selection for 3 days. Surviving cells were collected, and Rab27a knockdown efficiency was validated by qPCR.

Rab27a shRNA target sequence: CCAGTACACTGATGGCAAGTT.

Antisense oligonucleotides

Identification of the Mtdh-specific ASO (Mtdh ASO) and control ASO used in the present study has been described in our previous study³⁴. Triantennary N-Acetylgalactosamine-conjugated variants of Mtdh ASO (Sequence: +A*+G*+T*A*T*T*A*A*T*A*T*A*G*C*+G*+G*+T. +: locked-nucleic acid (LNA) modification; *: phosphorothioate linkage) and control ASO (Sequence: +C*+G*+T*T*T*A*G*G*C*T*A*T*G*T*A*+C*+T*+T) were ordered from WuXi AppTech. For targeting hepatocytes, the GalNAc ligand, consisting of three N-Acetylgalactosamine units arranged in a branched triantennary structure, was covalently attached to the 5′-end of the ASO via a phosphodiester-linked aminohexyl linker. For in vivo experiments, ASOs were resuspended in PBS.

Construction of MTDH mutant mice

MTDH W391A/W398A knock-in mice were made using CRISPR-Cas9. C57BL/6J embryos were microinjected with a mixture containing Cas9 protein (Integrated DNA Technologies), an sgRNA (single guide RNA) (Sigma-Aldrich) and a ssODN (single-stranded oligodeoxynucleotides) (Integrated DNA Technologies) which contained homology arms and the W391A, W398A mutations. The sgRNA used was CTGCTGACCCTAGCTCAGACTGG (PAM is underlined) and the donor oligo sequence was GGACTTCAGAAGTGAAGCTCTATCTTCATCTACCCAGTTCCCCgcCTCCTCTGCTGGTGCATTCgcGTCTGAGCTAGGGTCAGCAGAAGACAAACCATCTAAAAAATAATAATTTC (mutant sequences change in lower case bold). Founders were screened by PCR amplification of relevant DNA fragments, which were then digested separately with restriction enzymes BstUI (W391A) and AciI (W398A) (NEB) then further confirmed by NGS (Azenta Life Sciences) or Sanger sequencing. Confirmed founders were bred and progeny were screened using PCR primers MTDHA: TCTTGTGCTTAGCAACATTTGGAC and MTDHB: CTTTCTGAGACTCAATAGTGCCTG and presence of either or both restriction sites or by Sanger sequencing of PCR products.

Primary tumor models

For mammary tumor models,10 μl of cells (Py8119, 10⁴ or 2×10⁴ cells; EO771, 5×10⁴ cells; PyMT-FVB, 10⁵ cells; 4T1, 10³ cells; MCF7, 10⁵ cells; SUM159-M1a, 10⁵ cells; Py8119-Ova, 2×10⁴ cells) were injected into the left or both fourth mammary glands of 6–8-week-old female mice. Py8119 and EO711 are for C57BL/6J mice; PyMT-FVB is for FVB mice; 4T1 is for BALB/c mice; MCF7 and SUM159-M1a are for NSG mice. At the endpoint, lungs from Py8119 or 4T1 tumor-bearing mice were collected for spontaneous lung metastasis analysis after fixation in Bouin’s solution (Sigma-Aldrich, HT101128).

For subcutaneous tumor models, 100 μl of cells (MC38, 2×10⁵ or 3.5×10⁶ cells; Acsl4-WT and Acsl4-KO MC38, 3.5×10⁶ cells; LLC1, 2×10⁵ cells; MC38-Ova, 5×10⁵ cells) were subcutaneously injected into the right flank or both flanks of 6–8-week-old C57BL/6J male mice. For tumor rechallenge experiments, mice that remained tumor-free for 5 months following complete tumor regression were reinjected with MC38 cells. Age-matched naïve WT mice were used as controls.

Mice were examined weekly or twice a week for tumor development. Tumors were measured by calipers for the calculation of tumor volumes (length × width² /2).

Experimental lung metastasis model

For experimental lung metastasis models, 100 μl of cells (Py8119, 3×10⁵ cells; MC38, 10⁶ cells) were injected into 6–8-week-old female (Py8119) or male (MC38) mice via tail vein. For luciferase-labeled cell lines, lung metastasis was monitored by bioluminescent imaging (BLI), and images were processed with Living Image 3D Analysis (v.1.0). For the MC38 model, lungs were collected three weeks post-injection and fixed in Bouin’s solution for metastatic nodule counting.

Human PDX model

For human breast cancer PDX models (HCI-001 and HCI-002), implantation was performed as previously described⁵³. A single fragment of frozen tumor (~8 mm³) was implanted into cleared inguinal mammary fat pads of 3–4-week-old female NSG mice. Tumors were collected for experiments when they reached approximately 1 cm in diameter.

Cell depletion and neutralization antibody treatment

For the depletion of T cells and macrophages, 200 μg anti-CD4 (BioXcell; clone YST117), 100 μg anti-CD8 (BioXcell; clone 2.43), 300 μg anti-CSF1R (BioXcell; clone AFS98) or the same amount of isotype control (BioXcell; clones 2A3 and LTF-2) were injected intraperitoneally every 3 days after tumor cell injection.

For the depletion of B cells, 200 μg anti-CD20 (BioXcell; clone MB20–11) or the same amount of isotype control (BioXcell; clone DV5–1) were injected intraperitoneally weekly after tumor cell injection.

For Kupffer cell depletion, the Standard Macrophage Depletion Kit (Encapsula NanoSciences; SKU# CLD-8901) was used. Briefly, 200 μl of clodronate liposomes (Clodrosome^®) or control liposomes (Encapsome^®) was administered to female C57BL/6J mice via tail vein injection every 2 or 3 days for a total of 4 doses.

For cytokine neutralization, 200 μg of anti-TNFα antibody (BioXcell; clone XT3.11) and 300 μg of anti-TGF-β neutralizing antibody, or the corresponding isotype controls at the same doses (BioXcell; clones HRPN and MOPC-21), were administered by intraperitoneal injection every 2 days for a total of 7 doses.

For anti-PD-1 treatment, 200 μg anti-PD-1 (BioXcell; clone RMP1–14) or the same amount of isotype control (BioXcell; clones 2A3) were injected intraperitoneally every 3 days after tumor cell injection or once tumors were well established. In the MC38 experimental lung metastasis model, mice only received three treatments after tumor injection.

Construction of bone marrow chimeras

For the construction of bone marrow chimeras, 6–8-week-old recipient mice were lethally irradiated with two 6 Gy doses (separated by a 2-hour interval) or one 9 Gy dose (only for Mtdh KO mice, due to higher sensitivity to irradiation) by X-ray, and then intravenously transferred with a combination of 2×10⁶ bone marrow leukocytes from indicated donor mice. Mice were maintained on antibiotic water containing 0.3 mg/ml Baytril (Enrofloxacin) and 2.5% sucrose (Sigma-Aldrich, S0389) for one week prior to irradiation and 2 weeks post-irradiation. Bone marrow was obtained from age- and sex-matched mice. Bone marrow was flushed from femurs and tibiae with PBS, lysed for red blood cells, and resuspended in PBS at a concentration of 2×10⁷ cells/ml. Transferring of bone marrow cells was completed within 24 hours post-irradiation. Chimeras were used for tumor experiments 8 weeks after the initial reconstitution.

Adoptive T cell transfer

For the adoptive transfer of CD8⁺ T cells, a total of 1×10⁶ CD8⁺ T cells isolated from spleen were intravenously injected into recipient mice. Recipient mice were used for tumor experiments after 24 hours. For the adoptive transfer of retrovirus infected OT-1 T cells, a total of sorted 1×10⁶ GFP positive cells were intravenously injected into recipient mice 10 days after Py8119-Ova tumor injection.

ASO treatment and combination therapy

For ASO treatment, ASOs in PBS were injected into mice intraperitoneally at a dose of 2.5 mg/kg. For hepatocytes isolation, mice were treated with ASOs on day 1, day 3, day 5, then primary hepatocytes were isolated on day 7 for downstream experiments. For in vivo therapeutic experiments, the treatment schemes are indicated in the figures.

Briefly, Rag1 KO mice (male for MC38-Ova model; female for Py8119-Ova model) were pretreated with three doses ASOs one week prior to tumor injection. After MFP injection of Py8119-Ova cells or subcutaneous injection of MC38-Ova cells, ASO were treated twice a week. One week (Py8119 model) or six days (MC38 model) after tumor cell injection (tumors were well established), 3×10⁶ purified OT-1 T cells were transferred into mice via tail vein. One day after T cell transfer, 200 μg anti-PD-1 (Bio X cell; clone RMP1–14) were injected intraperitoneally 3 times per week.

HFD treatment

For HFD treatment, 6–8-week-old female mice were fed a diet containing 60% calories from fat (Research Diets, #D22061308i) for 7 weeks before tumor injection and maintained on the HFD throughout tumor development. Standard chow diet (PicoLab Rodent Diet, #5053) was used as the control.

EVPs depletion and isolation

The protocol was described previously¹². Briefly, when collecting conditioned media for EVP isolation, FBS was depleted of EVPs by ultracentrifugation at 100,000g for 4 hours. Cells were cultured in EVP-depleted media for 3 days and supernatant was collected for EVP isolation. For tumor explants, tumors were cut into small pieces and cultured overnight in RPMI media supplemented with 1x penicillin/streptomycin, and the supernatant was collected and filtered with a 70 μm strainer and then centrifugated at 3,000g for 30 minutes. The supernatant was used for EVP isolation. EVPs from cell culture media, tumor explant culture media or mouse serum were pelleted by ultracentrifugation at 100,000g for 70 minutes or following the manufacturer’s instructions of isolation kit (Thermo Fisher, 4478359 or 4478360). EVPs pellets were resuspended in PBS and concentration of EVPs was determined by BCA Protein Assay Kits (Thermo Fisher, 55865).

In vivo EVP tracing

EVP tracing was performed as previously described¹². Briefly, 10 μg of Py8119 tumor cell-derived EVPs was mixed with 0.4 μL of CellVue Burgundy dye (Thermo Fisher, 88–0872-16), pre-diluted in 50 μl of Diluent C, and incubated for 5 min in the dark. Labeling was quenched by adding 50 μl of 35% BSA (Sigma-Aldrich, A7979). The labeled EVPs were then washed with PBS and ultracentrifuged at 100,000 × g for 70 min. The EVP pellet was resuspended in 100 μl PBS and injected into mice via tail vein injection. PBS alone was used as a control. After 24 h, livers were collected for flow cytometry analysis.

Primary cell isolation and culture

For CD8⁺ T cells isolation, spleens were smashed and passed through sterile mesh filters using 10 ml of medium to wash the screen. Dissociated cells were resuspended with RBC lysis buffer (BioLegend, #420302) for red blood cell lysis. Single-cell suspensions were used for total CD8⁺ T cell isolation (Miltenyi Biotec, #130–104-075) or naïve CD8⁺ T cell isolation (Miltenyi Biotec, #130–096-543) following the manufacturer’s protocols. Purity of isolated cells was confirmed by flow cytometry. Naïve CD8⁺ T cells were seeded into 2 μg/ml anti-CD3/CD28 (BioXcell, clones 145–2C11 and PV-1) pre-coated plates plus 10 ng/ml IL-2 (Thermo fisher, 212–12-50UG) for culture. After 48 hours of culture, activated CD8⁺ T cells were used for further analysis.

For hepatocytes isolation, the largest mouse liver lobe was perfused using the liver perfusion kit (Miltenyi Biotec, #130–128-030). Following the manufacture’s protocol, the perfusion was performed on the gentleMACS Octo Dissociator (Miltenyi Biotec, #130–096-427) and hepatocyte were collected for downstream analysis or cultured in media (Fisher Scientific, NC9538906) for treatment. Plates for primary hepatocytes culture were coated with 30 μg/ml collagen I (Thermo fisher, A1048301) in sterile water overnight prior to seeding cells.

For Kupffer cell isolation, liver tissues were directly processed into single cell suspensions using liver dissociation kit (Miltenyi Biotec, # 130–105-807). The cell pellet was processed to macrophage isolation following the manufacture’s protocol of Anti-F4/80 MicroBeads UltraPure (Miltenyi Biotec, # 130–110-443). Kupffer cells were cultured in DMEM media with 10% EVP-depleted FBS onto collagen I-coated plates.

Treatment of primary Kupffer cells

Kupffer cells were treated with 200 μM palmitic acid (Cayman chemical, 29558) or 10 μg/ml of EVPs derived from cell lines culture, tumor explants culture or mouse serum. After 4 hours treatment, Kupffer cells were harvested for RNA extraction. For conditioned media collection and ELISA analysis, culture media were harvested after 48 hours or 72 hours treatment by centrifugation at 3,000g for 30 minutes and supernatant was used for further experiments.

Treatment of primary hepatocytes

Primary hepatocytes were treated with 20 ng/ml TNFα (Peprotech, 300–01A), 5 ng/ml TGF-β (R&D Systems, 7666-MB), or conditioned media from EVP-treated Kupffer cells. For NF-κB inhibtion, hepatocytes were pretreated with 10 μM NF-κB inhibitor BAY 11–7082 (Selleckchem, S2913) for 1 hour, followed by TNFα or TGF-β treatment for 24h. For PPARα activation, hepatocytes were treated with 100 μM PPARα agonist GW7647 (Selleckchem, S8046) together with TNFα or TGF-β for 24 hours. For endpoint analyses, hepatocytes were treated for 3 hours to assess nuclear p65 expression, 24 hours to assess nuclear PPARα expression or for RNA extraction and RNA-seq, and 24 to 48 hours for BODIPY staining. For immunoblot analysis of NF-κB signaling, hepatocytes were treated for 0,15, or 30 minutes.

Flow cytometry analysis

Tumors were dissociated with dissociation kits (Miltenyi Biotec, #130–096-730) on a gentleMACS Octo Dissociator (Miltenyi Biotec, #130–096-427), following the manufacturer’s instructions. Dissociated tumor cells were resuspended and centrifugated on a discontinuous Percoll gradient (Sigma-Aldrich, GE17–0891-01) to enrich immune cells. Spleens, pLN and thymus were smashed and passed through sterile mesh filters using 10 ml of medium to wash the screen. Dissociated cells were resuspended with RBC lysis buffer (BioLegend, #420302) for red blood cell lysis. Single-cell suspensions were first incubated with Fc block (BD biosciences, # 553142), and then stained with live dye and antibodies against surface molecules.

For intracellular cytokines staining, cells were stimulated with 100 ng/mL Phorbol myristate acetate (PMA) (Sigma-Aldrich, 79346) and 400 ng/mL ionomycin (Sigma-Aldrich,10634) in the presence of GolgiStop (BD Biosciences, 554715) for 4 hours prior to staining with antibodies against surface molecules followed by fixation and permeabilization (BD biosciences, # 554714), and then staining with antibodies against intracellular cytokines at 4°C overnight. After incubation, cells were washed and analyzed by flow cytometry.

Lipid peroxidation was measured according to the manufacturer’s instructions and previously published protocols²¹. Specifically, cells were incubated with a lipid peroxidation sensor (BODIPY 581/591 C11, Thermo Fisher, C10445) in the cell culture incubator for 30 minutes. After incubation, cells were washed, stained with live dye and surface markers, and analyzed by flow cytometry. Signals from both reduced C11 (PE channel) and oxidized C11 (FITC channel) were detected. Relative lipid ROS is calculated as the ratio of oxidized to reduced BODIPY-C11 mean fluorescence intensity (MFI) in cells, and the data were normalized to control samples. For cellular ROS measurement, cells were incubated with Cellular ROS Assay Kit (Deep Red) (Abcam, ab186029) at 37°C for 30 minutes, and analyzed by flow cytometry in the APC channel.

For glucose uptake and mitochondrial mass measurements, single-cell suspensions from tumors were pulsed with 20 μM 2-NBDG (Cayman Chemical, #186689–07-6) or 50 nM MitoTracker Deep Red (Thermo Fisher, M22426) in the cell culture incubator for 30 minutes. After incubation, cells were washed, stained with live dye and surface markers, and analyzed by flow cytometry.

For lipid droplet staining, cultured hepatocytes were trypsinized, washed with PBS, and collected by centrifugation. Cell pellets were resuspended in 2 μM of BODIPY 493/503(Thermo Fisher, D3922) and incubated at 37°C for 15 minutes. Following staining, cells were washed with PBS, resuspended in a viability dye and incubated at room temperature for 10 minutes. After incubation, cells were washed and subjected to flow cytometry analysis.

For p65 and PPARα nuclear staining, cultured hepatocytes were trypsinized, washed with PBS, and collected by centrifugation. The cell pellets were resuspended in a viability dye and incubated at room temperature for 10 minutes. Cells were then washed with PBS and processed for permeabilization and fixation by using the Transcription Factor Staining Kit (Thermo Fisher, 00–5523-00). For p65 staining, cells were incubated with p65 antibody (Biolegend, 653004) at 4°C for 30 minutes. For PPARα staining, cells were incubated with PPARα antibody (Thermo Fisher, MA1–822) at room temperature for 30 minutes, washed, then further incubated with anti-mouse IgG secondary antibody at 4 °C for 30 minutes. After incubation, cells were washed and analyzed by flow cytometry.

For Foxp3 nuclear staining, single cell suspensions were incubated with a viability dye and surface marker antibodies at 4°C for 30 minutes. Cells were then washed with PBS and processed for permeabilization and fixation by using the Foxp3/Transcription Factor Staining Kit (Thermo Fisher, 00–5523-00). Fixed cells were incubated with Foxp3 antibody (Thermo Fisher, 11–5773-82) at 4°C for 30 minutes. After incubation, cells were washed and analyzed by flow cytometry.

For p-STAT3 staining, single cells from dissociated tumors were stained with surface marker antibodies and a viability dye for 30 min on ice, fixed with 2% paraformaldehyde for 10 min at room temperature. Fixed cells were permeabilized with ice-cold BD PhosFlow Perm Buffer III (BD Biosciences, 558050) for 60 min, then stained with p-STAT3 (Y705) antibody (BD Biosciences, 612569) in BD Perm/Wash buffer (BD Biosciences, 554723) for 1 h at room temperature, as per the manufacturer’s protocol. After incubation, cells were washed and analyzed by flow cytometry.

For SCENITH assay, it was performed as described previously⁵⁹. Briefly, freshly isolated tumor single-cell suspensions (1×10⁶/well) were seeded in 48-well plates with 0.5 ml RPMI medium supplemented with 10% FBS. Freshly prepared metabolic inhibitors, including 2-DG (100 mM; Sigma-Aldrich, D6134), oligomycin (1 μM; Sigma-Aldrich, 75351), or the combination of 2-DG and oligomycin, were added to the cells and incubated for 15 min. Puromycin (10 μg/ml; Sigma-Aldrich, P7255) was then added for 25–30 min at 37°C. Cells were subsequently stained with Live/Dead dye, anti-CD45, anti-TCRβ and anti-CD8, followed by fixation and permeabilization using Foxp3/Transcription Factor Staining Kit (Thermo Fisher, 00–5523-00). Intracellular puromycin incorporation was detected using anti-puromycin-AF647 (Sigma-Aldrich, MABE343-AF647). Samples were analyzed by flow cytometry. Puromycin MFI was quantified in live CD8⁺ T cells for each treatment condition. Basal level of translation was calculated as MFI_Ct- MFI_DGO. Mitochondrial dependency (%) was calculated as: 100* [(MFI_Ct-MFI_O) / (MFI_Ct-MFI_DGO)]. Ct: Control; DG: 2-Deoxy-Glucose; O: Oligomycin; DGO: 2-Deoxy-Glucose plus Oligomycin.

All flow cytometry analyses were performed on an Attune NxT (Thermo Fisher) flow cytometer and data generated was processed by Flowjo (version 10.10.0). Cell sorting was performed on FACSAria (BD biosciences). Representative gating strategy is shown in Figure S12.

Seahorse Assay

The Seahorse XF Cell Mito Stress Test Kit (Agilent, #103015–100) was used for this assay. Isolated T cells or macrophages (50,000 or 100,000 cells/well) were plated on a Poly-D-lysine coated 96-well plate (Agilent, 103799–100) in Seahorse XF RPMI-1640 media supplemented with 10 mM glucose and 2 mM glutamine. OCR and ECAR of cells were then measured by the Agilent Seahorse XFe96 flux analyzer. Oligomycin (2 μM for T cells, 1 μM for macrophages) was loaded in port A, FCCP (1 μM) in port B, and Rot/AA (0.5 μM) in port C of the cartridge. For ex vivo activated CD8⁺ T cell, naïve CD8⁺ T cells were seeded into anti-CD3/CD28 pre-coated plates plus IL-2 for culture. After 3 days, activated CD8⁺ T cells were collected for Seahorse analysis. For lipid-depletion treatment, activated CD8⁺ T cells were cultured in medium containing 10% lipoprotein deficient serum FBS (Sigma-Aldrich, S5394) for 2h or 12h before Seahorse analysis, with normal FBS (Nucleus Biologics, 1824–003) used as the control.

Tumor and T cell co-culture

Py8119-Ova tumor cells expressing firefly luciferase were seeded in 96-well plates at 20,000 cells per well. After 24h, flow cytometry-sorted tumor-infiltrating CD8⁺ T cells or magnetically isolated splenic CD8⁺ T cells were added to tumor cells at the indicated tumor cell:CD8⁺ T cell ratios of 1:1 and 1:10, respectively. Cells were treated with 1 μM STAT3 inhibitor Stattic (MCE, HY-13818) or an equivalent volume of DMSO as vehicle control. After 24h of co-culture, tumor cell survival was assessed by measuring luciferase activity. Briefly, D-luciferin (Gold Biotechnology, #eLUCK) was added to each well at a final concentration of 100 μg/ml, and luminescence was measured using a Promega GloMax Microplate Luminometer. Relative tumor cell survival was normalized to control wells.

Immunoblotting

Hepatocytes were extracted with RIPA buffer, subjected to SDS-PAGE, and immunoblotted with standard protocols. Membrane imaging was done using the Licor Odyssey CLX system supplied with IRDye 800CW secondary antibodies (Licor, #926–32212 for anti-mouse; # 926–32211 for anti-rabbit).

Quantitative PCR Analysis

Total RNA was isolated from cells or tissues using the RNeasy Micro kit (Qiagen, #74004) in accordance with the manufacturer’s instructions. Reverse transcription into cDNA was performed with SuperScript IV first-strand synthesis system (Thermo Fisher, # 18091050). cDNA was used in subsequent qPCR using SYBR green master mix (Applied Biosystems, #A25777) following the manual. Sequences of primers used for RT-qPCR are listed in Table S1.

Quantification of AST and ALT

AST and ALT activity in serum were measured using the Aspartate Aminotransferase Colorimetric Activity Assay Kit (Cayman chemical, 701640) and Alanine Transaminase Colorimetric Activity Assay Kit (Cayman chemical, 700260) according to the manufacturer’s instructions. Briefly, serum samples were incubated with substrate and cofactors at 37°C for 15 minutes, then adding the initiators and measuring the absorbance at 340nm. The values of AST and ALT activity were calculated following the protocol. Serum from tumor-bearing mice was collected 3 weeks after tumor injection.

Quantification of Lipids

Lipids in tumor tissues were extracted by using the lipid extraction kit (Abcam, ab211044) following the manual. Free fatty acid (Abcam, ab65341), triglyceride (Abcam, ab65336), and cholesterol (Cayman chemical, 10007640) in extracted lipids or serum were detected using respective kits, following the manufacturer’s instructions. Serum from tumor-bearing mice was collected 4 weeks after tumor injection.

ELISA analysis

TNFα (Thermo Fisher 88–7324-88) or TGF-β1 (Thermo Fisher, 88–8350-22) ELISA Kits were utilized to measure the concentration in conditioned cell culture media or mouse serum according to the manufacturer’s instructions. Serum from tumor-bearing mice was collected 4 weeks after tumor injection.

Tissue processing and Lipid droplet staining

Fresh liver tissues were collected and fixed in 4% PFA overnight at room temperature, followed by incubation in 30% sucrose overnight at 4°C. The tissues were then embedded in OCT compound, frozen in -80°C, and cryosectioned at a thickness of 7 μm. Tissue slides were washed with cold PBS, incubated with autofluorescence quenching reagents (Vector laboratories, SP-8400–15) for 10 minutes, and washed again with PBS for 5 minutes. The slides were stained with 2 μM of BODIPY 493/503 (Thermo Fisher Scientific, D3922) at 37°C for 15 minutes, followed by washing with PBS for 5 minutes. Nuclei were counterstained with 1μg/ml DAPI for 5 minutes, washed with PBS for 5 minutes, and mounted using Prolong Diamond Antifade Mountant (Thermo Fisher Scientific, P36965). Lipid droplets were visualized by a Zeiss fluorescence microscope, and BODIPY fluorescence intensities were quantified using ImageJ.

Immunohistochemistry (IHC) and Immunofluorescence (IF) staining

Paraffin-embedded primary Py8119 tumors were processed as previously noted³⁴. Slide were incubated at 4°C overnight with Ki67(Cell signaling technology, #12202S), CD8 (Cell signaling technology, #98941) and cleaved caspase-3 (Cell signaling technology, #9661S) antibody at a 1:100 dilution. Following washes with PBS, slide stained as described before³⁴. Image was taken by Leica slide scanner. Positive staining areas were quantified using ImageJ.

Metabolomics and Lipidomics analysis

Metabolite extraction for LC-MS (Liquid chromatography-mass spectrometry).

Tumors were collected 4 weeks after tumor injection. Serum from Py8119 tumor-bearing mice was collected 4 weeks after tumor injection for analyses shown in Figures 4 and S5, or 3 weeks after tumor injection for analyses shown in Figure S6. Serum from normal mice was collected from age-matched female mice. For serum samples, 3 μL serum was mixed with 117 μL methanol, vortexed for 30 seconds, and centrifuged at 17,000g for 10 minutes at 4°C. For tumor samples, 10–20 mg of tissues was weighed and pulverized under liquid nitrogen using a cryomill (Restch, Newtown, PA). Next, 40x volume (40 μL extraction solvent per 1 mg tissue) of 40:40:20 methanol: acetonitrile: water with 0.5% formic acid pre-cooled to -20°C was added to the tissue powder, vortexed, and incubated on ice for 10 minutes. Acid was then neutralized by adding 15% ammonium bicarbonate (NH4HCO3) aqueous solution (8.75% v/v of extraction buffer), and samples were vortexed and centrifuged at 17,000 g for 10 minutes at 4°C. Clear supernatant was transferred to MS vials and loaded on the mass spectrometer autosampler.

LC-MS.

Metabolite measurements were obtained by running samples on orbitrap Exploris 240 or Q Exactive Plus orbitrap mass spectrometer (Thermo Scientific) coupled with hydrophilic interaction chromatography (HILIC) with electrospray ionization. LC separation was performed on an XBridge BEH Amide column (2.1×150 mm, 2.5 μm particle size, 130 Å pore size; Waters Corporation) using a gradient of solvent A (95:5 water: acetonitrile with 20 mM of ammonium acetate and 20 mM of ammonium hydroxide, pH 9.45) and solvent B (acetonitrile). Flow rate was 150 μL/min. The LC gradient was: 0 min,90% B; 2 min,90% B; 3 min,75%; 7 min, 75% B; 8 min,70%,9 min, 70% B; 10 min, 50% B; 12 min, 50% B; 13 min, 25% B; 14 min, 25% B; 16 min, 0.5% B, 20.5 min, 0.5% B; 21 min, 90% B; 25 min, 90% B. Injection volume was 5–10 μL and autosampler temperature was 4°C. MS scans were in negative and positive ion switching mode with a resolution of 120,000 at m/z 200 and scan range of m/z 70–1000 for neg mode and m/z 58–116.5 and m/z 120–1000 for pos mode. Mass spec data was analyzed using El-Maven (v0.12.0, Elucidata).

Lipid measurement.

Serum (3 μl) or tumor powder (10–20 mg) was dissolved in 117 μl or 40x volume of isopropanol, centrifuged at 17,000g for 20 minutes at 4 °C. The supernatant (100 μl) was transferred to a glass vial for LC-MS analysis. Lipid species were detected with a quadrupole orbitrap mass spectrometer (Q Exactive) operating in positive and negative ion switching mode with a resolution of 140,000 at m/z 200 and scan range of m/z 250–1000 for pos mode and a resolution of 70,000 at m/z 200 and scan range of m/z 200–1000 for neg mode. LC separation was achieved on an Agilent Poroshell 120 EC-C18 column (150 × 2.1 m2, 2.7 μm particle size) using a gradient of solvent A (90:10 water: methanol with 1 mM of ammonium acetate and 0.2% acetic acid) and solvent B (2:98 methanol: isopropanol with 1 mM of ammonium acetate and 0.2% acetic acid). Flow rate was 150 μl / min. The LC gradient was 0 min, 25% B; 2 min, 25% B; 4 min, 65% B; 16 min, 100% B; 20 min, 100% B; 21 min, 25% B; 25 min, 25% B. Mass spec data was analyzed using El-Maven (v0.12.0, Elucidata).

Data plotting.

Raw counts data were uploaded to MetaboAnalyst (version 5.0) for further analysis⁸⁹. Raw counts are shown in Data S1. The figures of heatmap, volcano plots, and principal component (PC) analysis were generated accordingly.

Proteomics for CD8⁺ T cells by LC-MS

Sample preparation for LC-MS Analysis.

Tumor-infiltrating CD8⁺ T cells were isolated from Py8119 tumors 4 weeks after tumor injection by flow cytometry sorting. Splenic CD8⁺ T cells from normal mice were isolated using a MACS kit (Miltenyi Biotec, #130–104-075). Samples were prepared as described previously⁹⁰. Briefly, samples were sequentially reduced and alkylated with 25 mM TCEP and 50 mM chloroacetamide. Afterwards, samples were acidified with phosphoric acid prior to the addition of S-Trap resuspension buffer and loading onto S-Trap micro spin columns (Protifi). S-Trap columns were washed 8 times with the resuspension buffer, then digested with trypsin at 37°C for 16 hours. Peptides were eluted with increasing concentrations of acetonitrile (ACN) and then dried using a Savant SpeedVac. Dried peptides were resuspended in 0.1% formic acid and 4% ACN to a final concentration of 0.1 μg/μl. Samples were stored at 4°C until ready for MS analysis.

MS Data Acquisition (timsTOF Ultra).

LC-MS/MS analyses were performed using a NanoElute 2 LC system coupled to a timsToF Ultra mass spectrometer (Bruker). 100 ng of peptides (1μl of a 0.1 μg/μl solution) were injected onto a PepSep C18 25 cm × 75 μm column with 1.5 μm particle diameter. Separation was achieved with a linear gradient of mobile phases A (0.1% formic acid in water) and B (0.1% formic acid in 99.9% acetonitrile) from 4 to 35% B over 20 or 30 minutes at a flow rate of 200 nL/min. The MS ion source was configured with a 10 μm emitter. MS/MS acquisition was performed using a 16 × 3 dia-pasef method (50 ms ramp time) that covered precursors in an m/z range of 300 to 1300 and an ion mobility range of 0.65–1.46 Vs/cm2 (1/K0). DIA windows were optimized based on data-dependent acquisition analyses. Estimated cycle time was 0.95 s, and the collision energy was adjusted linearly with the ion mobility, from 20 eV at 0.6 Vs/cm2 to 59 eV at 1.6 Vs/cm2.

DIA Analysis.

DIA-NN was used for protein identification and quantification⁹¹. First, an in silico spectral library was generated from a UniProt mouse sequence database (21,857 sequences from the UniProt reference proteome database appended with common contaminants, 2023–11) using Bruker’s library generation tool. In silico library generation allowed for 1 missed cleavage, N-terminal M excision, and static Cys carbamidomethylation, peptide length 7–30 amino acids, precursor charges 1–4, and precursor and fragment m/z values from 300–1800 and 200–1800, respectively. Raw.d files were searched against this library with default settings including enabling of Match between runs (MBR). MS1 and MS2 mass accuracies were fixed at 15–20 ppm. The report matrix files were uploaded to FragPipe Analyst for further analysis⁹². Normalized protein abundances were taken from the pg.matrix output from DIA-NN. The data are shown in Data S1.

Pathway analysis.

The results of fold changes and p-values obtained from FragPipe Analyst were ranked, and then the protein list was submitted to the STRING database webtool (https://string-db.org/). Significantly changed pathways were identified and used for plotting. The results were also mapped to proteins in oxidative phosphorylation gene signature (Gene Ontology), or other proteins related to mitochondrial function from previous studies^93,94.

Bulk RNA sequencing and GSEA

Tumor-infiltrating CD8⁺ T cells were isolated from Py8119 tumors 4 weeks after tumor injection. Liver tissues were collected from Py8119-tumor bearing mice 4 weeks after tumor injection or from age-matched normal female mice. Primary hepatocytes were collected after 24h treatment with conditioned media from Py8119-EVP-treated Kupffer cells, 20 ng/ml TNFα, or 5 ng/ml TGF-β. Total RNA was extracted using the RNeasy Micro kit (Qiagen, #74004). Bulk RNA sequencing was performed at the Genomics Core Facility of Lewis-Sigler Institute at Princeton University. The integrity of total RNA samples was assessed on Bioanalyzer 2100 using RNA 6000 Pico chip (Agilent Technologies, CA). Messenger RNA was enriched from these samples using QIAseq FastSelect-rRNA HMR Kit (Qiagen, CA), fragmented and converted to cDNA and Illumina sequencing library using the PrepX RNA-seq library preparation protocol on the Apollo 324TM NGS Library Prep System (Takara Bio, CA). Unique DNA barcode was incorporated into each library for sample demultiplexing. These RNA-seq libraries were examined on Agilent Bioanalyzer DNA High Sensitivity chips for size distribution, quantified by Qubit fluorometer (Invitrogen, CA), and pooled at equal molar amount. The library pools were denatured and sequenced on Illumina NovaSeq 6000 using the S Prime flowcells as pair-end 65 nt reads according to the manufacturer’s protocol. Raw sequencing reads were filtered by Illumina NovaSeq Control Software, only the Pass-Filter (PF) reads were de-multiplexed allowing 1 mismatch and used for further analysis.

Demultiplexed raw sequence file (FASTQ) were analyzed using Partek^™ Flow^™ software package (version 12.0). Raw reads were aligned to the Mus musculus (mouse) - mm39 assembly using STAR 2.7.8a. Gene counts were obtained by quantifying against the mm39-Ensembl Transcripts release 104 or 108 annotation model. After noise filtering, counts were normalized using the DESeq2 package. GSEA (version 4.3.3) was used for pre-ranked analysis performed with log2 (fold-change) as rank list. Alternatively, gene lists were tested for gene set enrichment using the EnrichR and PreRank functionalities implemented in the gseapy package (v.0.10.3). The following gene set repositories were used for enrichment analysis: MSigDB Hallmark (2020) and KEGG mouse (2019). Top significantly enriched gene signatures were sorted out for further figure plotting. The gene set for ‘PPARα activates gene expression’ was downloaded from Reactome Pathways.

Single-cell RNA sequencing and analysis

Live CD45⁺ cells were sorted from Py8119 tumors 4 weeks post MFP injection by flow cytometry. Single cell suspension samples were prepared (see details in ‘Flow cytometry analysis’) and examined by flow cytometry for cell density and viability. Around 15,000 cells from each sample were loaded into each channel of the 10X Genomics Chromium X system using the Chromium Single Cell 3’ v3.1 Reagent Kits (10X Genomics Inc., CA) to generate and amplify cDNA from individual cells. The amplified cDNA samples were purified with Ampure XP magnetic beads (Beckman Coulter, CA), quantified by Qubit fluorometer (Invitrogen, CA), and examined on Bioanalyzer with High Sensitivity DNA chips (Agilent, CA) for size distribution. Illumina sequencing libraries were generated from the amplified cDNA samples using the Illumina Tagment DNA Enzyme and Buffer kit (Illumina, CA). These libraries were examined by Qubit and Bioanalyzer, then pooled at equal molar amount and sequenced on Illumina NovaSeq 6000 S Prime flowcells as 28+94 nt pair-end reads following the standard Illumina protocol. Raw sequencing reads were filtered by Illumina NovaSeq Control Software and only the Pass-Filter (PF) reads were used for further analysis. After sample demultiplexing, raw PF reads were processed through the CellRanger software (10X Genomics Inc., CA) to obtain the gene expression levels in single cells and quality control matrices for each sample.

As a basic preprocessing step, low-quality cells were removed by filtering out cells that did not express at least 200 genes. Similarly, genes not expressed in at least three cells were excluded. Additionally, cells with more than 25% mitochondrial gene expression were filtered out. To detect doublet cells, scDblFinder⁷⁷ was used, and cells marked as doublets were consequently removed. Each sample retained approximately 90% singlets, signifying the quality of the experiment. After preprocessing, each single-cell dataset was normalized and integrated using the 5000 most variable genes per sample. Seurat (version 5.0.0) was used for batch correction and integration. Unsupervised clustering was performed using the Louvain algorithm with a resolution of 0.5. Upon clustering, marker genes for the clusters were identified using the ‘FindAllMarkers’ function in Seurat with a log fold change threshold of 0.25. These markers were later used to determine the cell types of the clusters. The lower-dimensional representation of the single-cell clusters was computed using UMAP embedding.

ATAC-seq process and data analysis

Live tumor-infiltrating CD8⁺ T cells were sorted from Py8119 tumors 4 weeks after tumor injection using a flow cytometry sorter, with a total of 150,000 cells collected. ATAC-seq libraries were generated from dissociated cells according to the Omni-ATAC protocol⁹⁵, as described in our previous study⁹⁶. Briefly, 50,000–100,000 cells per sample were used for nuclei isolation, during which cells were lysed on ice for 3 minutes. Permeabilized nuclei were subjected to transposition using TDE1 transposase (Illumina, CA) at 37°C for 60 minutes, followed by purification with the Zymo DNA Clean and Concentrator-5 kit. The transposed DNA fragments were then amplified by PCR to incorporate Illumina sequencing adapters and barcodes. Purified ATAC-seq libraries were evaluated for fragment size distribution using Agilent Bioanalyzer DNA High Sensitivity chips, quantified with a Qubit fluorometer (Invitrogen, CA), and pooled in equimolar amounts. The pooled libraries were sequenced on Illumina NovaSeq 6000 S Prime 100-cycle flow cells. Raw sequencing reads were processed by the NovaSeq control software, and only Pass-Filter (PF) reads were retained for downstream analysis.

Reads were aligned to the mouse genome (mm10) using Bowtie2 with default parameters. BAM file conversion, sorting, duplicate removal, and indexing were performed using samtools v1.6. Peaks were called with MACS2 using the following parameters: “–shift -75 –extsize 150 –nomodel –call-summits –nolambda –keep-dup all -p 0.01”. For visualization, filtered BAM files were converted to BigWig format and normalized to counts per million (CPM) using the bamCoverage function in deepTools. Profile plots were generated using plotCoverage from the same package.

pSTAT3-binding peaks were defined using publicly available pSTAT3 ChIP-seq datasets from mouse CD8⁺ T cells stimulated with IL-6, IL-10, or IL-21 (GSE217374⁶³). Peaks shared across all three datasets were designated as pSTAT3-binding peaks. ATAC-seq signals at these regions were quantified using the featureCounts package and normalized to reads per million (RPM). These shared pSTAT3-binding peaks were also used for GSEA analysis. For this analysis, merged WT and Mtdh KO ATAC-seq peaks from CD8⁺ T cells were ranked by fold change comparing Mtdh KO versus WT and used as the reference dataset.

Motif enrichment analysis

ATAC-seq peaks gained in Mtdh KO cells and used for motif analysis were defined as peaks with a fold change >1.5 (Mtdh KO versus WT). Motif enrichment analysis was performed using the findMotifsGenome.pl module from the HOMER package with the parameter “-size given”.

QUANTIFICATION AND STATISTICAL ANALYSIS

GraphPad Prism v10.2.3 (GraphPad Software Inc., La Jolla, CA) was employed for data analysis. Detailed statistical methods are indicated in the figure legends. Data are shown as mean ± SEM. Exact sample sizes, statistical test methods, and p-values are specified in the figure legends or figures. p values <0.05 were considered statistically significant.

Supplementary Material

Document S1. Supplemental table S1 and Figures S1–S12.

Data S1. Uncropped scans of all the blots; raw metabolomics, lipidomics, and proteomics data; and unprocessed data underlying the display items in the manuscript, related to Figures 1–7 and S1–S11.

Key resources table.

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Rat anti-CD45 PerCP-Cy5.5 (clone 30-F11)	Thermo Fisher Scientific	Cat#45-0451-82; RRID: AB_1107002
Hamster anti-CD3e PE (clone 145-2C11)	BioLegend	Cat#100308; RRID: AB_312673
Rat anti-CD4 PE/Cy7(clone GK1.5)	BioLegend	Cat#100422; RRID: AB_312707
Rat anti-CD4 PE (clone GK1.5)	BioLegend	Cat#100408; RRID: AB_312693
Rat anti-CD8α APC/Cy7 (clone 53-6.7)	BioLegend	Cat#100714; RRID: AB_312753
Rat anti-CD8α APC (clone 53-6.7)	BioLegend	Cat#100712; RRID: AB_312751
Rat anti-CD8α FITC (clone 53-6.7)	Thermo Fisher Scientific	Cat#11-0081-85; RRID: AB_464916
Hamster anti-TCRβ PE (clone H57-597)	BioLegend	Cat#109208; RRID: AB_313431
Rat anti-CD25 APC (clone PC61)	BioLegend	Cat#102012; RRID: AB_312861
Rat anti-CD44 APC (clone IM7)	BioLegend	Cat#103030; RRID: AB_830787
Rat anti-CD62L APC-eFluor 780 (clone MEL-14)	Thermo Fisher Scientific	Cat#47-0621-82; RRID: AB_1603256
Rat anti-Foxp3 FITC (clone FJK-16s)	Thermo Fisher Scientific	Cat#11-5773-82; RRID: AB_465243
Rat anti-CD19 PE-Cy5.5 (clone eBio1D3 (1D3))	Thermo Fisher Scientific	Cat#35-0193-82; RRID: AB_891395
Rat anti-CD279(PD-1) APC (clone 29F.1A12)	BioLegend	Cat#135210; RRID: AB_2159183
Hamster anti-CD279(PD-1) FITC (clone J43)	Thermo Fisher Scientific	Cat#11-9985-82, RRID: AB_465472
Rat anti-CD366 (Tim-3) BV711 (clone RMT3-23)	BioLegend	Cat#119727; RRID: AB_2716208
Rat anti-IFNγ APC (clone XMG1.2)	BioLegend	Cat#505810; RRID: AB_315404
Rat anti-TNFα PE-Cy7 (clone MP6-XT22)	BD Biosciences	Cat#557644; RRID: AB_396761
Rat anti-granzyme B PE (clone NGZB)	Thermo Fisher Scientific	Cat#12-8898-82; RRID: AB_10870787
Rat anti-Perforin PE (clone S16009B)	BioLegend	Cat#154406; RRID: AB_2721641
Rat anti-CD11b PE/Cy7 (clone M1/70)	BioLegend	Cat#101216; RRID: AB_312799
Rat anti-F4/80 APC/Cy7 (clone BM8)	BioLegend	Cat#123118; RRID: AB_893477
Rat anti-CD206 FITC (clone C068C2)	BioLegend	Cat#141704; RRID: AB_10901166
Mouse anti-H-2Kb/H-2Db FITC (clone 28-8-6)	BioLegend	Cat#114605; RRID: AB_313596
Rat anti-MHC II (I-A/I-E) FITC (clone M5/114.15.2)	Thermo Fisher Scientific	Cat#11-5321-82; RRID: AB_465232
Mouse anti-CD36 PE (clone CRF D-2712)	BD Biosciences	Cat#562702; RRID: AB_2737732
Mouse anti-STAT3 (pY705) PE (clone 4/P-STAT3)	BD Biosciences	Cat#612569; RRID: AB_399860
Mouse anti-NF-κB p65 PE (clone 14G10A21)	BioLegend	Cat#653004; RRID: AB_2562769
Goat anti-Mouse IgG(H+L) APC	Thermo Fisher Scientific	Cat#17-4010-82, RRID: AB_2573203
Rat anti-mouse CD16/CD32 (clone 2.4G2) (Fc block)	BD Biosciences	Cat#553142; RRID: AB_394657
Mouse anti-puromycin-AF647 (clone 12D10)	Sigma-Aldrich	Cat#MABE343-AF647, RRID: AB_2736876
Mouse PPARα Antibody (clone 3B6/PPAR)	Thermo Fisher Scientific	Cat#MA1-822; RRID: AB_2165745
Rabbit Phospho-IKKα/β (Ser176/180) (clone 16A6)	Cell Signaling Technology	Cat#2697; RRID: AB_2079382
Rabbit Phospho-IkappaB alpha (Ser32) (clone 14D4)	Cell Signaling Technology	Cat#2859; RRID: AB_561111
Rabbit Phospho-NF-κB p65 (Ser536) (clone 93H1)	Cell Signaling Technology	Cat#3033; RRID: AB_331284
Rabbit NF-kappaB p65 (clone D14E12)	Cell Signaling Technology	Cat#8242, RRID: AB_10859369
Ki-67(D3B5) Rabbit mAb	Cell Signaling Technology	Cat#12202, RRID: AB_2620142
CD8 alpha(D4W2Z) Rabbit Monoclonal Antibody	Cell Signaling Technology	Cat#98941, RRID: AB_2756376
Cleaved Caspase-3 (Asp175) Antibody	Cell Signaling Technology	Cat#9661, RRID: AB_2341188
Mouse Anti-β-Actin antibody (clone AC-15)	Sigma-Aldrich	Cat#A1978; RRID: AB_476692
Rabbit Metadherin Polyclonal Antibody	Thermo Fisher Scientific	Cat#40-6500; RRID: AB_2533475
InVivoMab anti-mouse CD4 (clone YTS 177)	Bio X Cell	Cat#BE0003-3; RRID: AB_1107642
InVivoMab anti-mouse CD8α (clone 2.43)	Bio X Cell	Cat# BE0061; RRID: AB_1125541
InVivoMab anti-mouse CSF1R (clone AFS98)	Bio X Cell	Cat#BE0213; RRID: AB_2687699
InVivoMAb rat IgG2a isotype control (clone 2A3)	Bio X Cell	Cat#BE0089; RRID: AB_1107769
InVivoMAb rat IgG2b isotype control (clone LTF-2)	Bio X Cell	Cat#BE0090; RRID: AB_1107780
InVivoPlus anti-mouse PD-1 (CD279) (clone RMP1-14)	Bio X Cell	Cat#BE0146; RRID: AB_10949053
InVivoPlus anti-mouse CD20 (clone MB20-11)	Bio X Cell	Cat#BP0356; RRID: AB_2894820
InVivoPlus mouse IgG2c isotype control (clone DV5-1)	Bio X Cell	Cat#BP0366; RRID: AB_3720026
InVivoPlus anti-mouse TNFα (clone XT3.11)	Bio X Cell	Cat#BP0058; RRID: AB_2894798
InVivoPlus anti-mouse TGF-β (clone 1D11.16.8)	Bio X Cell	Cat#BP0057; RRID: AB_2894797
InVivoPlus mouse IgG1 isotype control (clone MOPC-21)	Bio X Cell	Cat#BP0083; RRID: AB_2894739
InVivoMab anti-mouse CD3ε (clone 145-2C11)	Bio X Cell	Cat#BE0001-1; RRID: AB_1107634
InVivoMab anti-mouse CD28 (clone PV-1)	Bio X Cell	Cat#BE0015-5; RRID: AB_1107628
IRDye 800CW Donkey anti-mouse Secondary Antibody	Licor	Cat#926-32212; RRID: AB_621847
IRDye 800CW goat anti-rabbit Secondary Antibody	Licor	Cat#926-32211; RRID: AB_621843
Biological samples
Human PDXs	Gift from Dr. Alana L. Welm53	N/A
Chemicals, peptides, and recombinant proteins
eBioscience™ Fixable Viability Dye eFluor™ 506	Thermo Fisher Scientific	Cat#65-0866-18
LIVE/DEAD™ Fixable Near-IR Dead Cell Stain Kit	Thermo Fisher Scientific	Cat#L34976
DAPI solution	Thermo Fisher Scientific	Cat#564907
BAY 11-7082	Selleckchem	Cat#S2913
GW7647	Selleckchem	Cat#S8046
Percoll	Sigma-Aldrich	Cat#GE17-0891-01
Phorbol myristate acetate (PMA)	Sigma-Aldrich	Cat#79346
Ionomycin	Sigma-Aldrich	Cat#10634
2-NBDG	Cayman Chemical	Cat#186689-07-6
BODIPY 581/591 C11	Thermo Fisher Scientific	Cat#C10445
MitoTracker Deep Red	Thermo Fisher Scientific	Cat#M22426
BODIPY 493/503	Thermo Fisher Scientific	Cat#D3922
Cellular ROS Assay Kit (Deep Red)	Abcam	Cat#ab186029
D-Luciferin Firefly	Gold Biotechnology	Cat#eLUCK
Palmitic acid	Cayman chemical	Cat#29558
Oligomycin	Sigma-Aldrich	Cat# 75351
2-DG	Sigma-Aldrich	Cat# D6134
Puromycin	Sigma-Aldrich	Cat#P7255
Stattic	MedChemExpress	Cat#HY-13818
Human TNF-alpha Recombinant Protein	PeproTech	Cat#300-01A
Mouse IL-2 Recombinant Protein	Thermo Fisher Scientific	Cat#212-12-20UG
Recombinant Mouse TGF-beta 1 Protein	R&D Systems	Cat#7666-MB
Bouin’s solution	Sigma-Aldrich	Cat#HT101128
Critical commercial assays
Standard Macrophage Depletion Kit (Clodrosome + Encapsome)	Encapsula NanoSciences	Cat#CLD-8901
Naive CD8a+ T Cell Isolation Kit, mouse	Miltenyi Biotec	Cat#130-096-543
CD8a+ T Cell Isolation Kit, mouse	Miltenyi Biotec	Cat#130-104-075
Tumor Dissociation Kit, mouse	Miltenyi Biotec	Cat#130-096-730
Anti-F4/80-MicroBeads UltraPure, mouse	Miltenyi Biotec	Cat#130-110-443
liver dissociation kit	Miltenyi Biotec	Cat#130-105-807
Liver Perfusion Kit, mouse and rat	Miltenyi Biotec	Cat#130-128-030
Total Exosome Isolation Reagent (from cell culture media)	Thermo Fisher Scientific	Cat#4478359
Total Exosome Isolation Reagent (from serum)	Thermo Fisher Scientific	Cat#4478360
RNeasy Micro Kit	Qiagen	Cat#74004
SuperScript IV First-Strand Synthesis System	Thermo Fisher Scientific	Cat#18091050
PowerUp SYBR Green Master Mix for qPCR	Applied Biosystems	Cat#A25777
BD Cytofix/Cytoperm™ Plus Fixation/Permeabilization Solution Kit with BD GolgiStop™	BD Biosciences	Cat#554715
eBioscience™ Foxp3 / Transcription Factor Staining Buffer Set	Thermo Fisher Scientific	Cat#00-5523-00
BD Phosflow™ Perm Buffer III	BD Biosciences	Cat#558050
BD Perm/Wash™ Perm/Wash Buffer	BD Biosciences	Cat#554723
Seahorse XF Cell Mito Stress Test Kit	Agilent	Cat#103015-100
Lipofectamine™ 3000 Transfection Reagent	Thermo Fisher Scientific	Cat#L3000015
Lenti-X™ Concentrator	Takara	Cat#631231
Aspartate Aminotransferase Colorimetric Activity Assay Kit	Cayman chemical	Cat#701640
Alanine Transaminase Colorimetric Activity Assay Kit	Cayman chemical	Cat#700260
Lipid Extraction Kit	Abcam	Cat#ab211044
Free Fatty Acid Assay Kit	Abcam	Cat#ab65341
Triglyceride Assay Kit	Abcam	Cat#ab65336
Cholesterol Fluorometric/Colorimetric Assay Kit	Cayman chemical	Cat#10007640
Mouse TNF alpha Uncoated ELISA Kit	Thermo Fisher Scientific	Cat#88-7324-88
Human/Mouse TGF beta-1 Uncoated ELISA Kit	Thermo Fisher Scientific	Cat#88-8350-22
BCA Protein Assay Kits	Thermo Fisher Scientific	Cat#55865
CellVue Burgundy dye	Thermo Fisher Scientific	Cat#88-0872-16
Deposited data
Raw and processing data of bulk-RNAseq, scRNA-seq, and ATAC-seq	This paper	GEO: GSE276107
Raw and processing data of proteomics data	This paper	ProteomeXchange Consortium: PXD064796
Raw and processing data of metabolomics data	This paper	Metabolomics Workbench:ST004024; ST004025
pSTAT3 ChIP-seq data	Sun et al.63	GEO: GSE217374
Data S1	This paper	N/A
Experimental models: Cell lines
Py8119	Gift from Dr. Weizhou Zhang	ATCC, Cat#CRL-3278; RRID: CVCL_AQ09
Py8119-Ova	Shen et al.37	N/A
E0771	Gift from Dr. Xiang H.-F. Zhang	CH3 Biosystems, Cat#94A001; RRID: CVCL_GR23
PyMT-FVB	Wan et al.27	N/A
MC38	Kerafast	Kerafast, Cat#ENH204-FP; RRID: CVCL_B288
MC38-Ova	This paper	N/A
SUM159-M1a	Esposito et al.71	N/A
LLC1	ATCC	ATCC, Cat#CRL-1642; RRID: CVCL_4358
4T1	ATCC	ATCC, Cat#CRL-2539; RRID: CVCL_0125
Platinum-E	Fisher Scientific	Fisher Scientific, Cat#NC0066908; RRID: CVCL_B488
MCF7	ATCC	ATCC, Cat#HTB-22™; RRID: CVCL_0031
MCF10A	Gift from Dr. Fred Miller	RRID: CVCL_0598
MCF10AT-cA1a	Gift from Dr. Fred Miller	RRID: CVCL_6676
HEK293T	ATCC	ATCC, Cat#CRL-3216; RRID: CVCL_0063
Acsl4-WT MC38	Gift from Dr. Weiping Zou50	N/A
Acsl4-KO MC38	Gift from Dr. Weiping Zou50	N/A
Experimental models: Organisms/strains
Mouse: C57BL/6J	The Jackson Laboratory	RRID: IMSR_JAX:000664
Mouse: BALB/c	The Jackson Laboratory	RRID: IMSR_JAX:000651
Mouse: FVB/NJ	The Jackson Laboratory	RRID: IMSR_JAX:001800
Mouse: NSG: NOD.Cg-Prkdcscid Il2rgtm1Wjl/SzJ	The Jackson Laboratory	RRID: IMSR_JAX:005557
Mouse: Mtdh KO	Wan et al.27	N/A
Mouse: Mtdh-flox mice	Wan et al.27	N/A
Mouse: Mtdh W391A/W398A mut	This paper	N/A
Mouse: CD4-Cre: B6. Cg-Tg (Cd4-cre)1Cwi/BfluJ	The Jackson Laboratory	RRID: IMSR_JAX:022071
Mouse: Albumin-Cre: B6. Cg-Speer6-ps1Tg (Alb-cre)21Mgn/J	The Jackson Laboratory	RRID: IMSR_JAX:003574
Mouse: VAV1-Cre: B6. Cg-Tg(VAV1-cre)1Graf/MdfJ	The Jackson Laboratory	RRID: IMSR_JAX:035670
Mouse: OT-1:C57BL/6-Tg (TcraTcrb)1100Mjb/J	The Jackson Laboratory	RRID: IMSR_JAX:003831
Mouse: PPARα KO: B6;129S4-Pparatm1Gonz/J	The Jackson Laboratory	RRID: IMSR_JAX:008154
Mouse: RAG1 KO: C57BL/6J-Rag1em10Lutzy/J	The Jackson Laboratory	RRID: IMSR_JAX:034159
Oligonucleotides
Control and Mtdh Antisense oligonucleotides	WuXi AppTec	See Method details
sgRNA oligo for Mtdh mutant mice	Sigma-Aldrich	See Method details
Donor oligo for Mtdh mutant mice	IDT	See Method details
PCR primers for Mtdh mutant mice	IDT	See Method details
PCR primers for CD36	IDT	See Method details
List of qPCR primers	IDT	See Table S1
Recombinant DNA
pMSCV-IRES-GFP II	Addgene	Cat#52107; RRID: Addgene_52107
pMSCV-IRES-GFP II-CD36	This paper	N/A
pCL-Eco	Addgene	Cat#12371; RRID: Addgene_12371
psPAX2	Addgene	Cat#12260; RRID: Addgene_12260
pMD2.G	Addgene	Cat#12259; RRID: Addgene_12259
pLKO.1-puro non-mammalian shRNA Control Plasmid DNA	Sigma-Aldrich	Cat#shc002
PLKO.1 shRNA for Rab27a	Sigma-Aldrich	Cat#TRCN0000100578
Software and algorithms
FlowJo v10.10.0	FlowJo	FlowJo (RRID:SCR_008520)
GraphPad Prism v10.2.3	GraphPad Software, Inc.	GraphPad Prism (RRID:SCR_002798)
Living Image v4.5.5	PerkinElmer	RRID:SCR_014247
PartekTM FlowTM software v12.0	Partek Genomics Suite	RRID:SCR_011860
ImageJ	National Institute of Health	ImageJ (RRID:SCR_003070)
Adobe Illustrator	Adobe	Adobe Illustrator (RRID:SCR_010279)
Biorender	Biorender	Biorender (RRID:SCR_018361)
GSEA v4.3.3	Mootha72 and Subramanian73 et al.	RRID:SCR_003199
MetaboAnalyst v5.0	Pan et al.74	RRID:SCR_015539
R v4.3.2	R Core Team	https://www.r-project.org/
R package: Seurat v5.4.0	Hao et al.75	https://satijalab.org/seurat/
R package: ggplot2 v4.0.2	Wickham76	https://ggplot2.tidyverse.org
R package: ggthemes v5.2.0	Arnold	https://jrnold.github.io/ggthemes/
R package: ggridges v0.5.7	Wilke	https://wilkelab.org/ggridges/
R package: ggrepel v0.9.6	Slowikowski	https://ggrepel.slowkow.com/
R package: ggpubr v0.6.3	Kassambara	https://CRAN.R-project.org/package=ggpubr
R package: scDblFinder v1.24.0	Germain et al.77	https://github.com/plger/scDblFinder
R package: clusterProfiler v4.18.4	Wu et al.78	https://github.com/YuLab-SMU/clusterProfiler
R package: purrr v1.2.1	Wickham and Henry	https://purrr.tidyverse.org/
R package: openxlsx v4.2.8	Schauberger and Walker	https://ycphs.github.io/openxlsx/index.html
R package: DOSE v4.4.0	Yu et al79	https://www.bioconductor.org/packages//2.12/bioc/html/DOSE.html
R package: stringr v1.6.0	Wickham	https://stringr.tidyverse.org
R package: scales v1.4.0	Wickham et al.	https://scales.r-lib.org
R package: GSVA v2.4.4	Hänzelmann et al.80	https://github.com/rcastelo/GSVA
R package: Rcpp v1.1.1	Eddelbuettel81	https://github.com/cran/Rcpp
R package: gtools v3.9.5	Warnes et al.	https://rscholar.com/pkg/gtools
R package: hdf5r v1.3.12	Hoefling and Annau	https://github.com/hhoeflin/hdf5r/
R package: sccore v1.0.6	Petukhov et al.	https://github.com/kharchenkolab/sccore
R package: conos v1.5.2	Petukhov et al.	https://github.com/kharchenkolab/conos
R package: dplyr v1.2.0	Wickham et al.	https://dplyr.tidyverse.org
R package: patchwork v1.3.2	Pedersen	https://patchwork.data-imaginist.com
R package:reshape2 v1.4.5	Wickham82	https://cran.r-project.org/web/packages/reshape2/index.html
R package: RColorBrewer v1.1.3	Neuwirth	https://CRAN.R-project.org/package=RColorBrewer
R package: viridis 0.6.5	Garnier et al.	https://sjmgarnier.github.io/viridis/
R package: KEGGREST v1.50.0	Tenenbaum and Maintainer	https://bioconductor.org/packages/KEGGREST
R package: org.Mm.eg.db v3.17.0	Carlson	https://bioconductor.org/packages/org.Mm.eg.db/
R package: scMetabolism v0.2.1	Wu et al.83	https://github.com/wu-yc/scMetabolism
MACS2 v2.2.9.1	Zhang et al.84	https://github.com/macs3-project/MACS
deepTools v3.5.6	Ramírez et al.85	https://deeptools.readthedocs.io/en/latest/
featureCounts v2.0.2	Liao et al.86	https://subread.sourceforge.net/featureCounts.html
HOMER v5.1	Heinz et al.87	http://homer.ucsd.edu/homer/
Other
LS Separation columns	Miltenyi Biotec	Cat#130-042-401
gentleMACS 25 C Tubes	Miltenyi Biotec	Cat#130-093-237
gentleMACS Perfusers	Miltenyi Biotec	Cat#130-128-151
RBC lysis buffer	BioLegend	Cat#420302
gentleMACS Octo Dissociator	Miltenyi Biotec	Cat#130-096-427
Collagen I	Thermo Fisher Scientific	Cat#A1048301
Lipoprotein Deficient Serum from fetal calf	Sigma-Aldrich	Cat#S5394
Fetal Bovine Serum	Nucleus Biologics	Cat#1824-003

Highlights.

• Host MTDH links hepatic lipid metabolism to CD8⁺ T cells dysfunction

• Tumor EVPs trigger hepatic MTDH-dependent systemic lipid metabolism reprogramming

• Dual MTDH loss in hepatocytes and CD8⁺ T cells enhances antitumor immunity

• Targeting host MTDH improves anti-PD-1 therapy across tumor models