Cancer suppresses mitochondrial chaperone activity in macrophages to drive immune evasion

Abstract

Contrary to tumor-infiltrating T cells with dysfunctional mitochondria, tumor-associated macrophages (TAMs) preserve their mitochondrial activity in the nutrient-limited tumor microenvironment (TME) to sustain immunosuppression. We identify TNF-receptor associated protein-1 (TRAP1), a mitochondrial HSP90 chaperone, as a metabolic checkpoint that restrains oxidative respiration and limits macrophage suppressive function. In the TME, TRAP1 is downregulated through TIM4-AMPK signaling, and its loss enhances immunoinhibitory activity, limits pro-inflammatory capacity, and promotes tumor immune escape. Mechanistically, TRAP1 suppression augments electron transport chain activity and elevates the α-ketoglutarate/succinate ratio, remodeling mitochondrial homeostasis. The resulting accumulation of α-ketoglutarate further potentiates JMJD3-mediated histone demethylation, establishing transcriptional programs that reinforce an immunosuppressive state. Restoring TRAP1 by targeting TIM4 and JMJD3 reprograms TAMs, disrupts the immune-evasive TME, and bolsters anti-tumor immunity. These findings establish TRAP1 as a critical regulator integrating metabolic and epigenetic control of suppressive TAM function and position the TRAP1 pathway as a promising target for cancer immunotherapy.

Introduction

Mitochondria play a pivotal role in modulating the differentiation and function of immune cells within the inflammatory microenvironment^1,2. During tumorigenesis, immune cells that infiltrate the tumor undergo profound metabolic adaptations to withstand the nutrient-limited tumor microenvironment (TME)³. This metabolic reprogramming is essential, as tumors place a considerable bioenergetic burden on immune cells, which can potentially compromise their ability to mount an effective anti-tumor response^4,5. Unlike tumor-infiltrating lymphocytes (TILs) which exhibit compromised mitochondrial function and immunological exhaustion^6-8, tumor-associated macrophages (TAMs) display intact and enhanced mitochondrial function to support their metabolic demands, characterized by elevated respiration and heightened tricarboxylic acid (TCA) cycle metabolism^9,10. This augmented mitochondrial function in TAMs engages in a dialogue with epigenetic elements, shifting macrophage transcriptional landscape toward an immunosuppressive phenotype^10-12, which favors tumor immune evasion and limits the effectiveness of immunotherapies. However, the underlying mechanisms through which tumor macrophages fine-tune their mitochondrial metabolic processes to adapt to the diverse challenges of the TME, and how these adaptations are linked to epigenetic modifications that contribute to immunological differentiation and function, remain elusive.

TNF-receptor associated protein-1 (TRAP1), a mitochondrial chaperone belonging to the heat shock protein (HSP) 90 family, is involved in regulating the bioenergetic metabolism in cancer cells through direct interactions with mitochondrial respiratory complexes¹³. Despite its significant interest due to its implications in neoplastic progression, TRAP1 displays a fascinating dualistic nature, functioning either as an oncogene or as a tumor suppressor depending on the cancer type and stage. Consequently, outcomes from studies involving TRAP1-targeted therapy for cancer treatment show only partial overlap^14,15. This intriguing observation hints at the existence of functions associated with TRAP1 that are far more intricate than currently understood. The gap in our knowledge of TRAP1 in immune cells, coupled with the lack of TRAP1-specific pharmacological inhibitors or suitable animal models, further obfuscates its roles in tumorigenesis.

Here, we uncover that TRAP1 expression is markedly curbed in immunosuppressive TAMs within the TME, as demonstrated by analyses of both cancer patient samples and tumor-bearing mice. We generated TRAP1-deficient mice, revealing that TRAP1 ablation significantly enhances macrophage immunoinhibitory function, promoting tumorigenesis. Mechanistically, we found that TRAP1 repression is driven by phosphatidylserine (PS) exposed in the TME, which engages the T cell immunoglobulin and mucin domain containing 4 (TIM4) on macrophages¹⁶. This interaction activates AMP-activated protein kinase (AMPK), leading to downstream repression of TRAP1 transcription. While the TIM4-AMPK axis has been linked to immunoinhibitory activity and poor cancer prognosis¹⁷, our findings uncover a novel connection between this pathway and the regulation of mitochondrial chaperone TRAP1 in TAMs. Importantly, TRAP1 inhibition enhances mitochondrial function, which not only supports biogenesis but also increases the production of α-ketoglutarate (α-KG), a key metabolite for Jumonji domain containing 3 (JMJD3)-dependent histone demethylation. This metabolic circuit reinforces the epigenetic programming that drives the immunosuppressive differentiation and function of TAMs. Taken together, our study highlights a previously uncharacterized role of TRAP1 as a novel metabolic checkpoint that orchestrates mitochondrial metabolic program and epigenetic modification in TAMs, offering crucial insights for developing therapeutic strategies to overcome the immunosuppressive effects of macrophages and potentially enhance the efficacy of cancer immunotherapy.

Results

Loss of TRAP1 associates with immunosuppression in cancer patients and mice.

Given the limited information on TRAP1 in immune cells, we first examined its expression profile and found that TRAP1 is broadly expressed across multiple immune cell types in various mouse tissues (Extended Data Fig. 1a,b). Notably, TRAP1 expression was relatively higher in tumoral macrophages from tumor-bearing mice (Extended Data Fig. 1c). We confirmed its mitochondrial localization in macrophages (Extended Data Fig. 1d), similar to its known distribution in cancer cells¹⁸. To further investigate the relationship between TRAP1 and macrophage function in tumors, we analyzed colon cancer patient samples stained for CD68, TRAP1, CD163, and CD86. Immunosuppressive marker CD163 was highly expressed in low TRAP1-expressing (TRAP1^lo) macrophages, while the pro-inflammatory marker CD86 was highly expressed in high TRAP1-expressing (TRAP1^hi) macrophages (Fig. 1a). Similar results were observed in single-cell RNA sequencing (scRNA-seq) data from colon cancer patients (GSE146771) (Extended Data Fig. 1e). Moreover, TIMER2.0 analysis (https://cistrome.shinyapps.io/timer/)^19,20 revealed a negative correlation between TRAP1 expression and myeloid suppressive genes (MRC1, CD163, CD274, PDCD1LG2), as well as T cell exhaustion markers (TIGIT, LAG3, TIM3, PDCD1, CTLA4) in cutaneous melanoma patients (Extended Data Fig. 1f). We next compared TRAP1 expression in tumoral macrophages versus splenic macrophages (non-tumoral) from YUMM1.7 and B16-F10 melanoma-bearing mice and found a significant reduced TRAP1 in tumoral macrophages (Fig. 1b and Extended Data Fig. 1g). A similar decrease was also observed relative to peritoneal macrophages, and was accompanied by a marked increase in the immunosuppressive marker CD206 (Extended Data Fig. 1h, i). Stratifying tumoral macrophages into TRAP1⁺ and TRAP1⁻ groups (Fig. 1c) revealed that TRAP1⁻ macrophages expressed higher levels of immunosuppressive markers CD206 and ARG1, whereas TRAP1⁺ macrophages displayed elevated pro-inflammatory markers CD86 and MHCII (Fig. 1d-g), mirroring the pattern observed in human tumors (Fig. 1a). Consistently, mouse bone marrow-derived macrophages (BMDMs) cultured in tumor conditioned medium (TCM) in vitro showed significant downregulation of TRAP1 and robust induction of immunosuppressive differentiation (Fig. 1h-j). To assess transcriptional changs, we conducted unbiased RNA-seq and real-time quantitative PCR (RT-qPCR) analyses on TCM-treated macrophages. These cells exhibited reduced Trap1 expression, coupled with upregulation of immunosuppressive genes and a downregulation of pro-inflammatory genes (Fig. 1k and Extended Data Fig. 1j). Together, these findings indicate that reduced TRAP1 expression is closely associated with enhanced immunosuppression in macrophages in both human and murine tumors.

Fig. 1. TRAP1 expression negatively correlates with macrophage immunosuppression in human and murine tumors.

(a) Representative immunofluorescent images of human colon cancer tissue stained for CD68, TRAP1, CD86, and CD163. White arrow indicates CD68⁺ macrophage with TRAP1^hiCD86^hiCD163^lo expression, while yellow arrow indicates CD68⁺ macrophage with TRAP1^loCD86^loCD163^hi expression. Scale bar = 50 μm. Quantitative plots of mean fluorescence intensity for CD163 and CD86 in TRAP1^lo and TRAP1^hi macrophages are shown (n = 17). A.U., arbitrary units. (b) Representative flow cytometry plots (left) and quantitative analysis (right) of TRAP1⁺ macrophages in the spleen and tumor from YUMM1.7 tumor-bearing mice on day 16 (n = 5 mice per group). Data are representative of two independent experiments. (c-g) Representative macrophage population based on TRAP1 expression (TRAP1⁺ and TRAP1⁻) in tumor from YUMM1.7 tumor-bearing mice on day 16 (c). Quantitative plots of geometric mean fluorescence intensity (gMFI) of CD206 (d), ARG1 (e), CD86 (f) and MHC-II (g) in TRAP1⁻ and TRAP1⁺ macrophages in tumor tissues from YUMM1.7 tumor-bearing mice on day 16 (n = 5 mice per group). Data are representative of two independent experiments. (h-j) Representative histogram (left) and quantitative analysis (right) of TRAP1 (h), CD206 (i) and ARG1 (j) from BMDMs in the presence or absence of YUMM1.7 TCM or LLC TCM for 24 h (n = 4). Data are representative of four independent experiments. (k) Expression of immunosuppressive or pro-inflammatory genes in BMDMs cultured either in the absence (Ctrl) or presence of YUMM1.7 tumor cell-conditioned medium (TCM) for 24 h (n = 3), assessed by RNA-seq analysis. All data are mean ± s.e.m. and were analyzed using a two-tailed, paired Student’s t-test (a-g) or one-way ANOVA with Sidak’s multiple comparison test (h-j).

Loss of TRAP1 in TAMs drives tumor immune escape

To elucidate the role of TRAP1 in macrophages, we generated Trap1^−/− mice on a C57BL/6 genetic background and performed unbiased RNA-seq on TCM-treated macrophages derived from Trap1^+/+ and Trap1^−/− mice. Pathway enrichment and gene ontology (GO) analysis showed that TRAP1 ablation markedly enhanced pathways related to angiogenesis, macrophage differentiation, cell growth, lipid metabolism, and mitochondrial respiratory function (Fig. 2a), processes well-recognized to underline immunoinhibitory macrophage phenotypes^9-11,21. TRAP1 deletion increased both transcript (Fig. 2b) and protein expression (Fig. 2c, d and Extended Data Fig. 2a) of immunosuppressive markers (CD206, ARG1, and CD301) in macrophages exposed to TCM or tumor cells, while concomitantly reducing their pro-inflammatory capacity (Extended Data Fig. 2b, c). Notably, overexpression of TRAP1 in Trap1^−/− macrophages reversed this hyper-immunosuppressive phenotype (Extended Data Fig. 2d-f), demonstrating that TRAP1 restrains macrophage suppressive function in tumors.

Fig. 2. TRAP1 deficiency enhances immunosuppressive activity in TAMs and promotes tumor escape.

(a) Gene set enrichment analysis was performed, and enrichment scores are shown for gene ontology (GO) pathway enrichment in TRAP1-deficient BMDMs treated with YUMM1.7 TCM compared with wild-type BMDMs treated with YUMM1.7 TCM. NES, normalized enrichment score. (b) Expression of genes encoding molecules associated with macrophage immunosuppressive function in TCM-treated BMDMs from Trap1^+/+ or Trap1^−/− mice, assessed by RNA-seq analysis (n = 3 in Trap1^+/+ group; n = 4 in Trap1^−/− group). (c and d) Representative histogram (left) and quantitative analysis (right) of CD206 (c) and ARG1 (d) in TRAP1 wild-type and knockout BMDMs in the absence (−) or presence (+) of YUMM1.7 TCM (n = 4). Data are representative of three independent experiments. (e-g) Tumor growth (e), tumor weight (f), and representative tumor picture (g) of YUMM1.7 melanoma from Trap1^wt and Trap1^cKO tumor-bearing mice. Data were taken from tumors harvested on day 16 post-tumor transplantation (n = 5 mice per group). Each symbol represents one individual. Data are representative of two independent experiments. (h) Absolute number of TAMs, frequencies of CD206⁺ TAMs, ARG1⁺ TAMs, PD-L1⁺ TAMs, PD-L2⁺ TAMs, CD86⁺ TAMs, and MHC-II⁺ TAMs, absolute number of TILs, frequencies of IFNγ⁺ CD8 and IFNγ⁺ CD4 T cells in Trap1^wt and Trap1^cKO tumor-bearing mice from day 16 tumors (n = 5 mice per group). Each symbol represents one individual mouse. (i) Survival analysis between Trap1^wt and Trap1^cKO mice bearing YUMM1.7 melanoma (n = 5 mice per group). Data are representative of two independent experiments. (j) Proliferation of CTV-labeled CD8⁺ T cells activated with anti-CD3 and anti-CD28, and co-cultured with tumoral macrophages (TAMs) isolated from Trap1^wt or Trap1^cKO YUMM1.7 tumor-bearing mice at a 1:10 ratio for 72 h (n = 4). Data are representative of two independent experiments. (k) Production of IFNγ (top) and granzyme B (GZMB; bottom) in CD8⁺ T cells activated with anti-CD3 and anti-CD28, and co-cultured with tumoral macrophages (TAMs) isolated from Trap1^wt or Trap1^cKO YUMM1.7 tumor-bearing mice at a 1:10 ratio for 72 h (n = 4). Data are representative of two independent experiments. All data are mean ± s.e.m. and were analyzed using a one-way ANOVA with Sidak’s multiple comparison test (c, d, j, k), two-way ANOVA with Tukey’s multiple comparison test (e), or Mantel-Cox test for survival (i), two-tailed, unpaired Student’s t-test (f, h).

To assess the intrinsic role of TRAP1 in myeloid cells during tumor progression, we generated Trap1^−/− mice on a Rag1^−/− genetic background. In the context of NK cell depletion (Extended Data Fig. 3a), TRAP1 loss enhanced tumor development (Extended Data Fig. 3b-d) and increased intratumoral CD206⁺ and ARG1⁺ TAMs (Extended Data Fig. 3e, f). We therefore hypothesized that TRAP1-deficient macrophages reshape the TME by fostering a strongly immunosuppressive milieu that promotes tumor progression. Supporting this, co-injection of tumor cells with Trap1^−/− BMDMs into wild-type mice accelerated tumor growth and mass, whereas TRAP1 overexpression (Trap1^O/E) reversed these effects (Extended Data Fig. 3g, h). Tumors with TRAP1-deficient macrophages exhibited greater infiltration of immunosuppressive TAMs and fewer tumor-infiltrating lymphocytes (TILs) with reduced frequencies of IFNγ⁺ CD8⁺ T cells, all of which were rescued by TRAP1 overexpression (Extended Data Fig. 3i-n). Moreover, TRAP1-deficient macrophages increased regulatory T cells (Tregs) frequencies, and this scenario was similarly reversed by restoring TRAP1 (Extended Data Fig. 3o). Altogether, these findings indicate that TRAP1 loss intrinsically reprograms macrophages into a hyper-immunosuppressive TAM phenotype, thereby facilitating tumor immune evasion.

We next developed Trap1 conditional knockout mice (Trap1^cKO) by crossing LysMCre with Trap1^fl/fl mice. Upon YUMM1.7 melanoma challenge, Trap1^cKO mice exhibited significantly increased tumor growth and weight, indicating impaired tumor control (Fig. 2e-g). Tumors from Trap1^cKO mice exhibited more TAMs expressing CD206, ARG1, PD-L1, and PD-L2, but fewer co-stimulatory molecules CD86 and MHC-II, alongside with reduced TILs and lower frequencies of IFNγ-producing CD8⁺ and CD4⁺ T cells (Fig. 2h and Extended Data Fig. 3p). TRAP1 deficiency also decreased tumor-infiltrating dendritic cells (DCs), particularly cross-presenting DC1s, potentially limiting TIL activation. Meanwhile, the frequencies of DC2s, neutrophils, monocytes, and NK cells were unaffected (Extended Data Fig. 3p). This excessive presence of immunosuppressive macrophages together with reduced anti-tumor T responses correlated with decreased overall survival in Trap1^cKO tumor-bearing mice (Fig. 2i). Importantly, the pro-tumorigenic effect of TRAP1 deficiency extended beyond melanoma, as Trap1^cKO mice also developed larger tumors following with Lewis lung carcinoma (LLC) challenge (Extended Data Fig. 3q, r).

To determine whether TRAP1-deficinet macrophages functionally suppress T cells, we performed ex vivo co-culture assays. TAMs isolated from Trap1^cKO tumor-bearing mice exhibited greater capacity to attenuate CD8⁺ T cell proliferation and IFNγ/granzyme B production (Fig. 2j, k). Similar suppression was observed when CD8⁺ T cells were co-cultured with TCM-treated BMDMs (Extended Data Fig. 3s).

Although LysMCre is commonly used to target macrophages, it can also drive recombination in neutrophils²², raising the possibility that TRAP1 loss in neutrophils contributes to the phenotype. To test this, we depleted neutrophils in Trap1^wt and Trap1^cKO mice. Neutrophil depletion had minimal impact on tumor burden, TAM numbers or phenotype, TIL frequency, and the proportion of IFNγ⁺ CD8⁺ and IFNγ⁺ CD4⁺ T cells (Supplementary Data Fig. 1a-i), indicating that the pro-tumorigenic effects in Trap1^cKO mice are not attributable to neutrophils. Collectively, our data demonstrate that TRAP1 deficiency in macrophages promotes their accumulation and immunosuppressive function, impairs T cell-mediated anti-tumor immunity, and facilitates tumor immune evasion.

TRAP1 deficiency promotes macrophage alternative activation

TAMs are generally considered to resemble alternatively activated macrophages (commonly described as M2), induced by the Th2 cytokine IL-4^10,23,24. To systematically examine how TRAP1 regulates this anti-inflammatory program, we performed RNA-seq on IL-4-stimulated macrophages from Trap1^+/+ and Trap1^−/− mice. GO analysis showed that TRAP1 loss enhanced pathways involved in cytokinesis, cell proliferation, mitochondrion morphogenesis, lysosomal transport, fatty acid oxidation, serine family amino acid metabolic process, and amino acid transport by TRAP1 deletion (Supplementary Data Fig. 2a), all of which are crucial for macrophage alternative activation^10,21,24. In contrast, MAPKK activity, TLR4 signaling, and Th1 immune response pathways were downregulated (Supplementary Data Fig. 2a). Consistently, TRAP1-deficient macrophages displayed higher levels of anti-inflammatory genes in response to IL-4 (Supplementary Data Fig. 2b), supported by RT-qPCR and flow cytometric analyses both in vitro (Supplementary Data Fig. 2c-e) and in vivo (Supplementary Data Fig. 2f, g). TRAP1 loss also increased Ki67 expression in IL-4 stimulated peritoneal macrophages, indicating enhanced proliferation (Supplementary Data Fig. 2h). Conversely, TRAP1 overexpression in Trap1^−/− macrophages suppressed anti-inflammatory markers, including CD206, CD301, and RELMα (Supplementary Data Fig. 2i, j). Together, these results demonstrate that TRAP1 negatively regulates macrophage alternative activation, and its absence promotes polarization toward anti-inflammatory differentiation and function.

TIM4-AMPK signaling modulates TRAP1 expression in TAMs

We next sought to define the mechanism underlying TRAP1 reduction in immunosuppressive TAMs. AMP-activated protein kinase (AMPK) is a central energy sensor activated by AMP/ATP imbalance and other stress signals in the TME²⁵, and has been implicated in mitochondrial biogenesis²⁶ and anti-inflammatory programming in macrophages during tissue regeneration^27,28. We therefore tested whether AMPK activation mediates TRAP1 downregulation in TAMs. Consistent with prior studies²⁹, tumoral macrophages displayed significantly higher AMPK activation than non-tumoral splenic and peritoneal macrophages, as indicated by elevated p-AMPKα levels in vivo and in vitro (Fig. 3a and Extended Data Fig. 4a, b). Genetic knockdown or pharmacological inhibition of AMPK with Compound C (CC) (ref.²⁹), effectively suppressed AMPK activation (Fig. 3b, c and Extended Data Fig. 4c), which counteracted TRAP1 suppression (Fig. 3d and Extended Data Fig. 4d) and curtailed immunosuppressive marker expression in TAMs (Fig. 3d-f and Extended Data Fig. 4e-g), while increasing pro-inflammatory genes (Tnf, Cd86, Nos2, Il1b; Extended Data Fig. 4h, i). By contrast, AMPK inhibition failed to reverse the immunosuppressive phenotype in TRAP1-deficient macrophages (Extended Data Fig. 4j, k), indicating that AMPK regulates macrophage immunosuppressive function in a TRAP1-dependent manner.

Fig. 3. TIM4-AMPKα signaling axis controls TRAP1 expression in TAMs.

(a) Representative histogram (left) and quantitative analysis (right) of p-AMPKα⁺ macrophages in the tumor and spleen of YUMM1.7 tumor-bearing mice on day 16 (n = 5). Data are representative of two independent experiments. (b-f) BMDMs were transduced with retrovirus-expressing Prkaa1 (Prkaa1^sh) or luciferase (Luc^sh) shRNA, and treated with (+) or without (−) YUMM1.7 TCM for 24 h. Representative immunoblot analysis of p-AMPKα, AMPKα, TRAP1 and β-Actin (b), quantification of p-AMPKα to total AMPKα ratio (n = 3) (c), RT-qPCR analysis of Trap1, Mrc1, Arg1, and Vegfa in TCM-treated BMDMs (n = 4, normalized to Luc^sh controls) (d), and flow cytometry analysis of CD206 (e) and ARG1 (f) expression (n = 3). Data are representative of two independent experiments. (g) Representative histogram (left) and quantitative analysis (right) of TIM4⁺ macrophages in the tumor and spleen from YUMM1.7 tumor-bearing mice on day 16 (n = 5). Data are representative of two independent experiments. (h) RNA-seq data (GSE157673) analysis for the expression of TRAP1 in TIM4⁺ and TIM4⁻ tumor-associated peritoneal macrophages from tumor-bearing mice. (i, j) Immunoblot analysis of p-AMPKα, AMPKα, TRAP1, and β-Actin in Timd4^+/+ or Timd4^−/− BMDMs treated with (+) or without (−) YUMM1.7 TCM for 24 h (i), quantification of p-AMPKα to total AMPKα ratio (n = 3) (j). Data are representative of two independent experiments. (k) Expression of Trap1, Mrc1, Arg1 and Vegfa in Timd4^+/+ or Timd4^−/− BMDMs treated with YUMM1.7 TCM for 24 h, assessed by RT-qPCR (n = 4, normalized to naive wild-type macrophages). Data are representative of two independent experiments. (l, m) Representative histogram (left) and quantitative analysis (right) of CD206 (l) and ARG1 (m) in Timd4^+/+ or Timd4^−/− BMDMs in the presence or absence of YUMM1.7 TCM for 24 h, assessed by flow cytometry (n = 4). Data are representative of two independent experiments. Box plots display data distributions, with box limits representing the first and third quartiles, the center line denoting the median, and whiskers extending to 1.5 times the interquartile range. All data are mean ± s.e.m. and were analyzed using a two-tailed, unpaired Student’s t-test (d, h, k) or a two-tailed, paired t-test (a, g) or one-way ANOVA with Sidak’s multiple comparison test (c, e, f, j, l, m).

T cell immunoglobulin and mucin domain containing 4 (TIM4), a phosphatidylserine (PS) receptor highly expressed on TAMs and associated with unfavorable cancer prognosis^17,30, has been implicated in activating AMPK signaling to regulate antigen presentation and inflammation in macrophages^17,31. We found TIM4 significantly upregulated in tumoral macrophages from YUMM1.7 tumor-bearing mice (Fig. 3g). Notably, although non-tumoral peritoneal macrophages exhibited high basal TIM4 expression, they neither activated AMPK (Extended Data Fig. 4a) nor acquired an immunosuppressive phenotype (Extended Data Fig. 4l), indicating that TIM4 expression alone is insufficient to trigger this signaling cascade outside the tumor context. We analyzed a publicly available RNA-seq dataset (GSE157673)³² and found that TIM4⁺ TAMs exhibited reduced TRAP1 expression (Fig. 3h). Genetic TIM4 knockout and TIM4 blockade decreased AMPK activation and restored TRAP1 expression in tumoral macrophages (Fig. 3i-k and Extended Data Fig. 4m, n). Functionally, TIM4 inhibition restricted immunosuppressive marker expression (Fig. 3l, m and Extended Data Fig. 4o) and increased pro-inflammatory gene expression (Extended Data Fig. 4p), phenocopying the effects of AMPK inhibition. These findings suggest that TIM4 coordinates with AMPK signaling to regulate TRAP1 abundance and immunoinhibitory capacity in TAMs.

Apoptotic cells (AC) are major source of PS and contribute to immunosuppressive signals within the TME³³. We found PS levels significantly enriched in YUMM1.7-derived TCM (Extended Data Fig. 5a), leading us to hypothesize that AC exposure could recapitulate the effects of TCM by activating TIM4-AMPK signaling in macrophages. Indeed, AC treatment of naïve BMDMs induced TIM4 expression and AMPK activation (Extended Data Fig. 5b, c), resulting in TRAP1 downregulation and enhanced immunosuppression (Extended Data Fig. 5d, e), closely resembling TCM effects. TIM4 blockade abrogated the AC-induced AMPK activation and TRAP1 repression (Extended Data Fig. 5b-e), underscoring a central role for TIM4 in these responses. Similarly, direct PS exposure activated TIM4-AMPK signaling, reduced TRAP1 expression, and promoted immunosuppressive polarization (Extended Data Fig. 5f-i). Hypoxia, a hallmark of the TME, also enhanced TIM4-AMPK signaling (Extended Data Fig. 6a, b), further suppressed downstream TRAP1 (Extended Data Fig. 6c), and potentiated the immunosuppressive phenotype of TCM-treated BMDMs (Extended Data Fig. 6d, e). Together, these findings demonstrate that multiple TME-associated signals, including PS and hypoxia, converge on TIM4-AMPK signaling to repress TRAP1 and drive immunosuppressive programming of TAMs.

TIM4-AMPKα axis suppresses TRAP1 expression via HSF1 inactivation

TRAP1 transcription is regulated by heat shock factor-1 (HSF1) in response to pathological challenges³⁴. HSF1 activation requires a multi-step process involving phosphorylation and nuclear translocation³⁵. AMPK has been shown to phosphorylate HSF1 at serine 121 (Ser121), retaining it in the cytosol and preventing activation (Fig. 4a)³⁶. We hypothesized that TIM4-AMPK signaling suppresses TRAP1 by restricting HSF1 activation. Supporting this, tumoral macrophages exhibited elevated p-HSF1(Ser121) compared with splenic macrophages (Fig. 4b), as well as in vitro models (Fig. 4c, d). AMPK inhibition or upstream TIM4 blockade reduced p-HSF1(Ser121) and restored TRAP1 expression (Fig. 4c, d), whereas DTHIB, a selective HSF1 inhibitor, abrogated this recovery (Fig. 4e). DTHIB also increased CD206 and ARG1 expression (Fig. 4f, g), indicating reinforced immunosuppressive programming. Taken together, these results illustrate that the activation of TIM4-AMPK signaling suppresses TRAP1 expression through HSF1 inactivation.

Fig. 4. AMPKα signaling suppresses TRAP1 expression through HSF1 inactivation.

(a) Schematic illustrating AMPK phosphorylation of HSF1 at Ser121, retaining HSF1 in the cytosol and blocking TRAP1 transcription. Figure created with BioRender. (b) Representative histogram (left) and quantitative analysis (right) of p-HSF1(S121)⁺ macrophages in the tumor and spleen of YUMM1.7 tumor-bearing mice on day 16 (n = 5). Data are representative of two independent experiments. (c) Immunoblot analysis of p-AMPKα, AMPKα, p-HSF1(S121), HSF1, and β-Actin from BMDMs treated with (+) or without (−) YUMM1.7 TCM in the presence or absence of Compound C (CC). Data are representative of three independent experiments. (d) Immunoblot analysis of p-AMPKα, AMPKα, p-HSF1(S121), HSF1, and β-Actin from BMDMs treated with (+) or without (−) YUMM1.7 TCM in the presence of αTIM4 or IgG2b. Data are representative of two independent experiments. (e-g) Representative histogram (left) and quantitative analysis (right) of TRAP1 (e), CD206 (f), and ARG1 (g) in BMDMs treated with YUMM1.7 TCM for 24 h in the presence or absence of CC, αTIM4, or DTHIB (n = 3). Data are representative of two independent experiments. All data are mean ± s.e.m. and were analyzed using a two-tailed, paired Student’s t-test (b) or one-way ANOVA with Sidak’s multiple comparison test (e-g).

TRAP1 regulates mitochondrial homeostasis in suppressive macrophages

Our RNA-seq analysis revealed that TRAP1 deletion markedly increased the expression of genes associated with mitochondrial metabolism in macrophages responding to the TME and IL-4 stimulation (Fig. 2a and Supplementary Data Fig. 2a). Given that mitochondrial respiration and fitness are essential for sustaining the immunoinhibitory function of macrophages in tumor and anti-inflammatory environments^9-11, we hypothesized that TRAP1 may act as a metabolic checkpoint regulating mitochondrial homeostasis. Analysis of publicly available single-cell RNA-seq datasets revealed upregulation of oxidative phosphorylation (OXPHOS) and the electron transport chain (ETC) genes in tumoral macrophages from patients with lung and liver cancer (Supplementary Data Fig. 3a)^37,38. Consistently, tumoral macrophages from Trap1^cKO mice displayed higher mitochondrial respiratory activity than WT controls (Fig. 5a). Macrophages exposed to TCM showed similarly enhanced mitochondrial biogenesis, indicated by increased ratio of mitochondrial DNA to nucleic DNA (mtDNA/nDNA; Fig. 5b), mitochondrial mass (Fig. 5c), membrane potential (Fig. 5d), and respiratory activity (Fig. 5e-g), all of which were further augmented by TRAP1 deficiency. SCENITH analysis, a technique that links translational output with metabolic activity³⁹, confirmed elevated protein synthesis in Trap1^−/− macrophages, indicating intensified bioenergetic activity driven by TRAP1-mediated metabolic rewiring (Supplementary Data Fig. 3b). TRAP1 deficiency also upregulated ETC gene and protein expression (Fig. 5h, i), with elevated activities of complex I-IV (Fig. 5j). TRAP1 overexpression (Trap1^O/E) in Trap1^−/− macrophages, however, halted enhancement in mitochondrial respiration (Fig. 5k).

Fig. 5. TRAP1 deficiency enhances mitochondrial respiration and function in immunosuppressive macrophages.

(a) The basal OCR of tumoral macrophages (TAMs) isolated from Trap1^wt or Trap1^cKO tumor-bearing mice, assessed using Seahorse Flux Analyzer (n = 6). Data are representative of three independent experiments. (b) The ratio of mtDNA/nDNA in TRAP1 wild-type or knockout BMDMs cultured with YUMM1.7 TCM for 24 h, assessed by RT-qPCR analysis (n = 6, normalized to naïve wild-type macrophages). Data are representative of three independent experiments. (c, d) Representative histogram (left) and quantitative analysis (right) of MitoTracker Green (MTG) staining (c) and MitoTracker Orange (MTO) staining (d) in TRAP1 wild-type or knockout BMDMs cultured with (+) or without (−) YUMM1.7 TCM for 24 h (n = 3). Data are representative of two independent experiments. (e) The real-time changes of oxygen consumption rate (OCR) of TRAP1 wild-type or knockout BMDMs treated with (+) or without (−) YUMM1.7 TCM, followed by the sequential treatment with oligomycin (Oligo), FCCP, and rotenone plus antimycin (Rot + Ant) (n = 6). Data are representative of four independent experiments. (f, g) The basal OCR (f) and mitochondrial spare respiratory capacity (SRC) (g) of TRAP1 wild-type or knockout BMDMs treated with (+) or without (−) YUMM1.7 TCM (n = 6), assessed by using Seahorse Flux Analyzer. Data are representative of four independent experiments. (h) Immunoblot analysis of mitochondrial ETC complexes (ATP5A, UQCRC2, MTCO1, SDHB, and NDUFB8), PHB1, and β-Actin from Trap1^+/+ or Trap1^−/− BMDMs treated with (+) or without (−) YUMM1.7 TCM. Data are representative of three independent experiments. (i) Expression of mitochondrial ETC genes in Trap1^+/+ or Trap1^−/− BMDMs cultured with YUMM1.7 TCM for 24 h, assessed by RNA-seq analysis (n = 3 in Trap1^+/+ group; n = 4 in Trap1^−/− group). (j) Activity of complex I, Complex II, Complex III, Complex IV in YUMM1.7 TCM-treated BMDMs from Trap1^+/+ or Trap1^−/− mice (n = 4). Data are representative of two independent experiments. (k) TRAP1 wild-type or knockout BMDMs were transduced with either retrovirus overexpressing a control reporter gene (EV) or a reporter gene plus the Trap1 sequence (Trap1^O/E) and treated with YUMM1.7 TCM, and basal OCR (left) and ratio of OCR to extracellular acidification rate (ECAR) (right) were measured using Seahorse Flux Analyzer (n = 6). Data are representative of two independent experiments. All data are mean ± s.e.m. and were analyzed using a two-tailed, unpaired Student’s t-test (a) or one-way ANOVA with Sidak’s multiple comparison test (b-d, f, g, j, k).

Similarly, upon IL-4 stimulation, TRAP1-null macrophages exhibited a higher mtDNA/nDNA ratio, elevated mitochondrial oxygen consumption rate (OCR) and higher ATP production (Supplementary Data Fig. 3c-e). Mitochondria are the primary energy source of the cell, with energy produced in cristae, tightly organized cristae increase the surface area for the ETC, thereby enhancing energy production⁴⁰. Transmission electron microscopy (TEM) revealed that IL-4-stimulated TRAP1-deficient macrophages contained more mitochondria (Supplementary Data Fig. 3f, g) with greater length and denser cristae (Supplementary Data Fig. 3f, h). Similar to tumoral macrophages, IL-4-stimulated TRAP1-deficient macrophages also displayed elevated ETC gene and protein expression (Supplementary Data Fig. 3i, j), which was counteracted by TRAP1 overexpression (Supplementary Data Fig. 3k). Taken together, these data indicate that TRAP1 deficiency enhances mitochondrial operational efficiency to sustain high metabolic demands in both tumoral and IL-4-stimulated macrophages.

TRAP1 deficiency reprograms glutamine metabolism and alters the α-KG/succinate ratio

Glutamine serves as a key anaplerotic source of α-KG for the TCA cycle, supporting anti-inflammatory polarization of macrophages^11,41. TRAP1 has been implicated in regulating glutamine metabolism in HEK293T cells⁴², prompting us to test whether TRAP1 deficiency alters α-KG production. Transcriptomic analysis showed enrichment of genes involved in glutamine transport and metabolic pathways in TRAP1-deficient macrophages responding to tumors and IL-4 (Fig. 6a and Extended Data Fig. 7a). In line with previous studies⁴³, TCM exposure enhanced glutamine utilization and α-KG production (Fig. 6b), both of which were further elevated by TRAP1 deletion, including under IL-4 stimulation (Extended Data Fig. 7b, c). TRAP1 has been suggested to attenuate succinate dehydrogenase (SDH; ETC complex II) activity in cancer cells, leading to succinate accumulation⁴⁴. We observed increased SDH activity in TAMs from Trap1^cKO tumor-bearing mice (Extended Data Fig. 7d), while α-KG dehydrogenase (αKGDH) activity remained unchanged (Extended Data Fig. 7e), indicating selective enhancement of the TCA cycle at the SDH node. This inhibitory effect of TRAP1 on SDH activity was recapitulated in TCM- and IL-4-treated BMDMs (Fig. 5j and Extended Data Fig. 7f). Consequently, TRAP1 ablation reduced intracellular succinate while increasing in the α-KG/succinate ratio (Fig. 6b and Extended Data Fig. 7g, h), a metabolic state linked to macrophage anti-inflammatory function¹¹. TRAP1 loss also increased fumarate levels and the fumarate/succinate ratio (Fig. 6b), suggestive of augmented succinate oxidation through SDH activity. Together, these findings indicate that TRAP1 deficiency promotes glutaminolysis and SDH activity, thereby boosting mitochondrial respiratory metabolism and elevating the α-KG/succinate ratio to favor immunosuppressive programming. Supporting this, pharmacologic inhibition of SDH with DMM (ref.²) or glutaminase (GLS) with BPTES (ref.⁴⁵ ; Extended Data Fig. 7i) reduced mitochondrial activity (Fig. 6c-f and Extended Data Fig. 7j), lowered α-KG production (Extended Data Fig. 7k), and suppressed immunosuppressive marker expression in Trap1^−/− macrophages exposed to TCM or IL-4 (Fig. 6g-j and Extended Data Fig. 7l-n). Moreover, exogenous supplementation with succinate phenocopied the inhibitory effects of DMM and BPTES (Fig. 6k and Extended Data Fig. 10o-p), underscoring that maintaining an appropriate α-KG/succinate balance is essential for TRAP1-mediated immunosuppressive function.

Fig. 6. TRAP1 deficiency reprograms glutaminolysis and α-KG metabolism.

(a) Expression of Gls, Slc7a7, Slc7a8, Slc38a1, and Slc38a7 in TRAP1 wild-type or knockout BMDMs treated with YUMM1.7 TCM for 24 h, assessed by RNA-seq (n = 3 in Trap1^+/+ group; n = 4 in Trap1^−/− group). (b) Quantification of glutamine consumption, intracellular levels of α-KG and succinate, α-KG/succinate ratio, fumarate levels and fumarate/succinate ratio from TRAP1 wild-type or knockout BMDMs treated with YUMM1.7 TCM (n = 3 or 4). Data are representative of two independent experiments. (c, d) Basal OCR (c) and OCR/ECAR ratio (d) of TRAP1 wild-type or knockout BMDMs treated with YUMM1.7 TCM in the presence or absence (Veh) of DMM (n = 6). Data are representative of two independent experiments. (e, f) Basal OCR (e) and OCR/ECAR ratio (f) of TRAP1 wild-type or knockout BMDMs treated with YUMM1.7 TCM in the presence or absence of BPTES (n = 6). Data are representative of two independent experiments. (g, h) Representative histogram (left) and quantitative plots (right) of CD206 (L) and ARG1 (M) in TRAP1 wild-type and knockout BMDMs treated with YUMM1.7 TCM in the presence or absence of DMM (n = 3). Data are representative of three independent experiments. (i, j) Representative histogram (left) and quantitative plots (right) of CD206 (i) and ARG1 (j) in TRAP1 wild-type and knockout BMDMs treated with YUMM1.7 TCM in the presence or absence of BPTES (n = 3). Data are representative of three independent experiments. (k) Expression of Mrc1, Arg1, and Vegfa genes in TRAP1 wild-type or knockout BMDMs treated with YUMM1.7 TCM for 6 h in the presence or absence of diethyl succinate (Suc), assessed by RT-qPCR analysis (n = 4, normalized to naïve wild-type macrophages). Data are representative of three independent experiments. All data are mean ± s.e.m. and were analyzed using a one-way ANOVA with Sidak’s multiple comparison test (b-k).

Given that TIM4-AMPK signaling negatively regulates TRAP1, we next examined its role in TAM metabolism. AMPK inhibition or knockdown impaired mitochondrial respiration (Extended Data Fig. 8a, b), decreased SDH activity, increased succinate (Extended Data Fig. 8c, d), and reduced α-KG production and the α-KG/succinate ratio (Extended Data Fig. 8e, f). Similarly, TIM4 inhibition reduced mitochondrial respiration and α-KG levels (Extended Data Fig. 8g, h). Considering the established link between fatty acids and mitochondrial OXPHOS in immunosuppressive macrophages²¹, the observed reduction in key genes involved in lipid metabolism further supports the notion of impaired mitochondrial respiration following TIM4 inhibition (Extended Data Fig. 8i, j). Overall, these results underscore the strong dependency on the TIM4-AMPK axis in macrophage, which, together with TRAP1, regulates the α-KG/succinate ratio in the TCA cycle and sustains mitochondrial respiration, essential for immunosuppressive activity.

TRAP1-driven α-KG metabolism facilitates JMJD3 epigenetic regulation

α-KG not only fuels the mitochondrial TCA cycle but also serves as a cofactor for JMJD3-mediated histone demethylation, a process linked to macrophage immunosuppressive polarization^11,12,46. We reasoned that elevated α-KG in TRAP1-deficient macrophages could augment JMJD3 activity. Indeed, TRAP1 loss decreased histone methylation on H3 Lys27 (H3K27) (Fig. 7a), whereas TIM4 inhibition, which restored TRAP1 expression (Fig. 4e) and lowered α-KG levels (Extended Data Fig. 8h), resulted in H3K27 hypermethylation (Extended Data Fig. 9a), independent of JMJD3 expression (Extended Data Fig. 9a, b). CUT&Tag analysis showed that, despite reduced global level of H3K27 trimethylation (H3K27me3), its genomic distribution remained largely unchanged, with 99.5% of peaks overlapping between Trap1^+/+ and Trap1^−/− macrophages (Fig. 7b). Thus, TRAP1 deficiency caused an indiscriminate global reduction in H3K27me3 (Fig. 7b and Extended Data Fig. 9c), reminiscent of previous observations where the inhibition of the H3K27 methyltransferase EZH2 decreased the global H3K27me3 at most peaks without affecting the distribution pattern^47,48. We hypothesized that increased α-KG may enhance the histone demethylase activity of JMJD3, which could, in turn, modulate tumoral macrophage immunity. To address this, we generated a macrophage-specific JMJD3 conditional knockout mice (Jmjd3^cKO) by crossing LysMCre with Kdm6b^fl/fl mice (Extended Data Fig. 9d). JMJD3 deletion led to H3K27 hypermethylation (Fig. 7c) and suppressed the expression of genes and proteins associated with immunosuppressive differentiation in macrophages responding to TCM or tumor cells (Fig. 7d and Extended Data Fig. 9e), while upregulating pro-inflammatory genes (Extended Data Fig. 9f). Importantly, JMJD3 knockdown or pharmacological inhibition with GSK-J4 (ref.⁴⁹), abolished the increased CD206 and ARG1 expression in TRAP1-deficient macrophages (Fig. 7e, f and Extended Data Fig. 9g, h). JMJD3 deficiency also reduced mitochondrial respiration in TAMs from tumor-bearing mice in vivo and decreased ETC gene expression in TCM-stimulated BMDMs in vitro (Extended Data Fig. 9i, j). Collectively, these data suggest that TRAP1-mediated immunosuppression critically depends on JMJD3 activity in macrophages.

Fig. 7. TRAP1-mediated immunosuppressive effect hinges on JMJD3 histone demethylase function in tumor macrophages.

(a) Immunoblot analysis of histone methyl mark H3K27me3 and histone H3 from TRAP1 wild-type and knockout BMDMs treated with YUMM1.7 TCM. Data are representative of three independent experiments. (b) CUT&Tag analysis of H3K27me3. Left. Relative enrichment of H3K27me3 at the combined peaks (n=20,769) in TRAP1 WT and KO is plotted. Note that the enrichment is normalized with sequenced read depth and does not account for the decreased H3K27me3 level in KO (a). The majority of the peak distribution is similar between WT and KO. Two independent biological replicates. Middle. Heatmaps showing H3K27me3 enrichment across all peaks. The right column (“Normalized Trap1”) shows the signals that were normalized based on the abundance of H3K27me3 from (a). Right. Histograms showing the mean signal from the heatmap shown in the Middle. (c) Immunoblot analysis of histone methyl mark H3K27me3 and histone H3 from JMJD3 wild-type (Jmjd3^wt) and knockout (Jmjd3^cKO) BMDMs treated with YUMM1.7 TCM. Data are representative of three independent experiments. (d) Expression of indicated genes in JMJD3 wild-type and knockout BMDMs treated with YUMM1.7 TCM for 24 h, assessed by RNA-seq (n = 3). (e, f) Representative histogram (left) and quantitative plots (right) of CD206 (e) and ARG1 (f) in TRAP1 wild-type or knockout BMDMs transduced with either retrovirus-expressing luciferase shRNA (Luc^sh) or JMJD3 shRNA (Jmjd3^sh) and treated with YUMM1.7 TCM for 24 h (n = 3). Data are representative of two independent experiments. (g-i) Tumor growth (g), tumor weight (h), and representative tumor image (i) of YUMM1.7 melanoma from Jmjd3^wt or Jmjd3^cKO tumor-bearing mice (n = 5 mice per group). Data are representative of two independent experiments. (j) Absolute number of TAMs, frequencies of CD206⁺ TAMs, ARG1⁺ TAMs, PD-L1⁺ TAMs, PD-L2⁺ TAMs, CD86⁺ TAMs, and MHC-II⁺ TAMs, absolute number of TILs, and frequencies of IFNγ⁺ CD8 and CD4 T cells from Jmjd3^wt or Jmjd3^cKO mice. Data were obtained from tumors harvested on day 16 post-tumor transplantation (n = 5 mice per group). Each symbol represents one individual mouse. Data are representative of four independent experiments. (k) Survival analysis from Jmjd3^wt or Jmjd3^cKO mice bearing YUMM1.7 melanoma (n = 16 mice per group). Data were collected from three independent experiments. (l) Proliferation of CTV-labeled CD8⁺ T cells activated with anti-CD3 and anti-CD28, and co-cultured with tumoral macrophages (TAMs) isolated from Jmjd3^wt or Jmjd3^cKO YUMM1.7 tumor-bearing mice at a 1:10 ratio for 72 h (n = 4). Data are representative of two independent experiments. All data are mean ± s.e.m. and were analyzed using a two-tailed, unpaired Student’s t-test (a, c, h, j), one-way ANOVA with Sidak’s multiple comparison test (e, f, l), two-way ANOVA with Tukey’s multiple comparison test (g), or a Mantel-Cox test for survival (k).

Functionally, JMJD3 loss delayed tumor progression. YUMM1.7 melanoma growth and weight were significantly reduced in Jmjd3^cKO mice compared with controls (Fig. 7g-i). This antitumor effect was associated with decreased TAM infiltration and lower expression of CD206, ARG1, PD-L1, and PD-L2, alongside increased CD86 and MHC-II (Fig. 7j and Extended Data Fig. 9k). Jmjd3^cKO also potentiated anti-tumor T cell responses, with increased infiltration of TILs and higher IFNγ production by CD8⁺ and CD4⁺ T cells (Fig. 7j and Extended Data Fig. 9k). Neutrophils, monocytes, NK cells, DC subsets remained unaffected (Extended Data Fig. 9k). These immunological changes correlated with significantly prolonged survival (Fig. 7k). Importantly, Jmjd3^cKO mice also displayed reduced tumor burden following LLC implantation (Extended Data Fig. 9l, m), indicating broad anti-tumor activity.

To determine whether the enhanced anti-tumor response was mediated by macrophages, TAMs from Jmjd3^cKO tumors were isolated and shown to promote CD8⁺ T cell proliferation and effector function, increasing IFNγ and granzyme B production (Fig. 7l, Extended Data Fig. 9n). Neutrophil depletion in WT and Jmjd3^cKO mice had minimal effect on tumor growth, TAM phenotype, TIL abundance, or T cell IFNγ production (Supplementary Data Fig. 4a-i), confirming a macrophage-intrinsic role for JMJD3. Collectively, these findings establish JMJD3 as a key epigenetic regulator downstream of TRAP1-mediated metabolic rewiring.

Targeting TIM4 and JMJD3 boosts anti-tumor immunity

Our data indicate that TRAP1, negatively regulated by TIM4 signaling, functions as a central mediator of macrophage immunosuppression in a JMJD3-dependent manner. To translate these findings, we assessed the therapeutic effects of an αTIM4 monoclonal antibody and the JMJD3 inhibitor GSK-J4 in vivo (Extended Data Fig. 10a). In YUMM1.7 melanoma models, αTIM4 delayed tumor growth (Fig. 8a, b), reduced TAM numbers and immunosuppressive activity (Fig. 8c, d and Extended Data Fig. 10b), and increased TIL infiltration with expanded IFNγ-producing CD8⁺ and CD4⁺ T cells (Fig. 8e-g and Extended Data Fig. 10c). Importantly, the anti-tumor response was significantly enhanced by combining αTIM4 and GSK-J4 (Fig. 8a-g and Extended Data Fig. 10b, c). While neither treatment substantially altered other tumor-infiltrating immune subsets (Extended Data Fig. 10d-f), both extended survival in tumor-bearing mice (Fig. 8h). Importantly, αTIM4 plus GSK-J4 restored TRAP1 expression in TAMs (Fig. 8i), correlating with improved therapeutic outcomes.

Fig. 8. Therapeutic inhibition of TIM4 and JMJD3 signaling curtails tumorigenesis.

(a, b) Tumor growth (a), tumor weight and representative tumor image (b) from mice bearing YUMM1.7 melanoma treated with either IgG2b, GSK-J4, αTIM4, or αTIM4 + GSK-J4. Drug administration is indicated by arrows (n = 10 mice per group). Data were collected from two independent experiments. (c-g) Absolute number of TAMs (c), frequency of CD206⁺ TAMs (d), absolute number of TILs (e), frequency of IFNγ⁺ CD8 (f) or IFNγ⁺ CD4 (g) T cells in treatment groups on day 16 (n = 5 mice per group). Data are representative of two independent experiments. (h) Survival analysis of tumor-bearing mice treated with IgG2b control, GSK-J4, αTIM4, or αTIM4 + GSK-J4 (n = 10 mice per group). Data were collected from two independent experiments. (i) Expression of TRAP1 in TAMs from YUMM1.7 tumor-bearing mice treated with IgG2b, GSK-J4, αTIM4 or αTIM4 + GSK-J4 on day 16 (n = 4 mice per group). Data are representative of two independent experiments. All data are mean ± s.e.m. and were analyzed using a one-way ANOVA with Sidak’s multiple comparison test (b-g, i), two-way ANOVA with Tukey’s multiple comparison test (a) or the Mantel-Cox test for survival (h).

To directly assess the extent to which TRAP1 mediates the effects of TIM4 signaling, we administrated αTIM4 treatment to Trap1^cKO mice. In the absence of TRAP1, TIM4 blockade failed to suppress tumor growth and limit TAM immunosuppression (Extended Data Fig. 10g-j), as TRAP1 expression could not be reinstated (Extended Data Fig. 10k). These results establish TRAP1 as a critical downstream effector of TIM4 signaling, indispensable for TAM immunosuppressive differentiation. Restoring TRAP1 expression through TIM4 blockade and JMJD3 inhibition represents a promising therapeutic strategy to reprogram the TME and reinvigorate anti-tumor immunity.

Discussion

Mitochondrial metabolism directs the differentiation and function of immune cells within the TME^1,12. Unlike tumor-infiltrating T cells, where mitochondrial dysfunction drives exhaustion^6-8, TAMs manifest increased mitochondrial biogenesis^9,10. This includes an active OXPHOS and TCA cycle, supporting macrophage suppressive functions. Yet, the molecular drivers underlying this metabolic adaptation have remained elusive. Here, we identified the mitochondrial chaperone TRAP1 as a regulator of TAM metabolism, where its loss enhanced mitochondrial respiration, increased α-KG metabolism, and skewed macrophages toward an immunosuppressive phenotype through JMJD3-dependent histone demethylation (Extended Data Fig. 10l). Thus, TRAP1 acts as a metabolic checkpoint orchestrating mitochondrial metabolism and epigenetic modification to regulate immunosuppressive function in TAMs.

To investigate the role of TRAP1 in TAMs, we used Trap1^−/− mice on a Rag1^−/− genetic background, an immunodeficient strain lacking adaptive immunity. These mice developed significantly larger tumors with increased TAM infiltration and an enhanced immunosuppressive phenotype. Notably, this effect may be partly attributable to increased expression of Mmp9 and Tgfb1 (Fig. 2b), genes that drive angiogenesis, extracellular matrix remodeling and tumor progression^50,51, likely reinforces the pro-tumorigenic activity of TRAP1-deficient macrophages in vivo. Although the LysMCre system is widely used for gene deletion in macrophages, it may not fully exclude contributions from other myeloid populations, such as neutrophils. However, anti-Ly6G-mediated neutrophil depletion had minimal impact on tumor progression and was consistent with unchanged neutrophil abundance in Trap1^cKO tumor-bearing mice, supporting macrophage-intrinsic role for TRAP1 loss. This interpretation aligns with previous studies showing that absence of neutrophils does not alter primary tumor progression⁵², although their involvement may vary with tumor context.

Mitochondrial SDH activity is critical for macrophage inflammatory programming by controlling succinate oxidation². Despite a previous report suggesting that TAMs harbor lower succinate levels⁹, the regulation of SDH activity in TAMs remains unclear. We observed elevated SDH activity in macrophages in the TME, which coincided with reduced succinate accumulation. Notably, SDH activity was substantially inhibited by TRAP1. This observation aligns with prior findings demonstrating that TRAP1 interacts with and inhibits SDH in tumor cells⁴⁴. The mitochondria-associated enzyme immune responsive gene 1 (IRG1) produces itaconate, which suppresses SDH activity⁵³. We found that TRAP1 loss did not alter IRG1 expression in TAMs, suggesting that TRAP1 modulates SDH activity independent of IRG1. Given the importance of the α-KG/succinate balance in macrophage alternative activation¹¹, TRAP1-deficient macrophages showed increased glutamine utilization and α-KG production with reduced succinate, while αKGDH activity (which converts α-KG to succinyl-CoA) remained unchanged, indicating that α-KG accumulation was not due to impaired catabolism.

Elevated α-KG has been shown to potentiate JMJD3 activity, driving IL-4-induced macrophage polarization through H3K27 demethylation at M2-associated loci^11,46. Employing Jmjd3^cKO mice, we observed that JMJD3 deletion curtailed TAM immunosuppressive capacity and limited tumor progression. These findings hint at a potential therapeutic strategy, specifically targeting glutamine metabolism in myeloid cells to bolster anti-tumor immunity^54,55. It has been suggested that DNA methylcytosine dioxygenase ten-eleven translocation (TET)2, an α-KG-dependent epigenetic regulator, drives the immunological function of tumor-infiltrating myeloid cells, promoting tumor progression⁵⁶. However, whether TET2 or other dioxygenases contribute to TRAP1-mediated immunosuppression within TAMs warrants further study.

AMPK, responsive to energy and nutrient availability, displays distinct roles across tumor-infiltrating immune subsets. While AMPK activation aids CD8⁺ T cells to endure nutrient starvation and stimulate anti-tumor activity⁵⁷, it drives immunosuppressive and pro-tumoral function in myeloid cells²⁹. Yet, the mechanisms through which AMPK enacts these diverse programs remain unclear. We found that activated AMPK signaling interfaces with mitochondria by phosphorylating HSF1 at Ser121, thereby preventing nuclear translocation³⁶ and inhibiting HSF1-dependent TRAP1 transcription. Given that HSF1 inhibits inflammatory response in LPS-primed macrophages^58,59, its role in TAM functional reprogramming uncovers a previously unrecognized axis by which AMPK activation shapes mitochondrial metabolism and immune behavior in tumors.

Upstream of this axis, we identified PS and its receptor TIM4 as key regulators. TIM4, abundantly expressed on antigen-presenting cells, has been linked to suppression of CD8⁺ T cell anti-tumor immunity and activation of LKB1/AMPK to suppress inflammasome activation in macrophages during nonalcoholic fatty liver disease^17,30-32. However, the role of TIM4 in facilitating metabolic rewiring that dictates TAM immune response remains elusive. We found that PS, enriched in TCM derived from YUMM1.7 melanoma cells, activated AMPK in a TIM4-dependent manner. TIM4 blockade attenuated AMPK activation and restored TRAP1 expression, resulted in lessened mitochondrial respiration, reduced α-KG production, and impaired JMJD3-mediated H3K27 demethylation. These results suggest that TIM4-driven AMPK signaling facilitates metabolic adaptation and epigenetic remodeling in TAMs through suppression of TRAP1. PS can also engage other receptors, including Tyro3, MerTK, Axl, and integrin αvβ5⁶⁰, and whether these alternative PS receptors contribute to TRAP1-mediated immunosuppression remains to be determined. Moreover, hypoxia potentiated TIM4-AMPK signaling, further reinforcing TRAP1 suppression and TAM immunosuppressive programming. Collectively, our findings delineate a novel TIM4-AMPK-HSF1 signaling axis that links to multiple TME-associated cues, including apoptotic cell-derived PS and hypoxia, to drive mitochondrial TRAP1 chaperon-mediated metabolic rewiring and epigenetic remodeling in TAMs.

Building on these insights, we propose that TRAP1 acts as a critical metabolic checkpoint downstream of TIM4-AMPK signaling, enabling TAMs to meet elevated metabolic demands while maintaining sustained suppressive functions. Although the role of mitochondrial TRAP1 in immune regulation during tumorigenesis has remained understudied, our work identifies TRAP1 signaling as a potential therapeutic target. Specifically, co-targeting TIM4 and the epigenetic enzyme JMJD3 may remodel the TME into a more pro-inflammatory state. Therapeutic strategies aimed at enhancing TRAP1 expression present a potent immunotherapeutic approach to restrict the suppressive capacities of TAMs and improve cancer immunotherapy outcomes.

Materials and Methods

Mice and in vivo experiments

C57BL/6J, Trap1^−/−, Rag1^−/−, LysMCre, and Kdm6b^fl/fl mice were purchased from the Jackson Laboratory. C57BL/6JGpt-Trap1^em1Cflox/Gpt (Trap1^fl/fl) mice were purchased from GemPharmatech. All mice were bred and maintained in specific pathogen-free conditions under protocols approved by institutional animal care at the Ohio State University, and both male and female mice were used at age 8-12 weeks. Mice were housed in animal rooms for 12 h light-dark cycle, temperature and humidity controlled, between 68–74 ℉ and 30–70%, respectively. Bones of Timd4^−/− mice was kindly provided by T. Merghoub and A. Chow (ref.³⁰). For tumor experiments, 5 × 10⁵ of YUMM1.7 tumor cells were resuspended in 100 μl of ice-cold phosphate-buffered saline (PBS) for subcutaneous injection into the right flank of a mouse. In some experiments, mice were injected intraperitoneally with 250 μg anti-NK 1.1 antibody (Clone PK136; BioXCell) 3 days before and 1 day after tumor transplant (shown as Extended Data Fig. 3h). For neutrophil depletion, mice were administered 50 μg of anti-Ly6G monoclonal antibody (clone 1A8; BioXCell) via intraperitoneal injection on days 7, 10, and 13 after tumor cell transplantation. For co-injection experiments, C57BL/6J mice were injected subcutaneously with 5 × 10⁵ of B16-F10 tumor cells alone, or together with 2 × 10⁵ of either Trap1^+/+ or Trap1^−/− BMDMs. Tumors were monitored and measured every 2 days using calipers starting on day 7 post-tumor injection until day 16. Tumor volume was calculated using the formula $\left[\right(\text{length}\times \text{width}\times {(\text{length}\times \text{width})}^{0.5})\times \pi ∕6]$. For the antibody and/or drug treatments, mice were given an intraperitoneal injection of either 100 μl of immunoglobulin (Ig)G2b (250μg per mouse; Clone LTF-2, BioXCell), anti-TIM4 (250 μg per mouse; Clone RMT4-53, BioXCell), GSK-J4 (2 mg/kg; Selleck Chem), or combination therapy of αTIM4 + GSK-J4 on days 8, 10 and 12 post-tumor injection. Tumors were measured every 2 days starting on day 8 and until day 16. Maximal tumor size cut-off was determined to be 2 cm for non-endpoint studies. For IL-4c experiments, mice were injected intraperitoneally with 300 μl of 3% thioglycolate (Sigma-Aldrich) immediately before IL-4 complexed to the monoclonal antibody anti-IL-4 (IL-4c; containing 5 μg of IL-4 (PeproTech) and 25 μg of anti-IL-4, clone 11B11, BioXcell).

Cell lines

The YUMM1.7 melanoma cell line was obtained as described previously⁶¹. B16-F10 (CRL-6475) melanoma cells and LL/2 (LLC; CRL-1642) were purchased from American Type Culture Collection and maintained using complete medium (RPMI 1640 containing 2 mM L-glutamine, 100 U/ml of penicillin-streptomycin and 10% fetal bovine serum (FBS)). Cell lines were passaged for a maximum of 20 passages.

Bone marrow-derived macrophages (BMDMs) preparation and in vitro experiments

BMDMs were differentiated and cultured in complete RPMI 1640 in the presence of recombinant mouse macrophage colony-stimulating factor (M-CSF; 20 ng/ml; PeproTech). Fresh complete medium with M-CSF was supplemented on day 3. After 7 days of culture, macrophages were washed and stimulated with 70% TCM, co-cultured with tumor cells (1:3 macrophage : tumor cell ratio), or treated with PS liposomes (20 μM; Sigma-Aldrich), IL-4 (20 ng/mL; PeproTech), or LPS (10 ng/mL; Sigma-Aldrich) plus IFNγ (50 ng/mL; PeproTech). Treatments were conducted in the absence or presence of DMM (20 mM; Sigma-Aldrich), BPTES (10 μM; Cayman), GSK-J4 (10 μM; Selleck), diethyl succinate (5 mM; Sigma-Aldrich), Compound C (5 μM; Cayman), DTHIB (5 μM; MedChemExpress), IgG2b (100 μg/mL; Bio X Cell), or αTIM4 (100 μg/mL; Bio X Cell) for the durations indicated in figure legends.

Induction of tumor cell apoptosis

YUMM1.7 tumor cells (1 × 10⁶) were seeded in 6-well plates and treated with dexamethasone (500 μM; Sigma-Aldrich) for 48 hours to induce apoptosis. Following treatment, cells were rinsed once with sterile PBS and resuspended in complete RPMI medium. Apoptosis was confirmed by Annexin V and 7-AAD staining using flow cytometry. Apoptotic cells (AC) were then used for co-culture experiments with bone marrow-derived macrophages (BMDMs) at a ratio of 3:1 (AC:BMDMs).

Tumor-conditioned medium (TCM) preparation

Tumor cells were counted and plated into four 25 cm² BD Falcon flasks (6 × 10⁶ cells in 9 ml complete medium). Supernatants were collected on day 1, 2, 3 and 4 by centrifuge at 450 g for 3 min. The supernatants were filtered using a 0.22 μm filter and stored at −80°C for future experiments.

Flow cytometry

Tumor digestion and cell isolation were performed as described previously¹⁰. For surface staining, cells were blocked with 5 μg/ml of anti-CD16/32 (clone 93, eBiosciences) before staining to CD45 (clone 30-F11; BioLegend; 103116; B444732; 1:200), Ly-6G (clone 1A8, BioLegend; B431781; 1:200), Ly-6C (clone HK1.4, BioLegend; B451308; 1:200) NK1.1 (clone PK136; BioLegend; 108753;B347472; 1:200), CD3 (clone 145-2C11; BioLegend; 100312; B291091; 1:200), CD4 (clone GK1.5; BioLegend; 100433; B439113, 1:200), CD8 (clone 53-6.7; BioLegend; 100706; B253704; 1:200), F4/80 (clone BM8; BioLegend; 123124; B439607), CD11b (clone M1/70; BioLegend; 101212; B290572), CD206 (clone C068C2; BioLegend; 141706; B427532; 1:200), CD301 (clone ER-MP23; BioRad; MCA2392A647; 148196; 1:200), PD-L2 (clone TY25; eBiosciences; 12-5986-82; 2970916; 1:200), PD-L1 (clone 10F.9G2, BioLegend; 124319; B427068; 1:200), CD172a (clone P84, BioLegend, 144015; B392230: 1:200), XCR1 (clone ZET, BioLegend; 148210; B402856; 1:200), TIM4 (clone RMT4-54; BioLegend, 130005; B434676; 1:200), CD86 (clone GL1; eBiosciences, 12-0862-82; 2324704; 1:200), MHC-II (clone M5/114.15.2; BioLegend 107622; B433647; 1:200), For intracellular staining of ARG1 (clone A1exF5; eBioscience; 17-3697-82; 3023992; 1:200), Relmα (PeproTech; 500-P214-100UG; 0806M263RBK1120; 1:200), Ki67 (BD; 556027; 4018886; 1:200), FoxP3 (MF-14; BioLegend; 136410; B366022; 1:200) Phospho-AMPKα (T172) (40H9; Cell Signaling Technology; 2535S; 27; 1:200), Phospho-HSF1(S121) (Thermo Fisher Scientific; PA5-39718; YL4148601; 1:200) or TRAP1 (Proteintech; 10325-1-AP; 10017473; 1:200), cells were fixed with BD Cytofix/Cytoperm buffer (BD Biosciences) and stained with appropriate primary antibody followed by incubation with appropriate fluorochrome-conjugated anti-rabbit IgG (Jackson Immunoresearch; 711-095-152; 105546; 1:200) or anti-mouse IgG (clone Poly4053; BioLegend; 405307; 1:200). For IFNγ and GZMB staining, cells were cultured in complete RPMI medium containing phorbol 12-myristate 13-acetate (50 ng/ml; Sigma-Aldrich), ionomycin (750 ng/ml; Sigma-Aldrich) and GolgiStop (1,000×; BD Biosciences) for 2 h at 37 °C. After surface staining, the cells were stained with IFNγ antibody (clone XMG1.2; BioLegend; 505808; B249122; 1:200) or GZMB (clone QA16A02; BioLegend; 372212; B357972; 1:200) according to intracellular staining protocol. Data were acquired on a BD LSRFortessa or Cytek Aurora high-dimensional flow cytometers and were analyzed with FlowJo software version 9.5.2 (TreeStar).

Isolation of tumoral macrophages (TAMs) from tumor-bearing mice

TAMs were isolated using magnetic-activated cell sorting (MACS). Tumor tissues were processed into single-cell suspensions, followed by red blood cell lysis. To reduce non-specific binding, cells were incubated with murine CD16/32 Fc block. F4/80⁺ TAMs were then enriched using anti-F4/80 microbeads (Miltenyi Biotec) and separated via MACS columns according to the manufacturer’s protocol.

Mitochondrial and cytosol extraction

Mitochondrial and cytosol fractions from 2 × 10⁷ macrophage cells were obtained using the Mitochondria Isolation Kit for Cultured Cells (Thermo Fisher Scientific), following the manufacture’s protocol.

Immunoblot

For immunoblotting, cells were lysed in radioimmunoprecipitation assay buffer (Thermo Fisher Scientific) with protease and phosphatase inhibitors (Cell Signaling Technology); Protein concentration was determined using the Pierce BCA protein assay kit (Thermo Fisher Scientific). Primary antibodies were anti-TRAP1 (Proteintech; 10325-1-AP; 10017473; 1:1000), anti-VDAC1/2 (Proteintech; 10866-1-AP; 1; 1:1000), anti-phospho-AMPKα(T172) (40H9; Cell Signaling Technology; 2535S; 27; 1:1000), anti-AMPKα (23A3; Cell Signaling Technology; 2603; 10; 1:1000), anti-phospho-HSF1(S121) (Thermo Fisher Scientific; PA5-39718; YL4148601; 1:1000), anti-HSF1 (D3L8I; Cell Signaling Technology; 12972S; 1; 1:1000), anti-JMJD3 (Abcepta, AP1022A; 1:1000), anti-H3K27me3 (C36B11; Cell Signaling Technology; 9733; 19; 1:1000), anti-H3 (D1H2; Cell Signaling Technology; 4499; 9; 1:1000), anti-GAPDH (14C10; Cell Signaling Technology; 2118; 9; 1:1000), anti-β-Actin (Sigma-Aldrich; A5316; 059M4770V; 1:1000), anti-α-Tubulin (11H10; Cell Signaling Technology; 2125S; 3; 1:1000), anti-PHB1 (Cell Signaling Technology; 2426; 1; 1:1000), total OXPHOS rodent antibody cocktail (Abcam; ab110413; P5685; 1000). Primary antibody staining was followed by peroxidase-linked secondary antibody (Cell Signaling Technology) and ECL immunoblot detection (BioRad).

Immunofluorescent staining and quantification

Human tissue samples from colon cancer patients were obtained from Lausanne University Hospital (CHUV) between 2015 to 2020 with informed consent. The study was approved by institutional review board of CHUV under PRE-IT 2016-02094 (material collection) and TRP_0001 2017-00359 (research purposes). Paraffin sections from 17 colon cancer patients were deparaffinized, rehydrated and subjected to antigen retrieval in 10 mM trisodium citrate dihydrate buffer (pH 6.0). Following the antigen retrieval, tissue sections were incubated with blocking buffer (10% normal goat serum and 0.3% Triton X-100 in PBS) for 1 h at 25°C. Human CD86 antibody (91882S, Cell Signaling) and TRAP1 antibody (TA504202, Invitrogen) were incubated for overnight at 4°C, which was followed by incubation of secondary antibodies conjugated with AF405 or AF647 at 25°C for 2 h. Tissue sections were further incubated with human CD68 AF488-conjugated antibody (ab277276, Abcam) and human CD163 AF555-conjugated antibody (ab281746, Abcam) for 2 hours at 25°C. Sections were then mounted in ProLong Gold Antifade mounting solution (P36930, Invitrogen). All antibodies were diluted in the blocking buffer. Images for each patient sample were randomly acquired using LSM880 confocal microscope imaging system (Zeiss), and subjected for quantification of mean fluorescence intensity by ImageJ software.

Metabolic analysis

For real-time analysis of oxygen consumption rate (OCR), extracellular acidification rate (ECAR), and spare respiratory capacity (SRC), macrophages were analyzed using a Seahorse XFe96 Flux Analyzer (Agilent). 5 × 10⁴ macrophages were plated in the Seahorse XF Assay Medium. Oligomycin (1 μM), fluorocarbonyl cyanide phenylhydrazone (FCCP, 3 μM) and rotenone (100 nM)/antimycin A (1 μM) were sequentially injected, following the protocol outlined in our previous work ¹⁰. All drugs were purchased from Sigma-Aldrich. ATP production was measured using the ATP Determination Kit (Invitrogen). Glutamine consumption was measured using the Glutamine Assay Kit (Abnova). Succinate and fumarate were measured using the succinate assay kit (Biovision) or fumarate assay kit (Abcam), respectively. α-KG levels were measured using the alpha ketoglutarate assay kit following the manufacturer’s instruction (Abcam). Succinate dehydrogenase (SDH) activity and α-ketoglutarate dehydrogenase (αKGDH) were measured using the or SDH activity assay kit (Biovision) or αKGDH activity assay kit (Elabscience), respectively. Phosphatidylserine (PS) levels in YUMM1.7 TCM were measured using a phosphatidylserine ELISA kit (Aviva Systems Biology) according to the manufacturer’s instructions. SCENITH analysis was performed as previously described³⁹. BMDMs (2 × 10⁵) were seeded in 96-well plates and treated with TCM for 24 h. Cells were then exposed to control, 2-Deoxy-D-Glucose (100 mM), Oligomycin (2 μM), or both sequentially for 30-45 min, with puromycin (10 μg/mL) added during the final 15 min to assess protein synthesis. After washing, cells were stained with viability dye, Fc receptor blockade, and surface markers, followed by fixation, permeabilization, and intracellular staining using PE-conjugated anti-puromycin antibody (clone 2A4, BioLegend). Data were acquired using a Cytek Aurora cytometer and analyzed with FlowJo v9.5.2.

Electron transport chain complex activity assay

For Complex I activity assay, 2 × 10⁷ macrophages were plated and treated with or without TCM for 24 h. Samples were collected and the enzyme activity was quantified by the Complex I Enzyme Activity Assay Kit (Colorimetric) according to the manufacturer’s protocol (ab109721, Abcam). For Complex II (SDH) activity assay, 2 × 10⁶ macrophages were plated and treated with or without TCM or IL-4 for 24 h. Samples were collected and the enzyme activity was quantified by the Succinate Dehydrogenase Activity Assay Kit (Colorimetric) according to the manufacturer’s protocol (ab228560, Abcam). For Complex III activity assay, 2 × 10⁷ macrophages were plated and treated with or without TCM for 24 h. Samples were collected and the enzyme activity was quantified by the Mitochondrial Complex III Activity Assay Kit according to the manufacturer’s protocol (ab287844, Abcam). For Complex IV activity assay, 2 × 10⁷ macrophages were plated and treated with or without TCM for 24 h. Samples were collected, and the enzyme activity was quantified by the Complex IV Rodent Enzyme Activity Microplate Assay Kit according to the manufacturer’s protocol (ab109911, Abcam).

RNA extraction and RT-qPCR

RNA was extracted using TRIzol reagent (Life Technologies). Complementary DNA was generated using PrimeScript RT Reagent Kit with gDNA Eraser (Takara Bio) according to the manufacturer’s instruction. The TaqMan or SYBR Green method was used for RT-qPCR with primers from Thermo Fisher Scientific or IDT. The assay was performed on a BioRad CFX96 machine. Relative expression was normalized by β-actin in each sample. The following probes or primers were used: for Taqman, Actb (Mm00607939_s1), Arg1 (Mm00475988_m1), Vegfa (Mm00437306_m1), Mgl2 (Mm00460844_m1), Retnla (Mm00445109_m1), Chil3 (Mm00657889_mH), Tgfb1 (Mm00441724_m1), Mrc1 (Mm00485148_m1), Irf4 (Mm00516431_m1), Trap1 (Mm00446003_m1), Nos2 (Mm00440502_m1), Cd86 (Mm00444540_m1), Il1b (Mm00434228_m1), Tnf (Mm00443260_g1), Il12b (Mm00434174_m1), Cd36 (Mm00432399_m1), Lipa (Mm00498820_m1), Pparg (Mm01184322_m1), Ppargc1b (Mm00504723_m1); for SYBR Green, Jmjd3 (FW: 5’-CCTGTTTACCGCTTCGT-3’; RV: 5’-TCACCTCGTTCCACTCATA-3’).

gDNA extraction and mtDNA/nDNA ratio measurement

Genomic DNA (gDNA) extraction was performed using QIAamp DNA Mini Kit (QIAGEN) according to the manufacturer’s instruction. The SYBR green method was used for qPCR with the following primers: mtDNA (FW: 5’-TGCTAGCCGCAGGCATTAC-3’; RV: 5’-GGGTGCCCAAAGAATCAGAAC-3’) nDNA (FW: 5’-CTTCCCCACTGGCCTCAAG -3’; RV: 5’-CCAAAACCCAGTGATCCAGC -3’). The mtDNA/nDNA ratio was calculated using following formula: $\left(2^{\text{CtnDNA}}\right)∕\left(2^{\text{CtmtDNA}}\right)$.

Ex vivo T cell proliferation & IFNγ and Granzyme B assay

Mouse splenic CD8⁺ T cells from C57BL/6J mice were isolated using EasySep Mouse CD8α Positive Selection Kit (STEMCELL Technologies). Isolated CD8⁺ T cells were labeled with CellTrace Violet (CTV) Cell proliferation Kit (Thermo Fisher Scientific) according to the manufacturer’s instructions. The CTV-labeled CD8⁺ T cells were washed and cultured (0.5 × 10⁵ per well) in plate-bound 1 μg/ml of anti-CD3 (clone 145-2C11; Thermo Fisher Scientific) and 5 μg/ml of anti-CD28 (Clone 37.51; Thermo Fisher Scientific) with complete RPMI medium containing 55 μM β-mercaptoethanol and 10 ng/ml of IL-2. TCM- or IL-4- or LPS + IFNγ-treated BMDMs, or TAMs isolated from tumor-bearing mice, were then added into T cell cultures at a ratio of 1:10 (macrophages : T cells) for 72 h. Cells were then harvested and stained with FITC-anti-CD8 (clone 53-6.7, BioLegend). For intracellular cytokine staining, cells were restimulated for 2 h at 37°C with PMA (50 ng/mL; Sigma-Aldrich), ionomycin (750 ng/mL; Sigma-Aldrich), and GolgiStop (1,000×; BD Biosciences), followed by staining with PE-anti-IFNγ (clone XMG1.2, BioLegend) and PerCP-Cy5.5-anti-Granzyme B (clone 2A4; ) according to intracellular staining protocol. Samples were acquired using a BD LSRFortessa or Cytek Aurora cytometer and analyzed with FlowJo v9.5.2.

Retroviral transduction

Retroviral transduction of macrophages was accomplished using protocols that we have used previously¹⁰. Sequences for shRNA targeting luciferase, Prkaa1 or Jmjd3 were obtained from Open Biosystems and cloned into the MSCV-LTRmir30-PI8 retroviral vector, which encodes human CD8 (huCD8) as a reporter. For overexpression, the TRAP1 sequence was cloned into MSCV–IRES retroviral vector, encoding huCD8 as a reporter. Day 3 BMDM cultures were spin infected with retrovirus at 800 g. On day 7 of culture, macrophages were harvested and transduced cells were identified by huCD8 expression.

RNA-seq and bioinformatics analysis.

The mRNA was extracted from lysates of macrophages that had been stimulated for 24 h. The synthesis, sequencing and sequence analysis of cDNA were performed as described¹⁰. The differential expression test, GSEA and visualization were performed using R (v.5.2.0), ggplot2 (v3.2.1), edgeR (v.3.32.1), pheatmap (v.1.0.12), clusterProfiler (v.3.10) and MSigDB (v.7.0).

Cleavage Under Targets & Tagmentation (CUT&Tag)

CUT&Tag was performed using the CUTANA kit (Epicypher). Buffers were supplemented with 1x protease inhibitor (Sigma-Aldrich). Cells were resuspended in 1mL PBS and nuclei were extracted with Nuclear Extraction buffer (20 mM HEPES-KOH pH 7.9, 10 mM KCl, 0.1% Triton X-100, 20% glycerol, 0.5 mM spermidine). Released nuclei (100,000) were bound to 10 μl concavalin A beads (BioMag Plus conA, Bangs Laboratories). Primary antibody against H3K27me3 (Cell Signaling Technology; C36B11) was applied at a 1:50 ratio to samples with 50 μl cold antibody buffer (1x BSA, 0.01% digitonin, 0.5 mM spermidine, 2 mM EDTA, 20 mM HEPES pH 7.5, 150 mM NaCl) and incubated at 4°C overnight on nutators. The secondary antibody (Epicypher; 13-0047) was mixed at a 1:100 ratio with Digitonin Wash 150 Buffer (0.01% digitonin, 0.5 mM spermidine, 20 mM HEPES pH 7.5, 150 mM NaCl) and then 50 μl of the mixture was applied to the samples at 25°C on rotators. Samples were twice washed with 200 μl Digitonin 150 wash and then resuspended in 50 μl Digitonin 150 wash and mixed with 2.5 μl Cutana pAG-Tn5 (Epicypher; 15-1017). After 1 h of incubation on a rotator, samples were twice washed with 200 μl Digitonin 300 wash (1x BSA, 0.01% digitonin, 0.5 mM spermidine, 2 mM EDTA, 20 mM HEPES pH 7.5, 300mM NaCl). Samples were resuspended in 50 μl of Tagmentation buffer (Digitonin 300 wash and 1% MgCl₂) and then incubated in a thermocycler for 1 h at 37°C for tagmentation. Reaction was washed with 50 μl TAPS wash (10 mM TAPS pH 8.5, 0.2 mM EDTA) and then resuspended in 5 μl SDS release buffer (10 mM TAPS pH 8.5, 0.1% SDS). Samples were incubated at 58°C for 1 h to release Tn5, then SDS was denatured by 15 μl Triton neutralization solution (0.67% Triton-X). CUT&Tag libraries were generated with the NEBNext Ultra II Q5 master mix (NEB) and barcoded with Truseq DNA CD Indexes (Illumina). qPCR was conducted for sample quality control using 2X SYBR Select Master Mix for CFX (Applied Biosystems). The assay was performed using the CFX96 real-time PCR detection system (Bio-Rad). DNA was quantified with Qubit (Thermo Fisher Scientific) and size distributions of the final libraries were measured using the Agilent 4200 TapeStation System (Agilent). Libraries were sequenced using NovaSeq 6000 (Illumina) at 2 x 150 bp paired-end.

Transmission electron microscopy

For TEM analysis, cells were seeded onto 6-well plates with 12-mm diameter inserts (Corning Snapwell inserts) at a density of 1 - 2 × 10⁵ cells per well. Cells were stimulated as described previously. After 24 h, the membrane filter with their attached cells was immersed in fixative. The initial fixative was 2.5% glutaraldehyde in cacodylate buffer, pH 7.3. The specimen was postfixed in ferrocyanide-reduced 1% osmium tetroxide. After a soak in acidified uranyl acetate, the specimen was dehydrated in ethanol, passed through propylene oxide and embedded in Embed-812 (Electron Microscopy Sciences). Thin sections (70 nm) were cut on an RMC MT6000-XL ultramicrotome. Sections were cut in a horizontal plane parallel to that of the membrane to provide panoramic views of the cells. These were mounted on Gilder square 300-mesh nickel grids (Electron Microscopy Sciences) and then sequentially stained with acidified methanolic uranyl acetate and stable lead-staining solution. These were coated on a Denton DV-401 carbon coater (Denton Vacuum LLC), and examined in an FEI Tecnai Spirit (T12) with a Gatan US4000 4kx4k CCD.

Data analysis and statistics

Data were analyzed with Graphpad Prism software (version 9). Comparisons of three or more groups were calculated by one-way ANOVA and, where indicated, unpaired or paired two-tailed Student's t-tests. Differences with P values <0.05 were considered significant. No randomization or exclusion of data points was used. Pilot in vivo studies were used for estimation of the sample size required to ensure adequate power. No statistical methods were used to predetermine sample size, but our sample sizes are similar to those reported in previous publications¹⁰. Data distribution was assumed to be normal, but this was not formally tested. Age and sex-matched animals were randomly assigned to experimental conditions. Data collection and analysis were not performed blind to the conditions of the experiments. No data exclusion was performed.

Extended Data

Extended Data Fig. 1. TRAP1 expression negatively correlates with immunosuppression in human and murine tumors.

(a) Representative gating strategy used for flow cytometry analysis of immune cells. (b) Heatmap showing TRAP1 expression in macrophages, monocytes, neutrophils (Neu), DC, NK, CD8⁺ T and CD4⁺ T cells isolated from the brain, liver, peritoneal cavity (PEC), lymph nodes (LN), spleen (SP), blood, and bone marrow (BM) tissues of naïve wild-type mice (n = 5 mice per group). Data are representative of three independent experiments. (c) Frequencies of TRAP1⁺ macrophages (Mac), monocytes (Mono), neutrophils (Neu), DC, NK, CD8⁺ T and CD4⁺ T cells in tumors from YUMM1.7 tumor-bearing mice on day 16 (n = 5 mice per group). Data are representative of two independent experiments. (d) Immunoblot analysis for TRAP1, VDAC1/2, prohibitin 1 (PHB1), GAPDH, and α-Tubulin in the mitochondrial and cytosolic fractions of naïve wild-type macrophages. Data are representative of two independent experiments. (e) Single-cell RNA-seq dataset (GSE146771) analysis of CD86 and CD163 expression in TRAP1^hi and TRAP1^lo macrophages from patients with colon cancer. (f) Pearson correlation of TRAP1 expression with genes encoding molecules associated with immune response in patients with skin cutaneous melanoma from the TCGA. (g) Representative flow cytometry plots (left) and quantitative analysis (right) of TRAP1⁺ macrophages in the spleen and tumor from B16-F10 tumor-bearing mice on day 16 (n = 5 mice per group). Data are representative of two independent experiments. (h, i) Quantitative analysis of TRAP1⁺ macrophages (h) and CD206⁺ macrophages (i) in the peritoneum, spleen, and tumor of YUMM1.7 tumor-bearing mice on day 16 (n = 5 mice per group). Data are representative of two independent experiments. (j) Expression of indicated genes in BMDMs treated without (Ctrl) or with YUMM1.7 TCM for 24 h, assessed by RT-qPCR analysis (n = 4). Data are representative of four independent experiments. All data are mean ± s.e.m. and were analyzed using a two-tailed, paired Student’s t-test (g), unpaired t-test (e), or one-way ANOVA with Sidak’s multiple comparison test (h, i).

Extended Data Fig. 2. TRAP1 deficiency enhances immunosuppressive capacity of TAMs.

(a) Expression of CD206 and CD301 by BMDMs co-cultured with either YUMM1.7 (top), B16-F10 (middle), or LLC (bottom) tumor cells for 72 h (n = 4). Data are representative of three independent experiments. (b) Expression of indicated genes in Trap1^+/+ or Trap1^−/− BMDMs treated with YUMM1.7 TCM for 24 h, assessed by RNA-seq (n = 3 in Trap1^+/+ group; n = 4 in Trap1^−/− group). (c) Expression of pro-inflammatory genes in TRAP1 wild-type or knockout BMDMs treated with YUMM1.7 TCM for 24 h, assessed by RT-qPCR analysis (n = 4, normalized to TCM-treated wild-type macrophages). (d) TRAP1 wild-type or null BMDMs were transduced with either retrovirus overexpressing a control reporter gene (EV) or a reporter gene plus the Trap1 sequence (Trap1^O/E), and co-cultured with B16-F10 cells for 72 h. Expression of CD206 and CD301 was determined by flow cytometry analysis (n = 3). (e, f) Representative histogram (left) and quantitative plots (right) of CD206 (e) and ARG1 (f) in TRAP1 wild-type and knockout BMDMs transduced with EV or Trap1^O/E, treated with YUMM1.7 TCM for 24 h (n = 3). Data are representative of two independent experiments. All data are mean ± s.e.m. and were analyzed using a two-tailed, unpaired Student’s t-test (c) or one-way ANOVA with Sidak’s multiple comparison test (a, d-f).

Extended Data Fig. 3. TRAP1 deletion in macrophages advances immune evasion.

(a) Schematic of the experimental design for YUMM1.7 tumor inoculation. Trap1^+/+ x Rag1^−/− and Trap1^−/− x Rag1^−/− mice were administered i.p. with anti-NK1.1 antibody followed by subcutaneous injection 5 × 10⁵ YUMM1.7 cells. (b-f) Tumor growth (b), tumor weight (c), representative tumor image (d), and frequency of CD206⁺ TAMs (e) and ARG1⁺ TAMs (f) from Trap1^+/+ x Rag1^−/− and Trap1^−/− x Rag1^−/− tumor-bearing mice on day 16 (n = 4 mice per group). Data are representative of two independent experiments. (g-o) Wild-type mice were injected subcutaneously with B16-F10 cells together with either Trap1^+/+ or Trap1^−/− BMDMs, or Trap1^−/− BMDMs overexpressing TRAP1 (Trap1^O/E). Tumor growth (g) and tumor weight (h) from mice given the indicated treatment. Absolute number of TAMs (i), frequencies of CD206⁺ TAMs (j) and ARG1⁺ TAMs (k), absolute number of TILs (l), frequencies of IFNγ⁺ CD8 (m), IFNγ⁺ CD4 (n) and FoxP3⁺ CD4 (o) T cells (n = 5 mice per group). Data are representative of two independent experiments. (p) Frequencies of TAMs, TILs, total DCs, DC1s, DC2s, neutrophils, monocytes, NKs in Trap1^wt and Trap1^cKO tumor-bearing mice on day 16 (n = 5 mice per group). Data are representative of two independent experiments. (q, r) Tumor growth (q) and tumor weight (r) in Trap1^wt and Trap1^cKO mice subcutaneously injected with 5 × 10⁵ Lewis lung carcinoma (LLC) cells. Tumors were harvested on day 16 post-tumor transplantation (n = 4 mice per group). Each symbol represents one individual. Data are representative of two independent experiments. (s) Proliferation of CTV-labeled CD8⁺ T cells activated with anti-CD3 and anti-CD28, and co-cultured with Trap1^+/+ or Trap1^−/− BMDMs treated with YUMM1.7 TCM or LPS+IFNγ (M1) at a 1:10 ratio for 72 h (n = 4). Data are representative of two independent experiments. All data are mean ± s.e.m. and were analyzed using a two-tailed, unpaired Student’s t-test (c, e, f, p, r), one-way ANOVA with Sidak’s multiple comparison test (h-o, s) or two-way ANOVA with Tukey’s multiple comparison test (b, g, q).

Extended Data Fig. 4. TIM4-AMPKα signaling axis regulates TRAP1 expression in TAMs.

(a) Expression of p-AMPKα⁺ macrophages in the peritoneum, spleen, and tumor of YUMM1.7 tumor-bearing mice on day 16 (n = 5). Data are representative of two independent experiments. (b) Immunoblot of p-AMPKα (short and long exposure), AMPKα and β-Actin in BMDMs treated with or without YUMM1.7 TCM for 24 h. Data are representative of two independent experiments. (c-f) Expression of p-AMPKα (c), TRAP1 (d), CD206 (e) and ARG1 (f) in BMDMs treated with TCM in the presence or absence of compound c (CC) for 24 h (n = 3). Data are representative of two independent experiments. (g) Indicated gene expression in BMDMs treated with TCM in the presence or absence of CC for 24 h, assessed by RT-qPCR (n = 4). Data are representative of three independent experiments. (h-i) Expression of pro-inflammatory genes in Prkaa1^sh or Luc^sh transduced BMDMs treated with TCM (h) or in wild-type BMDMs treated with TCM in the presence or absence of CC (i) for 24 h, assessed by RT-qPCR (n = 4). Data are representative of three independent experiments. (j, k) Expression of CD206 (j) and ARG1 (k) in Trap1^+/+ or Trap1^−/− BMDMs treated with TCM in the presence or absence of CC for 24 h (n = 3). Data are representative of two independent experiments. (l) Quantitation of TIM4⁺ macrophages (left Y-axis) and CD206 expression (right Y-axis) in the peritoneum, spleen, and tumor of YUMM1.7 tumor-bearing mice on day 16 (n = 5). Data are representative of two independent experiments. (m-o) BMDMs co-cultured with YUMM1.7 cells for 24 h in the presence or absence (IgG2b) of αTIM4, and expression of p-AMPKα (m), TRAP1 (n) and CD206 (o) was determined by flow cytometry (n = 3). Data are representative of two independent experiments. (p) Expression of pro-inflammatory genes in Timd4^+/+ and Timd4^−/− macrophages treated with TCM for 24 h, assessed by RT-qPCR (n = 4). Data are representative of two independent experiments. All data are mean ± s.e.m. and were analyzed using a two-tailed, unpaired Student’s t-test (g to i, p) or one-way ANOVA with Sidak’s multiple comparison test (a, c-f, j-o).

Extended Data Fig. 5. TCM-derived PS activates TIM4-AMPKα signaling to suppress TRAP1 expression.

(a) Phosphatidylserine (PS) levels in YUMM1.7 TCM were measured by ELISA (n = 4). Data are representative of three independent experiments. (b-e) BMDMs were cultured with either YUMM1.7 TCM or apoptotic YUMM1.7 cells (AC) for 24 h in the presence of IgG2b or αTIM4, and the expression levels of TIM4 (b), p-AMPKα (c), TRAP1 (d), and ARG1 (e) were analyzed by flow cytometry (n = 4). Data are representative of three independent experiments. (f-i) BMDMs were co-cultured with either PS or TCM for 24 h, and expression levels of TIM4 (f), p-AMPKα (g), TRAP1 (h) and ARG1 (i) were determined by flow cytometry (n = 4). Data are representative of three independent experiments. All data are mean ± s.e.m. and were analyzed using a two-tailed, unpaired Student’s t-test (a) or one-way ANOVA with Sidak’s multiple comparison test (b-i).

Extended Data Fig. 6. Hypoxia enhances TIM4-AMPKα signaling pathway.

(a-e) BMDMs were treated with TCM for 24 h under either normoxic or hypoxic conditions. Expression levels of TIM4 (a), p-AMPKα (b), TRAP1 (c), CD206 (d) and ARG1 (e) were assessed by flow cytometry (n = 4). Data are representative of two independent experiments. All data are mean ± s.e.m. and were analyzed using one-way ANOVA with Sidak’s multiple comparison test (a-e).

Extended Data Fig. 7. TRAP1 deficiency reprograms glutamine and α-KG metabolism in IL-4-stimulated macrophages.

(a) GSEA of glutamine metabolism in Trap1^+/+ or Trap1^−/− BMDMs stimulated with IL-4 (n = 2). GOBP, Gene Ontology Biological Process. HPO, Human Pathway Ontology. (b, c) Quantification of glutamine consumption (n = 6) (b) and intracellular α-KG (n = 5) (c) in IL-4-stimulated Trap1^+/+ or Trap1^−/− BMDMs. Data from three independent experiments. (d, e) Activity of SDH (d; n = 3) and αKGDH (e; n = 3) in TAMs from Trap1^wt and Trap1^cKO tumor-bearing mice. AU, arbitrary units. Data from two independent experiments. (f-h) SDH activity (n = 3) (f), intracellular succinate (n = 4) (g) and α-KG/succinate ratio (n = 4) (h) of IL-4-stimulated Trap1^+/+ or Trap1^−/− BMDMs. Data are representative of two independent experiments. (i) Schematic of SDH-mediated succinate oxidation and GLS-mediated glutaminolysis. (j) Basal OCR of Trap1^+/+ or Trap1^−/− BMDMs treated with IL-4 in the presence or absence of DMM or BPTES (n = 6). Data are representative of two independent experiments. (k) α-KG levels in TCM-treated Trap1^+/+ or Trap1^−/− BMDMs with or without DMM for 24 h (n = 3). Data are representative of two independent experiments. (l, m) Expression of CD206 and CD301 (l), and PD-L2 and RELMα (m) in Trap1^+/+ or Trap1^−/− BMDMs stimulated with IL-4 in the presence or absence of DMM, measured by flow cytometry (n = 4). Data are representative of three independent experiments. (n) Gene expression in Trap1^+/+ or Trap1^−/− BMDMs stimulated with IL-4 in the presence or absence of BPTES for 6 h, assessed by RT-qPCR (n = 4). Data are representative of two independent experiments. (o) α-KG levels in TCM-treated Trap1^+/+ or Trap1^−/− BMDMs with or without diethyl succinate (Suc) for 24 h (n = 3). Data are representative of two independent experiments. (p) Gene expression in Trap1^+/+ or Trap1^−/− BMDMs stimulated with IL-4 under conditions with or without Suc for 6 h, assessed by RT-qPCR (n = 4). Data are representative of two independent experiments. All data are mean ± s.e.m. and were analyzed using a two-tailed, unpaired Student’s t-test (b-h) or one-way ANOVA with Sidak’s multiple comparison test (j-p).

Extended Data Fig. 8. TIM4-AMPKα signaling regulates mitochondrial metabolism.

(a) Basal OCR of BMDMs treated with YUMM1.7 TCM in the presence or absence of CC (n = 6). Data are representative of two independent experiments. (b) Basal OCR of Prkaa1^sh or Luc^sh transduced BMDMs treated with YUMM1.7 TCM (n = 6). Data are representative of two independent experiments. (c-f) BMDMs were transduced with Prkaa1^sh or Luc^sh, and treated with YUMM1.7 TCM for 24 h. SDH (Complex II) activity (c), levels of intracellular succinate (d), α-KG (e), and α-KG/succinate ratio (f) were measured (n = 4). Data are representative of two independent experiments. (g) BMDMs were treated with αTIM4 or IgG2b, and cultured with YUMM1.7 TCM for 24 h. Basal OCR was measured by Seahorse Flux Analyzer (n = 6). Data are representative of two independent experiments. (h) Intracellular α-KG levels of YUMM1.7 TCM-treated BMDMs in the presence of αTIM4 or IgG2b (n = 4). Data are representative of two independent experiments. (i) Expression of indicated genes in BMDMs treated with YUMM1.7 TCM for 24 h in the presence of αTIM4 or IgG2b, assessed by RT-qPCR (n = 4, normalized to naïve macrophages). Data are representative of two independent experiments. (j) Expression of indicated genes in TIM4 wild-type (Timd4^+/+) or knockout (Timd4^−/−) BMDMs treated with YUMM1.7 TCM for 24 h, assessed by RT-qPCR analysis (n = 4, normalized to naive wild-type macrophages). Data are representative of two independent experiments. All data are mean ± s.e.m. and were analyzed using a two-tailed, unpaired Student’s t-test (i-j) or one-way ANOVA with Sidak’s multiple comparison test (a-h).

Extended Data Fig. 9. Histone demethylase JMJD3 is essential for TRAP1-mediated immunosuppression.

(a) Immunoblot analysis of JMJD3, H3K27me3, H3 and β-Actin in YUMM1.7 TCM-treated BMDMs with αTIM4 or IgG2b; representative of two independent experiments. (b) Jmjd3 expression in Trap1^+/+ and Trap1^−/− naïve BMDMs (n = 5); representative of two independent experiments. (c) CUT&Tag analysis of H3K27me3 enrichment in TCM-treated Trap1^+/+ and Trap1^−/− BMDMs. To account for the globally decreased H3K27me3 in the absence of TRAP1, Trap1^−/− track was normalized to the H3K27me3/H3 ratio from Fig. 7a, which shows the relative H3K27me3/H3 ratios are 0.817 and 0.475 for Trap1^+/+ and Trap1^−/−, respectively. Key genes shown include Arg1, Mrc1, Pparg, and Irf4. (d) Jmjd3 expression in Jmjd3^wt and Jmjd3^cKO naïve BMDMs (n = 4); representative of three independent experiments. (e) Expression of CD206 and CD301 in Jmjd3^wt and Jmjd3^cKO BMDMs co-cultured with YUMM1.7 cells for 72 h (n = 4); representative of three independent experiments. (f) Pro-inflammatory genes in TCM-treated Jmjd3^wt and Jmjd3^cKO BMDMs by RNA-seq (n = 3). (g, h) CD206 (g) and ARG1 (h) in TCM-treated Trap1^+/+ and Trap1^−/− BMDMs with or without GSK-J4 (n = 3); representative of two independent experiments. (i) Basal OCR of TAMs from Jmjd3^wt and Jmjd3^cKO tumor-bearing mice (n = 6); representative of three independent experiments. (j) ETC genes in TCM-treated Jmjd3^wt and Jmjd3^cKO BMDMs by RNA-seq (n = 3). (k) Frequencies of TAMs, TILs, neutrophils, monocytes, NKs, total DCs, DC1s, and DC2s in Jmjd3^wt and Jmjd3^cKO tumor-bearing mice on day 16 (n = 5); representative of two independent experiments. (l, m) Tumor growth (l) and weight (m) in Jmjd3^wt and Jmjd3^cKO mice subcutaneously injected with LLC cells (n = 4); representative of two independent experiments. (n) IFNγ and GZMB production in CD8⁺ T cells activated with anti-CD3/CD28, and co-cultured with TAMs from Jmjd3^wt or Jmjd3^cKO tumor-bearing mice at a 1:10 ratio for 72 h (n = 4); representative of two independent experiments. All data are mean ± s.e.m. and were analyzed using a two-tailed, unpaired Student’s t-test (b, d, k, m), one-way ANOVA with Sidak’s multiple comparison test (e, g, h, n) or two-way ANOVA with Tukey’s multiple comparison test (l).

Extended Data Fig. 10. Restoring TRAP1 expression in TAMs promotes anti-tumor immunity.

(a) Schematic of the experimental design for treatments. (b-f) Frequencies of TAMs (b), TILs (c), MDSCs (d), DCs (e) and NK cells (f) from YUMM1.7 tumor-bearing mice treated with IgG2b control, GSK-J4, αTIM4 or αTIM4 + GSK-J4 on day 16 (n = 5 mice per group). Data are representative of two independent experiments. Each symbol represents one individual. (g-k) Tumor growth (g), tumor weight (h), and frequencies of CD206⁺ TAMs (i), ARG1⁺ TAMs (j), and TRAP1⁺ TAMs (k) in Trap1^wt or Trap1^cKO YUMM1.7 tumor-bearing mice treated with either IgG2b or αTIM4. Drug administration time points are indicated by arrows (n = 3 mice for Trap1^wt + IgG2b and Trap1^wt + αTIM4 group; n = 4 mice for Trap1^cKO + IgG2b and Trap1^cKO + αTIM4 group). Data are representative of two independent experiments. Each symbol represents one individual. (l) Schematic illustrating how TIM4-AMPK signaling regulates mitochondrial TRAP1 activity, supporting mitochondrial metabolism and facilitating α-KG-JMJD3 epigenetic modifications that promote immunosuppressive activity in TAMs. Figure created with BioRender. All data are mean ± s.e.m. and were analyzed using one-way ANOVA with Sidak’s multiple comparison test (b-f, h-k) or a two-way ANOVA with Tukey’s multiple comparison test (g).

Data availability

RNA-Seq and CUT&Tag data are available in the NCBI database under accession numbers GSE279754 and PRJNA1172334, respectively. Source data are provided with this paper.