Targeting IRG1 reverses the immunosuppressive function of tumor-associated macrophages and enhances cancer immunotherapy

1Cancer immunotherapy has revolutionized cancer treatment. Exemplified by the success of immune checkpoint inhibition using antibodies against PD-1 and PD-L1 and chimeric antigen receptor-T (CAR-T) cell therapies, cancer immunotherapy has achieved remarkably effective and durable responses in a subset of patients with different cancer types. However, most patients receiving these therapies, even in combination, do not derive clinical benefit. Further development of agents, including small molecules, and strategies targeting additional immune checkpoints is critical.

2Macrophages are crucial innate immune effector cells that perform a multitude of functions in response to pathogens and tissue inflammatory signals, including tumor development. Tumor-associated macrophages (TAMs) represent one of the most abundant immune cell types in the tumor stroma, and infiltration of macrophages is associated with poor prognosis and correlates with chemotherapy resistance in most cancers. In a widely used model, TAMs undergo polarization into two forms: the “classical” (or M1-polarized) form that is involved in the responses of T helper 1 (TH1) cells and produces proinflammatory antitumor effects, and the “alternative” (or M2-polarized) form that is involved in TH2-type responses and produces anti-inflammatory mediators to contribute many protumorigenic outcomes in cancer. A better understanding of the regulation of TAM polarization may identify additional immune checkpoint targets to achieve more durable antitumor immunity in broader patient populations.

3The rapid proliferation of cancer cells in growing tumors results in a local change of oxygen, nutrients, and metabolites in the tumor microenvironment (TME). The alterations of TME cause a metabolic reprogramming of tumor-associated immune cells, such as T cells and macrophages. One marked metabolic change is the rapid accumulation of a previously understudied metabolite, itaconic acid (ITA), from a barely detectable low level in resting cells to 3 to 4 mM in classically activated macrophages. This is resulted from transcriptional activation of the immune-responsive gene (IRG1), which encodes the metabolic enzyme aconitate decarboxylase (ACOD1) that catalyzes the production of ITA. Genetic analyses of Irg1-deficient mice have established the anti-inflammatory function of IRG1/ITA in multiple pathogen models, including sepsis, viral infections, psoriasis, gout, ischemia/reperfusion injury, and pulmonary fibrosis. Emerging evidence suggests that IRG1 exhibits tumor-promoting properties by activating MAPK cascade in tumor cells, but very little is known about its biological function in tumor immunity in vivo.

4In this study, we present evidence that tumor cells induce IRG1 expression in TAM subsets by nuclear factor κB (NF-κB) activation, underlying the cross-talk between tumor cells and macrophages. Irg1 regulates TAM polarization and promotes their immune suppressive function to potentiate tumor outgrowth. Irg1-deficient mice are resistant to tumor growth and have enhanced responses to anti–PD-(L)1 immunotherapy. Infusion of Irg1-deficient macrophages into wild-type mice inhibits syngeneic tumor growth. These results thus identify ACOD1 as a potential target for cancer immunotherapy and IRG1-deficient macrophages as a potent cell therapy strategy for cancer treatment.

5The broad induction of IRG1 transcription by various pathogens and the anti-inflammatory activity of ITA prompted us to explore the function of IRG1 in tumorigenesis. We first analyzed the expression of IRG1 gene in patients from The Cancer Genome Atlas (TCGA) cohorts and found that IRG1 mRNA expression was up-regulated in multiple types of tumors as compared to the corresponding para-carcinoma normal tissues, including BLCA (bladder urothelial carcinoma), BRCA (breast invasive carcinoma), COAD (colon adenocarcinoma), ESCA (esophageal carcinoma), HNSC (head and neck squamous cell carcinoma), KICH (kidney chromophobe), KIRP (kidney renal papillary cell carcinoma), LUSC (lung squamous cell carcinoma), PAAD (pancreatic adenocarcinoma), STAD (stomach adenocarcinoma), and UCEC (uterine corpus endometrial carcinoma) (Fig. 1A). Likewise, IRG1 mRNA expression was up-regulated in patients with SKCM (skin cutaneous melanoma) compared with healthy skin tissues (Fig. 1A). By using the computational method EPIC (estimating the proportions of immune and cancer cells), we found that IRG1 mRNA expression was positively correlated with the TAM fraction in a broad spectrum of human cancers (fig. S1). In accord, single-cell RNA sequencing (scRNA-seq) in samples from patients with uveal melanoma revealed that IRG1 was highly expressed in TAMs, but not other types of immune cells (fig. S2A). In addition, this expression pattern of IRG1 was also observed in monocytes and macrophages in the TME of 12 other types of human cancers (fig. S2B).

6To decipher how IRG1 expression is induced in TAMs, we cocultured mouse bone marrow–derived macrophages (BMDMs) with different types of syngeneic tumor cells or treated BMDMs with tumor cell–conditioned medium (TCM) and found that either treatment led to Irg1 induction and ITA accumulation (to millimolar level) in BMDMs (fig. S3, A and B). It has previously been reported that IRG1 expression is regulated by signaling pathways under inflammatory conditions, such as NF-κB, glucocorticoid receptor (GR), and Janus kinase (JAK)/signal transducer and activator of transcription (STAT) signaling pathways. We found that the effect of B16-F10-TCM on up-regulating Irg1 expression was diminished by the treatment with pyrrolidine dithiocarbamate (an NF-κB inhibitor), but not Ruxolitinib (Ruxo, a JAK1/2 inhibitor) (fig. S3C). Il-6 and Il-1b were included as two positive control genes (fig. S3C). In addition, we found that occupancy of RelA/P65 at the promoter region of Irg1 was increased in BMDMs by the treatment of B16-F10-TCM (fig. S3D), further supporting that tumor cells induce Irg1 expression in macrophages through NF-κB activation.

7To explore the biological function of Irg1 during tumor development, we generated different syngeneic tumor models in which B16-F10 (melanoma), MC38 (colon cancer), and E0771 (breast cancer) cells were inoculated subcutaneously into Irg1−/− and Irg1+/+ mice. We found that, in all the abovementioned tumor models, the tumor growth was significantly suppressed in Irg1−/− mice (fig. S4A), and the survival of tumor-bearing mice was prolonged (Fig. 1B). Furthermore, pancreatic cancer cells derived from the genetically engineered mouse tumor model (LSL-KrasG12D; LSL-Trp53R172H; Pdx1-Cre mice; KPC) were implanted into the pancreas of Irg1−/− and Irg1+/+ mice, and the KPC tumor growth was also remarkably inhibited in Irg1−/− mice as monitored by bioluminescent imaging (Fig. 1C).

8It is known that Acod1 produces ITA under inflammatory conditions, principally in cells of myeloid lineage. We found that, in the TME of B10-F10 tumors, the accumulation of ITA was readily detected in myeloid cells (CD45+CD11b+) including TAMs (F4/80+Cd45+), but not in lymphocytes (CD45+CD11b−) or nonimmune cells (CD45−CD11b−) (fig. S4, B and C). We then generated myeloid-specific Irg1-deficient mice (Irg1f/fLyz2-cre+) and verified the proper deletion of Irg1 in cells of myeloid lineage in these animals (fig. S5, A to C). Consistent with the whole-body knockout (KO), myeloid-specific KO of Irg1 inhibited the growth of B16-F10, MC38, and KPC tumors in mice (Fig. 1, D to F, and fig. S5, D and E). Collectively, these results suggest that Irg1 in myeloid cells potentiates tumor growth in immunocompetent mice.

9To evaluate the impact of Irg1 on the TME, immune cells (Cd45+) were purified from B16-F10 tumors grown in Irg1+/+ and Irg1−/− mice and then subjected to scRNA-seq (Fig. 2A and table S1). Using Uniform Manifold Approximation and Projection (UMAP) clustering analyses, 19 clusters of major immune populations were distinguished (fig. S6A), and these clusters were assigned to eight different types of immune cells, including macrophages (MФ) and monocytes, T cells, B cells, natural killer (NK) cells, plasmacytoid dendritic cells, dendritic cells, neutrophils, and limited leaked melanocytes (fig. S6, B and C). As expected, Irg1 mRNA expression was readily detected in myeloid cells characterized by high expression of Itgam (Cd11b), particularly in macrophages characterized by high expression of Adgre1 (F4/80) and neutrophils characterized by high expression of Retnlg or S100a9 in the TME of B16-F10 tumors from Irg1+/+ mice (Fig. 2B and fig. S6D).

10Next, we used UMAP clustering and divided the population of “MФ and monocytes” into six distinct clusters. We found that Irg1 was mainly expressed in a few subpopulations, including Vcan+ monocytes, proinflammatory macrophages, and Vegfa+ macrophages (Fig. 2, C and D). Among these clusters, Vcan+ monocytes exhibited high expression of Ly6c2, Cxcl10, and Vcan and enriched in type I and II interferon (IFN) pathways (fig. S7, A and B), which are similar with hemopoietic system-derived classical inflammatory monocytes as previously termed. Proinflammatory macrophages exhibited high expression of major histocompatibility complex class II (MHC-II) (e.g., H2-Aa, H2-Ab1, H2-DMb1, and H2-Eb1) (fig. S7A), which are similar with the classical activated and M1-like macrophages. Vegfa+ macrophages exhibited high expression of Spp1, Vegfa, and Mmp12, and enriched in anti-inflammatory pathways, sharing the similarity with the previously termed SPP1+ macrophages, which have an enrichment of tumor angiogenesis and protumorigenic role. Mrc1+ macrophages were found to show high expression of Mrc1, Retnla, and Gatm (fig. S7A), sharing the similarity with alternatively activated and M2-like macrophages. According to the pseudotime-organized sequence of “differentiation/activation” events, Vegfa+ macrophages, together with proinflammatory macrophages and Mrc1+ macrophages, were differentiated from Vcan+ monocytes and Itgal+ monocytes (fig. S7C). We found that Irg1 deficiency in mice led to the increase of Vcan+ monocytes and proinflammatory macrophages and a decrease of Vegfa+ macrophages in the TME of B16-F10 tumors (Fig. 2C), indicating that Irg1 may regulate TAM polarization during tumor development.

11To validate the abovementioned observations, we carried out flow cytometric analysis and demonstrated that the populations of tumor-infiltrating monocytes (Ly6C+Cd11b+Cd45+) and M1-like TAMs (iNOS+F4/80+Cd11b+Cd45+) were significantly increased in the TME of B16-F10 tumors from Irg1−/− mice compared with Irg1+/+ controls (Fig. 2E and fig. S8). No significant changes in the population of M2-like TAMs (Cd206+F4/80+Cd11b+Cd45+) and myeloid-derived suppressor cells (MDSCs) (Gr1+Cd11b+Cd45+) were observed in the TME of B16-F10 tumors from Irg1−/− mice compared with Irg1+/+ controls (Fig. 2E). In addition, measurement of tumor-infiltrating lymphocytes revealed a significant increase in the population of T cells (Cd3+Cd45+), especially CD8+ T (Cd8+Cd3+Cd45+), in the TME of B16-F10 tumors from Irg1−/− mice compared with Irg1+/+ controls (Fig. 2E). Immunofluorescent staining confirmed more iNOS+ or CD8+ cells in the TME of B16-F10 tumors growing in Irg1−/− mice than Irg1+/+ controls (Fig. 2F). Together, these findings indicate that Irg1 deficiency in mice is associated with reduced immunosuppressive TME and thus supports the immunosuppressive function of IRG1 during tumor development.

12We recently reported that ITA is structurally similar to α-ketoglutaric acid (α-KG) and binds to and inhibits α-KG–dependent TET (ten-eleven translocation) DNA demethylases, thereby down-regulating NF-κB and STAT target genes to dampen the inflammatory responses. The expression of IRG1 in macrophages in both human and mouse tumors and the association of IRG1 with immunosuppressive TME led us to further characterize the effect of IRG1 on the transcriptomic landscape in TAMs. To this end, gene set enrichment analysis (GSEA) and single-sample GSEA were conducted to calculate gene signature–based scores. We found that intratumoral monocytes/macrophages in melanoma from Irg1−/− mice exhibited higher gene signature scores of cytokine production, antigen processing and presentation, chemokine signaling pathway, and positive regulation of T cell migration as compared with those cells from Irg1+/+ controls (Fig. 3A and fig. S9A). Among the population of MФ and monocytes, Vcan+ monocytes exhibited up-regulation of IFN-stimulated genes (e.g., Irf7, Isg15, Ifi44, and Ifi27l2a) in the TME of B16-F10 tumors from Irg1−/− mice compared with Irg1+/+ controls, and genes associated with antigen presentation and T cell migration (e.g., Aif1, Timd4, and Tap1) were also up-regulated in proinflammatory macrophages from tumor-bearing Irg1−/− mice (Fig. 3B). In accord, flow cytometry analysis illustrated that TAMs from tumor-bearing Irg1−/− mice manifested enhanced immunogenic antigen presentation, as evidenced by higher expression of cell surface markers, such as MHC-I, MHC-II, or Cd40 (Fig. 3C).

13In agreement with the aforementioned observation (Fig. 2E), transcriptional analysis by quantitative reverse transcription polymerase chain reaction (qRT-PCR) showed that TAMs from tumors grown in Irg1−/− mice exhibited more of proinflammatory phenotypes, with increased Tnfα, Il-6, and Il-1β but decreased Arg1, Cd206, and Il-10 expression (Fig. 3D). The mRNA expression of proinflammatory cytokines (e.g., Tnfα and Il-6) was also up-regulated in Irg1−/− BMDMs compared to Irg1+/+ controls after B16-F10-TCM treatment (fig. S9B). In accord, Tnfα or Il-6 protein secretion was increased in Irg1−/− BMDMs compared to Irg1+/+ controls after B16-F10-TCM treatment (fig. S9C). Flow cytometry analysis revealed that Irg1−/− BMDMs exhibited higher levels of iNOS than Irg1+/+ cells after either coculture with B16-F10 tumor cells or treatment with B16-F10-TCM (fig. S9, D and E). The M2-like macrophage marker Cd206, however, did not differ significantly between Irg1+/+ and Irg1−/− BMDMs after either coculture with B16-F10 tumor cells or treatment with B16-F10-TCM (fig. S9, D and E). On the other hand, genes enriched in Vegfa+ macrophages as identified by scRNA-seq (fig. S7A), were commonly down-regulated in Irg1−/− BMDMs compared to Irg1+/+ controls after B16-F10-TCM treatment (fig. S9F). These results thus suggest that Irg1-deficient macrophages exhibit altered characteristics, including more proinflammatory features but reduced proangiogenic potential, which are associated with tumor suppressive functions.

14Of note, Irg1 mRNA expression in wild-type BMDMs was induced at 6 hours after coculture with B16-F10 tumor cells, and higher mRNA expression of proinflammatory genes and those involved in antigen presentation and chemotaxis was observed in Irg1−/− BMDMs at 12 hours after B16-F10 coculture (Fig. 3E). These results thus indicate an association between IRG1 induction and IRG1/ITA-mediated suppression of genes linked to proinflammatory response and antigen presentation, reaffirming the anti-inflammatory role of IRG1.

15Chemokines play a paramount role in regulating the infiltration of different immune cells into tumors, and, as such, these molecules affect tumor immunity and influence therapeutic outcomes in patients. The enriched chemokine signaling pathways in intratumoral monocytes/macrophages in Irg1−/− mice attracted our attention (Fig. 3A). Irg1 deficiency led to up-regulation of T cell chemotaxis genes (e.g., Cxcl9 and Cxcl10) in isolated TAMs or TCM-treated BMDMs (Fig. 3, D and E). To determine whether Irg1 affects the ability of macrophages to facilitate T cell chemotaxis, we developed a transwell system in which BMDMs were cocultured with B16-F10 tumor cells or treated with B16-F10-TCM (Fig. 3F). In this coculture system, Irg1−/− BMDMs promoted the migration of CD8+ T cells, which were isolated from the spleen of wild-type mice, more efficiently than Irg1+/+ macrophages. This effect was, however, abrogated by the CXCR3-specific antagonist SCH546738, which blocks the CXCL9, CXCL10, CXCL11/CXCR3 axis and thus T cell migration. Administration with SCH546738 reversed the effect of Irg1 deficiency on facilitating T cell chemotaxis and diminished the tumoricidal effect in vivo (Fig. 3, G and H), suggesting the potential role of IRG1 in limiting T cell trafficking into the tumor.

16We previously showed that CXCL9, CXCL10, and CXCL11, three chemokines that attract T cells through binding to their receptor CXCR3 on T cells, are direct targets of TET2. When Tet-catalyzed 5-hydroxymethylcytosine (5hmC) production was examined in engrafted tumor tissues, we found that 5hmC staining intensity was increased in tumors from Irg1−/− than Irg1+/+ mice (Fig. 4A). Furthermore, 5hmC mapping showed that its enrichment at the promoter regions of chemokine genes (e.g., Cxcl9, Cxcl10, Ccl2, Ccl3, Ccl4, and Ccl5) was increased in TAMs from Irg1−/− tumor-bearing mice than Irg1+/+ controls (Fig. 4B and fig. S10A). Increased 5hmC at these gene promoters was also observed in Irg1−/− BMDMs treated with B16-F10-TCM (Fig. 4C and fig. S10B). We isolated BMDMs from catalytically inactive Tet2HxD knock-in (KI) mutant or wild-type mice and and then treated these macrophages with B16-F10-TCM either alone or together with ITA. The Tet2HxD KI mutant contains H1295Y and D1297A double substitution mutations in mouse Tet2, which are equivalent to H1382 and H1384 in human TET2, respectively, and disrupt the binding with the essential cofactor Fe2+. We found that ITA suppressed the effect of B16-F10-TCM on up-regulating Cxcl9/10 mRNA expression in wild-type BMDMs, but not in catalytically inactive Tet2HxD macrophages (Fig. 4D). Consistently, 5hmC enrichment at the promoter of Cxcl9/10 and selected chemokine genes was inhibited by ITA treatment in wild-type, but not in Tet2HxD macrophages (Fig. 4E and fig. S10C). As a result, ITA prevented the effect of wild-type, but not Tet2HxD BMDMs on promoting CD8+ T cell migration in the coculture transwell system (Fig. 4F).

17In agreement with our findings in murine macrophages, human IRG1 protein expression was induced in THP1-derived macrophages upon stimulation with TCM from breast cancer cell line MDA-MB-231, with intracellular ITA reaching as high as 300 μM (Fig. 4G and fig. S10D). In these MDA-MB-231-TCM stimulated human macrophages, deletion of IRG1 (fig. S10E) led to up-regulation of CXCL9/10 and other chemotaxis genes (Fig. 4H and fig. S10F), thereby promoting the migration of Jurkat T cells in the transwell system (Fig. 4I). Furthermore, we also isolated human peripheral blood mononuclear cells (PBMCs) and differentiated them into macrophages ex vivo (Fig. 4J). As expected, IRG1 mRNA expression was remarkably induced by MDA-MB-231-TCM in primary human macrophages (Fig. 4K). IRG1 depletion by small interfering RNA (siRNA) significantly up-regulated CXCL9/10 expression in primary human macrophages upon stimulation with MDA-MB-231-TCM (Fig. 4K). In accord, the conditioned medium (CM) from IRG1-deficient human macrophages facilitated the migration of CD8+ T cells in the transwell system (Fig. 4L). Together, these results in mouse and human cells suggest that IRG1/ITA affect the role of TAMs in influencing T cell trafficking through regulating TET2-CXCL9/10/11-CXCR3 axis.

18Supporting the notion that targeting IRG1 may skew TAMs into an antitumor mode that enhances T cell–mediated tumor killing, the percentage of tumor-infiltrating CD8+ T cells was increased in Irg1−/− mice. However, their cellular production of IFN-γ or tumor necrosis factor–α (TNF-α) did not differ in the TME of B16-F10 tumors between Irg1−/− and Irg1+/+ mice (fig. S11), indicating that Irg1 may not directly affect the tumor killing function of T cells. Next, we investigated the functional consequences of Irg1 loss on cancer responses to T cell–based immunotherapy. To this end, B16-F10 melanoma cells were inoculated subcutaneously into Irg1+/+ and Irg1−/− mice that were then subjected to PD-L1 blockade. We found that the growth of B16-F10 tumor in Irg1−/− mice (without PD-L1 blockade) was reduced to the extent achieved by the anti–PD-L1 in Irg1+/+ mice, extending the life span from 14 days in untreated Irg1+/+ mice to 18 days in Irg1−/− mice or Irg1+/+ mice treated with PD-L1 antibody (Fig. 5, A and B). Notably, the combination of anti–PD-L1 treatment and Irg1 deficiency achieved the most significant antitumor effect and longest survival, with the mean life span extended to 27 days (Fig. 5, A and B). In another syngeneic tumor model, the E0771 tumor growth in Irg1−/− mice (without PD-1 blockade) was comparable to that in anti–PD-1–treated Irg1+/+ mice (Fig. 5, C and D). The combination of anti–PD-1 and Irg1 deficiency showed the most effective tumor inhibition and prolonged survival, with the mean life span being extended from 22 and 23 days in Irg1−/− mice (without PD-1 blockade) or anti–PD-1–treated Irg1+/+ mice, respectively, to 31 days (Fig. 5, C and D).

19To gain more insight into human relevance, we examined and found that IRG1 expression was readily detected in patients with hepatocellular carcinoma (HCC) after immunotherapy (Fig. 5E). In a study cohort consisting of 34 patients with HCC (table S2), we found that IRG1 staining signal was higher in αPD-1–resistant HCC patients (n = 12) than responsive patients (n = 22) (Fig. 5, E and F). In agreement with our data in mouse tumor models, IRG1 expression exhibited a negative correlation with tumor-infiltrating CD8+ T cells in clinical samples of patients with HCC after immunotherapy (Fig. 5, E and G), further suggesting that targeting IRG1/ACOD1 represents a promising strategy for enhancing the efficacy of cancer immunotherapy.

20Beside TAM, our scRNA-seq results revealed that Irg1 was also expressed in tumor-associated neutrophils (TANs) in the TME of B16-F10 tumors (fig. S6D). To decipher which Irg1-expressing cell population, macrophages or neutrophils, is primarily responsible for the observed tumor phenotype, we applied the antibody against CSF1R and Ly6G to deplete TAMs and TANs, respectively, in tumor-bearing mice as previously described. The injection of αCSF1R antibody led to a depletion of TAMs by ~70% in Irg1f/fLyz2-cre+ mice (fig. S12A) and abrogated the antitumor effect of Irg1-KO myeloid cells (Fig. 6, A and B). The injection of αLy6G antibody led to a depletion of TANs by ~80% in Irg1f/fLyz2-cre+ mice (fig. S12B) but had little effect on tumor phenotype (Figs. 6, C and D). Furthermore, we conducted adoptive transfer experiments by intratumoral injection of macrophages or neutrophils into B16-F10 tumor-bearing mice (fig. S12C) as previously described. The tumor-killing effect was observed in mice receiving Irg1−/− BMDMs, but not in those animals receiving Irg1−/− neutrophils (Fig. 6, D and E). Collectively, these results indicate that targeting Irg1 in macrophages, but not in neutrophils, inhibits tumor growth.

21To further enhance the tumor-killing effect of Irg1−/− BMDMs, we combined the macrophage adoptive transfer assay together with anti–PD-L1 treatment in wild-type mice bearing B16-F10 melanoma. Notably, the tumor-killing effect of Irg1−/− BMDM adoptive transfer was almost identical to anti–PD-L1 treatment in control mice receiving Irg1+/+ BMDMs. The adoptive transfer of Irg1−/− BMDMs in combination with PD-L1 blockade exhibited the most effective tumor inhibition and prolonged survival in wild-type recipients (Fig. 7, A and B).

22Last, we examined the effect of Irg1 deficiency in macrophages on the growth of pancreatic tumors, which are known to be resistant to T cell–based immunotherapy. Anti–PD-1 treatment failed to suppress KPC tumor growth in Irg1+/+ mice (fig. S13A). In agreement with our earlier finding (Fig. 1C), KPC tumor growth was remarkably retarded in Irg1−/− mice compared with Irg1+/+ mice. PD-1 blockade exhibited a tendency to further decrease KPC tumor growth in Irg1−/− mice but failed to reach statistical significance, even with more proinflammatory TAMs and increased infiltration of CD8+ T cells (fig. S13, A and B). Then, the macrophage adoptive transfer was conducted by intravenous injection of Irg1−/− or Irg1+/+ BMDMs into mice bearing KPC tumors at different stages (Fig. 7C). To compare the recruitment of macrophages into engrafted KPC tumors, Irg1−/− or Irg1+/+ BMDMs were labeled with fluorescein isothiocyanate (FITC)–conjugated beads, as previously reported. These FITC-labeled macrophages were then intravenously injected into mice implanted with KPC tumor cells (1 × 107 per mouse) (fig. S14A). We found that the transferred macrophages comprised ~40% of total intratumoral macrophages in the TME of KPC tumors receiving either Irg1−/− or Irg1+/+ macrophages (fig. S14B), suggesting that Irg1 does not affect the macrophage trafficking to and infiltration into KPC tumor sites.

23Flow cytometry analysis revealed that the populations of M1-like TAMs (iNOS+F4/80+Cd11b+) and CD8+ T cells (Cd8+Cd3+) were significantly increased in KPC tumors receiving Irg1−/− macrophages compared to Irg1+/+ controls (Fig. 7D). Immunofluorescent staining in KPC tumors in mice infused with Irg1−/− macrophages confirmed more iNOS+ or CD8+ cells (Fig. 7E and fig. S14C) and terminal deoxynucleotidyl transferase–mediated deoxyuridine triphosphate nick end labeling (TUNEL)–positive apoptotic cells (Fig. 7, E and F). Moreover, the population of MDSCs (Gr1+Cd11b+Cd45+), which are known to be significantly expanded and contribute to ineffective therapies in pancreatic cancer, was decreased in KPC tumors receiving Irg1−/− macrophages (Fig. 7D). In addition, we also checked the tumor vasculature and found a decrease in CD31 or VCAM1 staining in KPC tumors receiving Irg1−/− macrophages compared to Irg1+/+ controls (fig. S15, A and B), suggesting that targeting Irg1 in macrophages may suppress angiogenesis that is consistent with the scRNA analysis (Fig 2C and fig. S7) and decrease pancreatic tumor malignancy. In accord, the number of Ki67+ proliferating cells was decreased in KPC tumors receiving Irg1−/− macrophages compared to Irg1+/+ controls (fig. S15C). As a result, the infusion of Irg1−/− macrophages suppressed the growth of KPC tumors at both early (10 days after inoculation) and later stages (21 days after inoculation), resulting in a 58.8 and 50.0% decrease in tumor weight, respectively, as compared to those receiving Irg1+/+ macrophages (Fig. 7, G and H). Together, these data suggest that Irg1-deficient macrophages represent a potent cell therapy strategy for cancer treatment even in pancreatic tumors that are resistant to T cell–based immunotherapy.

24The present study reveals a mechanism by which tumor cells escape the immunosurveillance through inducing IRG1 expression in TAMs to create an immunosuppressive environment. This is evidenced by the expression of IRG1 in TAMs of human and mouse tumors, by cell (co)culture studies demonstrating that tumor cells or tumor-secreted factors induce IRG1 expression through NF-κB activation, by genetic studies in mice showing that deletion of Irg1 in TAMs reduces tumor growth and enhances the efficacy of PD-1/PD-L1 blockade, and by human data showing a negative correlation of IRG1 expression with tumor-infiltrating CD8+ T cells and HCC patient response to immunotherapy. We believe that the IRG1-mediated escape of immunosurveillance is a broad mechanism, as deletion of Irg1 suppresses tumor growth in multiple syngeneic mouse models, including melanoma, colon, breast, and pancreatic cancers.

25We provide ex vivo and in vivo evidence that IRG1 can change TAM functionality toward proangiogenic features and limit proinflammatory features (fig. S16). This is demonstrated by reduced expression of proangiogenic genes in Irg1-deficient macrophages stimulated with tumor CM, as well as decreased angiogenesis in syngeneic pancreatic tumors after receiving Irg1-deficient macrophage therapy. On the other hand, increased expression of proinflammatory genes in TAMs from tumor grown in Irg1-deficient mice, such as genes involved in immunogenic antigen presentation and T cell chemotaxis, further supports the notion that IRG1 induction favors tumor cells to escape the antitumor immunity via skewing TAM polarization. At the molecular level, ITA achieves its anti-inflammatory function and suppresses proinflammatory features in macrophages, in part, through inhibition of Tet2-catalyzed DNA demethylation. Besides Tet inhibition, ITA is known to cause SDH inhibition that results in succinate accumulation and mitochondrial respiration, protein alkylation at cysteine residues that induces the electrophilic stress response mediated by NRF2 (nuclear factor erythroid 2-related factor 2) and IκBζ [inhibitor of NF-κB (nuclear factor κB) ζ], impairment of aerobic glycolysis, and increased lysosomal biogenesis. The contribution of each of these mechanisms to the role of IRG1 in mediating tumor escape of immunosurveillance needs further clarification.

26Zhao et al. recently reported that intracellular ITA was higher in polymorphonuclear myeloid-derived suppressor cells (PMN-MDSCs) than that in naïve bone marrow cells, and they proposed non–cell-autonomous mechanism in which ITA produced by myeloid cells is secreted out, up-taken by T cells, and then attenuates CD8+ T cell proliferation and function Differently, the data presented here support a cell-autonomous mechanism in which ITA produced by myeloid cells regulates gene expression within TAMs, in part, via inhibiting Tet2, thereby changing TAM inflammatory features and their role in recruiting CD8+ T cells into tumor sites. In our mouse tumor models, we cannot detect the intracellular accumulation of ITA in tumor-infiltrating lymphocytes (CD45+CD11b−) (fig. S4B). Moreover, we provide in vivo evidence that Irg1 deficiency may not directly affect the tumor killing function of T cells (fig. S11), whereas the study by Zhao et al. relied on mostly in vitro study using exogenously added supraphysiological high concentration of ITA. It therefore remains to be determined whether ITA is sufficiently uptake by the T cells and regulate T cell function in vivo. Despite different cell types and mechanisms of ITA action, non–cell-autonomous T cell proliferation versus cell-autonomous macrophage polarization, the study by Zhao et al. and our current study show that deletion of Irg1 in mice enhances antitumor immunity.

27There are two clinical implications of this study. First, it identifies ACOD1 as a potential target for developing immunotherapy drugs. The current antibody-based immune checkpoint blockades, in addition to their limited efficacy in most cancer types, are less convenient to administer because of its intravenous injection and take a long time to resolve adverse effect due to the long half-life of antibody. We show that Irg1-deficient mice, which develop normally, are resistant to several types of synergetic tumors. The antitumor effect conferred by Irg1 loss is similar to that from administration of anti–PD-(L)1 immunotherapy. The combination of Irg1 deficiency and PD-1/PD-L1 blockade can collaborate to enhance the antitumor immunity, suggesting that targeting ACOD1 can be effective and convenient for monotherapy as well as adjuvant therapy with current PD-(L)1 blockades for treating different cancer types. Mechanistically, the coordination between an ACOD1 inhibitor and PD-(L)1 blockades can be explained by the orthogonality of their respect target cells and mechanisms of action. While ACOD1 inhibitors target monocytes/macrophages to reprogram TAM polarization and promote antigen presentation and T cell trafficking, PD-(L)1 blockades target and reenergize CD8+ T cells and restore their antitumor effector functionality.

28Another clinical implication of our study is that IRG1-deficient macrophages could potentially be used for cell therapy for cancer. Although the adoptive transfer of T cells and CAR-T therapy have produced clinical efficacy in the treatment of blood malignancies, their applications have so far failed to demonstrate a similar efficacy in solid tumors. Much of the failures can be attributed to the lack of T cell trafficking into tumor sites and the immunosuppressive milieu suppressing the cytotoxic activity of T cells. Increasing evidence suggests that the adoptive therapy of genetically engineered monocytes and macrophages can overcome some of these limitations, as they penetrate tumors more efficiently and, in their proinflammatory state, exhibit tumoricidal activity in solid tumors. In this study, we show that the infusion of Irg1-deficient macrophages, but not neutrophils, into tumor-bearing mice exerts potent tumor-killing effects and also enhances the efficacy of anti–PD-(L)1 immunotherapy in the melanoma model (Fig. 6, E and F, and Fig. 7, A and B), in which most of TAMs should be from recruited monocytes. Differently, both recruited monocytes and tissue-resident macrophages are major sources of TAMs in KPC pancreatic tumors. It is known that monocyte-derived TAMs (recruited TAMs) play potent roles in antigen presentation, while resident TAMs are responsible for remodeling extracellular matrix. We think that Irg1 deletion should predominantly occur in “recruited” TAMs but not “resident” TAMs in the orthotopic KPC model (Fig. 7C). Whether transferred TAMs would affect resident TAMs and thereby change the extracellular matrix to affect the recruitment of MDSCs, as seen in the KPC pancreatic tumor model but not in the melanoma model, still needs further investigation. Nevertheless, our current findings, together with the recent development in CAR-macrophage engineering and induced pluripotent stem cell–derived macrophages, suggest that adoptive transfer of IRG1-deficient macrophages merits exploration as a cell therapy strategy for solid tumors.

29The genome-wide transcriptome data quantified as fragments per kilobase per million were downloaded from TCGA. We used the data from 18 cancer types, all of which had profiled enough numbers of samples for both cancer and normal tissues. Data include BLCA, BRCA, cholangiocarcinoma (CHOL), COAD, ESCA, HNSC, KICH, kidney clear cell carcinoma (KIRC), KIRP, liver hepatocellular carcinoma (LIHC), lung adenocarcinoma (LUAD), LUSC, PAAD, prostate adenocarcinoma (PRAD), rectum adenocarcinoma (READ), STAD, thyroid carcinoma (THCA), and UCEC. For the SKCM, the cancer samples were collected from TCGA, whereas the normal control samples were collected from Genotype-Tissue Expression (GTEx). The mRNA expression of IRG1 gene was compared between normal and cancer samples. Wilcoxon rank sum test was used to determine the significance of difference for these 19 cancer types. EPIC was used to calculate the immune cell fractions for all the samples. The cell fraction of TAMs was identified, and Spearman rank correlation test was used to determine the significance of correlation between immune cell fractions and IRG1 expression level.

30Samples from patients with HCC were collected by the Department of Liver Surgery, Zhongshan Hospital of Fudan University, and by the Department of General Surgery, Huashan Hospital of Fudan University. Informed consent was received from each patient before surgery. The procedures related to human subjects were approved by the Ethics Committee of Zhongshan Hospital (project approval number Y2020-622) and Huashan Hospital (project approval number 2020 M-010). Information about clinical samples used in this study is listed in table S2.

31Irg1−/− mice (JAX stock no. 029340) were purchased from the Jackson Laboratory. Animals were backcrossed for more than seven generations onto the C57BL/6J background and were maintained at Shanghai Research Center for Model Organisms. Myeloid cell–specific Irg1-deficient mice (Irg1f/f, Lyz-Cre) on C57BL/6J background was generated by Shanghai Model Organisms Center Inc. and were maintained at the Center for New Drug Evaluation and Research, China Pharmaceutical University (Nanjing, China). Tet2HxD KI–mutant mice on C57BL/6J background were constructed by using CRISPR-Cas9 system, which were provided by J. Chen and S. Gao (School of Life Sciences and Technology, Tongji University, Shanghai) and were maintained at GemPharmatech Co. Ltd. Analysis was performed on 6- to 10-week-old female mice obtained from the abovementioned breeding. Animals were given unrestricted access to a standard diet and tap water. Animal experiments were performed at Fudan Animal Center in accordance with animal welfare guidelines, and the procedures were approved by the Ethics Committee of the Institutes of Biomedical Sciences (IBS), Fudan University.

32For the genotyping of Irg1−/− mice, we followed the instruction from the Jackson Laboratory. Briefly, three primers were used for PCR followed by agarose gel electrophoresis: GTG GGG AGG GGA ACT ATG AG (forward primer for both wild-type and mutant Irg1), ATT TGG AGG AAC CCC ATG AC (reverse primer for wild-type), and CAG CCT CTA AGC CAG ACA GC (reverse primer for mutant). The sizes of PCR fragments are 157 and 220 base pairs (bp) for the wild-type and mutant Irg1, respectively.

33For the genotyping of myeloid cell–specific Irg1-deficient mice (Irg1f/f, Lyz2-Cre), two primers were used for the genotyping of flox: CAGGAAGGCCAGTGCTCA GTAATC (forward primer) and ACCTCCTCGCACCCCCTTTGTATG (reverse primer). The sizes of PCR fragments are 331 and 388 bp for the wild-type and mutant Irg1, respectively. Moreover, four primers were used to genotype Lyz2-Cre: GGAGGATGC TTAAATAGCAGG (transgene forward primer), CATCACTCGTTGCATCGACC (transgene reverse primer), AGTGGCCTCTTCCAGAAATG (internal positive control forward primer), and TGCGACTGTGTCTGATTTCC (internal positive control reverse primer). The sizes of PCR fragments are 150 and 521 bp for Cre+ and control, respectively.

34For the genotyping of Tet2HxD mice, two primers were used: CTGGATATCTGC AGAATTCACCTTACTACACCGAGAAGCCT and GGTACCGAGCTCGGATCCACCCTTCCTAG GTG ACAGGTGAC. The PCR products were cloned to pCDNA3.1 digested by EcoR I and BamH I. The constructs were sequenced and compared with wild-type Tet2 DNA sequence to confirm the two mutant sites as previously reported.

35HEK293T cells [American Type Culture Collection (ATCC), CRL-3216], B16-F10 cells (ATCC, CRL-6475), MC38 cells (Kerafast, catalog no. ENH204), and E0771 cells (ATCC, CRL-3461) were commercially purchased. FC1242 KPC cells were gifted from L. Fei (Fudan University). Tumor cells were maintained in Dulbecco's modified Eagle's medium (DMEM) containing 10% fetal bovine serum (FBS) (ExCell Bio, FSD500), 1% penicillin/streptomycin antibiotics.

36For differentiation of BMDMs, mice were euthanized in a 5% CO2 chamber and death was confirmed by cervical dislocation. Bone marrow was harvested from the femur and the tibia and differentiated in DMEM [containing 10% fetal calf serum (FCS), 1% penicillin/streptomycin, and macrophage colony-stimulating factor (M-CSF) (100 ng/ml)] for 7 days. Then, 1 × 106 BMDMs/ml were cocultured with tumor cells or treated with tumor-derived CM for different time points.

37To prepare for TCM, B16-F10, E0771, MC38, or FC1242 tumor cells were seeded with 10 ml of medium in 10-cm dish, and the culture medium were collected after 24 hours followed by filtration using 0.45-μm filter (Corning) and were then stored at −20°C.

38For generation of the syngeneic mouse models of melanoma, colon, and breast tumor, 2 × 105 or 1 × 105 B16-F10 cells, 1 × 105 MC38 cells, and 1 × 105 E0771 cells were kept in 25 μl of DMEM, respectively. The same volume of Matrigel (BD Biosciences, 356234) was added and mixed. The cells were subcutaneously injected into Irg1+/+ or Irg1−/− mice (female, 8 to 10 weeks). Tumors were measured every 2 or 3 days, and the volume was calculated using the formula V = as (width by width by length)/2 as previously described. The survival rate was determined as previously reported. Briefly, mice were removed from the study if they were found dead in their cage, declining health necessitated killing, or tumor diameter exceeded 15 mm. These parameters were used to construct a Kaplan-Meier survival curve.

39For the syngeneic mouse models of pancreatic tumor, 1 × 105 FC1242 cells were kept in 25 μl of DMEM and were mixed with 25 μl of Matrigel (BD Biosciences). The cell suspension was then injected orthotopically into the pancreas of Irg1+/+ or Irg1−/− mice (female, 8 to 10 weeks). At day 15 after FC1242 tumor inoculation, the pancreatic tumors were dissected, weighted, and subjected to further analysis.

40For cancer immunotherapy, anti–PD-L1 (200 μg per mouse; BioXcell, BP0101) was intraperitoneally injected at days 6, 8, 10, 13, 15, and 17 after B16-F10 tumor inoculation; anti–PD-1 (200 μg per mouse; BioXcell, BE0146; or Biointron, B176401) was intraperitoneally delivered at days 10, 12, and 14 after E0771 tumor inoculation or intraperitoneally injected at days 7, 10, 13, and 16 after FC1242 KPC tumor inoculation. At indicated days after anti–PD-(L)1 treatment, tumor size was determined.

41Mice were intratumorally injected with the CXCR3 antagonist SCH546738 (600 mg per mouse; MedChem Express, HY-10017), starting at day 6 after tumor inoculation and then maintained by daily injection for 1 week.

42Human peripheral blood was collected from healthy women donors (20 to 40 years old) following the protocol approved by the Ethics Committee of IBS, Fudan University (project approval number 2021-23). PBMCs were isolated from human peripheral blood following the manufacturer’s instructions. Briefly, whole blood (10 ml) was layered on 15 ml of Ficoll (Cytiva, 17144002) and spun for 22 min at 2300 rpm with the brake off to prevent disrupting the density gradient during deceleration. PBMCs were then isolated from the middle layer.

43PBMCs were counted, and the cell density was adjusted to 2.5 × 106 cells/ml and was then maintained in DMEM supplemented with 10% (v/v) FCS, 2 mM l-glutamine, and 1% penicillin/streptomycin. Cells were seeded into six-well plates in culture medium containing M-CSF (100 ng/ml; Novoprotein, C417) and maintained at 37°C, 5% CO2, for 9 days, to allow the differentiation into macrophages. The proper differentiation into macrophages was verified by the staining of CD14 and CD163. In these human macrophages, siRNA against IRG1 (sequence: CACAGGUGAGAGAGCUGCUCAGUAA) was transfected using RNAiMAX reagent (Life Technologies, 13778150) for 12 hours, followed by another 12 hours of treatment with MDA-MB-231-TCM. The culture medium and challenged cells were collected for further analysis.

44For the macrophage adoptive transfer experiment, tumor volume of 1 × 104 BMDMs/mm3 was injected intratumorally when the tumor reached ~200 mm3 after B16-F10 inoculation with or without simultaneous intraperitoneal injection of anti–PD-L1 (200 μg per mouse).

45Neutrophils were isolated from mouse bone marrow by percoll gradient, and the cell purity was confirmed by flow cytometry. For the neutrophil adoptive transfer experiment, 1.5 × 106 neutrophils were intratumorally injected into B16-F10 tumor bearing mice at days 8 and 11 after inoculation. Later, the tumors were harvested and weighted at day 14. The same experimental setting was applied for the macrophage adoptive transfer experiment in Fig. 6 (E and F).

46For macrophage adoptive transfer in KPC model, BMDMs at 1 × 107/50 μl were intravenously injected at days 10 and 13 after FC1242 KPC tumor inoculation. Mice were euthanized at day 15 after tumor implantation, and the pancreatic tumors were dissected, weighted, and subjected to further analysis. At a later stage in KPC model, BMDMs at 1 × 107/50 μl were intravenously injected at days 21 and 23 after KPC tumor inoculation. Mice were euthanized at day 26 after tumor implantation, and the samples were collected as described above.

47Cell number and cell diameter were measured using an automated cell counter (Countstar). C13-labeled ITA was added as an internal standard. Cells were fixed by immediate addition of 1 ml of 80% (v/v) chilled (−80°C) methanol. Cell extracts were analyzed by ultrahigh performance liquid chromatograph (Acquity UPLC I-Class, Waters) coupled to a Triple Quadrupole Mass Spectrometer (Xevo TQ-XS, Waters). Cells were assumed to have a spherical shape, and intracellular metabolite concentrations were calculated using external standard curves and taking into account cellular diameter (d, micrometers) and cell number using the following equation: (metabolite) = metabolite quantity (moles)/[(4/3000)π(d/2)3*cell number].

48Tumors were harvested at indicated time. Briefly, 100 mm3 of each tumor was chopped, digested, and filtered through a 70-μm cell strainer to generate a single-cell suspension. After red blood cell lysis (Beyotime, C3702), cells were counted and plated in phosphate-buffered saline (PBS). Cell surface molecule staining was performed at 4°C for 15 to 30 min in PBS in the dark.

49For intracellular staining, cells were fixed/permeabilized in 50 μl of a saponin-containing buffer (BD Biosciences, 554722) and incubated at 4°C for 30 min in the dark. Cells were then washed two times with saponin-containing buffer (BD Biosciences, 554722) and resuspended in staining buffer followed by antibody staining. Stained samples were acquired on a LSRFortessa flow cytometer (BD Biosciences). Collected data were analyzed using FlowJo software (Tree Star Inc.). The antibodies used in this study are Fixable Viability Dye eFluor 506 (Invitrogen, 65-0866-14); phycoerythrin (PE) anti-Cd45 (BD Biosciences, 553081), FITC anti-CD45 (BD Biosciences, 553079), FITC anti-Cd11b (BD Biosciences, 561688), allophycocyanin (APC) anti-Cd11b (BD Biosciences, 553312), PE anti-F4/80 (BD Biosciences, 565410), APC anti-F4/80 antibody (Invitrogen, 17–4801-82); APC anti-mouse Ly-6G (BD Biosciences, 553129), APC-Cy7 anti–Ly-6C (BD Biosciences, 560596), PE-Cy7 anti-Cd3e (BD Biosciences, 552774), APC anti-Cd3e (BD Biosciences, 561826), FITC anti-Cd8a (BD Biosciences, 553030), APC anti-Cd8a (BD Biosciences, 553035), Peridinin-Chlorophyll-Protein (PerCP)/cyanine5.5 (Cy5.5) anti-Cd80 (BioLegend, 104722), PE-Cy7 anti-Cd86 (BD Biosciences, 560582), PE anti-Cd4 (BD Biosciences, 557308), PE anti–Nk-1.1 (BD Biosciences, 557391), PE-Cy7 anti-NOS2 (type 2 nitric oxide synthase) (eBioscience, 25-5920-80), APC anti-CD206 (BioLegend, E-AB-F1135E), APC anti–Ly-6G(GR1) (MACS, 130-098-047), APC anti-Cd40 (BD Biosciences, 558695), BV421 anti–MHC-I (BD Biosciences, 749711), APC anti–MHC-II (BioLegend, 107613), PE anti–Ifn-γ (BD Biosciences, 554412), PE anti–TNF-α (BioLegend, 506305), PE-Cy7 anti–PD-1 (BioLegend, 135215), FITC anti- CD14(BioLegend, 325604), and APC anti-CD163 antibody (BioLegend, 333610).

50For visualizing mouse immune cells in the TME, tumor samples were collected and then fixed in 4% paraformaldehyde. Paraffin-embedded sections (4 μm) were deparaffinized in xylene, then rehydrated through graded alcohol, and washed briefly in tap water. To retrieve antigenicity, sections were boiled in 10 mM citrate buffer (pH 5.8) for 30 min in microwave (800 W). Sections were left to room temperature and washed briefly in PBS. Following that, sections were incubated with goat serum diluted in PBS (pH 7.4) for 1 hour at room temperature and were washed briefly in PBS. Subsequently, sections were incubated at 4°C overnight with the primary antibodies specific for CD8 [1:200 dilution; Cell Signaling Technology (CST), 98941S] or iNOS (1:200 dilution; CST, 13120), rinsed with fresh PBS, and incubated with the secondary antibody Alexa Fluor 647 goat anti-rabbit immunoglobulin G (IgG) (1:200 dilution; Thermo Fisher Scientific, A-21245) at room temperature for 1 hour in darkness, and then washed briefly in PBS. Last, sections were incubated by ProLongGold antifade reagent with 4′,6-diamidino-2-phenylindole (Invitrogen, p36931) and mounted. Immunofluorescence pictures were taken by a Leica SP8 laser scanning confocal microscope.

51To examine IRG1 expression and immune cells in samples from patients with HCC after immunotherapy, the following antibodies were used: IRG1 (D6H2Y) rabbit mAb (1:200 dilution; CST, 77510s); CD8 (1:1000 dilution; Abcam, ab4055); CD45 (1:500 dilution; Abcam, ab30470); donkey anti-rabbit IgG (H+L) secondary antibody, Alexa Fluor 488 conjugate (1:200 dilution; Thermo Fisher Scientific, A-21206); and donkey anti-mouse IgG (H+L), Alexa Fluor 594 (1:200 dilution; Thermo Fisher Scientific, A-21203). Immunofluorescence pictures of stained slides were taken by OLYMPUS SLIDEVIEW VS200 and analyzed by ImageJ.

52Total RNA was extracted from cells using TRIzol (Invitrogen) as recommended by the manufacturer’s protocol. RNA was reversely transcribed using oligo(dT) primers. Diluted complementary DNA (cDNA) was then used in qRT-PCR reactions containing SYBR Premix ExTaq (TaKaRa) and gene-specific primers. The reactions were performed in the QuantStudioTM 6 Flex Real-Time PCR System (Applied Biosystems). 18S RNA was used as a housekeeping control. Primers used in this study are listed in table S3.

53Irg1−/− and Irg1+/+ mice were injected subcutaneously with 2 × 105 B16-F10 cells (in 50 μl of PBS). Ten days after inoculation, tumors were harvested, and 100 mm3 of each tumor was collected. Tumor tissues from three mice of the same group were randomly combined into one mixed sample, and then, tissues were digested and went through 70-μm filters (BD Biosciences, 352350) to achieve single-cell suspensions. After treatment with red blood cell lysis buffer (Beyotime Biotechnology, C3702) for 5 min at room temperature, all samples were washed and resuspended in flow cytometry buffer (PBS/0.5% albumin/2 Mm EDTA). In the end, Cd45+ immune cells (n = 6000 per sample) were sorted by flow cytometry and subjected to scRNA-seq (NovelBio Bio-Pharm Technology Co.).

54BD Rhapsody system was used to capture the transcriptomic information of the single cells. Single-cell capture was achieved by random distribution of a single-cell suspension across >200,000 microwells through a limited dilution approach. Beads with oligonucleotide barcodes were added to saturation so that a bead was paired with a cell in a microwell. The cells were lysed in the microwell to hybridize mRNA molecules to barcoded capture oligos on the beads. Beads were collected into a single tube for reverse transcription and ExoI digestion (New England Biolabs, M0293S). Upon synthesis, each cDNA molecule was tagged on the 5′ end (that is, the 3′ end of mRNA transcript) with a unique molecular identifier (UMI) and cell barcode indicating its cell of origin. Whole transcriptome libraries were prepared using the BD Rhapsody single-cell whole-transcriptome amplification (WTA) workflow including random priming and extension (RPE), RPE amplification PCR, and WTA index PCR. The libraries were quantified using a High Sensitivity DNA chip (Agilent) on a Bioanalyzer 2200 and the Qubit High Sensitivity DNA assay (Thermo Fisher Scientific). Sequencing was performed by sequencer (Illumina, San Diego, CA) on a 150-bp paired-end run.

55Next, scRNA-seq data analysis was performed by using the NovelBrain Cloud Analysis Platform. We applied fastp with default parameter filtering the adaptor sequence and removed the low-quality reads to achieve the clean data. UMI-tools were applied to identify the cell barcode whitelist. The UMI-based clean data were mapped to mouse genome (Ensemble version 100) using STAR mapping with customized parameter from UMI-tools standard pipeline to obtain the UMI counts of each sample. Cells contained over 200 expressed genes and mitochondria UMI rate below 10% passed the cell quality filtering, and mitochondria genes were removed in the expression table. Seurat package (version 3.1.4, https://satijalab.org/seurat/) was used for cell normalization and regression, based on the expression table according to the UMI counts of each sample and percent of mitochondria rate to obtain the scaled data. Principal components analysis (PCA) was constructed on the basis of the scaled data with top 2000 high variable genes, and top 10 principals were used for t-distributed Stochastic Neighbor Embedding (t-SNE) construction and UMAP construction.

56By using graph-based cluster method (resolution, 0.8), we acquired the unsupervised cell cluster result based on the PCA top 10 principal and calculated the marker genes by FindAllMarkers function with Wilcoxon rank sum test algorithm under the following criteria: (i) Ln (Fold Change) (LnFC) > 0.25, (ii) P < 0.05, and (iii) minimum percentage (Min.pct) > 0.1. To identify the cell type detailed, the clusters of same cell type were selected for re-t-SNE analysis, graph-based clustering, and marker analysis. To identify differentially expressed genes among samples, the function FindMarkers with Wilcoxon rank sum test algorithm was used under following criteria: (i) LnFC > 0.25, (ii) P < 0.05, and (iii) Min.pct > 0.1.

57Chromatin immunoprecipitation (ChIP) assay was performed as described previously. Briefly, cells were cross-linked with 1% paraformaldehyde and sonicated at 4°C for 20 min (Bioruptor, low mode). Chromatin was immunoprecipitated at 4°C for 3 hours with indicated antibodies. Antibody-chromatin complexes were pulled down using protein A-Sepharose (Millipore, 16-125), washed, and eluted. After cross-link reversal and proteinase K treatment (TaKaRa, 9034), immunoprecipitated DNA was extracted using the PCR Purification Kit (QIAGEN, 28006). The following primer sequences were used: Acod1 (forward): AATGGGCTGTCTGTGAGA; Acod1 (reverse): CCAGGTCTGTCCCTTTCAC.

58Supernatants of 5 × 105 BMDM cells in six-well plates were collected at 12 hours after stimulation of B16-F10-TCM. Cytokines were assessed by enzyme-linked immunosorbent assay by using the kits of Tnf-α (DY410-05) and IL-6 (DY406-05), purchased from R&D Systems (Minuteneapolis, MN, USA), according to the manufacturer’s instructions.

59Myeloid-specific Irg1-KO mice and wild-type controls were inoculated with 1 × 105 B16-F10. Then, the antibody of anti-CSF1R (500 μg per mouse, BioXcell, BE0213) or anti-Ly6G (400 μg per mouse, BioXcell, BE0075) was intraperitoneally injected from day 3 after inoculation for every 2 days. The tumors were harvested to define the depletion efficiency at day 14 by flow cytometry analysis.

60T cell migration assay was performed by using 24-well 6.5-mm transwell with 5.0 μm Pore Polycarbonate Membrane Insert. In brief, BMDMs (5 × 105) were cocultured with B16-F10 tumor cells (5 × 105) or challenged with B16-F10-TCM for 12 hours at the bottom chamber. Meanwhile, CD8+ T cells (1 × 106) were isolated from the spleen of wild-type mice and suspended in 200 μl of 1% FBS–RPMI 1640 and were then placed at the top chamber with or without 100 nM SCH546738 (MCE, HY-10017). The migration of CD8+ T cells from the top to the bottom chamber was evaluated after incubation for 4 hours at 37°C in a 5% CO2 atmosphere. Cells in the bottom chambers were collected and the number of CD8+ T cells was counted.

61In addition, PMA-differentiated THP1-derived macrophages (1 × 106) were challenged with MDA-MB-231-TCM for 12 hours at the bottom chamber. Meanwhile, Jurkat T cells (1 × 106) were suspended in 200 μl of 1% FBS–RPMI 1640 and were placed at the top chamber. The migration of Jurkat T cells from the top to the bottom chamber was evaluated after 4-hour incubation at 37°C in a 5% CO2 atmosphere.

62Furthermore, human peripheral blood was subjected to EasySep Human CD8+ T Cell Enrichment Kit (STEMCELL Technologies, 17953) to isolate primary CD8+ T cells, and these T cells were placed at the top chamber. Meanwhile, the CM from primary human macrophages after stimulation with MDA-MB-231-TCM was placed at the bottom chamber. The migration of CD8+ T cells from the top to the bottom chamber was evaluated after 6-hour incubation at 37°C in a 5% CO2 atmosphere, as previously reported.

63Statistical analyses were performed using GraphPad Prism software version 8.0. Summarized views on data that underlie the statistical tests as well as the exact P values are available in the Source Data. The statistical details for each experiment are also provided in the figure legends. Generally, the statistical analyses were performed using a two-tailed Student’s t test for paired comparisons or one/two-way analysis of variance (ANOVA) for multiple comparisons. The data shown (unless indicated otherwise) represent the results obtained from at least triplicate independent experiments with the means ± SD or SEM. Where indicated, *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001, and n.s. denotes nonsignificant.

64In this study, in vivo experiments were repeated at least two or three times, and the sample size was defined empirically or referred to previous literatures. In vitro experiments were performed including at least three biological replicates, with two to three technical replicates for each biological replicate, to confirm the reproducibility.