DJ-1 inhibition reshapes tumor microenvironment and potentiates immune checkpoint inhibitors

Abstract

Immune checkpoint inhibitors (ICIs) can essentially treat cancer but only in a small subset of patients. Treatment strategies capable of effectively and robustly sensitizing refractory patients to ICIs represent a highly coveted yet unmet clinical need. In this study, we identified DJ-1 as a negative T cell regulator. DJ-1 knockout boosts antitumor immunity and significantly potentiates PD-1 and TIM-3 blockades in murine cancer models. Single-cell sequencing of tumor-infiltrating CD45⁺ cells revealed that DJ-1 deficiency indirectly activates T cells by reprogramming macrophages. Mechanistically, loss of DJ-1 increases reactive oxygen species (ROS) in macrophages, activating NF-κB/STAT3 signaling to promote differentiation into Cxcl9⁺ immune-stimulatory phenotypes while reducing immune-suppressive Spp1⁺ macrophages. Notably, this reprogramming may be stable across tumor microenvironments because the transplanted DJ-1–deficient macrophages maintain T cell–activating capacity. Pharmacological inhibition of DJ-1 by disulfiram markedly potentiated antitumor efficacy of PD-1 blockade. This designates DJ-1 as a promising target for overcoming immune checkpoint resistance and optimize combination therapies.

DJ-1 inhibition reprograms macrophages to activate T cells and enhance antitumor immunity.

INTRODUCTION

Cancer immunotherapies have revolutionized the treatment landscape for various malignancies, with immune checkpoint inhibitors (ICIs) showing significant clinical efficacy but only in 15 to 40% patients across different cancer types (1, 2). One of the key factors influencing this variability in response is the tumor microenvironment (TME), which plays a central role in regulating antitumor immunity. The TME is a complex and dynamic network of immune, stromal, and tumor cells, which can either promote or suppress immune responses (3). Identifying and suspending immunosuppressive mechanisms within the TME are therefore critical to improving the efficacy of immunotherapy and making it benefit more patients with cancer and cancer indications.

Immunological heterogeneity of TME dictates immunotherapy outcomes. Hot tumors, highly infiltrated by effector immune subsets such as cytotoxic T cells, tend to respond better to ICIs that reinvigorate the preexisting exhausted immune cells (4, 5). In contrast, cold tumors with limited immune cell infiltration are often resistant to ICIs. This emphasizes the need for strategies to reprogram TME and enhance immune activation (6, 7). The activation of T cells is mainly carried out by professional antigen-presenting cells (APCs) such as tumor-associated macrophages (TAMs) that are remarkably adaptable and can be reeducated by TME into diverse subsets (8, 9). TAM polarization toward the proinflammatory M1 phenotype by epigenetic modification, lipid metabolism, and cell cross-talk can activate CD8⁺ T cells but is often less effective due to competing TME signals (10–12). Another strategy is directly inducing proinflammatory cytokines by targeting the effectors on key innate immune-regulatory pathways including the stimulator of interferon gene (13, 14). However, these strategies face significant challenges including toxicity risks like cytokine storms and off-target effects and require further refinement to ensure clinical safety and efficacy (15, 16).

Recent studies have highlighted the impact of oxidative stress on immune modulation within TME. Reactive oxygen species (ROS) may result in activation or shutdown of diverse cellular processes such as triggers aberrant DNA methylation and histone modifications (17). But in many cases, the production of the hydrogen peroxide is a critical mediator of monocyte recruitment, and oxidative burst is also required in the clearance of apoptotic cells by alternatively activated macrophages (18–20). Therefore, targeting oxidative stress–related genes may be a promising strategy to enhance immunotherapy efficacy. Previous researches have confirmed that DJ-1, known as an oxidative stress sensor, participates in the onset of oxidative stress–related diseases such as cancer, neurodegenerative disorders, and type 2 diabetes (21–23). Evidence has recently emerged that it plays a key role in immune and inflammatory disorders (24, 25), emphasizing its potential as a therapeutic target for inflammatory conditions. However, the role of DJ-1 in the tumor immune microenvironment, particularly in tumor-infiltrating immune cells, is still largely unexplored.

In this study, we found that both systemic knockout (KO) and pharmacological inhibition of a key regulator of oxidative stress called Park7 (encodes DJ-1 protein) in mice boost antitumor immunity and synergistically enhance the antitumor effects of various ICIs including programmed cell death protein 1 (PD-1) and T-cell immunoglobulin and mucin-domain containing protein 3 (TIM-3) inhibitors. Mechanistically, Park7 deficiency promotes proinflammatory Cxcl9⁺ TAMs as opposed to immune-suppressive Spp1⁺ TAMs, thereby augmenting T cell recruitment and activation. DJ-1 loss in macrophages elevates ROS that up-regulate proinflammatory genes and cytokine production by modulating transcription factors such as signal transducer and activator of transcription 3 (STAT3) and nuclear factor κB (NF-κB). In sum, DJ-1 blockade represents a promising strategy to reprogram TME, enhance immune cell infiltration and activation, and has potential to sensitize more patients with cancer to curative ICIs.

RESULTS

DJ-1 deficiency enhances antitumor activity of ICIs

Accumulating evidence suggests that oxidative stress plays an important role in immune regulation during tumor progression (26, 27). Both our group and others have identified DJ-1, a protein encoded by the PARK7 gene, as a crucial oxidative stress sensor (23, 28). To determine whether DJ-1 associates with immune regulation, we first tested antitumor efficacy of ICIs in DJ-1 KO (Park7^−/−) and wild-type (WT; Park7^+/+) littermates on a C57BL/6 background (Fig. 1A). As anticipated, PD-1 blockade moderately reduced (by 35%) the growth of subcutaneous immunogenic MC38 colorectal tumors in WT mice. Notably, Park7-KO reduced tumor growth comparably to PD-1 inhibition, and PD-1 blockade further impeded MC38 growth in Park7-KO mice to 75% relative to isotype-treated WT littermates. These data indicate that Park7 knockout mice exhibit an enhanced response to PD-1 blockade (Fig. 1, B and C).

Fig. 1. DJ-1 deficiency enhances the antitumor activity of ICIs.

(A) MC38 colorectal cancer cells were subcutaneously inoculated into DJ-1 KO (Park7^−/−) and WT (Park7^+/+) littermates on a C57BL/6 background. Once tumors reached a size of ~50 to 100 mm³, mice were treated with either α-PD-1/α-TIM-3 antibodies or an IgG2a isotype control antibody. (B and F) Tumor size was monitored every 2 to 3 days, and tumor weights were measured at the end of the experiment. (C and G) Final tumor weights in each group are shown. (D, E, H, and I) Representative flow cytometry plots and quantification of CD8⁺ T cell populations in MC38 tumors. DJ-1 KO mice showed a significant increase in CD8⁺ T cell infiltration compared to WT controls. Bar graphs represent data as mean ± SD. Statistical significance: *P < 0.05 and ***P < 0.001.

To further investigate the immune-phenotype within the TME, we analyzed CD8⁺ T cell infiltration in subcutaneous tumors. As shown in Fig. 1 (D and E), tumors from WT mice treated with immunoglobulin G2a (IgG2a) exhibited only 0.5% CD8⁺ T cell infiltration, while those treated with α-PD-1 showed a modest increase to 1.0%. In contrast, tumors from KO mice demonstrated significantly higher CD8⁺ T cell infiltration, with 2.2% observed in the IgG2a-treated group and 4.3% in the α-PD-1–treated group. These results indicate that tumors in KO mice consistently contain more CD8⁺ T cells compared to those in WT mice irrespective of the treatment suggestive of enhanced immune response within the TME.

To further validate the robust negative role of DJ-1 in shaping the tumor immune microenvironment, we conducted additional experiments using another ICI, the α-TIM-3 antibody, in the MC38 tumor model, as well as the α-PD-1 antibody in a second Lewis lung carcinoma (LLC) tumor model. In MC38 tumor–bearing mice, systemic DJ-1 KO significantly enhanced the antitumor efficacy of α-TIM-3 antibody, accompanied by an increased percentage of CD8⁺ T cells within the tumor tissues (Fig. 1, F to I), resulting in rapid tumor regression. Similarly, in the LLC model, DJ-1 deficiency substantially potentiated the therapeutic response to α-PD-1 treatment, further demonstrating a consistent synergistic effect (fig. S1, A to D). Together, these results suggest that DJ-1 negatively regulates the efficacy of α-PD-1 and α-TIM-3 therapies within the TME.

scRNA-seq reveals that DJ-1 negatively regulates intratumoral T cell differentiation, activation, and recruitment

To investigate the mechanisms underlying the enhanced sensitivity to ICIs driven by DJ-1 deficiency, we performed single-cell RNA sequencing (scRNA-seq) on intratumoral immune cells. Tumors were dissociated into single-cell suspensions, and CD45⁺ immune cells were sorted for scRNA-seq analysis (Fig. 2, A and B). A total of 16,529 CD45⁺ cells were analyzed, comprising 9226 cells from WT (Park7^+/+) mice and 7303 cells from DJ-1 KO (Park7^−/−) mice. The flow cytometry and subcluster analysis identified overall higher lymphocyte infiltration or survival of each distinct immune cell population including myeloid cells, T cells, and natural killer (NK) cells (Fig. 2C) in DJ-1 KO tumors. Comparative transcriptomic analysis of signaling pathways revealed that anti–PD-1 treatment up-regulated numerous inflammatory processes such as leukocyte migration, defense responses, and cell adhesion in DJ-1–deficient tumors. Notably, key genes associated with T cell cytotoxicity, such as Nkg7 and Gzmb, were significantly up-regulated in Park7^−/− mice compared to Park7^+/+ controls. Further, APCs in DJ-1–deficient tumors exhibited elevated expression of major histocompatibility complex molecules, suggesting enhanced antigen presentation (fig. S2, A and B).

Fig. 2. DJ-1 negatively regulates intratumoral T cell differentiation, activation, and recruitment.

(A and B) Representative plots and quantification of tumor-infiltrating CD45⁺ cells from MC38 tumors. CD45⁺ immune cells were sorted from DJ-1 WT (Park7^+/+) and KO (Park7^−/−) mice for scRNA-seq and T cell receptor sequencing (TCR-seq). FACS, fluorescence-activated cell sorting. (C) Uniform Manifold Approximation and Projection (UMAP) visualization of 16,529 CD45⁺ cells from WT and Park7^−/− mice, revealing five distinct immune cell clusters: myeloid cells, epithelial cells, T cells, NK cells, and other minor populations. (D) UMAP visualization of T cell subclusters within the TME, with six distinct T cell subpopulations identified: T naïve, T memory, T circulating, T cytotoxic, T transitory, and T suppressive (each represented by a unique color). (E and F) Gene expression profiles of marker genes defining the six T cell subclusters (e.g., Sell, Cd4, Gzmb, and Foxp3) and quantification of the relative proportions of each subcluster in WT and KO groups. (G) Analysis of the top 20 most frequently observed TRA/TRB CDR3 clonotypes across WT and Park7^−/− tumor-infiltrating lymphocytes (TILs). Clonotype abundance was measured, and T cell receptor (TCR) clones were tracked.

Given the critical role of T cells in ICI-mediated antitumor activity, we next focused our analysis on tumor-infiltrating T cells in the MC38 colorectal cancer model. Using scRNA-seq data, we defined T cell subpopulations based on the expression of key marker genes (29, 30): Sell (T naïve cells), Cd4 and Cxcr6 (T memory cells), Mki67 and Hmgb2 (T circulating cells), Cd8a and Gzmb (T cytotoxic cells), Cx3cr1 and Ccl5 (T transitory cells), and Foxp3 (T suppressive cells) (Fig. 2, D and E). Quantitative analysis of these subpopulations revealed notable differences between DJ-1 KO (Park7^−/−) and WT (Park7^+/+) T cells. Notably, the proportions of memory, effector, and cycling T cell subsets increased exponentially in tumors from DJ-1 KO mice, while the proportions of transitory and suppressive T cell populations decreased markedly (Fig. 2F). These findings suggest that DJ-1 deficiency sustains T cell cytotoxicity, limits T cell exhaustion, and promotes the recruitment of cycling T cells, collectively enhancing antitumor immunity.

To further explore the impact of DJ-1 deficiency on T cell dynamics, we conducted single-cell T cell receptor sequencing (scTCR-seq) in combination with scRNA-seq, preserving barcode information to link transcriptomic profiles with immune repertoire data (31, 32). Clonal T cells were identified on the basis of TRA/TRB-CDR3 sequence pairs, and the top 20 most frequently expanded clonotypes were analyzed. In Park7^+/+ T cells, the most abundant clonotype accounted for only 10.88% of detected sequences, reflecting relatively low clonal expansion. In contrast, Park7^−/− T cells exhibited a remarkable increase in clonal expansion, with the top clonotype (CASSLDNYAEQFF) observed more than 100 times and representing 26.12% of total detected sequences (Fig. 2G). These observations imply the distinct influence of DJ-1 deficiency on the proliferation of tumor-infiltrating lymphocytes (TILs).

DJ-1 indirectly modulates T cell function

Our data thus far suggest a key role of T cells in tumor control. To confirm the involvement of CD8⁺ T cells, we injected MC38-bearing DJ-1 KO and WT mice with neutralizing α-CD8 antibodies prior to α–PD-1 treatment. As expected, depletion of CD8⁺ T cells neutralized antitumor activity of α–PD-1 antibody, strongly suggesting that CD8⁺ T cells are key tumor-killing effectors (Fig. 3A).

Fig. 3. DJ-1 indirectly modulates T cell function.

(A) Impact of CD8⁺ T cell depletion on α-PD-1 therapy in WT and KO mice. WT and KO mice inoculated with MC38 cells received intraperitoneal injections of α-CD8 antibody or isotype control, with tumor sizes measured every 2 days. (B) Evaluation of isolated splenic T cell proliferation after coculture with anti-CD3 and anti-CD28. T cells were prestained with carboxyfluorescein diacetate succinimidyl ester (CFSE); fluorescence attenuation indicates cell division. (C and D) Proportions of CD69⁺ and CD25⁺ splenic T cells isolated from WT and KO mice. (E) Proportion of PD-1⁺ splenic T cells isolated from WT and KO mice. (F) Levels of IFN-γ secreted by activated splenic T cells. T cells were stimulated with anti-CD3 and anti-CD28 antibodies for 24 hours, followed by a 6-hour incubation with a cell activation cocktail. IFN-γ concentration in the culture supernatant was measured by enzyme-linked immunosorbent assay (ELISA). n.s. > 0.05 indicates no statistical difference.

We next investigated whether DJ-1 directly affects T cell function. The splenic T cells from Park7^+/+ and Park7^−/− mice were isolated, and their proliferation was further analyzed by using carboxyfluorescein diacetate succinimidyl ester–based flow cytometry. As shown in Fig. 3B, no significant differences in proliferation were observed between the two groups. Furthermore, we assessed the expression of T cell activation markers (CD25 and CD69) and the inhibitory receptor PD-1 to evaluate T cell activation and exhaustion states, respectively (33). Again, DJ-1 deletion did not significantly alter these markers in splenic T cells (Fig. 3, C to E).

To further examine the functional activity of T cells, splenic T cells were stimulated with anti-CD3 and anti-CD28 antibodies for 24 hours, followed by treatment with a cell activation cocktail for 6 hours. The expression of interferon-γ (IFN-γ), a key cytokine associated with T cell activation and cytotoxicity, was measured. As shown in Fig. 3F, acute stimulation of DJ-1 KO splenic T cells did not result in significantly altered IFN-γ expression compared to WT splenic T cells.

Overall, these findings suggest that DJ-1 deficiency does not directly affect T cell activity, activation, or proliferation. Instead, the enhanced antitumor immunity observed in DJ-1–deficient mice likely stems from indirect modulation of T cell function through interactions with other immune cell types or components of the TME. This highlights DJ-1 as an upstream regulator of immune responses, influencing T cell–mediated antitumor effects through indirect mechanisms. We posit that DJ-1 does not have any critical role in functionality of T cells.

DJ-1 deficiency triggered immune activation through antigen presentation

The above results suggest that DJ-1 may influence immune cell interactions to regulate T cell immunity. Therefore, we further investigated the global landscape of tumor-infiltrating immune cells (CD45⁺) in Park7^−/− and Park7^+/+ mice using scRNA-seq to explore intrinsic heterogeneity and cell-cell interactions. After quality filtering, we classified 17,367 Park7^+/+ and 18,643 Park7^−/− high-quality immune cells into 21 clusters, grouped into five major cell types: myeloid cells, epithelial cells, T cells, NK cells, and fibroblasts (Fig. 4A). To infer molecular interactions mediating intercellular communication, we computed the strength of ligand-receptor pairs in the scRNA-seq dataset using CellChat package (34) and observed an increased enrichment in the C-C motif chemokine ligand and C-X-C motif chemokine ligand pathways in Park7^−/− immune cells. On the basis of these observations, we propose that DJ-1 plays a key role in enhancing the intercellular cross-talk among immune cells (Fig. 4, B to D) through chemokine signaling leading to potentially more effective antigen presentation and T cell activation.

Fig. 4. DJ-1 deficiency triggered immune activation through antigen presentation.

(A) UMAP analysis of 17,367 WT and 18,643 Park7^−/− (KO) CD45⁺ cells isolated from tumor tissues. Cells were categorized into five major groups based on differential gene expression: epithelial cells, myeloid cells, T cells, NK cells, and fibroblasts. (B and C) Analysis of cell communication interaction strength. Circos plots illustrate putative ligand-receptor interactions between T cell, NK cell, and myeloid cell clusters within the TME in both WT and KO groups. (D) Summary of ligand-receptor interactions among different cell clusters, with blue representing the WT group and orange representing the KO group. (E) Steps for OVA immunization and MC38-EGFP-OVA subcutaneous injection. OVA protein mixed with incomplete Freund’s adjuvant at a 1:1 volume ratio was emulsified into a “water-in-oil” emulsion. Mice received a subcutaneous injection of emulsion (10 μg OVA/20 μl). MC38 cells were stably transfected with the pCDH-OVA-EGFP plasmid via lentiviral packaging and sorted by flow cytometry. FITC, fluorescein isothiocyanate. (F and G) Tumor size was measured every 5 days (F), and tumor weight was recorded at the end of the experiment (G). (H) TIL content was analyzed using flow cytometry. Single cells from tumor tissues were stained with fluorescent antibodies and analyzed. (I) IFN-γ levels in tumor tissues were determined. Tumor homogenates were prepared, and IFN-γ content was measured by ELISA.

Given the critical role of professional APCs in initiating T cell responses, we further assessed APC function using a highly immunogenic protein, ovalbumin (OVA), in Park7^−/− mice. C57BL/6J mice were immunized with OVA mixed with incomplete Freund’s adjuvant, and WT and KO mice were subcutaneously transplanted with MC38–enhanced green fluorescent protein (EGFP)–OVA cells (Fig. 4E). Continuous OVA protein immunization resulted in significantly slower growth of MC38-EGFP-OVA tumors in KO mice compared to WT mice, indicating a stronger immune response in Park7^−/− mice (Fig. 4, F and G). In line with this, elevated frequencies and IFN-γ secretion of TILs from tumor tissues were observed in KO mice postimmunization (Fig. 4, H and I).

These results suggest that Park7^−/− mice are more responsive to the antitumor effects of OVA immunization, possibly due to the enhanced presentation of OVA-derived antigenic peptides by professional APCs. This suggests that professional APCs may be the targets of DJ-1 immunomodulatory effects.

DJ-1 modulates Spp1/Cxcl9 dynamics in TAMs to drive immunosuppressive niche formation

To unravel how DJ-1 mediates its immune-regulatory function, we have identified differentially expressed genes (DEGs) between WT and Park7^−/− CD45⁺ cells from the scRNA-seq data. Tumor-infiltrating myeloid cells in Park7^−/− expressed markedly more Cxcl9, H2-Aa, and Ly6a and less Spp1, Ccl2, and Ccl12 (Fig. 5A) relative to their WT counterparts. Notably, Cxcl9 and Spp1 each enriched in different subsets of TAMs (Fig. 5B). On the basis of DEGs, we classified myeloid cells into two monocyte clusters (mono-Ly6c and mono-Cxcl10), two dendritic cell (DC) clusters (plasmacytoid DC and activated DC), and several macrophage clusters (Macro-C1qc, Macro-Cxcl9, Macro-Spp1, and Macro-Mki67) (Fig. 5C and fig. S3A). Park7^−/− mice had more Cxcl9⁺ TAMs and a lower abundance of Spp1⁺ TAMs compared to WT mice (Fig. 5D).

Fig. 5. DJ-1 modulates Spp1/Cxcl9 dynamics in TAMs to drive immunosuppressive niche formation.

(A) Differential gene expression analysis of intratumoral CD45⁺ cells between WT and Park7^−/− mice. Up-regulated genes in Park7^−/− cells are highlighted in orange, while down-regulated genes are in blue. FC, fold change. (B) UMAP plot displaying the expression levels of Cxcl9 and Spp1, with color intensity indicating gene expression values. (C) Identification of 10 distinct clusters from the UMAP projection of myeloid cells. (D) Proportional representation of each cluster in KO versus WT myeloid cells. (E) Developmental trajectories of cells within each cluster inferred using Monocle. (F) Pathway enrichment analysis in Macro_Cxcl9 and Macro_Spp1 clusters. MHC, major histocompatibility complex. (G) Kaplan-Meier survival analysis of patients stratified by CXCL9 and SPP1 expression levels.

Next, we analyzed myeloid cells in pseudo-time to elucidate their developmental stages and immune-regulatory functions. As shown in Fig. 5E and fig. S3B, there is a branching trajectory from early to late stages, where C1qc⁺ and Mki67⁺ TAMs represented a terminally differentiated state, exhibiting mutual exclusivity with Spp1⁺ TAMs. Cxcl9⁺ TAMs were dispersed along the trajectory, positioned closer to the end, and presented a distinct branch in Park7^−/− TILs compared to WT. Gene ontology (GO) enrichment analysis revealed that immune response and antigen processing pathways were enriched in Macro-Cxcl9, aligning with immune-stimulatory macrophage characteristics. In contrast, cell migration and angiogenesis pathways were enriched in Spp1⁺ TAMs, indicating their protumorigenic role (Fig. 5F).

Since Macro-Cxcl9 and Macro-Spp1 may have differences or even opposite effects in antitumor immunity, we next investigated whether the expression of these two genes is related to the clinical efficacy of immunotherapies. As shown in Fig. 5G, patients with higher SPP1 expression had poorer outcomes after anti–PD1 therapy, while those with higher CXCL9 expression fared better. CXCL9, a kind of chemokines, can recruit cytotoxic T lymphocytes (CTLs) via CXCR3, which has been demonstrated to promote antitumor T cell immunity. Similarly, we found that DJ-1 deficiency can induce T cells to secrete more chemokines and cytokines characteristic of CTLs, such as Ccl4, Ccl5, and Ifng (fig. S4A). Moreover, the tumor immune estimation resource (TIMER) database showed that the high expression of CXCL9 was positively correlated with the intratumoral abundance of CD8⁺ T cells (fig. S4B). Overall, these findings suggest that DJ-1 modulates the proportion of Spp1/Cxcl9 in TAMs toward immunosuppressive TME.

DJ-1 deficiency in macrophages induces ROS-associated proinflammatory signals to enhance T cell activation

We next sought to determine whether TAMs are the principal APC regulating T cell function via DJ-1 by performing macrophage transfer. Briefly, bone marrow cells from Park7^−/− or WT C57/BL6 mice were differentiated into bone marrow–derived macrophages (BMDMs) in vitro, mixed with MC38 tumor cells, and subcutaneously inoculated the pool into WT mice (Fig. 6A). Tumors inoculated with Park7^−/− BMDMs grew substantially slower compared to those carrying WT BMDMs when treated with anti–PD-1 antibody (Fig. 6, B and C). At the experimental endpoint, CD8⁺ T cell infiltration was on average 1.7-fold higher in tumors carrying Park7^−/− BMDMs (Fig. 6D). Increased CXCL9 protein expression in these tumors indicated that DJ-1 deficiency up-regulates Macro-Cxcl9 (Fig. 6E).

Fig. 6. DJ-1 deficiency in macrophages induces ROS-associated proinflammatory signals to enhance T cell activation.

(A) Construction of MC38 tumor models with macrophage adoptive transfer. Bone marrow cells from C57BL/6 mice were cultured in Dulbecco’s modified Eagle’s medium (DMEM) with 10% fetal bovine serum (FBS) and macrophage colony-stimulating factor (M-CSF; 50 ng/ml) for 3 days to obtain BMDM0. MC38 tumor cells mixed with WT or Park7^−/− BMDMs were subcutaneously inoculated into WT mice. (B and C) Tumor growth curves and tumor weights were recorded following α-PD-1 antibody treatment. (D) Quantitative analysis of intratumoral CD8⁺ T cells. Tumor tissues were digested into single cells and stained with a CD8⁺ antibody for flow cytometry. (E) Intratumoral CXCL9 concentration was quantified using ELISA. Soluble proteins from tumor homogenates were used to determine CXCL9 levels, normalized to total protein concentration. (F to H) In vitro analysis of BMDM-mediated T cell activation. BMDMs phagocytosed OVA protein in 12-well plates for 6 hours, after which residual OVA was cleared. Splenic lymphocytes from OT-1 mice were cocultured with OVA-treated BMDMs at a 1:1 ratio for 48 hours. Supernatants were quantified and analyzed for IFN-γ and CXCL9 concentrations via ELISA. (I) Phosphorylated STAT3 (Tyr⁷⁰⁵) and NF-κB p65-Ser⁵²⁹ in isolated BMDMs were detected by Western blot. Band intensity was normalized to β-actin after quantitative analysis. (J) Intracellular ROS was detected with the DCFH-DA probe. BMDMs were incubated with DCFH-DA for 20 min, and mean fluorescence intensity (MFI) was measured. Fluorescence microscopy demonstrated intracellular ROS differences. (K) Schematic of the experimental design showing DSF and NAC treatment of BMDMs and subsequent coculture with OT-1 T cells. (L) BMDMs were treated with 0.1 μM DSF with or without 1 mM N-acetylcysteine (NAC) for 24 hours. After drug removal, BMDMs were cocultured with splenic lymphocytes from OT-1 mice at a 1:1 ratio for 24 hours. Flow cytometry was performed to analyze CD3 and CD69 staining, and the percentage of CD3⁺CD69⁺ T cells was quantified. *P < 0.05 and **P < 0.01.

Notably, the inclusion of Park7^−/− bone marrow–derived dendritic cells (BMDCs), another classic APC, in MC38 tumors did not affect antitumor activity of α-PD-1 antibody. This suggests that TAMs, rather than DCs, are the primary APCs influencing T cell regulation via DJ-1 (fig. S5).

We further examined T cell activation by Park7^−/− macrophages. Intratumoral Park7^−/− or WT BMDMs were isolated, stimulated with lipopolysaccharide, and used to present OVA protein–derived peptides to T cells from OT-1 mice (Fig. 6F and fig. S6, A and B). As shown in Fig. 6 (G and H) and fig. S6C, activated T cells (CD25⁺) were up-regulated, and more IFN-γ and CXCL9 were expressed in the presence of Park7^−/− BMDMs, indicating proinflammatory characteristics and robust T cell activation. RNA-seq data revealed enriched pathways related to oxygen response and organic/carboxylic acid biosynthesis, suggesting a DJ-1 role in macrophage respiration and catabolic processes (fig. S7). Proinflammatory macrophages typically shift from oxidative phosphorylation to aerobic glycolysis, increasing ROS (35). As expected, ROS sensor 2′,7′-Dichlorodihydrofluorescein diacetate (DCFH-DA) showed enhanced ROS in Park7^−/− macrophages. Notably, phosphorylation levels of STAT3 and NF-κB were significantly elevated in Park7^−/− macrophages (Fig. 6, I and J). To further investigate the direct role of ROS in regulating macrophage-mediated T cell activation, the antioxidant N-acetylcysteine (NAC) was added before OVA antigen presentation. NAC treatment reduced the proportion of CD69⁺ T cells, effectively reversing the stimulatory effect observed with DJ-1 inhibitor disulfiram (DSF)–treated macrophages (Fig. 6, K and L). Given that Cys¹⁰⁶ is critical for the antioxidant activity of DJ-1, we overexpressed WT DJ-1 and the functional mutants C106A and C106S in DJ-1 KO BMDMs. We found that reintroduction of DJ-1 suppressed T cell activation, whereas the C106A and C106S mutants failed to do so, indicating that the redox-active Cys¹⁰⁶ site is essential for DJ-1–mediated regulation of macrophage function (fig. S8). Together, DJ-1 deficiency in macrophages induces ROS and promotes proinflammatory gene expression via transcription factors and enhancing T cell activation capacity.

Pharmacological inhibition of DJ-1 potentiates the immune checkpoint blockade

On the basis of these results, DJ-1 emerges as a potential drug target for enhancing the efficacy of ICIs. We have previously identified DSF as an effective small-molecule DJ-1 inhibitor from marketed drug library (36). Next, we asked whether DSF could phenocopy DJ-1 KO to potentiate ICIs and stimulate antitumor immunity in MC38 tumor–bearing C57BL/6 mice. As hypothesized, DSF significantly enhanced both the antitumor activity and intratumoral CD8⁺ T cell infiltration of PD-1 blockade, although the antitumor effect of DSF alone was not statistically significant (Fig. 7, A to C). Similarly, in the immunologically cold 4T1 tumor model, combined treatment with DSF and α-PD-1 resulted in pronounced tumor growth inhibition and markedly increased intratumoral CD8⁺ T cell infiltration (fig. S9).

Fig. 7. Pharmacological inhibition of DJ-1 potentiates immune checkpoint blockade.

(A) MC38 tumor growth curves following treatment with α-PD-1 antibodies and DSF. (B) Representative images and masses of MC38 tumors. (C) Quantitative analysis of intratumoral CD8⁺ T cells. Tumors were digested into single cells and stained with CD45 and CD8 antibodies for flow cytometry analysis. *P < 0.05 and ***P < 0.001.

Overall, both the pharmacological inhibition and KO of DJ-1 stimulate antitumor immunity and impede tumor growth. This further strengthens DJ-1 involvement in antitumor immune regulation and its potential as a TME-modifying target.

DISCUSSION

Here, we have found that DJ-1 reprograms TAMs to inhibit T cell immunity through proinflammatory ROS signaling pathways and validated DJ-1 as a drug target for potentiating cancer immunotherapy (Fig. 8). This approach contributes to overturning the immunosuppressive milieu and restoring T cell activation through modulation of multifaceted TAMs that can both support and restrain immunity depending on TME signals they receive. Tumors commonly reeducate TAMs into anti-inflammatory phenotype to reduce immune surveillance and facilitate tumor immune escape (37). Therefore, repolarizing macrophages toward an immune activation phenotype represents a prominent strategy for unleashing antitumor immunity and full potential of ICIs.

Fig. 8. Graphic model.

DJ-1 inhibition converts immunosuppressive tumors into immunostimulatory phenotypes and potentiates immunotherapy efficacy.

We have identified six subtypes of TAMs in the DJ-1 KO mice based on their gene expression profiles. Compared with WT mice, Cxcl9⁺ TAMs were found to be significantly enriched, while Spp1⁺ TAMs were notably down-regulated. Studies have shown that Spp1⁺ TAMs exhibit higher hypoxia scores to promote epithelial-to-mesenchymal transition and glycolysis in hypoxic TME (38), thereby facilitating tumor growth. Further, Spp1⁺ TAMs can inhibit the activity of CD8⁺ T cells through the adenosine signaling pathway (39). In contrast, CXCL9 is a classical chemokine involved in the recruitment of T cells. Thus, Cxcl9⁺ TAMs are likely to act as gatekeepers inviting cytotoxic tumor-specific T cells into tumors. While T cell trafficking into tumors is essential, recent studies have highlighted that the physical proximity of T cells and APCs is paramount for effective T cell activation (40). Consistent with these reports, our cell interaction analyses based on scRNA-seq data suggest that DJ-1 loss promotes interactions between T cells and TAMs in mice. This is further reinforced by up-regulation of pathways related to antigen presentation, cytokine-mediated T cell activation, and cell migration signaling in intratumoral CD45⁺ cells.

Repolarization of TAMs into proinflammatory Cxcl9⁺ phenotype by DJ-1 inactivation provides a potential avenue for therapeutic intervention. That is clearly demonstrated by our finding that pharmacological inhibition of DJ-1 potentiates PD-1 blockade. APCs such as TAMs are key protagonists in regulation and activation of T cell function. Therefore, modulating APC activity provides a promising strategy for indirectly activating T cells. Precise and durable polarization of highly adaptable cells such as TAMs by drugs is challenging as TME signals can override an originally induced polarization state (41). But notably, DJ-1–deficient macrophages generated ex vivo–retained T cell stimulating function even after transplantation into WT tumors. TAM polarization by DJ-1 deletion seems enormously stable, thus offering a superior therapeutic benefit. DJ-1 KO also provides an effective option for enhancing the functionality of chimeric antigen receptor macrophages (CAR-M) (42).

In our previous studies, we focused on exploring the role of DJ-1 (PARK7) as a redox sensor and therapeutic target in cancer. In addition, we have demonstrated that DJ-1 governs tumors fate decisions (apoptosis/ferroptosis) by regulating ASK1 activity upon specific oxidative conditions (43) and modulating the trans-sulfuration pathway (44). While ROS plays a critical role in the immune-activated phenotype of TAMs like classical M1 proinflammatory phenotype (TAM1). Here, we have found that DJ-1 KO in macrophages accumulates ROS and activates proinflammatory signaling pathways. Although ROS also plays a proinflammatory role in DCs, analysis of intratumoral CD45⁺ cells in systemic DJ-1 KO mice suggested that DJ-1 is not a key regulator for the ROS production of DCs, which was further verified in the adoptive experiment of BMDCs. Overall, DJ-1 deletion can promote ROS accumulation, thereby achieving a dual antitumor effect by suppressing tumor growth and promoting T cell activation by TAMs. Our earlier findings showed that DJ-1 KO tumors were more susceptible to ferroptosis (44), a form of immunogenic cell death (ICD) (45, 46). ICD releases a broad range of tumor antigens, nucleic acids, and other danger-associated molecular patterns that activate innate immune cells including macrophages (47). Thus, we propose that DJ-1 inhibition boosts macrophage polarization and ICD, hence producing a highly effective synergy system stimulating antitumor immunity.

Although we did not observe significant side effects in systemic DJ-1 KO mice including the life span, DJ-1 deficiency is known to associate with autosomal recessive familial Parkinson’s disease (22). Given that DJ-1 KO mice are housed in pathogen-free environments with limited exposure to pathogens, we were unable to determine whether DJ-1–induced macrophage polarization might affect susceptibility to some infectious diseases. Similarly, an effect of macrophage polarization induced by DJ-1 deletion on autoimmune diseases remains unclear. However, it is likely that these diseases require long-term DJ-1 deficiency or occur under specific conditions. Therefore, we believe that temporary pharmacological inhibition of DJ-1 function is likely to be safe and without any serious adverse events in mice and potentially also in humans.

In conclusion, we found that DJ-1 deficiency induces ROS accumulation to trigger TAM polarization toward an immuno-stimulatory phenotype, thereby enhancing T cell activity and ultimately promoting the antitumor efficacy of ICIs. Pharmacological inhibition of DJ-1 by DSF potentiates ICIs and establishes DJ-1 as a prospective drug target for sensitizing patients with cancer to immunotherapies. DJ-1 blockade or genetic deletion represents a unique approach for TAM polarization and CAR-M enhancement that introduces a previously unknown paradigm into the immune oncology arsenal.

MATERIALS AND METHODS

Cell culture

The MC38 colon adenocarcinoma cells (serial: SCSP-5431), LLC Lewis lung carcinoma cells (serial: TCM 7), and 4T1 breast cancer cells (serial: TCM32) were purchased from the Cell Bank of the China Science Academy (Shanghai, China), which were cultured in RPMI 1640 supplemented with 10% fetal bovine serum (FBS). Cells were grown in a 5% CO₂ humidified incubator at 37°C.

Plasmids and cell transfection

The lentiviral vector plasmid pCDH-EF1-Puro was purchased from System Biosciences. The lentiviral vector plasmid pMND-Puro was generated in our laboratory. pLV2-CMV-OVAL-EGFP-Fluc-Puro was purchased from the Miao Ling Biological Platform (P28082) and subsequently subcloned into pCDH plasmid with the following primers: forward_TGAGACTTCGTGGTGGTAGTATAATCAACT-TTGAAAAACTGGAATTCGAATTTAAATCGGATC and reverse_GATCCGATTTAAATTCGAATTCCAGTTTTTCAAAGTGATTATACTACCACCACGAAGTCTCA.

The Park7-WT, Park7-C106A, and Park7-C106S genes were synthesized by GENEWIZ (Suzhou, China) and subsequently cloned using the following primers: forward_ACAAAAAAGCAGGCTTCACCGCTAGCATGGCTTCCAAAAGAGC and reverse_AGATCTCGAGCTCAAGCTTCGAATTCCTAGTCTTTGAGAACAAGCGGTG.

The lentiviral vectors were transfected into 293FT cells for packaging. The virus particles were harvested about 36 hours after transfection. For stable infection, MC38 cells were grown in six-well plates at 30% confluency, and 1 ml of viral supernatant was added with polybrene at the final concentration of 6 μg/ml for 12 hours. The infected cells were cultured in fresh media for 72 hours and then sorted by flow cytometry using the fluorescein isothiocyanate channel to obtain a stable MC38-OVA cell line. For BMDMs, cells were seeded in 12-well plates at 40% confluency, and 1 ml of viral supernatant was added together with 1× lentiviral transduction enhancer for 24 hours. After 72 hours in fresh medium, the expression of DJ-1 (WT, C106A, and C106S) was verified by Western blot.

Mice

Park7^−/− mice on a C57/BL6 background were obtained from the Jackson Laboratory, and female OT-1 BALB/c mice were purchased from the National Rodent Laboratory Animal Resource Center (Shanghai, China). Offspring from Park7^+/− × Park7^+/− breeding included littermates that were either Park7^+/+ (WT controls) or Park7^−/− (DJ-1 KO). All experiments used age-matched (6 to 12 weeks old) and gender-matched (male and female) mice and were approved by the Institutional Animal Care and Use Committee (number: DW202510211520) of the Innovation Institute for Artificial Intelligence in Medicine, Zhejiang University.

Subcutaneous tumor model

Tumor cells (5 × 10⁵ MC38 cells or 1 × 10⁶ LLC cells) were subcutaneously injected into Park7^+/+ or Park7^−/− mice, respectively. Tumor volume was calculated using the formula (L × W × W)/2, where L is the tumor length and W is the tumor width. According to humane requirements, mice are euthanized when the tumor volume reaches 2000 mm³, the tumor exceeds 20 mm in one direction, or the tumor forms an ulcer.

ICI therapy

When the tumor grows to ~100 mm³, anti–PD-1 (10 mg/kg; clone RMP1-14, Bio X Cell) or anti-TIM-3 (clone RMT3-23, Bio X Cell) antibodies were intraperitoneally injected into mice every 3 days, with the control group receiving the corresponding isotype IgG. The sizes of tumor were monitored every 2 to 4 days.

CD8⁺ T cell depletion

To study the contribution of CD8⁺ T cells to the antitumor effect, mice in each group were intraperitoneally injected with 200 μg of anti-CD8a (clone 2.43, Bio X Cell) or rat IgG2a isotype control (clone 2A3, Bio X Cell) the first dose 7 days before tumor implantation and the second dose 1 day before tumor implantation. Before tumor implantation, venous blood was collected from each mouse, and flow cytometry was performed with CD8-specific antibodies to confirm successful depletion of CD8⁺ T cells after red blood cell lysis. The neutralizing anti-CD8a antibody was administered continuously every 6 days to maintain effective depletion.

OVA antigen loaded

The OVA protein was emulsified with incomplete Freund’s adjuvant at a 1:1 volume ratio to form a “water-in-oil” emulsion. A total of 10 μg/20 μl of this emulsion was subcutaneously administered to mice three times over a period of 10 days to induce in vivo antigen presentation by immune cells. Subsequently, 1 × 10⁶ MC38-OVA cells were inoculated into the axilla of the mice. Tumor growth was monitored and measured regularly.

BMDM/BMDC isolation and inoculation

BMDMs or BMDCs were generated from bone marrow cells collected from Park7^+/+ and Park7^−/− mice. The bone marrow cells were then seeded in Dulbecco’s modified Eagle’s medium (DMEM) supplemented with 1% antibiotics/antimycotics and 10% fetal calf serum containing macrophage colony-stimulating factor (50 ng/ml) to obtain BMDMs and granulocyte-macrophage colony-stimulating factor (20 ng/ml) to obtain BMDCs, respectively. On days 3 to 4, MC38 cells (5 × 10⁵) mixed with BMDMs (5 × 10⁵) or MC38 cells (4.5 × 10⁵) mixed with BMDCs (3 × 10⁵) were implanted subcutaneously into armpits and treated with monoclonal antibodies as same as above.

Combination therapy

To test the therapeutic effect of combinational blockade of DJ-1 and PD-1, control antibody or α-PD-1 (clone RMP1-14, Bio X Cell) was administered intraperitoneally at 10 mg/kg, and DSF was intravenously injected at a dosage of 20 mg/kg for either single or combination therapy.

Tumor tissue dissociation

The tumor samples were dissected into small pieces using forceps and scissors after removing adipose tissues. These pieces were then incubated at 37°C for 30 min with regular shaking in DMEM supplemented with type II collagenase (1 mg/ml), deoxyribonuclease I (0.1 mg/ml), and 0.25% trypsin-EDTA. Following incubation, the cell suspension was filtered through a 70-μm cell strainer, washed with phosphate-buffered saline (PBS), and resuspended in preparation for subsequent analyses.

Flow cytometry (fluorescence-activated cell sorting)

Single-cell suspensions were prepared as described above and stained with antibodies. Samples were first incubated with Fc-blocking solution (1:100 dilution of antibody) in 1× PBS containing 2% bovine serum albumin. Samples were incubated with the primary antibodies (CD45 clone: 30-F11, CD3 clone: 17A2, CD8 clone: 53-6.7, CD69 clone: H1.2F3, CD25 clone: 3C7, Cd11c clone: N418, F4/80 clone: BM8, I-A^b clone: AF6-120.1, PD-1 clone: RMP1-30) on 4°C for 45 min in the dark. Cells were then washed and analyzed using LSR II (BD Biosciences) flow cytometers. Data were analyzed using FlowJo software. Specific cells were sorted into designated populations using a MoFlo MLS high-speed cell sorter (Beckman Coulter).

In vitro T cell stimulation

The mouse spleens were mechanically dissociated and subsequently filtered through a 70-μm cell strainer. Lymphocytes were further isolated using lymphocyte separation medium (Dakewe). T cells were resuspended in RPMI 1640 medium supplemented with 10% FBS and 50 μM β-mercaptoethanol and added to 24-well plates. Plates were precoated with soluble anti-CD3 antibody (5 μg/ml) for 24 hours, then a total of 1 × 10⁶ cells per well were cultured in the presence of soluble anti-CD28 antibody (3 μg/ml) and recombinant interleukin-2 (10 ng/ml), and washed twice with PBS. For rapid stimulation, cells were incubated with a cell stimulation cocktail (Thermo Fisher Scientific) containing phorbol 12-myristate 13-acetate, ionomycin, and brefeldin A. Cells were harvested and analyzed by flow cytometry.

Enzyme-linked immunosorbent assay

Tissues were rinsed with precooled PBS to remove residual blood or impurities on the surface and homogenized thoroughly. After quantifying the protein by bicinchoninic acid (BCA) kit for normalization, the mouse IFN-γ (R&D Systems) and Cxcl9 (HuaAn Biotechnology) enzyme-linked immunosorbent assay kits were used to detect the protein levels of mouse IFN-γ and Cxcl9 in these supernatants according to the manufacturer’s protocol.

Immunoblotting

Immunoblotting was performed as previously described (44). The antibodies to p-STAT3-Tyr⁷⁰⁵ (ET1603-40) and p-NF-κB p65-Ser⁵²⁹ (ET1604-27) were obtained from HuaAn Biotechnology. The antibody to β-actin (db7283) was obtained from Diagnostic Biosystems.

scRNA-seq preprocessing

CD45-positive cells were sorted by flow cytometry, achieving a viability rate of more than 90%. Following the protocol of the GEXSCOPE Single Cell Sequencing Kit, the cell concentration was adjusted to approximately 100 cells/μl. The single-cell suspension was then mixed with nuclease-free water and 5′ single-cell RNA premix before being loaded onto the chip containing barcode gel beads and dispensing oil. cDNA was synthesized via reverse transcription and maintained at 4°C. Subsequently, the cDNA was amplified through polymerase chain reaction and stored at −80°C. The construction of the second-generation sequencing library for gene expression analysis was performed by Singleron Biotech. The raw barcode, feature, and matrix files were loaded into R (v4.2.1) and processed with Seurat (v4.3.0) for initial quality control. Cells with fewer than 500 or more than 6000 detected genes (fewer than 1000 or more than 5000 detected genes in T cell groups), greater than 5% mitochondrial genes, or greater than 3% erythrocyte genes were removed. Genes detected in fewer than three cells were also excluded. The data were normalized and log-transformed to correct for sequencing depth variations. After quality control, the processed data are ready for further analysis.

scRNA-seq data analysis

Cell annotation

Correlation analysis was performed using the top 2000 highly variable genes to identify anchor cells for unsupervised data integration. For cell clustering, the top 30 principal components (PCs) were used to construct a spiking neural network (SNN) and for Uniform Manifold Approximation and Projection (UMAP) visualization. Cell type–specific markers were used to define immune cell subsets. To achieve higher-resolution analysis of T cells and myeloid cells, these cell types were separately extracted from the total CD45-positive cells for further analysis. The top 10 significant PCs were used for both SNN map construction and UMAP clustering. Heatmaps, violin plots, and t-distributed stochastic neighbor embedding (t-SNE) plots visualized transcript expression in selected cell subsets, generated using built-in functions and the ggplot2 package.

Pseudotime analysis

An R object was constructed using the newCellDataSet function from Monocle2 (48), and the estimateDispersions function was used to estimate the empirical dispersion of each gene within a negative binomial model. Genes with an average log expression level less than 0.5 across samples were filtered out. The remaining genes were ranked on the basis of their overdispersion scores. Subsequently, the top-ranked genes were used as input for the Monocle2 algorithm to derive the pseudotime trajectory in a two-dimensional projection.

scTCR-seq analysis

TCR information was extracted from scRNA-seq data, and alignment along with de novo assembly was used to validate TCR-derived reads and reconstruct TCR chains. The reconstruction of the TCR repertoire was performed by Singleron. Subsequently, we integrated the scTCR data with the scRNA-seq data of T cell clusters using cell barcodes for further analysis. TCR clones associated with cell barcodes not present in the T cell clusters were excluded. In addition, we quantified the overlap of CDR3 amino acid sequences between the WT and KO groups.

GO enrichment analysis

For the DEGs, GO enrichment analysis was conducted using ClusterProfiler (49). Specifically, 100 DEGs with a P value cutoff of less than 0.05 were selected for statistical analysis and visualization to elucidate the functional profiles of genes and gene clusters.

Cell-cell interaction analysis

To investigate cell-cell communication, the CellChat package was used. The “secreted signaling” subset was selected as the communication database using the command subsetDB. Seurat objects from samples were used as inputs to generate CellChat objects via the createCellChat function. Communication analysis was performed according to the standard CellChat workflow.

Statistical analysis

Statistical analyses were performed using GraphPad Prism software. Data are expressed as means ± SD. For pairwise comparisons between two groups, two-tailed Student’s t tests were used to determine P values. For multiple group comparisons, one-way analysis of variance (ANOVA) was used, with significance levels defined as *P < 0.05, **P < 0.01, and ***P < 0.001.

Supplementary Materials

This PDF file includes:

Figs. S1 to S9