Targeting macrophage-derived SPP1 enhances CD8 T cell infiltration via ROS-DNA fragment/cGAS-STING/STAT1-mediated CXCL9/10 in tumor microenvironment.

Background
Cancer ranks as the second leading cause of global mortality and disease burden.1 While significant advances have been achieved in novel therapeutic strategies such as immune checkpoint inhibitors (ICIs), some patients remain unresponsive to immunotherapy due to the distinct immune microenvironment within tumors. Based on the infiltration of immune effector cells, particularly cytotoxic CD8 T cells in the tumor microenvironment (TME), tumors can be categorized into different immunophenotypes, notably “hot” and “cold” tumors. Hot tumors exhibit robust immune activity and demonstrate favorable responses to ICIs, whereas cold tumors, characterized by the absence of CD8 T cells, maintain therapeutic resistance.2 Consequently, strategies to convert immunologically “cold” tumors into “hot” tumors have emerged as a focal point in contemporary cancer immunotherapy research.
Tumor-associated macrophages (TAMs) play a pivotal role in shaping the immunosuppressive microenvironment characteristic of “cold tumors”.3 These cells are frequently abundant in solid malignancies, often comprising up to 50% of the total tumor cell population. They are tumorigenic and contribute to therapeutic resistance. Recent advances in single-cell RNA (scRNA) sequencing have enabled detailed characterization of TAM heterogeneity, leading to the identification of distinct phenotypic subsets, including SPP1 TAMs.4 Mounting evidence demonstrates that SPP1 TAMs represent a predominant subset within the tumor myeloid compartment across multiple cancer types, including pancreatic, head and neck, and colorectal cancers (CRC).5 6 These macrophages actively facilitate tumor progression and play a crucial role in tumorigenesis.379 A defining feature of SPP1 TAMs is their high expression of osteopontin (secreted phosphoprotein 1, SPP1), a multifunctional glycoprotein with diverse roles in cancer biology. Beyond its diagnostic utility, SPP1 serves as a key regulator of immunosuppressive macrophage polarization10 and functions as a central signaling molecule mediating crosstalk between immune cells and the TME.11 Recent investigations have further elucidated the mechanistic contributions of SPP1 macrophages to tumor progression such as impairment of CD8 T cell function,12 induction of CD8 T cell exhaustion,13 promotion of macrophage polarization,1416 enhancement of angiogenic processes,17 facilitation of desmoplastic reaction, and tumor immune barrier formation.18 19 Critically, elevated SPP1 expression in TAMs has been identified as a major determinant of clinical outcomes in cancer patients, underscoring its potential as both a prognostic indicator and therapeutic target.
Pan-cancer analyses have recently revealed a significant negative correlation between SPP1 expression levels and CD8 T cell infiltration.20 Notably, the CD8 T cell/SPP1 ratio demonstrates superior prognostic value over either marker alone, showing significant associations with patient survival across 14 cancer types.20 Supporting evidence indicates that SPP1 deficiency markedly enhances CD8 T cell activity within TMEs.21 Clinical observations further demonstrate that tumors with high proportions of SPP1TAMs exhibit reduced T cell infiltration and poorer clinical outcomes.7 Importantly, SPP1 TAMs have been implicated in CD8 T cell exclusion and resistance to ICI therapy.18 Despite these clinical associations, the mechanistic relationship between SPP1 TAMs and CD8 T cell infiltration remains poorly understood. In particular, the intracellular signaling pathways and molecular mediators through which macrophage-derived SPP1 regulates CD8 T cell trafficking require elucidation. In this study, we identify a novel mechanism whereby TAM-derived SPP1 impairs CD8 T cell infiltration by suppressing the production of chemokines CXCL9 and CXCL10 (CXCL9/10). Our findings demonstrate that SPP1 in TAMs disrupts the reactive oxygen species (ROS)-DNA fragment/cGAS-STING/STAT1 signaling axis, leading to downregulation of CXCL9/10 chemokines. This pathway ultimately results in impaired CD8 T cell recruitment, accelerated tumor progression, and diminished responsiveness to ICI-based immunotherapies.

Methods
Reagents and oligoes used in this study are listed in online supplemental table S1.

Animals and animal procedures
Construction and identification of Spp1fl/fl-Lyz2-Cre (Spp1cKO) mice. Spp1fl/fl and Lyz2-Cre mice were prepared by Gempharmatech (Nanjing, China). Spp1fl/fl mice were hybridized with Lyz2-Cre mice to generate myeloid-specific Spp1 knockout (Spp1fl/fl-Lyz2-Cre, Spp1cKO) mice. Spp1fl/fl mice were used as controls. There was no difference in the phenotypes of Spp1cKO and control Spp1fl/fl mice.
Establishment of a subcutaneous tumor model. MC38, Lewis lung carcinoma (LLC), and B16-OVA cell lines (obtained from ATCC) were cultured in RPMI 1640 medium supplemented with 10% fetal bovine serum (FBS) and 1% penicillin/streptomycin. Cells in logarithmic growth phase were harvested, washed with PBS, and resuspended at a concentration of 1×10⁷ cells/mL. 5–6-week-old Spp1fl/fl and Spp1cKO mice received subcutaneous injections of 100 µL cell suspension (containing 1×10⁶ tumor cells) in the right thigh region. Tumor growth was monitored by measuring tumor volume every 3 days postinjection.
For the mouse subcutaneous co-inoculation tumor model, bone marrow-derived macrophages (BMDMs) were generated by differentiating bone marrow cells isolated from Spp1fl/fl and Spp1cKO mice. The cells were cultured for 5 days in the presence of macrophage colony-stimulating factor (20 ng/mL) to induce macrophage differentiation. Subsequently, the BMDMs were co-inoculated subcutaneously with MC38 tumor cells into wild-type (WT) C57BL/6 mice at a ratio of 1:4 (macrophages to tumor cells).
For the CD8 T cell depletion model, mice were administered intraperitoneal injections of either anti-CD8 antibody (200 µg) or IgG isotype control (200 µg) 1 day prior to tumor inoculation, followed by twice-weekly maintenance doses throughout the experiment.
For anti-PD-1 immunotherapy experiments, Spp1fl/fl and Spp1cKO mice were randomly divided into two groups on day 4 post-tumor inoculation. Subsequently, the mice received intraperitoneal injections of either anti-PD-1 antibody (200 µg) or an IgG isotype control at the indicated time points. 
For in vivo inhibition and neutralization, mice received intraperitoneal administration of CXCR3 inhibitor AMG487 (5 mg/kg), SPP1 inhibitor (5 mg/kg), or BPS, and SPP1 (15 mg/kg), CXCL9/10 (5mg/kg) neutralizing antibodies or IgG vehicle control following tumor inoculation. The treatment was administered at the indicated time until the predetermined experimental endpoints.
For the macrophage adoptive transfer model, Spp1+ or Spp1– macrophages were induced with TCM (tumor conditioned medium) from WT or cKO mice. 1×106 macrophages in 100 µL were administered intravenously to indicated mice once a week following MC38 tumor inoculation.
Tumor volume was monitored every 3 days postinoculation using the formula: V=(long diameter×short diameter²)×0.5. The experimental endpoint was defined as tumor volume exceeding 2000 mm³, at which point tumor wet weight was recorded. All experiments were performed using age-matched and gender-matched mice maintained under specific pathogen-free conditions at Nankai University.

CRC tissue samples
Human CRC tissue specimens and associated clinical data were obtained from Tianjin Union Medical Center. All patients in our cohort were pathologically confirmed cases.

T cell migration assay
To evaluate the migratory capacity of CD8 T cells, we performed a transwell migration assay using lymphocytes and macrophage-conditioned medium. Briefly, splenic CD8 T cells were isolated by flow cytometry following staining with an anti-CD8α antibody. The sorted cells were then stimulated with CD3/CD28 antibodies for 3 days in the presence of recombinant murine IL-2 (20 ng/mL). BMDMs from Spp1cKO and control Spp1fl/fl mice were treated with the indicated siRNAs for 48 hours. In the transwell assay, 5 µm pore size inserts (24-well format, Labselect #14331) were used. The lower chambers were filled with 600 µL of migration medium containing either conditioned or control medium, while 3×105 splenic CD8 T cells in 100 µL of medium were seeded in the upper chamber. After 3-hour of incubation at 37°C, migrated CD8 T cells were collected from the lower chambers, stained with anti-CD8α antibody, and quantified by flow cytometry. To ensure accurate normalization, counting beads of known concentration were added to each sample prior to flow cytometric analysis. Migration ratios were calculated relative to the number of input CD8 T cells. For CXCR3 inhibiting experiments, CD8 T cells were pretreated with AMG487 (10 µM) at 37°C for 24 hours before being loaded into the transwell system.
For CXCL9 or CXCL10 neutralization, neutralizing antibodies against CXCL9 and CXCL10 (each at 1 µg/mL) were supplemented. For the co-culture assay between macrophages and T cells, SPP1 protein (100 ng/mL) and anti-CD44 antibodies (10 µg/mL) were supplemented.

Analyses of public scRNA-seq data
Processed scRNA sequencing (scRNA-seq) data and annotated cell populations from CRC tumor tissues were acquired from the scCRLM web portal (http://www.cancerdiversity.asia/scCRLM).22 From this dataset, we specifically extracted myeloid cells (including monocytes and macrophages) and CD8 T cells for downstream analysis. The data were processed using the Seurat R package (V.3.1.4) with the following analytical pipeline: First, we performed dimension reduction and unsupervised clustering to identify distinct cell populations. To ensure data quality, samples containing fewer than 10 myeloid cells were excluded from subsequent analyses. Cell population markers were identified using Seurat’s FindAllMarkers function with the following parameters: min.pct=0.2, while maintaining all other default settings. To evaluate the correlations among different cellular subsets, we employed the Pearson correlation coefficient. Specifically, in the correlation analysis of macrophages, the data underwent normalization and the expression matrix was filtered to remove zero values, followed by the computation of Pearson correlation coefficients for pairs of genes.

ScRNA-seq of immune cells in tumor
scRNA-seq of immune cells from MC38 tumors was performed as previously described.23 24 Briefly, tumors from Spp1fl/fl and Spp1cKO mice were dissected and enzymatically digested using a solution containing 0.2 mg/mL collagenase IV and 10 mg/mL deoxyribonuclease I. The resulting cell suspension was filtered through a 70 µm cell strainer and resuspended in Hanks’ balanced salt solution. Cells were then stained with a CD45 antibody for 30 min, followed by addition of 7-AAD immediately prior to fluorescence-activated cell sorting (FACS). Viable CD45+7-AAD− single cells were isolated for library preparation. Single-cell libraries were constructed using the 10×Genomics platform according to the manufacturer’s protocol, with a target capture of 8000 cells per library. Sequencing was performed on an Illumina HiSeq X Ten platform.
Data processing and analysis. Raw sequencing data were processed using Cell Ranger (V.3.0.2) to generate the gene expression matrix. Quality control filtering excluded cells with: (1) fewer than 500 or more than 7000 detected genes or (2) mitochondrial gene content exceeding 20%. After filtering, we retained 8059 cells from Spp1fl/fl mice and 12 121 cells from Spp1cKO mice. Downstream analysis was performed using the Seurat R package for additional quality control, data normalization, cell clustering, differential gene expression analysis, and marker gene identification. Cell types were annotated using SingleR based on cluster identities. Differentially expressed genes (DEGs) were identified using Seurat’s FindMarkers function, with significance defined as adjusted p<0.25. For cell–cell interaction analysis, we employed CellPhoneDB to identify significant ligand-receptor pairs between monocyte-macrophages and CD8 T cells, considering only interactions where both ligand and receptor were expressed in ≥10% of cells. Gene set enrichment analysis (GSEA) was performed with 1000 permutations to identify significantly enriched gene sets between clusters.

RNA-seq
BMDMs were isolated from Spp1fl/fl and Spp1cKO mice. Total RNA was extracted using Trizol reagent and subsequently submitted to Novo Zhiyuan for sequencing analysis. Libraries were prepared and quality-checked before being subjected to high-throughput sequencing on the Illumina HiSeq 2000 platform. Transcript abundance was quantified using RSEM software, which generated both raw read counts and FPKM values for each transcript. Differential gene expression analysis was performed using the DESeq2 package in R, with significance thresholds set at fold change >1.5 and an adjusted p<0.05. To elucidate the biological functions of the DEGs, we conducted Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analyses. Functional terms with a false discovery rate ≤0.05 were considered statistically significant and were further analyzed.

Detection of cytosolic DNA fragments
For the detection of cytosolic DNA fragments, cells were plated on poly-D-lysine-coated glass-bottom dishes and fixed with 4% paraformaldehyde for 20 min at room temperature. Following fixation, cells were permeabilized with 0.1% Triton X-100 in PBS for 30 min and subsequently blocked with 4% sheep serum in PBS for 1 hour at room temperature. Primary antibody incubation was performed overnight at 4°C using antibodies against double-stranded DNA (dsDNA) and HSP60 (as a mitochondrial marker). After washing, cells were incubated with appropriate fluorescently labeled secondary antibodies for 1 hour at room temperature. Nuclei were counterstained with DAPI. Imaging was conducted using super-resolution microscopy to ensure high-resolution detection of subcellular localization.

Cell immunofluorescence staining
Cells were cultured on 35 mm confocal dishes and fixed with 4% paraformaldehyde for 20 min at room temperature. Following fixation, cells were permeabilized with 0.1% Triton X-100 in PBS for 20 min and subsequently blocked with 5% sheep serum in PBS for 1 hour at room temperature. Primary antibody incubation was performed overnight at 4°C, followed by incubation with appropriate secondary antibodies for 1 hour at room temperature. Cell nuclei were counterstained with DAPI. Fluorescence images were acquired using a confocal microscope.

MitoSOX detection
BMDMs were seeded in six-well plates. After adherence, the cells were incubated with 1 mL of MitoSOX Red staining working solution at 37°C for 30 min in a humidified cell culture incubator (5% CO₂). Following incubation, the cells were washed twice with PBS to remove excess dye. Mitochondrial superoxide levels were then assessed by flow cytometry.

Immunohistochemistry analysis
Tumor tissue samples from CRC patients were processed for histochemical and immunohistochemical analysis as previously described.25 Briefly, paraffin-embedded sections were deparaffinized, rehydrated, and washed with PBS. Antigen retrieval was performed by boiling the sections in retrieval buffer for 15 min, followed by cooling to room temperature. Non-specific binding was blocked with 5% sheep serum, and sections were then incubated with primary antibodies (CD68, SPP1, or CD8) at 4°C overnight. Subsequently, the sections were treated with horseradish peroxidase-conjugated secondary antibodies, developed with DAB substrate, and counterstained with hematoxylin. Finally, the stained sections were dehydrated, cleared, and imaged under a microscope. For quantitative analysis, CD8-positive cells were counted in three randomly selected 20×fields per slide, with an average of three slides evaluated per sample.

Immunofluorescence analysis
Tumor tissue immunostaining was performed as previously described25 using multiplex fluorescent immunohistochemistry kits according to the manufacturer’s protocol. Briefly, tissue sections on slides were deparaffinized in xylene and rehydrated through a graded ethanol series. Following antigen retrieval and blocking, slides were incubated with primary antibody overnight at 4°C, then with appropriate secondary antibody. Tyramide signal amplification with single fluorescent dye was performed, followed by multiple rounds of repeated antigen retrieval and blocking for subsequent staining cycles. Finally, nuclei were counterstained with DAPI solution (37°C, 10 min, dark). For dual-labeling experiments, slides were incubated with primary antibodies from different species, followed by corresponding Alexa Fluor-conjugated secondary antibodies (1 hour, room temperature). Nuclei were visualized with DAPI staining. All fluorescent images were acquired using a fluorescence microscope.

Cell isolation and flow cytometry
The previous protocol was used in cell isolation and flow cytometry.25 Following euthanasia via carbon dioxide asphyxiation, tumors were excised and immediately placed on PBS-soaked laboratory tissue paper (calcium-free and magnesium-free PBS, room temperature). The tumors were then minced into small fragments and transferred to a 50 mL centrifuge tube containing 10 mL of digestion buffer (DMEM supplemented with 5% FBS, 1 mg/mL collagenase IV, hyaluronidase, and DNase I). After 30 min of enzymatic digestion, the reaction was quenched by adding 30 mL of ice-cold DMEM. The cell suspension was centrifuged at 400×g for 10 min at 4°C, resuspended in 3 mL PBS, and layered onto Ficoll in a 15 mL Falcon tube. Gradient separation was achieved by centrifugation at 1800 rpm for 20 min at room temperature. Lymphocytes were collected from the Ficoll interphase, washed, and resuspended in culture medium for subsequent staining. For extracellular staining, cells were washed with staining buffer (PBS containing 2% FBS, 1 mM EDTA, and 0.09% NaN₃) and incubated with fluorescently labeled antibodies against CD45, CD8, CD44, CD62L, CD4, CD11b, F4/80, NK1.1, Gr1, PD-1 (CD279), and Foxp3 (FITC, PE, APC, or PerCP/Cy5.5 conjugates). Viability was assessed by 7-AAD exclusion.
Intracellular staining. For cytokine detection, cells were stimulated with PMA for 4 hours, fixed with Cytofix/Cytoperm, and permeabilized using Perm/Wash buffer (BD Biosciences). Intracellular staining was performed with PE-conjugated, APC-conjugated, or PerCP/Cy5.5-conjugated anti-GZMB and anti-IFNγ antibodies, with dead cells excluded by 7-AAD staining. All samples were analyzed using FACScan flow cytometry.

RNA extraction, qRT‒PCR analysis, and Western blotting
RNA extraction, qRT‒PCR analysis, and Western blotting were performed according to previously reported method.2629

Statistical analyses
Statistical analysis was performed using GraphPad Prism V.8 software. Data are shown as means±SEM unless stated otherwise. The specific statistical tests are given in the respective figure legends. A 95% CI was considered significant and defined as p<0.05. * indicates p<0.05, ** p<0.01, *** p<0.001. Ns, no significance.

Results

Inverse relationship between SPP1 TAM abundance and CD8 T cell infiltration in CRC TME
CD8 cytotoxic T cells play a pivotal role in tumor cell eradication.30 To investigate the impact of TAM-derived SPP1 on CD8 T cell infiltration in tumor tissues, we first assessed the correlation between SPP1 expression and patient prognosis in CRC using publicly available datasets (UALCAN: https://ualcan.path.uab.edu/; GEPIA: http://gepia.cancer-pku.cn/). Our analysis revealed that elevated SPP1 expression levels in colon adenocarcinoma (COAD) and rectal adenocarcinoma (READ) were significantly associated with poorer patient survival (figure 1A,B), consistent with previous findings.14 Further analysis using TIMER 2.0 demonstrated a strong positive correlation between SPP1 expression and macrophage in COAD and READ (figure 1C). Notably, SPP1 expression was predominantly detected in TAMs but not in other immune cell populations (figure 1D). ScRNA-seq data22 revealed a significant increase in TAMs accompanied by a corresponding decrease in CD8 T cells within CRC tumor tissues, a phenomenon not observed in adjacent para-tumor tissues (figure 1E). This inverse relationship suggests that TAM accumulation may contribute to CD8 T cell exclusion in the TME. We further characterized TAM subpopulations and identified distinct SPP1 TAM clusters (figure 1F,G). To directly examine the association between SPP1 TAMs and CD8 T cell infiltration, we analyzed surgical specimens from 40 treatment-naïve CRC patients. Patients were stratified into high and low frequency of SPP1+ TAMs based on immunohistochemical staining scores. Intriguingly, tumors with high frequency of SPP1+ TAMs exhibited significantly reduced CD8 T cell infiltration compared with those with low frequency of SPP1+ TAMs (figure 1H). Moreover, fluorescence staining of CRC samples showed that SPP1 expression inversely correlated with the abundance of CD8 T cell infiltration (figure 1I). Furthermore, analysis revealed a close correlation between SPP1+ macrophages and CD8+ T cells in CRC patients (figure 1J), and a concurrent negative correlation between SPP1 expression and CD8+ T cell infiltration in CRCs (COAD and READ) (figure 1K). Taken together, these findings provide compelling evidence for an inverse relationship between SPP1 TAM abundance and CD8 T cell infiltration in the CRC TME.

SPP1 deletion in TAM promotes infiltration of CD8 T cells in tumor environment
To investigate the role of macrophage-derived SPP1 in CD8 T cell recruitment to TMEs, we generated macrophage-specific Spp1cKO mice by crossing Spp1fl/fl mice with Lyz2-Cre mice (online supplemental figure S1A–D). No differences in immune cell frequency or composition (including myeloid and T cell populations) were observed in the spleens of age-matched and sex-matched Spp1cKO and Spp1fl/fl mice (online supplemental figure S1E). We first established a CRC model by subcutaneously injecting MC38 cells (which lack SPP1 expression; online supplemental figure S2A) into these mice. By day 21 post-injection, Spp1cKO mice exhibited significantly reduced tumor volume and weight as compared with Spp1fl/fl controls (figure 2A and C). This antitumor effect also extended to other tumor models, including B16-OVA melanoma and LLC, where similar reductions in tumor growth were observed (figure 2B, E, G and I), consistent with previous reports.14 31 Immune profiling of MC38 tumors revealed striking differences in tumor-infiltrating leukocytes. Spp1cKO mice showed markedly increased CD8 T cell infiltration (figure 2D; online supplemental figure S2B), enhanced NK cell (NK1.1+) recruitment, and reduced frequencies of myeloid-derived suppressor cells (CD11B+Gr1+), macrophages (CD11B+F4/80+), and Tregs (CD4+CD25+Foxp3+) (online supplemental figure S2C). No significant differences were observed in CD4 T cell frequencies (online supplemental figure S2C). Similar CD8 T cell increases were also observed in B16-OVA and LLC models (figure 2F and J). To directly test macrophage effects on tumors, we co-inoculated MC38 cells with BMDMs from Spp1cKO or control mice. Tumors with SPP1-deficient BMDMs showed significant growth reduction (figure 2H and K) and increased CD8 T cell infiltration (both frequency and absolute numbers) (figure 2L). Moreover, reintroducing Spp1+ macrophages into the Spp1cKO mice also promoted tumor growth and reduced CD8+ T cell infiltration in tumor tissues (online supplemental figure S3). These findings demonstrate that macrophage SPP1 deficiency not only inhibits tumor progression but also promotes CD8 T cell accumulation in the TME across multiple cancer models.

Increased CD8 T cell infiltration in tumor tissues of Spp1cKO mice exhibits potent antitumor activity
Next, we characterized the enriched CD8 T cells in tumor tissues of Spp1cKO mice. ScRNA sequencing analysis of immune cells in MC38 colorectal tumors identified major populations, including monocytes/macrophages, granulocytes, T cells, NK cells, dendritic cells, plasma cells, neutrophils, and fibroblasts (figure 3A–C). compared with control Spp1fl/fl mice, Spp1cKO mice exhibited a significant increase in tumor-infiltrating CD8 T cells (figure 3D). Notably, these increased CD8 T cells displayed enhanced cytotoxic potential, as evidenced by high cytotoxicity and proliferation scores and reduced exhaustion signatures (figure 3E). Further analysis revealed upregulation of cytotoxic effector genes (GZMB, XCL1, IFNG)32 and downregulation of exhaustion-associated markers (TOX, STAT3, PDCD1)33 34 in these cells (figure 3F), which was validated by qPCR (figure 3G). These findings suggest that CD8 T cells in Spp1cKO tumors possess heightened antitumor functionality. KEGG and GO analyses corroborated these observations, showing enrichment in pathways related to T cell activation, migration, and cytokine-mediated signaling in CD8 T cells of Spp1cKO mice (online supplemental figure S4A). Consistent with this, these cells exhibited an effector/memory phenotype (Tem, CD62LlowCD44high) in MC38 tumor tissues (figure 3H) or in B16-OVA tumor tissues (online supplemental figure S5A) and higher frequencies of IFNγ and GZMB subsets compared with controls (figure 3I,J). Conversely, the proportion of PD-1 CD88 T cells was significantly reduced (figure 3I). Immunofluorescence (IF) further confirmed increased CD8 T cell infiltration in tumors of Spp1cKO mice (figure 3K). These phenotypes were recapitulated in the B16-OVA melanoma model (online supplemental figure S5B). Critically, enhanced antitumor response in Spp1cKO mice was completely abrogated on CD8 T cell depletion via anti-CD8 antibody administration (figure 3L,M). Collectively, these data demonstrate that tumors in Spp1cKO harbor CD8 T cells with superior cytotoxic capacity and reduced exhaustion, driving robust antitumor immunity.

Recruitment of CD8 T cells following SPP1 deletion in TAMs is mediated via CXCR3
The chemokine receptor CXCR3 and its ligands, including MIG/CXCL9 and IP-10/CXCL10, play a pivotal role in mediating cellular inflammation.35 Using CellPhoneDB, a cell–cell communication analysis tool, we identified unique interactions such as CXCL9-CXCR3 and CXCL10-CXCR3 between Spp1cKO macrophages and CD8 T cells (figure 4A). These findings suggest that CXCL9-CXCR3 and CXCL10-CXCR3 interactions may facilitate CD8 T cell infiltration into tumor tissues of Spp1cKO mice. Consistent with this hypothesis, immunostaining revealed fewer CXCR3+ CD8 T cells and CXCL9/10 macrophages in Spp1fl/fl tumor tissues compared with Spp1cKO tumor tissues (figure 4B; online supplemental figure S6B). Furthermore, co-staining of CXCR3/CD8 and CXCL9/F4/80 demonstrated enhanced crosstalk between these cell populations in tumor tissues of Spp1cKO mice (figure 4C). In vitro coculture experiments showed that Spp1cKO macrophages more effectively promoted CD8 T cell migration than their controls. This effect was attenuated by AMG487, a CXCR3 antagonist (figure 4D). Notably, while CXCR3 antagonists are known to influence T cell differentiation,36 37 our transwell assays demonstrated their specific impact on CD8 T cell migration (figure 4D), supporting a role for CXCR3 in SPP1 deletion-mediated CD8 T cell recruitment. In vivo, CXCR3 antagonism abolished the antitumor effects conferred by macrophage-specific SPP1 deletion (figure 4E,F), accompanied by reduced CD8 T cell infiltration as confirmed by flow cytometry (figure 4G). CXCR3 blockade also diminished the population of effector CD8 T cells in Spp1cKO mice (figure 4H), with reduced CD8 T cell infiltration further validated by IF (figure 4I). Clinically, tumor tissues from patients with high frequency of SPP1+ TAMs exhibited significantly fewer CXCR3+ CD8 T cells compared with those with low frequency of SPP1+ TAMs (figure 4J). Given that CXCR3 is not exclusive to CD8 T cells, we compared the effects of anti-CD8 and CXCR3 inhibitors on tumor growth in Spp1cKO mice. No significant differences were observed among anti-CD8, anti-CD8 plus CXCR3 inhibitor, or CXCR3 inhibitor alone treatments (online supplemental figure S7), suggesting that CXCR3+ CD8 T cells, rather than other CXCR3 cell populations, mediate the antitumor response. In addition, SPP1 levels were inversely correlated with the expression of CXCL9 and CXCL10 in tumor tissues from CRC patients (figure 4K,L). The neutralization of CXCL9/10 also blocked the increase in CD8+ T cell recruitment resulting from SPP1 deletion (online supplemental figure S8). Thus, these results demonstrate that SPP1 deletion in TAMs promotes CD8 T cell recruitment primarily through the CXCR3.

SPP1 deletion enhances STAT1-mediated CXCL9/10 production in macrophages
Since CD8 T cell recruitment is mediated by the chemokine receptor CXCR3 and its ligands CXCL9/10, we hypothesized that Spp1cKO macrophages might exhibit increased production of these chemokines. Supporting this notion, previous scRNA-seq data revealed mutually exclusive expression patterns of CXCL9 and SPP1 in TAMs [7]. RNA-seq analysis also demonstrated significantly elevated expression of both CXCL9 and CXCL10 in Spp1cKO macrophages compared with Spp1fl/fl controls (online supplemental figure S9A). Further clustering of MC38 tumor TAMs identified 11 distinct subpopulations (figure 5A; online supplemental figure S10A,B). Among these, clusters 5, 6, and 10 showed markedly higher CXCL9 and/or CXCL10 expression in Spp1cKO macrophages as compared with Spp1fl/fl macrophages (figure 5B; online supplemental figure S10B,C). Notably, the proportion distribution of some macrophage subsets was different between Spp1fl/fl and Spp1cKO mice (figure 5A; online supplemental figure S10), indicating that the absence of SPP1 may affect the differentiation of macrophages in tumor tissues. QRT-PCR validation confirmed the upregulation of CXCL9/10 in Spp1cKO BMDMs (figure 5C). RNA-seq profiling also revealed elevated STAT1 levels in Spp1cKO BMDMs (figure 5D), accompanied by increased phosphorylated STAT1 (p-STAT1) (figure 5E). Consistent findings were observed in SPP1-silenced THP1 macrophages (online supplemental figure S9B,C). Given the established role of STAT1 signaling in CXCL9/10 secretion,3840 these data suggest that SPP1 deficiency potentiates CXCL9/10 expression through STAT1 activation in macrophages. Indeed, STAT1 silencing abolished the differential expression of CXCL9/10 between WT and Spp1cKO BMDMs (figure 5F,G) and impaired CD8 T cell migration (figure 5H). Further supporting this mechanism, Spp1cKO BMDMs exhibited enhanced STAT1 phosphorylation and more rapid responses to IFNβ stimulation (a known STAT1 activator41 compared with controls, online supplemental figure S9D). Immunostaining additionally demonstrated increased CXCL9/10 macrophages in IFNβ-treated BMDMs (online supplemental figure S9E). In vivo experiments revealed that infusion of either STAT1-silenced or CXCL9/10-silenced macrophages similarly promoted tumor growth (figure 5I–K), indicating that STAT1 exerts its antitumor effects primarily through CXCL9/10. Notably, there was more CD74 expression in Spp1cKO macrophages (online supplemental figure S11), suggesting their polarization within TMEs.42 Taken together, these findings demonstrate that SPP1 deficiency amplifies STAT1-mediated CXCL9/10 secretion in macrophages, providing a mechanistic link between SPP1 signaling and CD8 T cell recruitment in tumors.

Deficiency of SPP1 leads to STAT1 activation via ROS-DNA fragment/cGAS-STING signaling pathway
Previous studies have demonstrated that cytosolic DNA fragments can co-localize with cyclic GMP-AMP synthase (cGAS), activating the cGAS-STING signaling pathway to promote STAT1 activation.43 44 To identify the pathway(s) responsible for STAT1 activation in Spp1cKO macrophages, we analyzed signaling pathways using scRNA-seq and bulk RNA-seq data. KEGG and GSEA analysis revealed significant enrichment of the cytosolic DNA-sensing pathway in Spp1cKO macrophages (figure 6A; online supplemental figure S12), suggesting potential involvement of cGAS-STING signaling in STAT1 activation. Consistent with this hypothesis, we observed elevated activation of the cGAS-STING pathway in both Spp1cKO TAMs and BMDMs by Western blot and IF (figure 6B). Increased phosphorylation of TBK1, a key marker of STING pathway activation,45 46 further supported enhanced cGAS-STING signaling in Spp1cKO macrophages (figure 6B,C). P-STAT1 could be detected in the cytoplasm and also in the nucleus (figure 6D,E). Given that cytosolic dsDNA fragments can trigger cGAS-STING signaling,42 43 we quantified these fragments and found significantly higher levels in Spp1cKO macrophages compared with controls (figure 6F).
We next investigated the source of increased cytosolic dsDNA. Since mitochondrial ROS can induce DNA damage47 and generate DNA fragments,48 49 we examined mitochondrial ROS levels in SPP1-deficient macrophages. Indeed, Spp1cKO macrophages exhibited elevated ROS (figure 6G), which was suppressed by the mitochondrial-targeted antioxidant mito-TEMPO50 (figure 6H). Importantly, mito-TEMPO also reduced cytosolic DNA accumulation (figure 6I) and STING phosphorylation (figure 6J) in Spp1cKO macrophages. Furthermore, treatment with CCCP (an oxidative phosphorylation inhibitor51) or mito-TEMPO diminished CXCL10 production and STAT1 activation in Spp1cKO macrophages. Conversely, stimulation with c-di-AMP, an agonist of the STING pathway, enhanced STAT1 activation and CXCL10 expression (figure 6K). Similar regulatory effects of SPP1 on CXCL9/10 were confirmed in THP1 cells (online supplemental figure S13).
We further explored the mechanism by which SPP1 inhibits CXCL9/10 production. SPP1 can function through the SPP1/CD44 signaling pathways.52 Our results demonstrate that SPP1 suppresses ROS generation, an effect that was attenuated by CD44 blockade (figure 6L,M). Furthermore, CD44 inhibition diminished SPP1-mediated suppression of STAT1 and STING phosphorylation (figure 6N,O). In addition, CD44 blockade improved macrophage-mediated CD8+ T cell infiltration following in vitro exposure to SPP1 (figure 6P). Similarly, pharmacological inhibition or neutralization of SPP1 in mice also suppressed tumor growth (online supplemental figure S14). These findings suggest that SPP1 modulates CXCL9/10 expression through the SPP1/CD44 signaling pathway.
The cGAS-STING pathway is a critical innate immune signaling cascade that induces type I interferons (IFN-I),53 which in turn promote CXCL9/10 expression and STAT1 activation.54 Consistent with this, Spp1cKO macrophages exhibited elevated IFN-I (IFNα/IFNβ) levels compared with Spp1fl/fl controls (online supplemental figure S15A,B). Blocking IFN-I signaling attenuated STAT1 activation and CXCL9/10 production (online supplemental figure S15C,D), meanwhile also affecting the STAT1 expression (online supplemental figure S15E–G), further supporting its role in this regulatory network.
Taken together, all of these suggest that activation of STAT1 mediated by SPP1 deletion is achieved through the ROS-DNA fragment /cGAS-STING signaling pathway.

SPP1 deficiency enhances CD8 T cell-dependent antitumor immunotherapy
ICIs, including anti-PD-1 antibodies, have revolutionized cancer treatment by harnessing the cytotoxic potential of CD8 T cells.55 Given the critical role of macrophage-derived SPP1 in modulating CD8 T cell infiltration into tumors, we assessed the impact of macrophage-specific SPP1 deletion on anti-PD-1 therapy efficacy using the MC38 tumor model. Consistent with previous reports,14 31 MC38 tumors responded robustly to anti-PD-1 treatment in control Spp1fl/fl mice. Strikingly, this therapeutic effect was also further enhanced in Spp1cKO mice, which exhibited significantly greater tumor growth inhibition (figure 7A–C). The diminished treatment response in Spp1cKO mice may be linked to reduced PD-1 expression in CD8 T cells. Notably, anti-PD-1 treatment also markedly enhanced CD8 T cell infiltration in tumors of Spp1cKO mice (figure 7D; online supplemental figure S16A). These infiltrating CD8+ T cells exhibited elevated cytotoxic (GZMB) and effector (IFNγ) activity alongside lower PD-1 expression (figure 7E; online supplemental figure S16B). SPP1-deficient macrophages also upregulated the T cell-recruiting chemokines CXCL9 and CXCL10 (figure 7F), which correlated with increased CD8 T cell accumulation, as confirmed by immunostaining (figure 7G,H). These findings were further validated in the B16 melanoma model (online supplemental figure S17). Consistent with the previous report,14 31 pharmacological blockade of CXCR3 abrogated the enhanced antitumor response in Spp1cKO mice (figure 7I–K), indicating the necessity of CD8 T cell recruitment for optimal immunotherapy efficacy. Thus, our results establish that SPP1 deficiency in macrophages augments CD8 T cell-dependent antitumor immunity, highlighting a promising strategy to ICI therapy.

Discussion
Our findings demonstrate that SPP1 ablation in macrophages promotes CD8 T cell infiltration into tumor tissues, suppresses tumor growth, and synergizes with ICI therapy. SPP1 deficiency induces mitochondrial ROS production, leading to the accumulation of cytoplasmic dsDNA fragments. This triggers activation of the cGAS-STING signaling pathway and subsequent upregulation of STAT1 phosphorylation. The enhanced STAT1 activation drives CXCL9/10 expression, creating a chemotactic gradient that recruits CD8 T cells into the TME. Together, these results establish macrophage SPP1 as a novel therapeutic target to potentiate antitumor immunity and improve immunotherapy outcomes.
We demonstrate that macrophage-derived SPP1 represents a promising target for tumor immunotherapy. While CD8 cytotoxic T cells play a crucial role in eliminating tumor cells,30 their limited infiltration in cold tumors (including immune-excluded and immune-desert subtypes) significantly restricts immunotherapy efficacy.56 57 This highlights the critical need to enhance CD8 T cell infiltration into tumor tissues to inhibit tumor growth and improve treatment response. In this study, we reveal that SPP1 deficiency in TAMs promotes the infiltration of cytotoxic CD8 T cells by upregulating CXCL9/10 secretion, effectively converting “cold tumors” into “hot tumors.” This transformation not only suppresses tumor growth but also potentiates the effects of immunotherapy. Supporting our observations, prior research has established a negative correlation between SPP1 TAMs and CD8 T cell infiltration in CRC, where a high abundance of SPP1 macrophages correlates with reduced T cell presence and unfavorable clinical outcomes.7 The importance of CXCL9-expressing TAMs in recruiting and positioning functional CD8 cytotoxic T cells has been increasingly recognized as a key mechanism in orchestrating effective antitumor immunity.58 Macrophage-secreted CXCL9/10 modulates tumor progression by directly engaging with CD8 T cells.5961 These chemokines act through CXCR3, a receptor predominantly expressed on CD8 T cells,34 to recruit tumor-suppressive CXCR3+ CD8 T cell populations.62 Notably, CXCL10 has shown therapeutic potential in cancer immunotherapy, both as a standalone treatment and in combination with ICIs.63 64
We found that macrophage-derived SPP1 suppresses CD8 T cell infiltration by attenuating CXCL9/10 chemokine production through the ROS-DNA fragment/cGAS-STING/STAT1 axis in the TME. While STAT1 activation is known to enhance chemokine expression38 and promote a proinflammatory and immune-supportive TME,65 our findings reveal that SPP1 macrophages disrupt this process. In rheumatoid arthritis, SPP1 macrophages have been shown to modulate JAK-STAT activation,66 and mechanistically, STAT1 phosphorylation at conserved tyrosine and serine residues induces dimerization and nuclear translocation, where it transcriptionally regulates CXCL9/10 expression.67 The cGAS-STING pathway, activated by extranuclear DNA fragments colocalizing with cGAS, plays a key role in STAT1 activation.43 44 This pathway integrates nucleic acid sensing with immune responses, contributing to cancer, autoimmune, and inflammatory disease regulation. Cytosolic DNA, whether self-derived or from foreign sources (eg, tumor cells, dead cells, viruses, or microbes), is detected by cGAS, triggering conformational changes that initiate cGAS-STING signaling.43 44 68 Our data suggest that targeting SPP1 in TAMs induces mitochondrial ROS production, leading to cytosolic DNA accumulation. Mitochondrial ROS exerts cytotoxic effects,47 including DNA damage and fragmentation.48 49 Interestingly, our data show that SPP1 from TAMs can affect mitochondrial ROS generation.
Recent studies have demonstrated that targeting SPP1 through RNA interference,69 small molecule inhibitors,70 or suppression of SPP1 expression4 can effectively inhibit tumor growth both in vitro and in vivo. Notably, denosumab, an SPP1-suppressing agent, has been shown to enhance antitumor immunity.20 Furthermore, the lead compound CANDI460, which downregulates SPP1 expression in both cellular and animal models, induces tumor remission across various murine cancer models.4 Our findings reveal that specifically targeting SPP1 in TAMs enhances CD8 T cell infiltration into the TME via the ROS-DNA fragment/cGAS-STING/STAT1 pathway, which subsequently upregulates CXCL9/10 chemokine production. These results strongly support the therapeutic potential of targeting SPP1 in TAMs as a promising strategy for cancer immunotherapy.

Supplementary material
10.1136/jitc-2025-013697online supplemental file 1