Scavenger Receptor CD36 in Tumor-Associated Macrophages Promotes Cancer Progression by Dampening Type-I IFN Signaling

Abstract

Tumor-associated macrophages (TAM) are a heterogeneous population of myeloid cells that dictate the inflammatory tone of the tumor microenvironment. In this study, we unveiled a mechanism by which scavenger receptor cluster of differentiation 36 (CD36) suppresses TAM inflammatory states. CD36 was upregulated in TAMs and associated with immunosuppressive features, and myeloid-specific deletion of CD36 significantly reduced tumor growth. Moreover, CD36-deficient TAMs acquired inflammatory signatures including elevated type-I IFN (IFNI) production, mirroring the inverse correlation between CD36 and IFNI response observed in patients with cancer. IFNI, especially IFNβ, produced by CD36-deficient TAMs directly induced tumor cell quiescence and delayed tumor growth. Mechanistically, CD36 acted as a natural suppressor of IFNI signaling in macrophages through p38 activation downstream of oxidized lipid signaling. These findings establish CD36 as a critical regulator of TAM function and the tumor inflammatory microenvironment, providing additional rationale for pharmacologic inhibition of CD36 to rejuvenate antitumor immunity. Significance: CD36 in tumor-associated macrophages mediates immunosuppression and can be targeted as a therapeutic avenue for stimulating interferon production and increasing the efficacy of immunotherapy.

Figure 1.. CD36 is highly expressed by TAMs and promotes tumor growth.

(A) Percentage of tumor-infiltrating myeloid cells expressing CD36 at different time points during progression of YUMMER melanoma tumors. Tumor-infiltrating myeloid cells were gated as live CD45⁺CD3⁻NK1.1⁻CD11b⁺Ly6G⁻ population by flow cytometry. (B) Representative gating of myeloid subsets in day-20 YUMMER melanoma tumors, and quantification of CD36 expression in different subsets. DN, double negative. Quantification pooled from 3 individual experiments (n=10 total), gated on live CD45⁺CD3⁻NK1.1⁻CD11b⁺Ly6G⁻ population. (C) Representative histograms of CD36 expression by myeloid subsets in day-20 YUMMER tumors, quantified by mean fluorescence intensity (MFI). Quantification pooled from 3 individual experiments (n=10 total) and normalized to DN population in each experiment. (D) Representative expression of CD36 by PD-L1, CD206 and ARG1 in TAMs from day-20 YUMMER tumor. TAMs were gated as live CD45⁺CD3⁻NK1.1⁻CD11b⁺Ly6G⁻ populations by flow cytometry. (E) Quantification of PD-L1, CD206 and ARG1 expression by CD36^hi and CD36^low TAM subsets. Data pooled from 4 individual experiments, n=8–12. CD36^hi and CD36^low TAM gating was indicated by grey and pink boxes, respectively, in (D). (F) Growth curve of YUMMER melanoma tumors in Cd36^f/f (WT) or Cd36^f/f Csf1r-Cre (CD36-cKO) animals. 2.5×10⁵ YUMMER cells were subcutaneously implanted, tumor sizes were measured every 3–4 days starting from day-7 post engraftment. Data pooled from more than 5 individual experiments, n=15–16. (G) Growth curve of MC38 colorectal tumors in WT or CD36-cKO animals. 5×10⁵ MC38 cells were subcutaneously implanted, tumor sizes were measured every 3–4 days starting from day-7 post engraftment. Data pooled from more than 5 individual experiments, n=10–14. (H) Percentage of indicated tumor-infiltrating immune cells in WT and CD36-cKO animals (day-20 YUMMER tumor). (I) Representative expression of CD36 by PD-L1, CD206 and ARG1 in CD36-cKO TAMs from day-20 YUMMER tumor. Note differences compared to WT TAMs in (D). (J) Quantification of PD-L1, CD206 and ARG1 expression by WT and CD36-cKO TAMs. Data pooled from 4 individual experiments, n=7–9. Statistical analyses were done using ordinary one-way ANOVA and Tukey’s multiple comparisons test in (A)(B)(C), multiple unpaired t test in (E)(H)(J), two-way ANOVA and Šídák’s multiple comparisons test at each timepoint for (F)(G). *, p<0.05; **, p < 0.01; ***, p<0.001, ****; p < 0.0001; ns, not significant.

Figure 2.. CD36-deficient TAMs express high ISGs and secrete inflammatory cytokines in tumor.

(A) Uniform manifold approximation and projection (UMAP) of myeloid clusters in single-cell RNA sequencing data generated from C57BL/6 (WT) and Cd36^−/− animals using B16 melanoma model. Data re-analyzed from GSE171194, myeloid cell clusters identified based on expression of Itgam or Itgax. (B) Gene set enrichment analysis (GSEA) of pathways up- or down-regulated in Cd36^−/− TAMs. Genes differentially expressed between WT and Cd36^−/− TAMs (all four TAM clusters) were used for the analysis. (C) Feature plot of single-cell expression of type-I interferon response gene (ISG) signature score. ISG signature score was generated using “AddModuleScore” function in Seurat, with gene list from “Interferome 2.0” database. (D) Violin plot of type-I ISG signature score in all myeloid clusters from (A), comparing WT and Cd36^−/−. ISG signature score was generated as described in (C). (E) Violin plot of ISG expression by WT and Cd36^−/− TAMs (all four TAM clusters). (F) (Upper panel) Expression of Ifnb1 and Ifit3 by WT and CD36-cKO TAMs measured by quantitative PCR (qPCR). TAMs were sorted from day-20 YUMMER tumors for qPCR analysis. (Lower panel) Protein expression of ISG15 and IRF7 in WT and CD36-cKO TAMs quantified by flow cytometry. n=3–7. (G) Secretion profile of 9 functional subsets identified from single-cell barcoded ship (SCBC) assay as previously described (reference 47). WT and CD36-cKO TAMs were sorted from day-20 YUMMER tumors for SCBC analysis, combined for identifying functional subsets using “PhenoGraph”. Clusters were named based on the primary secreted proteins in each cluster. (H) Percentage of WT or CD36-cKO cells in each functional subset from G, ranked by their prevalence of CD36-cKO TAMs. (I) Violin plot of single-cell secretion of TNF⍺, IL-12p40 and Chi3l3. Signal intensity was normalized to “background” within each chip based on empty wells lacking cells. Cells with signal >1 a.u. (arbitrary units) were considered as “secretors” and the percentage of secretors were labeled above each bar. (J) Intra-tumoral IFNβ level in YUMMER or MC38 tumor lysates from WT or CD36-cKO animals measured by ELISA. (K) Intra-tumoral cytokine level in day-20 YUMMER tumors from WT or CD36-cKO animals. Mouse Cytokine/Chemokine 44-Plex Discovery Assay^® Array was used for the multiplexed cytokine analysis. Significantly upregulated cytokines in CD36-cKO were indicated by pink stars. Statistical analyses were done using Wilcoxon test in (D)(E), unpaired t test in (F)(I)(J)(K). *, p<0.05; **, p < 0.01; ***, p<0.001; ****, p < 0.0001; ns, not significant.

Figure 3.. CD36 binds to oxidized LDL to suppress inflammatory response in mouse TAM and human macrophages.

(A) Representative flow plots (left) of fluorophore-labeled oxLDL binding to WT or CD36-cKO TAMs after 30 minutes of ex vivo incubation. Bar graph (right) shows percentage of oxLDL⁺ population in TAMs. Data pooled from 3 individual experiments, n=9–11. (B) Representative histogram (left) of CD36 surface expression after ex vivo incubation with 50 ug/ml oxLDL or media. Bar graph (right) shows cumulative CD36 MFI. TAMs were enriched using CD11b⁺ magnetic beads for the experiment. Data pooled from 2 individual experiments, n=6. (C) Ex vivo TNF production by WT or CD36-cKO TAMs, with or without the presence of oxLDL (50 ug/ml). n=5 per group. (D) Ex vivo IFNβ production by WT or CD36-cKO TAMs, with or without the presence of oxLDL (50 ug/ml). n=6 per group. (E) Ifnb1 induction in WT BMDMs stimulated with poly (I:C), with or without 50 ug/ml oxLDL. Data pooled from 2 individual experiments, n=5. (F) Representative histogram (left) of CD36 surface expression by human monocyte-derived macrophages after incubation with media, oxLDL (50 ug/ml), or oxLDL and sulfosuccinimidyl oleate (SSO, 100 μM). CD36 MFI was normalized to media control for each individual donor (n=4). (G) Expression of CD206 and production of TNF⍺ by human monocyte-derived macrophages after incubation with media, oxLDL (50 ug/ml), or oxLDL and sulfosuccinimidyl oleate (SSO, 100 μM). TNF⍺ production was measured after 4-hour LPS stimulation with or without oxLDL and SSO in the culture. Statistical analyses were done using unpaired t test in (A)(B)(E), two-way ANOVA and Šídák’s multiple comparisons test for (F)(G), one-way ANOVA and Fisher’s LSD test in (C)(D), ratio paired t test in (E)(F)(G). *, p<0.05; **, p < 0.01; ***, p<0.001; ****, p < 0.0001; ns, not significant.

Figure 4.. oxLDL-CD36 axis suppresses IFNβ production through phosphorylation of MAPK-p38 in mouse and human macrophages.

(A) Representative histogram of phospho-p38 level in WT and CD36-cKO TAMs. (B) Quantification of phospho-p38⁺ populations in WT and CD36-cKO TAMs (n=4 per group). (C) MFI of phospho-p38 in WT and CD36-cKO TAMs after ex vivo incubation with 50 ug/ml oxLDL or media for 30 minutes. TAMs were enriched using CD11b⁺ magnetic beads before treatment with oxLDL (n=3–4). (D) Ifnb1 expression by WT or p38⍺-KO BMDMs treated with oxLDL. BMDMs were transfected with Cas9 RNP containing scramble (Scb) or p38⍺-tartgeting (sgMapk14) guide RNAs. Data pooled from 4 individual experiments. (E) Western Blots of total- and phospho-p38 in human monocyte-derived macrophages treated with media control, oxLDL, or oxLDL and CD36 inhibitor SSO (n=4 healthy donors). (F) Quantification of phospho-p38 signal intensity from Western Blots in (D). Phospho-p38 signal was normalized to total-p38 for each donor (n=4). (G) Growth curves of YUMMER tumors in WT and CD36-cKO animals, treated with p38 inhibitor (p38i, SB203580) or vehicle control. 5×10⁵ YUMMER cells were subcutaneously implanted, p38i (10 mg/kg) or vehicle were given daily via oral gavage starting from day-8 post engraftment. Tumor sizes were measured every 2 days from day-8 post engraftment. n=4–12. (H) Growth curves of YUMMER-Ifnar1^−/− tumors in WT and CD36-cKO animals, treated with p38 inhibitor or vehicle control. 5×10⁵ tumor cells were subcutaneously, p38i or vehicle control were given as described in (F). n=5 per group. (I) Ifnb1 expression by TAMs sorted from WT animals treated in vivo with p38 inhibitor or vehicle control. n = 5–8, data pooled from 2 independent experiments. (J) IFNβ protein production by WT or CD36-cKO TAMs isolated from animals treated in vivo with p38 inhibitor or vehicle control. n = 4–6. (K) Intra-tumoral IFNβ level in YUMMER tumor lysates from WT or CD36-cKO animals treated in vivo with p38 inhibitor or vehicle control. n = 8–16, data pooled from 3 independent experiments. Statistical analyses were done using unpaired t test in (B)(C)(I), two-way ANOVA and uncorrected Fisher’s LSD test in (C), ratio paired t test in (E)(H), two-way ANOVA and Šídák’s multiple comparisons test in (H)(I), one-way ANOVA and Fisher’s LSD test in (J)(K). *, p<0.05; **, p < 0.01; ***, p<0.001; ****, p < 0.0001; ns, not significant.

Figure 5.. CD36-deficient TAMs promote tumor cell quiescence through production of IFNβ.

(A) Growth curves of YUMMER and MC38 tumors in WT or CD36-cKO animals, treated with anti-IFN^® mAb or IgG control intraperitoneally every 3–4 days from day-7 post engraftment. n=4–8 for YUMMER, n=7–9 for MC38. (B) Growth curves of YUMMER-Ifnar1^−/− and MC38-Ifnar1^−/− tumors in WT or CD36-cKO animals. 5×10⁵ tumor cells were subcutaneously implanted to the animals, tumor sizes were measured every 3–4 days starting from day-7 post engraftment. n=3–5. (C) Schematic of in vitro treatment of Fucci-expressing tumor cells with IFN^®. (D) Percentage of mKO2⁺⁺ (G0, left) and mAG⁺ (S/G2/M, right) populations in MC38-Fucci (circles with solid lines) and MC38-Fucci-Ifnar1^−/− (squares with dashed lines) cell lines when treated with different concentrations of IFN^® in vitro. Data was representative of 2 individual experiments. (E) Schematic of in vivo experiment using Fucci-expressing tumor cells. (F) Representative flow plots (left) and quantification (right) of mKO2 and mAG expression by MC38-Fucci tumor cells in WT or CD36-cKO animals. n=4. (G) Representative flow plot (left) and quantification (right) of mKO2 and mAG expression by MC38-Fucci tumor cells in WT and CD36-cKO animals treated with anti-IFN^® mAb. n = 3. (H) Representative flow plot (left) and quantification (right) of mKO2 and mAG expression by MC38-Fucci-Ifnar1^−/− tumor cells in WT and CD36-cKO animals. n = 3–4. Statistical analyses were done using two-way ANOVA and Šídák’s multiple comparisons test in (A)(B), unpaired multiple t test in (F)(G)(H). *, p<0.05; **, p < 0.01; ***, p<0.001; ****, p < 0.0001; ns, not significant.

Figure 6.. CD36 expression inversely correlates with type-I interferon signatures and overall survival in cancer patients.

(A) Expression of CD36 mRNA and selected ISGs by monocyte and macrophage clusters in patient pan-cancer single-cell RNA sequencing data set (GSE154763). (B) In-silico FACS analysis CD36 mRNA expression versus IFIT3 or ISG15 from pan-cancer myeloid single-cell RNA dataset (GSE154763). (C) Expression of CD36 and selected ISGs by bulk tumors in The Cancer Genome Atlas (TCGA) database. All patients (n = 1210) with RNA sequencing data were include in the heatmap. (D) Correlation of CD36 mRNA with selected ISGs (ISG20, IRF3) in the TCGA pan-cancer dataset (n = 1210). (E) Kaplan-Meier survival curve of skin cutaneous melanoma (SKCM), lung squamous cell carcinoma (LUSC) and acute myeloid leukemia (LAML) patients in TCGA dataset stratified by expression of CD36^hi TAM signature (CD36, MRC1, ARG1). High and low expression groups indicate the top 25^th and lower 25^th percentile in the dataset. Statistical analysis was done using Spearman and Pearson correlation in (D), log rank test in (E).