Elevated tumor NOS2/COX2 promotes immunosuppressive phenotypes associated with poor survival in ER- breast cancer

Abstract

Tumor immunosuppression affects survival and treatment efficacy. Tumor NOS2/COX2 coexpression strongly predicts poor outcome in estrogen receptor-negative (ER-) breast cancer by promoting metastasis, drug resistance, cancer stemness, and immune suppression. Herein, a spatially distinct NOS2/COX2 and CD3+CD8+PD1- T effector (TEff) cell landscape correlated with poor survival in ER- tumors. NOS2 was primarily expressed at the tumor margin, whereas COX2 together with B7H4 was associated with immune desert regions lacking TEff cells, where a higher ratio of tumor NOS2 or COX2 to TEff cells predicted poor survival. Also, programmed cell death ligand 1/programmed cell death 1, regulatory T cells (TRegs), and IDO1 were primarily associated with stroma-restricted TEff cells. Regardless of the survival outcome, CD4+ T cells and macrophages were primarily in stromal lymphoid aggregates. Finally, in a 4T1 model, COX2 inhibition led to increased CD8+ TEff/CD4+ TReg ratio and CD8+ TEff infiltration while Nos2 deficiency had no significant effect, thus reinforcing our observations that COX2 is an essential component of immunosuppression through CD8+ TEff cell exclusion from the tumor. Our study indicates that tumor NOS2/COX2 expression plays a central role in tumor immune evasion, suggesting that strategies combining clinically available NOS2/COX2 inhibitors with immune therapy could provide effective options for the treatment of aggressive and drug-resistant ER- breast tumors.

Keywords: Adaptive immunity; Inflammation; Oncology; T cells.

Figure 1. Survival analysis of tumor NOS2, COX2, and IFN-γ at 5 years after diagnosis.

Annotated viable tumor was analyzed for survival effects of (A) IFN-γ and (B) tumor NOS2 and COX2 normalized to CK-SOX10 tumor biomarker in tumors from deceased (n = 10) and alive (n = 10) ER^– breast cancer patients at 5 years after diagnosis. (C) Pearson’s correlation analysis of NOS2 and COX2 in tumors from deceased and alive ER^– breast cancer patients. (D) Effect of tumor NOS2 and CD8⁺ T cells on complete pathological response (non pCR n = 10, pCR n = 8) in biopsies from patients in the neoadjuvant KEYNOTE 522 (K522) clinical trial who received the PD1 inhibitor pembrolizumab and chemotherapy. *P < 0.05, **P = 0.0031, Mann-Whitney test.

Figure 2. Cytokine-induced NOS2 and COX2 expression.

(A) Gene expression (left) and Western blot (right) showing temporal NOS2 and COX2 expression in MDA-MB-231 breast cancer cells treated with IFN-γ + IL-1β + TNF-α (CM) at 24 and 48 hours. ****P < 0.0001 Mann-Whitney test for both NOS2 (top) and COX2 (bottom); 2-way ANOVA + Fisher’s least significant difference *P < 0.05, **P = 0.0012, or Šídák's multiple-comparison test ***P = 0.0006. MWM, molecular weight marker. (B) NOS2s/IFN-γ ratio is significantly elevated in tumors from deceased ER^– patients at 5 years after diagnosis. ***P = 0.0007 Mann-Whitney test. (C) Pearson’s correlation showing significance between NOS2s and COX2_all versus IFN-γ in tumors from deceased and alive patients at 5 years after diagnosis. (D) The significance of the NOS2/IFNG and COX2/IFNG relationships are validated in National Center for Biotechnology Information (NCBI) Gene Expression Omnibus (GEO) datasets found in https://www.ncbi.nlm.nih.gov/geo/ Gehan-Breslow-Wilcoxon survival analysis and Mantel-Haenszel hazard ratio analysis were used. Top graphs validate survival using cohort data from this study, while lower graphs are from combined GEO datasets (n = 796) generated using https://kmplot.com/analysis

Figure 3. Tumor COX2 promotes immune cold regions.

(A) Schematic showing immunosuppressive mediators and abated CD8⁺ T_Eff cell infiltration. Pearson’s correlation analysis showing significance between CD8⁺ T_Eff cells versus immunosuppressive phenotypes normalized to COX2 tumor expression, where (B) CD8⁺ T_Reg, CD4⁺ T_Reg, and CD8⁺ T_Ex cells, as well as (C) COX2-expressing macrophages but not PDL1-expressing tumor macrophages correlated significantly.

Figure 4. Unsupervised spatial S-UMAP analysis validates the predictive value of NOS2⁺/COX2⁺ tumor clusters.

(A) S-UMAP of the clustering cellular phenotypes revealing distinct clustering regions (yellow circles) and (B) the differential cluster distribution analysis of the of %deceased normalized to %alive (red and blue circles) patient tumors at 5 years after diagnosis. (C) Nearest neighborhood analyses show distance-dependent changes in cellular phenotypes that are predictive of poor patient survival. Importantly, density-dependent clustering gradients are demonstrated for NOS2_Tumor and COX2_Tumor phenotypes. (D) Moderately increased clustering of CD8⁺ T_Ex and PDL1_Macrophage phenotypes is predictive of improved patient survival. For C and D the first bar is 25 µm.

Figure 5. Regional placement of cellular phenotypes within the tumor.

The %cells of predictive phenotypes including (A) CD8⁺ and (B) CD4⁺ (C) tumor and (D) macrophage phenotypes are shown. Ordinary 1-way ANOVA *P < 0.05, **P ≤ 0.0061, ***P ≤ 0.0004, ****P < 0.0001. Sat, satellites.

Figure 6. Progression of CD8⁺ T cell exclusion during the formation of tumor immune deserts.

Distinct classes of CD8⁺ T cell exclusion were observed. (A) Type I: inflamed but stroma-restricted CD8⁺ T cells (orange) with NOS2 (red) expression at the tumor margin and high COX2 (green) expression deeper into tumor core. Type II: abated CD8⁺ T cells, low NOS2 expression at the tumor margin, and high COX2 expression deeper into tumor core. Type III: absence of CD8⁺ T cells with abated NOS2 expression at the tumor margin and low, sparse COX2 expression deeper into tumor core. (B) Differential penetration of CD4⁺ versus CD8⁺ T cells into high NOS2/COX2-expressing tumors of deceased patients where both CD4⁺ and CD8⁺ T cells remain in the tumor stroma, as shown in fluorescence image, spatial animation (middle), and density heatmap (right). (C) In contrast, CD4⁺ T cells remain in the tumor stroma while CD8⁺ T cells penetrate deep into low NOS2/COX2-expressing tumor core of alive patients.

Figure 7. Analysis of immune phenotypes in lymphoid aggregates of ER^– tumors.

(A) CD8⁺ T_Eff/CD4⁺ T_Eff and CD8⁺ T_Eff/PDL1_Macrophage ratios are significantly higher in tumors from alive patients. Statistical outliers were excluded based upon mathematical analysis using Prism software ROUT method. (B) CD8⁺ T_Eff/CD4⁺ T_Reg and CD8⁺ T_Eff/PDL1_Macrophage ratios are significantly higher in indomethacin-treated 4T1 tumors when compared with WT control. (C) Validation of the predictive value of elevated CD8a/FOXP3 using the GEO database. *P < 0.05, ***P = 0.001 Mann-Whitney test. Survival analysis was performed using Gehan-Breslow-Wilcoxon test, and hazard ratio was determined using Mantel-Haenszel test.

Figure 8. Density heatmap distributions of immunosuppressive T_Reg, PDL1, IDO1, B7H4, and tumor NOS2 and COX2 phenotypes.

Stroma (orange) as well as viable (green) and necrotic (purple) annotated tumor regions are shown in each panel (left). Density heatmap distributions with low-to-high scale are shown for CD3⁺CD8⁺ T cells, CD4⁺FOXP3⁺T_Reg, PDL1_Tumor, IDO1, B7H4, NOS2_Tumor, and COX2_Tumor in NOS2/COX2-high tumors from deceased patients with (A) type I stroma-restricted CD8⁺ T cells with high NOS2 expression at the tumor margin and high COX2 expression deeper into tumor core and (B) type II immune deserts with abated CD8⁺ T cells, low NOS2 expression at the tumor margin, and high COX2 expression deeper into tumor core. (C) NOS2/COX2-low tumor from alive patient with high CD3⁺CD8⁺ T cell infiltration into the tumor. B and C show the same tumors as in Figure 6A (type III immune desert) and Figure 6C (alive patient).

Figure 9. Summary of immunosuppressive signaling progression.

Stroma-restricted CD8⁺ T_Eff cells secrete IFN-γ/TNF-α that induces tumor NOS2 and COX2 expression near tumor margin, leading to type I T_Eff restricted regions. Tumor COX2-expressing type II immune deserts with limited T_Eff infiltration progress to type III immune desert regions with abated NOS2 and COX2 tumor expression. IDO1 and PDL1_Macrophage are expressed in tumor stroma along with increased T_Reg populations in lymphoid aggregates.