Co-targeting TGF-β and PD-L1 sensitizes triple-negative breast cancer to experimental immunogenic cisplatin-eribulin chemotherapy doublet

Abstract

In preclinical mouse models of triple-negative breast cancer (TNBC), we show that a combination of chemotherapy with cisplatin (CDDP) and eribulin (Eri) was additive from an immunological point of view and was accompanied by the induction of an intratumoral immune and inflammatory response favored by the immunogenic cell death induced by CDDP, as well as by the vascular and tumor stromal remodeling induced by each chemotherapy. Unexpectedly, despite the favorable immune context created by this immunomodulatory chemotherapy combination, our models remained refractory to the addition of anti-PD-L1 immunotherapy. These surprising observations led us to discover that CDDP chemotherapy was simultaneously responsible for the production of TGF-β by several populations of cells present in tumors, which favored the emergence of different subpopulations of immune cells and cancer-associated fibroblasts characterized by immunosuppressive properties. Accordingly, co-treatment with anti-TGF-β restored the immunological synergy between this immunogenic doublet of chemotherapy and anti-PD-L1 in a CD8-dependent manner. Translational studies revealed the unfavorable prognostic effect of the TGF-β pathway on the immune response in human TNBC, as well as the ability of CDDP to induce this cytokine also in human TNBC cell lines, thus highlighting the clinical relevance of targeting TGF-β in the context of human TNBC treated with chemoimmunotherapy.

Keywords: Breast cancer; Cancer immunotherapy; Immunology; Oncology.

Figure 1. The CDDP-Eri doublet increases tumor CTL infiltration and functionality.

(A) 4T1 and EMT6 tumor–bearing mice were treated with CDDP, Eri, or CDDP-Eri, and tumor volume was monitored for 30 days (n ≥10 for 4T1; n ≥5 for EMT6). (B) CTL proportions were assessed by flow cytometry in tumors collected at days 4, 8, and 14 after treatment (n ≥5/group). (C) At day 8, CTLs were visualized by IHC and quantified using QPath (scale bars: 200 μm; n ≥6 for 4T1; n ≥5 for EMT6). Images are shown again in Supplemental Figure 2C. (D and F) CTL functionality was evaluated by measuring GzmB, TNF-α, and IFN-γ expression in 4T1 (D) and EMT6 (F) tumors. Representative dot plots and mean values are shown. (E and G) Activation/proliferation markers (PD-1, TIM-3, and Ki67) were analyzed in 4T1 (E) and EMT6 (G) tumors using flow cytometry, with representative dot plots at day 8 (left) and mean values (right). (H) PD-L1 expression on CD45⁻ tumor cells was quantified by flow cytometry (n ≥4/group). (I) Immune-related gene expression was assessed by NanoString for total tumor mRNA. Heatmap shows normalized z scores, and volcano plots indicate statistical significance (n = 6/group). Box plots show the mean ± SEM. *P < 0.05, **P < 0.01, and ***P < 0.001, by 2-way ANOVA. orth., orthotopic.

Figure 2. The CDDP-Eri doublet induces ICD markers.

(A) 4T1 cells were treated with Eri (50 nM), CDDP (4 μM), DXR (500 nM), or CDDP-Eri. Drug concentrations correspond to the IC₅₀ at 48 hours. ICD markers were analyzed at 24 hours and 48 hours. Heatmap shows normalized marker expression and statistical significance. (B) Pie charts indicate the proportion of ICD marker positivity across independent experiments (1-way ANOVA). (C) EIF2α phosphorylation (Ser51) and LC3I/II levels were assessed by Western blotting. Heatmaps represent the densitometric phosphorylated/total ratio (1 representative experiment of 2). (D and E) Experimental design. 4T1 tumor–bearing mice received CDDP, Eri, or both. Four days later, tumor cells were isolated for NanoString analysis. Heatmap in D shows normalized gene expression; volcano plot in E indicates statistical significance (n = 4/group); 1-way ANOVA. (F) CD31⁺ endothelial cell proportions were measured by flow cytometry on post-treatment days 4, 8, and 14. Dot plots show CD31⁺ populations at day 8 (n ≥ 6/group). (G) CD31⁺ cells were quantified by IHC using QPath (scale bar: 200 μm). (H) CAF proportions were assessed according to PDPN expression using flow cytometry on day 8. Dot plots represent PDPN⁺ cells (n ≥ 6/group). (I) Correlations between CD31⁺, CAF, and CTL proportions were analyzed (n ≥ 11). (J) CTL proportions were evaluated on the basis of CAF and CD31⁺ levels (> median = high, < median = low). Box plots show the mean ± SEM. *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001, by 2-way ANOVA. Ctrl, control.

Figure 3. Combination therapy is CTL independent and resistant to anti–PD-L1 immunotherapy.

(A) 4T1 tumor–bearing (orth) BALB/c immunocompetent or nude mice received CDDP, Eri, or CDDP-Eri, with or without anti-CD8a or isotype control antibodies. The anti-CD8a antibody was injected twice before and 4 times after chemotherapy. Tumor volume was monitored for at least 3 weeks after treatment (n = at least 4 mice/group). (B) 4T1 tumor–bearing mice received CDDP, Eri, or both, with or without anti–PD-L1 or isotype control antibodies. Tumor volume was monitored for at least 3 weeks after treatment (n = at least 6 mice per group). Box plots show the mean ± SEM. *P < 0.05, ***P < 0.001, and ****P < 0.0001, by 2-way ANOVA. (C and D) 4T1 tumor–bearing mice received CDDP, Eri, or both. Four days later, the 4T1 tumor was recovered, and tumor cells and CD45⁺ cells were isolated by magnetic beads. Total cell mRNA was extracted, and immune-related gene expression was analyzed by NanoString for CD45⁺ TILS (C) and tumor cells (D). The heatmap corresponds to normalized marker expression, and the volcano plot indicates the P value from the statistical analysis (n = 4 mice/group). aCD8, anti-CD8 antibody; aPD-L1, anti–PD-L1 antibody.

Figure 4. The CDDP-Eri doublet induces TGF-β–mediated immunoresistance.

(A) 4T1 tumor–bearing mice were treated with CDDP, Eri, or both. Tumors were collected at days 4, 8, and 14. At day 8, total tumor mRNA was analyzed for immunosuppressive gene expression by NanoString. The heatmap on the left shows normalized expression, and the heatmap on the right shows statistical significance (n = 4/group). P values were determined by 2-way ANOVA. (B) Tgfb1 expression was measured by qPCR at days (D) 4, 8, and 14 (n ≥6/group). (C and D) 4T1 cells were treated with Eri (50 nM), CDDP (4 μM), CDDP-Eri, or left untreated. Tgfb1 (C) and Serpin1 (D) expression was analyzed by qPCR at 24 hours and 48 hours (2 experiments, n = 3/experiment). (E) 4T1 cells were pretreated with galunisertib (10, 100, and 1,000 nM) for 2 hours, and then treated with CDDP (4 μM), TGF-β (2 ng/mL), or left untreated for 48 hours. Serpin1 expression was analyzed by qPCR (2 experiments, n = 2/experiment). (F) Under the same conditions as in C, Smad2/3 phosphorylation [pSmad2 (Ser465/467), pSmad3 (Ser423/425)] was analyzed by Western blotting. Heatmap represents the phosphorylated/total ratio (results are from 1 of 2 experiments). (G and H) 4T1 tumor–bearing mice were treated as in A. At day 4, immunosuppressive cells were sorted, and Tgfb1 expression was measured by qPCR (G). MyCAF-related genes were analyzed in CAFs (H) (n ≥3/group). (I) At day 8, αSMA⁺ CAF proportions were measured by flow cytometry. (J and K) αSMA⁺ cells were quantified by IHC (J), and collagen deposition was analyzed by Masson’s trichrome (MTC) staining (K) (n ≥7/group). Images are shown again in Supplemental Figure 4I. (L–N) Mo-MDSC (L), TAM2 (M), and Treg (N) proportions were assessed by flow cytometry (n = 5/group). Box plots represent the mean ± SEM. *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001, by 1-way ANOVA (C–E) and 2-way ANOVA (B, G, H, and L–N).

Figure 5. TGF–β blockade restores chemoimmunotherapy efficacy.

(A–C) 4T1 (A), EMT6 (B), and TS/A (C) tumor–bearing mice were treated with the CDDP-Eri doublet, anti–PD-L1 mAb, and anti–TGF-β mAb, or the combination of the 4 molecules, or left untreated. Tumor volume was monitored (schema). Data represent the mean ± SEM. Mouse survival was evaluated (n = at least 5 mice/group). Data represent the median; log-rank test. (D) MMTV-PyMT mice were treated with CDDP-Eri doublet, anti–PD-L1, and anti–TGF-β mAbs, or the combination of the 4 molecules, or left untreated at the appearance of the first tumor target lesion. Mouse survival was evaluated for 120 days (control: n = 13, CDDP-Eri: n = 7, anti–PD-L1/anti–TGF-β: n = 16, quadritherapy: n = 10). Data represent the median; log-rank test. The number of tumors was assessed for each mouse at day 29 after the different chemotherapy treatments. (E and F) 4T1 (E) and EMT6 (F) tumor–bearing mice were treated with CDDP-Eri and CDDP-Eri plus anti–PD-L1 with or without anti–TGF-β mAbs with or without CTL depletion. Tumor volume was monitored for at least 4 weeks after treatment (n = at least 3 mice/group). Box plots show the mean ± SEM. *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001, by 2-way ANOVA.

Figure 6. TGF-β blockade reduces immunosuppression induced by chemotherapy combination.

(A) 4T1 tumor–bearing mice were treated with CDDP-Eri doublet, anti–TGF-β, anti–PD-L1, or a combination of these treatments. Tumors were recovered 8 days later. Total tumor mRNA was extracted, and immunosuppressive pathway gene expression was analyzed by NanoString. The heatmap on the left corresponds to normalized marker expression and the heatmap on the right indicates the P value from statistical analysis (at least n = 3 mice/group). Statistical significance was determined by 2-way ANOVA. (B–L) 4T1 tumor–bearing mice were treated as in A. (B) Proportions of TAM2s among total living cells were measured by flow cytometry. (C) Total CAFs were isolated using magnetic activated cell sorting (MACS), and MyCAF-related gene expression was analyzed by qPCR. (D) αSMA levels were measured by IHC in the tumor and automatically quantified with QPath software and (E) intratumoral collagen fibrosis fiber deposits were analyzed by Masson’s trichrome staining. Scale bars: 200 μm and 50 μm (enlarged insets). (F) Total tumor mRNA was extracted, and immune-related gene expression was analyzed by NanoString. The heatmap on the left corresponds to normalized marker expression, and the heatmap on the right indicates the P value from statistical analysis (2-way ANOVA). (G) Proportions of CTLs among total living cells were measured by flow cytometry. (H) CD8 was measured by IHC and automatically quantified with QPath software (scale bars: 200 μm). The CD8/TAM2 ratio (I) and CD8/CAF ratio (J) in tumors were calculated. (K) CD8⁺ activation was evaluated by analysis of PD-1, TIM-3, and Ki67 markers, and (L) the functionality of CD8⁺ was evaluated by analysis of GzmB, TNF-α, and IFN-γ expression. n = 5 mice/group. Data represent the mean ± SEM. *P < 0.05, **P < 0.01, ***P < 0.0001, and ****P < 0.0001, by 2-way ANOVA.

Figure 7. The CDDP-Eri combination induces TGF-β in human BC cell lines.

(A) Box plots of TGFB1 gene expression across standard pathological classifications (ER⁺HER2^–, HER2⁺, ER^–HER2^–) and normal breast tissue. (B) TGFB1-related metagene expression in TNBC subtypes (BLIA, BLIS, LAR, MES) from the Burstein classification. (C) Kaplan-Meier curves for PFI based on TGFB1 metagene expression. Green: low expression; red: high expression. Ticks denote censored data. (D) Heatmap of correlations between TGFB1 gene, TGFB1 metagene, and immune signatures (IFNG, CXCL9, STAT1, TILs, HLA, DRB1, CXCL10, IDO1) in TNBC. (E) Forest plots showing HRs for immune signatures and the TGFB1 metagene in relation to PFI (left). (F) Kaplan-Meier curves for PFI based on CYTOX and TGFB1 metagene expression in TCGA. Gray: CYTOX^loTGFB1^lo; red: CYTOX^loTGFB1^hi; green: CYTOX^hiTGFB1^lo; black: CYTOX^hiTGFB1^hi. (G) Kaplan-Meier curves for OS based on CYTOX and TGFB1 metagene expression in METABRIC (same color coding as in F). (H) Bar and pie charts depicting the distribution of patients among TNBC subtypes (BLIA, BLIS, LAR, MES) based on CYTOX-TGFB1 metagene combinations. (I) TNBC cell lines (MFM-223, HCC38, MDA-MB-231, MDA-MB-468, HCC1937, BT-549, DU4475) were treated with Eri, CDDP, or both (IC₅₀ at 48 hours). TGFB1 expression was measured by qPCR (3 independent experiments, n = 3/experiment). (J) MDA-MB-468 cells were treated with CDDP-Eri, and supernatant was applied to human macrophages and CCD-Lu19 fibroblasts pretreated with galunisertib (100 nM, 24 hours). SERPIN1 expression was analyzed by qPCR. *P < 0.05, **P < 0.01, ***P < 0.0001, and ****P < 0.0001, by 1-way ANOVA.