Bone morphogenetic protein 7 promotes resistance to immunotherapy

Abstract

Immunotherapies revolutionized cancer treatment by harnessing the immune system to target cancer cells. However, most patients are resistant to immunotherapies and the mechanisms underlying this resistant is still poorly understood. Here, we report that overexpression of BMP7, a member of the TGFB superfamily, represents a mechanism for resistance to anti-PD1 therapy in preclinical models and in patients with disease progression while on immunotherapies. BMP7 secreted by tumor cells acts on macrophages and CD4⁺ T cells in the tumor microenvironment, inhibiting MAPK14 expression and impairing pro-inflammatory responses. Knockdown of BMP7 or its neutralization via follistatin in combination with anti-PD1 re-sensitizes resistant tumors to immunotherapies. Thus, we identify the BMP7 signaling pathway as a potential immunotherapeutic target in cancer.

Fig. 1. BMP7 is upregulated in tumors resistant to immunotherapies.

a, b Reduced representation bisulfite sequencing (RRBS) results from 344SQP (parental) (n = 2 biologically independent samples) and 344SQR (anti-PD1-resistant) (n = 2 biologically independent samples) tumors treated with anti-PD1 (10 mg/kg). a Percentages of CpG sites methylation and b heatmap of top 10 hypomethylated (green) and 10 hypermethylated (red) genes in 344SQP (n = 2 biologically independent samples) and 344SQR (n = 2 biologically independent samples) tumors treated with anti-PD1(10 mg/kg). The methylation percentages for CpG sites were calculated by the bismark_methylation_extractor script from Bismark and an in-house Perl script. the differential methylation on CpG sites was statistically assessed by R/Bioconductor package methylKit (version 0.9.5). The CpG sites with read coverage ≥20 in all the samples were qualified for the test. The significance of differential methylation on gene level was calculated using Stouffer’s z score method by combining all the qualified CpG sites inside each gene’s promoter region (defined as −1000bp to +500 of TSS), and was corrected to FDR by Benjamini & Hochberg (BH) method. c Pyrosequencing methylation assay with specific primers for BMP7 CpG in 344SQP (n = 2 biologically independent samples with three technical replicates for each sample) and 344SQR (n = 2 biologically independent samples with three technical replicates for each sample) tumors treated with anti-PD1(10 mg/kg). Box-and-whisker plots show the minimum and maximum values. d Quantitative polymerase chain reaction (PCR) analysis of BMP7 expression in 344SQP (n = 4 biologically independent samples with two technical replicates for each sample) and 344SQR (n = 5 biologically independent samples with two technical replicates for each sample) tumors treated with anti-PD1 (10 mg/kg). ACTB expression was used as a housekeeping gene for quantitative PCR analysis. The comparative Ct method was used to calculate the relative abundance of mRNAs compared with ACTB expression. Box-and-whisker plots show the minimum and maximum. **p = 0.0159, two-sided Mann–Whitney test. e Enzyme-linked immunosorbent assay of BMP7 levels in serum from mice bearing 344SQR (n = 3 biologically independent samples with two technical replicates for each sample) or 344SQP (n = 3 biologically independent samples with two technical replicates for each sample) tumors treated with anti-PD1(10 mg/kg). Box-and-whisker plots show the minimum and maximum. **p = 0.0080 unpaired, two-sided t tests. f Enzyme-linked immunosorbent assay for BMP7 in pretreatment plasma collected from patients with disease that progressed (PD) (n = 5 biologically independent samples) on pembrolizumab (NCT02444741; NCT02402920) versus patients with progressive response (PR) or stable disease (SD) (n = 4 biologically independent samples). Box-and-whisker plots show the minimum and maximum. *p = 0.0317, two-sided Mann–Whitney test. g Representative images of immunohistochemical stains of BMP7 expression in formalin-fixed paraffin-embedded tissue sections from 344SQP and 344SQR tumors treated with anti-PD1. Data shown are representative of two reproducible independent experiments. Scale bar, 100 μm (×40 magnification). h–j Representative images of immunohistochemical stains of BMP7 expression in formalin-fixed paraffin-embedded tissue sections collected from patients with NSCLC, adrenocortical carcinoma, and mixed Mullerian carcinoma before and at the time of disease progression on pembrolizumab or ipimilumab. Data shown are representative of two reproducible independent experiments. Scale bar, 100 μm (×40magnification).

Fig. 2. BMP7 modulates MAPK14 in anti-PD1-resistant tumors and immune cells.

a Reverse phase protein array (RPPA) results on expression levels and activation status of 243 proteins in 344SQP (n = 3 biologically independent samples) and 344SQR (n = 3 biologically independent samples) tumors treated with anti-PD1(10 mg/kg). Normalized data were first log2-transformed (log2(x + 1)). Proteins expressed at different levels between groups (downregulated proteins in green and upregulated proteins in red) were identified by a P value of <0.05 obtained from LIMMA’s moderated t statistic (MAPK14, *p = 0.0107). b Representative images of immunohistochemical stains for MAPK14, SMAD1/5/9 phosphorylation, and SMAD1 in formalin-fixed paraffin-embedded tissue sections from 344SQP and 344SQR tumors treated with IgG (10 mg/kg). Scale bar, 100 μm (×40 magnification). Data shown are representative of two reproducible independent experiments of  three biologically independent samples. c, Representative images of immunohistochemical stains for MAPK14, SMAD1/5/9 phosphorylation, and SMAD1 in formalin-fixed paraffin-embedded tissue sections from BMP7-knockdown tumors treated with IgG (10 mg/kg) compared with control (scale bar, 100 μm) (×40 magnification). Data shown are representative of two reproducible independent experiments of three biologically independent samples. d Representative images of immunohistochemical stains for MAPK14, SMAD1/5/9 phosphorylation, and SMAD1 in formalin-fixed paraffin-embedded tissue sections from 344SQP and 344SQR tumors treated with anti-PD1(10 mg/kg). Scale bar, 100 μm (×40 magnification). Data shown are representative of two reproducible independent experiments of three biologically independent samples. e Representative images of immunohistochemical stains for MAPK14, SMAD1/5/9 phosphorylation, and SMAD1 in formalin-fixed paraffin-embedded tissue sections from BMP7-knockdown tumors treated with anti-PD1(10 mg/kg) compared with control (scale bar, 100 μm) (×40 magnification). Data shown are representative of two reproducible independent experiments of three biologically independent samples. f Representative images of immunohistochemical stains for MAPK14 and SMAD1/5/9 phosphorylation in formalin-fixed paraffin-embedded tissue sections from two patients collected before treatment and at the time of disease progression on pembrolizumab or ipimilumab. Scale bar, 100 μm (×40 magnification). Data shown are representative of two reproducible independent experiments. g Nanostring immune panel results for 770 genes in tumor-infiltrating leukocytes (TILs) collected from 344SQP (n = 2 biologically independent samples) and 344SQR (n = 3 biologically independent samples) tumors treated with anti-PD1(10 mg/kg). Genes expressed at different levels between groups (downregulated genes in green and upregulated genes in red) were identified by a P value of <0.05 obtained from LIMMA’s moderated t statistic (MAPK14, p = 0.0047). h Immunofluorescence analysis of MAPK14 and SMAD1/5/9 phosphorylation (green) in the macrophage cell line RAW 264.7 at 24 h after treatment with BMP7 (250 ng) or BMP7 plus follistatin (foll) (250 ng). DAPI (blue) was used to stain cellular nucleus. Data shown are representative of three reproducible independent experiments. Scale bar, 100 μm (×40 magnification).

Fig. 3. BMP7 reduced macrophage-mediated proinflammatory signaling via MAPK14.

a, b BMP7-knockdown and –control cells (0.5 × 10⁶) were injected into 129 Sv/Ev mice and treated with IgG (n = 3 biologically independent samples) or anti-PD1 (10 mg/Kg) (n = 2 biologically independent samples) twice a week for 2 weeks. A week after the final IgG and anti-PD1 treatment, tumor-infiltrating leukocytes (TILs) were collected, and expression of MAPK14, IL1A, IL1B, TNF, and CCL5 were analyzed by quantitative PCR. Box-and-whisker plots show the minimum and maximum values. P values are from unpaired, two-sided t tests. Statistical significance was defined as *,P < 0.05, **,P < 0.01, ***P < 0.001, and ****, P < 0.0001. c, d Quantitative PCR of MAPK14, IL1A, IL1B, TNF, and CCL5 expression in RAW 264.7 cells c and peritoneal macrophages d untreated or treated with BMP7 (250 ng) for 24 or 48 h. Box-and-whisker plots show the minimum and maximum values of two independent experiments. e Quantitative PCR of MAPK14, IL1A, IL1B, TNF, and CCL5 expression in RAW 264.7 cells untreated and treated with BMP7 (250 ng) or BMP7 plus follistatin (foll) (250 ng) for 24 h. Box-and-whisker plots show the minimum and maximum values of  two independent experiments. f BMP7 levels in cell culture supernatant from 344SQP, 344SQR, and 344SQR ctrl and 344SQR shBMP7 cells analyzed by enzyme-linked immunosorbent assay. Box-and-whisker plots show the minimum and maximum values of two biologically independent samples. g Quantitative PCR of MAPK14, IL1A, IL1B, TNF, and CCL5 expression in RAW 264.7 cells co-cultured with 344SQP or 344SQR, and 344SQR shBMP7 or 344SQR ctrl cells for 24 h. Box-and-whisker plots show the minimum and maximum values of two independent experiments. h, i Quantitative PCR of MAPK14, IL1A, IL1B, TNF, and CCL5 expression in RAW 264.7 cells and peritoneal macrophages co-cultured with 344SQR cells or 344SQR cells plus follistatin (foll) (250 ng) for 24 or 48 h. CD45 expression was used as a housekeeping gene for quantitative PCR analysis. The comparative Ct method was used to calculate the relative abundance of mRNAs compared with CD45 expression. Box-and-whisker plots show the minimum and maximum values of two independent experiments. P values are from unpaired, two-sided t tests. Statistical significance was defined as *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001.

Fig. 4. BMP7 regulates CD4⁺ T-cell production of IFNG and IL2 via MAPK14.

a Western blotting analysis of MAPK14 and SMAD1/5/9 phosphorylation in CD4⁺ T cells at 60 min after treatment with BMP7 (250 ng) or BMP7 plus follistatin (foll) (250 ng). ACTB expression was used for normalization in western blotting. b, c BMP7-knockdown and –control cells (0.5 × 10⁶) were injected into 129 Sv/Ev mice and treated with IgG (n = 3 biologically independent samples) anti-PD1 (10 mg/Kg) (n = 2 biologically independent samples) twice a week for 2 weeks. A week after the final IgG or anti-PD1 treatment, tumor-infiltrating leukocytes (TILs) were collected, and expression of IFNG, and IL2 were analyzed by quantitative PCR. Box-and-whisker plots show the minimum and maximum values. P values are from unpaired, two-sided t tests. Statistical significance was defined as *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001. d, e Quantitative PCR of MAPK14, IFNG, and IL2 expression in CD4 + T cells co-cultured with 344SQP or 344SQR, and 344SQR shBMP7 or 344SQR ctrl cells for 24 h. Box-and-whisker plots show the minimum and maximum values  of two independent experiments. f Quantitative PCR of MAPK14, IFNG, and IL2 expression in CD4⁺ T cells untreated and treated with BMP7 (250 ng) or BMP7 plus follistatin (foll) (250 ng) for 24 h. Box-and-whisker plots show the minimum and maximum values of two independent experiments. g Quantitative of MAPK14, IFNG, and IL2 expression in CD4⁺ T cells co-cultured with 344SQR cells or 344SQR cells plus follistatin (foll) (250 ng) for 24 h. Box-and-whisker plots show the minimum and maximum values of two independent experiments. CD45 expression was used as a housekeeping gene for quantitative PCR analysis. The comparative Ct method was used to calculate the relative abundance of mRNAs compared with CD45 expression. P values are from unpaired, two-sided t tests. Statistical significance was defined as *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001.

Fig. 5. BMP7 inhibition re-sensitizes resistant tumors to anti-PD1 therapy.

a Tumor growth and survival analysis of mice with 344SQR tumors treated with IgG ctrl (n = 5 animals) or anti-PD1 (10 mg/kg) (n = 5 animals) or 344SQR shBMP7 tumors treated with IgG (n = 5 animals) or anti-PD1 (10 mg/kg) (n = 5 animals) twice a week for 2 weeks. b Tumor growth and survival analysis of mice with 344SQR tumors (n = 5 animals) treated with IgG, anti-PD1 (10 mg/kg), follistatin (0.1 mg/kg), or follistatin (0.1 mg/kg) plus anti-PD1(10 mg/kg) for 2 weeks. For a, ctrl+ IgG vs. shBMP7 + αPD1, ****p < 0.0001, ctrl+ αPD1 vs. shBMP7 + αPD1, ****p < 0.0001 shBMP7+IgG vs. shBMP7 + αPD1, ****p < 0.0001, Two-way RM ANOVA. For b, IgG vs. foll+αPD1, **p = 0.0060, αPD1 vs. foll+αPD1, ***p = 0.0003, foll+IgG vs. foll+αPD1, ****p < 0.0001, Two-way RM ANOVA. Statistical significance was defined as *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001. Mouse survival rates were analyzed by the Kaplan–Meier method and compared with log-rank tests. c–e Flow cytometry analysis of CD8⁺(*p = 0.0421), CD8⁺IFNG⁺(*p = 0.0475, *p = 0.0121), F4/80⁺CD206⁺, CD4⁺ (*p = 0.0199), CD4⁺IFNG⁺ T cells (*p = 0.0303) in tumor-infiltrating leukocytes (TILs) from 344SQR ctrl (n = 3 biologically independent samples) and 344SQR shBMP7 (n = 3 biologically independent samples) tumors treated with IgG or anti-PD1 (10 mg/kg) twice a week for 2 weeks. Data are presented as mean values ±SD. P values are from unpaired, two-sided t tests. f, g Representative images of immunohistochemical stains for CD206 (M2 macrophage marker) and CD4 (brown dots) in formalin-fixed paraffin-embedded tissue sections from BMP7-knockdown tumors treated with IgG or anti-PD1 compared with control. A representative staining image from each cohort (n = 3 biologically independent samples) is displayed. Data shown are representative of two reproducible independent experiments. Scale bar, 100 μm (×40 magnification). h, i Survival analysis of mice with 344SQR ctrl tumors or 344SQR shBMP7 tumors treated with IgG or anti-PDL1 (10 mg/kg) (n = 5 animals) or anti-CTLA4 (10 mg/kg) (n = 8 animals) twice a week for 2 weeks. Mouse survival rates were analyzed with the Kaplan–Meier method, and curves compared with log-rank tests. Statistical significance was defined as *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001.

Fig. 6. Work model.

Tumors with acquired resistance to immunotherapies secretes BMP7, a member of the TGFβ superfamily, repressing macrophage-mediated inflammatory responses and Th1-associated cytokines in the tumor microenvironment. BMP7 downregulated MAPK14 and MAPK14-regulated cytokines and chemokines including IL1A, IL1B, TNF, and CCL5 via SMAD1 activation in macrophages. At the same time, BMP7 decreased CD4⁺ T-cell activation by downregulating IFNG and IL2 expression via SMAD1/ MAPK14 signaling. Taken together, these findings suggest that BMP7 may represent a potential target for therapeutic approaches to overcome resistance to immunotherapies.