Loss of tumor-derived SMAD4 enhances primary tumor growth but not metastasis following BMP4 signalling

Abstract

Background: Bone morphogenetic protein 4 (BMP4) is a potent inhibitor of breast cancer metastasis. However, a tumor-promoting effect of BMP4 is reported in other tumor types, especially when SMAD4 is inactive.

Methods: To assess the requirement for SMAD4 in BMP4-mediated suppression of metastasis, we knocked down SMAD4 in two different breast tumors and enforced SMAD4 expression in a third line with endogenous SMAD4 deletion. In addition, we assessed the requirement for SMAD4 in tumor cell-specific BMP signalling by expression of a constitutively active BMP receptor. Delineation of genes regulated by BMP4 in the presence or absence of SMAD4 was assessed by RNA sequencing and a BMP4-induced gene, MYO1F was assessed for its role in metastasis. Genes regulated by BMP4 and/or SMAD4 were assessed in a publicly available database of gene expression profiles of breast cancer patients.

Results: In the absence of SMAD4, BMP4 promotes primary tumor growth that is accompanied by increased expression of genes associated with DNA replication, cell cycle, and MYC signalling pathways. Despite increased primary tumor growth, BMP4 suppresses metastasis in the absence of tumor cell expression of SMAD4. Consistent with the anti-metastatic activity of BMP4, enforced signalling through the constitutively active receptor in SMAD4 positive tumors that lacked BMP4 expression still suppressed metastasis, but in the absence of SMAD4, the suppression of metastasis was largely prevented. Thus BMP4 is required for suppression of metastasis regardless of tumor SMAD4 status. The BMP4 upregulated gene, MYO1F, was shown to be a potent suppressor of breast cancer metastasis. Gene signature upregulated by BMP4 in the absence of SMAD4 was associated with poor prognosis in breast cancer patients, whereas gene signature upregulated by BMP4 in the presence of SMAD4 was associated with improved prognosis.

Conclusions: BMP4 expression is required for suppression of metastasis regardless of the SMAD4 status of the tumor cells. Since BMP4 is a secreted protein, we conclude that it can act both in an autocrine manner in SMAD4-expressing tumor cells and in a paracrine manner on stromal cells to suppress metastasis. Deletion of SMAD4 from tumor cells does not prevent BMP4 from suppressing metastasis via a paracrine mechanism.

Keywords: BMP4; Breast cancer; MYO1F; Metastasis; Patient outcomes; SMAD4.

Fig. 1

Modification of MDA-MB-231-HM (231-HM) breast cancer cells to investigate canonical and non-canonical BMP4 signalling. a Schematic diagram of canonical and non-canonical BMP4 signalling. b Western blotting validation of enforced BMP4 expression, SMAD1/5/8 signalling and knockdown of SMAD4 using two different short hairpin RNA constructs in 231-HM cells. c Quantitation of secreted BMP4 protein levels in 24-h conditioned medium. n = 3/group, mean ± SEM. d RT-qPCR analysis of the expression of canonical target genes in 231-HM cells with modified levels of BMP4 and/or SMAD4. n = 3/group, mean ± SEM. e Effect of enforced BMP4 expression and/or SMAD4 knockdown on the proliferation of cultured 231-HM cells. 500 cells were seeded on day 0 and proliferation was tracked for 5 days. n = 6/group, mean ± SEM. Statistical analysis was completed using the exponential growth curve equation function in Prism. ns, not significant. f Effect of enforced BMP4 expression and/or SMAD4 knockdown on colony formation of 231-HM cells. 60 cells were seeded on day 0 and colonies were counted on day 22. n = 3/group, mean ± SEM. For bar plots, statistical analysis was completed by Student’s t test. ns, not significant; *, p < 0.05; **, p < 0.01; ***, p < 0.001; ****, p < 0.0001. See also Fig. S1

Fig. 2

Non-canonical BMP4 signalling, but not canonical signalling, promotes tumor growth. a Timeframe for tumor growth and onset of metastatic disease of the 231-HM tumor model. Cells (1,000,000) were injected into the mammary glands of NSG mice. Tumor growth was tracked before resection until approximately 400 mm³. Metastatic burden was assessed 15 days after resection. Created with BioRender.com. b Effect of enforced BMP4 expression on SMAD4-expressing (top) and SMAD4-knockdown (middle and lower panels) 231-HM tumors. n = 15/group, mean ± SEM. c Immunohistochemical analysis of the levels of SMAD4 in resected 231-HM tumors, shown at both low and high magnification in different tumors

Fig. 3

BMP4 suppresses metastasis independent of tumor-intrinsic expression of SMAD4. a Visualisation of luciferase-tagged metastatic lesions in mice using the IVIS Spectrum imaging system. b Representative images of TurboGFP-tagged metastatic lesions in the lungs and livers visualized ex vivo using the Maestro imaging system. c Metastatic burden in the lungs (left panel), livers (middle panel) and spines (right panel) at endpoint, as quantitated by determining the levels of tumor-specific TurboGfp genomic DNA in each organ. n ≥ 9/group, mean ± SEM. Statistical analysis was completed by Student’s t test. ns, not significant; *, p < 0.05; **, p < 0.01; ***, p < 0.001; ****, p < 0.0001. See also Fig. S2

Fig. 4

Tumor cell-specific activation of BMP signalling suppresses metastasis in a SMAD4-dependent manner. a Western blotting validation of enforced expression of a constitutively active type I BMP receptor (caBMPR1a) in SMAD4 expressing and SMAD4 depleted 231-HM cells. b RT-qPCR analysis of the expression of canonical target genes in 231-HM cells with enforced expression of caBMPR1a and/or knockdown of SMAD4. n = 3/group, mean ± SEM. c Effect of enforced expression of caBMPR1a on the growth of SMAD4-expressing (left panel) or SMAD4-knockdown (right panel) 231-HM tumors. n = 9/group, mean ± SEM. d TurboGFP-tagged metastatic lesions in the lungs and livers were visualized using the Maestro imaging system at endpoint. e Metastatic burden in the lungs (left), liver (middle) and spine (right) at endpoint. n = 9/group, mean ± SEM. Statistical analysis was completed by Student’s t test. ns, not significant; *, p < 0.05; **, p < 0.01; ***, p < 0.001; ****, p < 0.0001

Fig. 5

Transcriptomic analysis reveals that non-canonical BMP4 signalling drives DNA replication and cell cycle progression. a Workflow of recovery of 231-HM cells from primary tumors for RNA sequencing. Compared to Fig. 2a, resected tumors were dissociated, and TurboGFP-tagged tumor cells were recovered by FACS for downstream RNA sequencing analysis. Created with BioRender.com. b Heatmap of differentially upregulated and downregulated genes in 231-HM tumor cells with modified expression of BMP4 and/or SMAD4. c XY plot of the correlation between the logFC induced by enforced BMP4 expression in SMAD4-expressing (x axis) and in SMAD4-knockdown (y axis) tumors. Genes that were uniquely upregulated or downregulated by BMP4 in the absence of SMAD4 are coloured in purple. MYO1F that was investigated in Fig. S6 is highlighted in bold. d Gene set enrichment analysis of gene ontology pathways regulated by BMP4 specifically in SMAD4-knockdown tumors compared to SMAD4-expressing tumors. Top 30 pathways by significance are shown. e Enrichment plots illustrating BMP4-induced gene expression changes in the E2F_targets, the G2M_checkpoint and the MYC_targets hallmark pathways in SMAD4-expressing (upper panel) and SMAD4-knockdown (lower panel) tumors. See also Fig. S5

Fig. 6

Prognostic value of canonical and non-canonical BMP4 signalling in breast cancer patients included in the Metabric dataset. a Expression of genes included in the canonical signature in 231-HM tumors with modified expression of BMP4 and/or SMAD4. b Canonical signature score was calculated based on the upregulated genes minus the down-regulated genes from (a). Average signature scores for tumors of different grades in the Metabric dataset are visualized. c Correlation between canonical signature score and overall survival in breast cancer patients. d Expression of genes included in the non-canonical signature in 231-HM tumors with modified expression of BMP4 and/or SMAD4. e Non-canonical signature score was calculated based on the upregulated genes minus the down-regulated genes from (d). Average signature scores for tumors of different grades in the Metabric dataset are visualized. f Correlation between non-canonical signature score and overall survival in breast cancer patients. Statistical analysis was completed by the stat_compare_means function in R for (b) and (e). N.S., not significant; *, p < 0.05; **, p < 0.01; ***, p < 0.001; ****, p < 0.0001. Statistical analysis was completed by the coxph function in R for (c) and (f). HR, hazard ratio; CI, confidence interval; Cox, Cox proportional-hazards model