A BMP7 Variant Inhibits Tumor Angiogenesis In Vitro and In Vivo through Direct Modulation of Endothelial Cell Biology

Abstract

Bone morphogenetic proteins (BMPs), members of the TGF-β superfamily, have numerous biological activities including control of growth, differentiation, and vascular development. Using an in vitro co-culture endothelial cord formation assay, we investigated the role of a BMP7 variant (BMP7v) in VEGF, bFGF, and tumor-driven angiogenesis. BMP7v treatment led to disruption of neo-endothelial cord formation and regression of existing VEGF and bFGF cords in vitro. Using a series of tumor cell models capable of driving angiogenesis in vitro, BMP7v treatment completely blocked cord formation. Pre-treatment of endothelial cells with BMP7v significantly reduced their cord forming ability, indicating a direct effect on endothelial cell function. BMP7v activated the canonical SMAD signaling pathway in endothelial cells but targeted gene knockdown using shRNA directed against SMAD4 suggests this pathway is not required to mediate the anti-angiogenic effect. In contrast to SMAD activation, BMP7v selectively decreased ERK and AKT activation, significantly decreased endothelial cell migration and down-regulated expression of critical RTKs involved in VEGF and FGF angiogenic signaling, VEGFR2 and FGFR1 respectively. Importantly, in an in vivo angiogenic plug assay that serves as a measurement of angiogenesis, BMP7v significantly decreased hemoglobin content indicating inhibition of neoangiogenesis. In addition, BMP7v significantly decreased angiogenesis in glioblastoma stem-like cell (GSLC) Matrigel plugs and significantly impaired in vivo growth of a GSLC xenograft with a concomitant reduction in microvessel density. These data support BMP7v as a potent anti-angiogenic molecule that is effective in the context of tumor angiogenesis.

Fig 1. BMP7v reduced VEGF, bFGF, and Matrigel-driven cord formation.

(A) The ADSC/ECFC co-culture was unstimulated (basal) or stimulated with 10 ng/ml VEGF or bFGF and treated simultaneously with PBS or 2 nM BMP4, BMP7, BMP7v, BMP9, or BMP10 for 96 hours prior to immunohistochemistry for CD31 (green), α-smooth muscle actin (red), and Hoechst 33342 to stain all nuclei (blue). Representative images (5X magnification) are shown, BMP7, BMP7v, BMP9, and BMP10 demonstrated statistically significant (***, p<0.001) differences in connected tube area compared to PBS controls. (B) HUVECs were pre-treated for 24, 48, or 72 hours with PBS, 2 nM BMP4, or 2 nM BMP7v then plated onto Matrigel for 5 hours. Representative images (10X magnification) are shown, graphs represent mean ± SEM from three independent experiments, and asterisks denote statistically significant (***, p<0.001) differences compared to PBS controls. (C) ECFCs were pre-treated with PBS or 100 ng/ml BMP7v for 24 hours then plated into the cord formation assay. The ADSC/ECFC co-culture was then unstimulated (basal) or stimulated with 10 ng/ml VEGF or bFGF for 96 hours prior to treatment with PBS or 100 ng/ml BMP7v for 72 hours and immunohistochemistry for CD31 (green), α-smooth muscle actin (red), and Hoechst 33342 to stain all nuclei (blue). Representative images (5X magnification) are shown, all samples treated with BMP7v displayed statistically significant (***, p<0.001) differences compared to PBS controls. (D) ADSC/ECFC cell mixture in Matrigel was treated with PBS, BMP7v (100 ng or 30 ng), or BMP4 (100 ng) and co-implanted subcutaneously into the flanks of athymic nude mice (8 mice per treatment group). Oral dosing of a subset of mice with sunitinib (25 mg/kg) began 4 hours prior to cell implantation and occurred twice daily. After 5 days, Matrigel plugs were removed and hemoglobin was quantified. Graph is representative of three independent experiments and indicates mean ± SEM from one experiment. Asterisks denote statistically significant (*, p<0.05; **p<0.01) differences compared to vehicle controls.

Fig 2. BMP7v reduced VEGF and bFGF established cords.

(A) The ADSC/ECFC co-culture was unstimulated (basal) or stimulated with 10 ng/ml VEGF or bFGF for 96 hours prior to treatment with PBS or 100 ng/ml BMP7v for 72 hours and immunohistochemistry for CD31 (green), α-smooth muscle actin (red), and Hoechst 33342 to stain all nuclei (blue). Representative images (4X magnification) from three independent experiments are shown, and graphs represent mean connected tube area ± SEM after normalization to growth factor-induced cord values; asterisks denote statistically significant (**, p<0.01) differences compared to PBS controls. (B) The ADSC/ECFC co-culture was unstimulated (basal) or stimulated with 10 ng/ml bFGF for 96 hours prior to treatment with PBS or 100 ng/ml BMP7v for 72 hours and immunohistochemistry for CD31 (green), nidogen or collagen IV (red) and Hoechst 33342 to stain all nuclei (blue). Representative images (5X magnification) are shown.

Fig 3. BMP7v reduced tumor-driven cord formation.

ADSC/ECFC co-cultures with permeable transwells containing media (no cells-basal) or indicated tumor cells (GSLC1, GSLC28, A-2780, SK-OV-3, LXFA-629, MDA-MB-231, PC-3, U-87-MG, and 786–0) were treated with PBS or 100 ng/ml BMP7v for 96 hours prior to immunohistochemistry for CD31 (green), α-smooth muscle actin, (red), and Hoechst 33342 to stain all nuclei (blue). Representative images (5X magnification) are shown. All tumor cell lines treated with BMP7v displayed statistically significant (***, p<0.001) differences in connected tube area compared to PBS controls.

Fig 4. BMP7v reduced VEGFR2 and FGFR1 expression in endothelial cells.

(A) Whole cell protein extracts were isolated following 15 or 60 minute PBS or 100 ng/ml BMP7v treatment and subjected to Western blot analysis using antiserum directed against phospho-SMAD1,5,8 (pSMAD1,5,8), phospho-ERK1/2 (pERK1/2), ERK1/2, and β-actin as a loading control. Results shown are representative of three independent experiments. (B) ECFCs were treated with PBS or 2nM BMP7v or BMP4 in ADSC conditioned defined co-culture media for the times indicated. Whole cell protein extracts were isolated following treatment and subjected to Western blot analysis using antiserum directed against phospho-CRAF (pCRAF), phospho-MEK1/2 (pMEK1/2), phospho-ERK1/2 (pERK1/2), and β-actin as a loading control. (C) Whole cell protein extracts were isolated from ECFCs following 72 hours of PBS or 100 ng/ml BMP7v treatment and subjected to Western blot analysis using antiserum directed against VEGFR2, FGFR1, and β-actin as a loading control. Graph represents mean densitometry ± SEM from three independent experiments, and asterisks denote statistically significant (***, p<0.001) differences compared to PBS controls. (D) The ADSC/ECFC co-culture was unstimulated (basal) or stimulated with 10 ng/ml VEGF or bFGF for 96 hours prior to treatment with PBS or 100 ng/ml BMP7v for 72 hours prior to immunohistochemistry for VEGFR2 (VEGF) or FGFR1 (bFGF) (red) and Hoechst 33342 to stain all nuclei (blue). Representative images (5X magnification) are shown. Upon BMP7v treatment, mean staining intensity for VEGFR2 and FGFR1 was statistically decreased (***, p<0.001) compared to PBS controls.

Fig 5. BMP7v inhibited cord formation in a SMAD4-independent manner.

(A) Whole cell protein extracts were isolated from ECFCs following non-target (NT) or SMAD4 shRNA treatment and stable selection with puromycin. Extracts were subjected to Western blot analysis using antiserum directed against SMAD4 and β-actin as a loading control. Graph represents mean densitometry ± SEM from three independent experiments, and asterisks denote statistically significant (***, p<0.001) differences compared to PBS controls. (B) ECFCs were treated with non-target (NT) or SMAD4 shRNA followed by stable selection prior to plating into the ADSC/ECFC co-culture. The co-culture was stimulated with 10 ng/ml VEGF or bFGF simultaneously with PBS or 100 ng/ml BMP7v for 72 hours prior to immunohistochemistry for CD31 (green), α-smooth muscle actin (red), and Hoechst 33342 to stain all nuclei (blue). Representative images (5X magnification) are shown, graphs represent mean connected tube area ± SEM after basal cord formation data was subtracted from three independent experiments, and asterisks denote statistically significant (***, p<0.001) differences compared to PBS controls.

Fig 6. BMP7v reduced vascularization in vivo.

(A) Assessment of vascularity with immunofluorescence microscopy in Matrigel implants containing either GSLC1 or GSLC28 GFP expressing tumor cells (green) from control (CNTR) or BMP7v treated mice. The peripheral and central regions of implants were assessed for CD31 (red) and α-SMA (blue). (B) Graphs represent mean vessel density (number of vessels per microscopic field), total vessel area, and relative vessel area (total vessel area/area of microscopic field) from one experiment, and asterisks denote statistically significant (*, p<0.05) differences compared to controls. (C) Mice developing subcutaneous xenografts in the range of 9–13 mm entered the treatment schedule (vehicle n, 5; BMP7v 1 μg, n, 7; BMP7v 0.1 μg, n, 4). Time point “0” corresponds to the beginning of treatment. Values are expressed as means ± SEM, and arrows depict compound treatment. BMP7v was well tolerated with no apparent loss of body weight or overt signs of toxicity. In vehicle injected mice, the tumors progressively increased their diameter from 9.9 + 0.74 mm (mean + SEM) at Week 1 to 21 + 1.6 mm at Week 5. Statistically significant (p<0.05) decreases in tumor size were observed at the 4 and 5 week measurements with 0.1 μg BMP7v treatment and (p<0.01) at the 2, 3, 4, and 5 week timepoints with 1 μg BMP7v compared to vehicle controls (CNTR). (D) Analysis of CD31 immunostaining from GSLC efficacy study tumors as described in Methods.