Introduction
Aberrant vascular endothelial growth factor 2 (VEGFR2) signaling and increased VEGF expression has been connected to pathological angiogenesis in various vascular diseases and cancer [
1‐
4]. VEGF-mediated gene transfer has also been used to induce vascular growth in myocardium and skeletal muscle to treat ischemia [
5,
6]. Although next-generation sequencing analyses of VEGF-induced effects in endothelial cells have been performed [
7‐
9], the crosstalk of multiple cell types, and other growth factor pathways regulating VEGF-induced angiogenesis are still poorly understood.
Emerging knowledge supports the role of bone morphogenetic proteins (BMPs) in vascular homeostasis and angiogenesis. Dysfunctional BMP signaling is involved in various vascular disorders, such as hereditary hemorrhagic telangiectasia, cerebral cavernous malformation, pulmonary arterial hypertension and atherosclerosis [
10‐
12]. Concomitantly, BMP2/4, BMP receptors ALK1, ALK2, ALK3, or BMPR2 mouse knockouts lead to severe cardiovascular defects and embryonic lethality [
13‐
16]. Specific BMP members have been shown to either stimulate or inhibit vessel formation. BMP2, -4, -6 and -7 are suggested to be pro-angiogenic, whereas BMP9 and BMP13 have anti-angiogenic effects [
10,
17−
21]. Generally, BMPs bind to two types of receptors: BMPRI (ALK1-3 and ALK6) and BMPRII (incl. BMPR2, ACVR2, ACVR2B), which form a heteromeric complex. Additionally, repulsive guidance molecules (RGMA-C), gremlin 1, BMP and activin membrane-bound inhibitor (BAMBI), and endoglin act as co-receptors [
22,
23]. BMP receptors can initiate Smad signaling cascades, as well as phosphatidylinositol 3-kinase (PI3K) and mitogen-activated protein kinase (MAPK) signaling pathways.
Our findings demonstrate that (i) several BMPs are regulated after systemic VEGF-induced angiogenesis and in normoxic endothelial cells, (ii) BMPs are regulated after acute myocardial ischemia, and in hypoxic endothelial cells, (iii) BMP2 and BMP6 synergistically modulate VEGF-induced endothelial cell sprouting via regulating VEGFR, Notch or TAZ-Hippo signaling, and (iv) BMP6 protein is pro-angiogenic in vivo. Thus, BMPs are potential targets to modulate formation of vasculature in pro- and anti-angiogenic therapies.
Materials and methods
Detailed methods section is available in the Supplemental Methods and Materials in the Major Resources Table.
Materials
Human umbilical vein endothelial cells (HUVECs) were isolated from umbilical cords as previously described [
24], and cultured in Endothelial Cell Growth medium (Promocell, Heidelberg, Germany) on fibronectin-gelatin coated surfaces (10 µg/ml, 0.05%; Sigma-Aldrich, St. Louis, MO). Passage < 6 was used for the experiments. Human lung primary fibroblasts (HPF) were purchased from Promocell and cultured in DMEM (10% FBS). Passage < 11 was used for the experiments. Silencer Select siRNAs against TAZ, TEAD2, BMP2, BMP6 and two control siRNAs were purchased from Thermo Fisher Scientific (Waltham, MA). For detailed information on the materials please see the Major Resources Table in the Supplemental Material.
Experimental animals
Gene transfer experiments were performed with 8–12 weeks old male C57/Bl6 mice (Harlan Laboratories, Indianapolis, IN). Mice were injected via tail vein under isoflurane anesthesia with adenovirus expressing VEGF-A165 under CMV promoter (serotype 5, 1.4 × 1010 vp). Empty adenovirus without a transgene containing only the CMV promoter was used as a control. Mice were sacrificed 6 days after the gene transfer. After PBS perfusion, tissues were harvested and snap frozen in liquid nitrogen for RT-qPCR and RNA-sequencing. For imaging purposes tissues were fixed in 4% PFA in PBS for 4 h, embedded in paraffin and sectioned for immunohistochemical stainings. Matrigel plug angiogenesis assay was performed on 6-week-old Hsd:Athymic Nude-Foxn1nu mice (Envigo, Indianapolis, IN). Mice were injected subcutaneously in their flank with 350 µl of growth factor reduced matrigel (Corning Life Sciences, Tewksbury, MA) containing 1 µM of sphingosine-1-phosphate (:S1P, Enzo Life Sciences, Farmingdale, NY) alone or together with 1.75 µg of human BMP6 recombinant protein (R&D Systems, Minneapolis, MN). Plugs contained 0.25 mg/ml fatty acid free bovine serum albumin (BSA, Biowest, Nuaillé, France) and DMEM with high glucose (Sigma Aldrich) leading to final matrigel protein concentration of 7 mg/ml. Each mouse had a control plug with only S1P and a plug with recombinant protein/s (n = 6 mice/treatment). Mice were sacrificed 7 days later, and the plugs were resected from surrounding tissues and fixed in 4% PFA in PBS for 4 h at RT. The plugs were embedded in paraffin in two parts to create cross-sections of both ends of the plug. The sections were labeled with HE or CD31-antibody. Imaging for the quantitation of cell nuclei area was performed with Leica Thunder 3D Tissue Imager from the whole plug area (× 20 objective) and of CD31-positive area with Eclipse Ni-E Nikon microscope from the plug edge areas (n = 8 images/plug, 20×/0.5 Plan Fluor objectives). Quantitative analysis were performed with NIS-Elements Analysis software.
RT-qPCR
Confluent cultures of HUVECs were washed with PBS, followed by starvation of cells for 16 h with EGM medium supplemented with 0.5% FBS. With siRNA experiments, HUVECs were transfected with 5 or 10 nM siRNA oligonucleotides using oligofectamine for 48 h (Life Technologies, Carlsbad, CA). Protein stimulations were performed with 50 ng/ml (0.5 h, 1 h, 2 h, 4 h, 7 h) or 100 ng/ml (7 h) of VEGF-A165 (VEGF) and 100 ng/ml of BMP2 or BMP6 (7 h; R&D Systems). Tissue samples were collected from VEGF or control virus-treated mice and homogenized with tissue homogenizer (Qiagen, Hilden, Germany). RNA was extracted with RNeasy Mini Kit (Qiagen) or with Tri-reagent (Molecular Research Center, Cincinnati, OH) according to the manufacturer’s instructions. Total RNA was reverse transcribed into cDNA using random hexamers and RevertAID reverse transcriptase (Thermo Fisher Scientific). Quantitative measurements of mRNA levels were performed using the Assays-on-Demand gene expression products (please see the Major Resources Table in the Supplemental Material) with StepOnePlus Real-Time PCR System (Applied Biosystems, Foster City, CA). Amplification of beta-2 microglobulin (B2M) was used as an endogenous control with human endothelial cells and peptidylprolyl isomerase A (PPIA) with mouse tissues. siRNA transfection efficiencies were 67.8% for TAZ, 74.6% for TEAD2, 68.3% for BMP2 and 83.0% for BMP6 detected by RT-qPCR. Decreased expression of TAZ was accordingly detected by western blot (Supplementary Fig. 4c).
NGS experiments
For detailed information see Supplemental Materials and Methods. Briefly, RNA was extracted from the mice liver tissue at d6. After enrichment, RNA was fragmented and purified. Poly(A)-tailing and cDNA synthesis was performed as described [
25]. For reverse transcription an oligo allowing custom barcoding during the final amplification was used. Exonuclease I (New England Biolabs, Ipswich, MA) was used to catalyze the removal of excess oligos. The DNA–RNA hybrid was further purified (Zymo Research Corporation, Irvine, CA), treated with RNaseH and circularized using CircLigase (Epicentre, Madison, WI). Libraries were amplified, purified and sequenced on HiSeq 2000 according to the manufacturer’s instructions (GeneCore, EMBL, Heidelberg, Germany). RNA-Seq data pre-processing and analysis were performed as described previously [
26].
RNA-Seq was mapped using TopHat (v2.0.7). Each sequencing experiment was normalized to the total of 107 uniquely mapped tags and visualized by preparing custom tracks for the UCSC Genome browser. The following thresholds were used: FDR < 0.05, RPKM > 0.5, log fold changes > 1.0 and < − 1.0. Gene expression tags were normalized, log transformed and centered to − 1 to 1 prior to clustering. Clustering and heatmaps were generated with Cluster 3.0 and Java Treeview softwares using hierarchical clustering with Euclidean distance for both genes and arrays as a similarity metric. Average linkage was used as a clustering method. Gene ontology analysis was performed using ‘findGO.pl’ program in HOMER 4.7 software. For analyzing distal regulatory elements near differentially regulated genes in mice, ENCODE ChIP-Seq data for H3K27ac and DNAse hypersensitive sites for mouse liver were used. Mice RNA-Seq data have been submitted to NCBI Gene Expression Omnibus under accession number GSE82106. A summary of the NGS samples used in the analysis and lists of genes in heatmaps are found in Supplemental Files: NGS experiments A–E.
Available global run-on sequencing (GRO-seq) data from (1) normoxic and hypoxic HUVECs (GEO: GSE94872) and (2) pig heart ischemia samples (GEO: GSE81155) were used for analysis. Single-cell sequencing data from mouse tissues are publicly available at tabula muris.ds.czbiohub.org/ [
27].
Immunohistochemistry
Paraffin embedded tissue samples were sectioned (4–10 µm), stained with primary antibodies CD31, VEGF-A, BMP2 and TAZ (please see the Major Resources Table in the Supplemental Material for details) and counterstained with Harris Hematoxylin. Imaging was performed using LSM700 Zeiss confocal microscope or Eclipse Ni-E Nikon microscope (10 ×/0.3 or 20 ×/0.5 Plan Fluor objectives, 4908 × 3264 frame size). In confocal microscope, 405/488/555 nm diode lasers were used together with the appropriate emission filters (10 ×/0.3 or 20 ×/0.5 PlanApo objectives, 512 × 512 and 1024 × 1024 frame sizes). Image processing was performed by ImageJ [
28] and quantitative analysis by NIS-Elements (
n = 4–6 images/tissue/animal).
Enzyme-linked immunosorbent assay
Plasma and liver VEGF concentrations were measured using ELISA for human VEGF (R&D Systems) according to manufacturer’s instructions.
Angiogenesis bead assay
A 3D in vitro model mimicking angiogenesis was performed as described previously [
29]. Shortly, HUVECs (p3) were seeded on top of collagen-coated Cytodex 3 beads (GE Healthcare, Little Chalfont, UK) and cultured in a fibrin gel (fibrinogen, aprotinin, thrombin; Merck KGaA, Darmstadt, Germany). HPF cells were cultured on top of the fibrin gel. Cell culture was maintained with EBM media with additives (Lonza, Basel, Switzerland), and stimulated with VEGF and/or BMP2/4/6 recombinant proteins. Media and protein stimulations (100 ng/ml) were replaced every other day during the 3–7 days follow up. After fixation, the cells were labeled with Phalloidin-A635 (Thermo Fisher Scientific), VEGFR2 antibody (Cell Signaling Technology, Danvers, MA), and DAPI. Images were taken with confocal laser scanning microscope (Zeiss LSM800). 405/555 nm diode lasers were used together with the appropriate emission filters (10 ×/0.3 PlanApo objective, 1024 × 1024 frame size). Image processing and analysis was performed by ImageJ [
28] or Angiosys softwares (Cellworks, Caltag Medsystems Ltd., Buckingham, UK). The angiogenic sprout analysis was performed from endothelial sprouts containing > 1 nuclei (29–43 beads/group). Segmented area of endothelial cells was detected by Angiosys.
HUVECs were seeded on 6-well plates and transfected with 5 or 10 nM siRNAs. After 24 h, cells were detached and transferred to growth factor reduced matrigel (Corning, Inc., New York, NY) coated 48-well plates (40,000 or 50,000 cells/well). Cells were imaged with IncuCyte® S3 Live-Cell Analysis System (Sartorius, Göttingen, Germany) or Olympus IX71 microscope (Tokyo, Japan) using a × 4 objective lens. After 16 h, the cells were fixed with 1% glutaraldehyde–2% PFA solution. Tube formation was analyzed with ImageJ Angiogenesis Analyzer.
CyQUANT cell proliferation assay
HUVECs were seeded on 96-well plates at 4000 cells/well and transfected with 10 nM siRNAs for 48 h. Detection of cellular DNA by CyQUANT cell proliferation assay was performed according to manufacturer’s instructions using absorbance of 530 nm (Thermo Fisher Scientific).
Western blot
Confluent cultures of HUVECs were washed with PBS, followed by starvation of cells for 16 h with EGM medium supplemented with 0.5% FBS. For whole cell protein extraction cells were either transfected with 5 nM siRNAs (siCTRL, siBMP2, siBMP6 or siTAZ) or treated with BMP6 (100 ng/ml; 7 h, 10 h, 14 h, 24 h) and after lysed with Tris–HCl buffer (50 nM Tris–HCl, 150 nM NaCl, 1 mM EDTA, 1% Triton X-100, Na-deoxycholate, 0.1% SDS, 10% glycerol). For cell compartment fragmentation HUVECs treated with BMP2 or BMP6 (100 ng/ml; 1 h, 2 h) were harvested, and cytoplasmic and nuclear proteins were extracted with NE-PER kit (Thermo Fisher Scientific) according to manufacturer’s instructions. Protease and phosphatase inhibitors (Roche, Basel, Switzerland) were added to lysing reagents. After determining the protein concentrations with Pierce™ BCA kit (Thermo Fisher Scientific) equal amounts of proteins were loaded on the gel from each sample (5 or 10 µg). Primary antibodies: YAP/TAZ, phospho-TAZ, VEGFR2, histone H3, β-actin (Cell Signaling Technology) were used. Horse radish peroxidase (HRP) conjugates were used as secondary antibodies. Detection of antigen–antibody complexes was performed with Pierce™ ECL western blotting substrate (Thermo Fisher Scientific) and ChemiDoc™ MP Imaging System (Bio-Rad, Hercules, CA). Quantitative analysis of adjusted volume intensities of the protein bands from immunoblots was performed by ImageLab 6.0 software. Intensity values were normalized to adjusted volume intensities of both loading control and untreated sample.
Statistical analysis
Statistical analyses were performed with GraphPad Prism software (San Diego, CA). Mann–Whitney U-test, One-way ANOVA followed by Dunnett’s multiple comparison test (*-indication) or Unpaired t-test (#-indication) were used. Ρ < 0.05 was used to define statistical significance.
Ethics statement
Animal experiments were approved by National Experimental Animal Board of Finland and carried out in accordance with guidelines of the Finnish Act on Animal Experimentation. Collection of umbilical cords for cell isolation was approved by Ethics Committee of the Kuopio University Hospital (Kuopio, Finland, 341/2015).
Discussion
BMP family members are important regulators of both vascular homeostasis and angiogenesis. Synergistic effect of VEGF and BMPs on vasculature have been previously detected in bone formation [
53] but their role in angiogenesis, particularly crosstalk with VEGFR2 signaling has remained elusive. Our data demonstrate that BMPs are widely expressed in endothelium of various tissues in hypoxia or normoxia and after VEGF-induced angiogenesis, and that BMP2 and BMP6 regulate VEGFR and Notch signaling. BMP6 was further demonstrated to induce neovessel formation in vivo. This is the first comprehensive data on BMPs in hypoxia, and in angiogenesis in various animal models.
Previously, BMPs have been connected to vascular development including endothelial cell differentiation and venous specification, and to various vascular disorders. For example, in hypoxia-induced pulmonary hypertension, BMP2 expression is increased, leading to upregulation of eNOS, as well as induction of endothelial cell survival and motility via Wnt pathways [
54]. Increased BMP6 or BMP2 expression have also been demonstrated in cerebral cavernous malformations and cancer [
11,
55,
56]. We show here that VEGF directly regulates transcription of BMP family members -2 and -4 in endothelial cells. Various BMP members, including BMP4 and BMP6 were also regulated in hypoxic endothelial cells. Both BMP2 and BMP6 modulated transcription of VEGFR2 and DLL4 mRNAs, thus regulating VEGF binding to its receptor and tip cell formation. BMP2 was shown to modify the endothelial sprout front, and to increase VEGF-mediated endothelial sprouting. Crosstalk of BMP2 and VEGF in regulating angiogenesis in vivo has been previously published. Synergistic effect of VEGF and BMP2 in angiogenesis was detected in a rabbit model using porous titanium scaffolds loaded with the growth factors [
57]. In a xenograft model of hepatoma carcinoma cells, overexpression of BMP2 by virus vectors also increased VEGF transcription and angiogenesis [
58]. As we demonstrate that BMP2 is upregulated after VEGF delivery into liver or primary endothelial cells, and that BMP2 modulates VEGFR2 expression and VEGF-mediated endothelial sprouting, our findings support the role of BMP2 in adjusting VEGF-mediated signaling. Since BMP2 was previously suggested to regulate lateral branching of neovessels [
59], BMP2 signaling may act as fine-tuning mechanism in VEGF-mediated angiogenesis.
Besides BMP2, other BMPs have been linked to VEGF-mediated angiogenesis prior to this study. Decreased VEGFR2 expression and VE-cadherin internalization was reported with BMP13, leading to stabilization of adherens junctions and increased vascular integrity [
21]. BMP4 instead was shown to induce expression and phosphorylation of VEGFR2 [
60], and BMP9 to reduce VEGF-mediated angiogenic events in bone-explant angiogenesis assay via an unknown mechanism [
61]. In contrast to BMP2, in our study BMP6 showed anti-angiogenic properties in primary endothelial cells, and time-point dependent fluctuation of VEGFR2 expression. To our surprise, instead of anti-angiogenic effects, BMP6 induced neovessel formation and cell infiltration in vivo. The pro-angiogenic properties of BMP6 protein in vivo has not been previously published. We hypothesize that the angiogenic effect caused by BMP6 occurs due to crosstalk with multiple cell types, as these were observed in the plugs, and is thus context dependent. Modulation of angiogenesis by fibroblasts and innate immune cells including macrophages, dendritic cells and mast cells has been previously demonstrated [
62,
63]. As BMP6 receptor ALK2 is expressed in multiple cells types besides endothelial cells e.g. in heart and skeletal muscle, BMP6 signaling and crosstalk with other cell types inducing angiogenesis warrants for further studies.
Recently, BMPR pathway was linked for the first time to dysfunctional Hippo-signaling, though the exact extracellular ligands, interaction mechanisms and end-responses remained unknown [
46,
48,
64]. We show here that BMP6 induces downstream signaling mediated by Hippo pathway. While active, Hippo signaling retains its downstream effectors YAP/TAZ in the cytoplasm, thus preventing expression of its target genes. While the pathway is inactive, YAP/TAZ is able to translocate to nucleus and induce expression of multiple downstream effectors via TEAD transcription factors, SMADs, p63, RUNX, and PAX [
45]. Hippo pathway has previously been shown to regulate multiple cellular functions such as cell proliferation, survival, differentiation, migration and apoptosis. Dysregulation of the pathway has also been linked to cancer metastasis, and to epithelioid hemangioendothelioma [
65,
66]. In our study, we show that BMP6 translocates TAZ to nucleus, and induces the expression of Hippo target genes, such as a transcriptional repressor SNAI2, a known regulator of angiogenesis [
9,
67]. Likewise, TAZ, which activation and expression levels were here regulated by BMP6, has been previously discovered to regulate angiogenic responses in endothelial cells [
48]. Based on our data we hypothesize that BMP6 regulates angiogenesis via Hippo/TAZ downstream factors such as pro-angiogenic growth factor AREG and SNAI2. This is the first time that a BMP family member has been shown to act as a direct mediator of TAZ/Hippo signaling pathway regulating shuttling of TAZ to nucleus. In accordance, ID1, a Hippo target protein, was earlier suggested to be upregulated by BMP6 in microvascular cells [
68]. TAZ/YAP was also demonstrated to regulate BMP4 expression in zebrafish [
69], and indirectly crosstalk with BMP2 signaling pathway [
70]. Additionally, BMP2 has been suggested to induce cytoplasmic retention of YAP [
71]. Thus, increasing data suggest crosstalk between BMPR and Hippo signaling pathways.
So far, the interplay between YAP/TAZ and the major signaling pathways regulating angiogenesis has remained poorly understood. Recently, YAP/TAZ was shown to induce VEGFR2 recycling to cell surface and to regulate VEGF-mediated developmental angiogenesis [
46,
48,
72]. In vitro, overexpression of YAP/TAZ was also shown to repress BMP and Notch target genes, such as BMP targets SMAD6, UNC5B, ID1; and BMP/Notch targets HES1, DLL4 and HEY1 [
48]. Our data demonstrate that regulation of VEGFR2 signaling occurs in part via BMP6/TAZ-Hippo signaling pathway. As YAP and TAZ have shown a differential effect on e.g. adherens junction modulation in endothelial cells, their signaling may differ in various tissue contexts and warrants for further studies.
To conclude, BMP6/Hippo signaling and BMP2 acts as regulators of VEGFR2 signaling pathway. Inhibition of BMP protein expression together with VEGF could be beneficial in the treatment of ocular diseases e.g. age-related macular degeneration, known to express excess amounts of VEGF. As BMP6 upregulation has also been shown to mediate onset and progression of cerebral cavernous malformations by inducing BMP and TGFβ-signaling [
11], Hippo pathway inhibitors may act as potential molecular targets for this disease.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.