Inflammation induces endothelial-to-mesenchymal transition and promotes vascular calcification through downregulation of BMPR2.

Endothelial-to-mesenchymal transition (EndMT) has been unveiled as a common cause for a multitude of human pathologies, including cancer and cardiovascular disease. Vascular calcification is a risk factor for ischemic vascular disorders and slowing calcification may reduce mortality in affected patients. The absence of early biomarkers hampers the identification of patients at risk. EndMT and vascular calcification are induced upon cooperation between distinct stimuli, including inflammatory cytokines and transforming growth factor beta (TGF-β) family members. However, how these signaling pathways interplay to promote cell differentiation and eventually vascular calcification is not well understood. Using in vitro and ex vivo analysis in animal models and patient-derived tissues, we have identified that the pro-inflammatory cytokines tumor necrosis factor alpha (TNF-α) and interleukin-1 beta (IL-1β) induce EndMT in human primary aortic endothelial cells, thereby sensitizing them for BMP-9-induced osteogenic differentiation. Downregulation of the BMP type II receptor BMPR2 is a key event in this process. Rather than compromising BMP canonical signal transduction, loss of BMPR2 results in decreased JNK signaling in ECs, thus enhancing BMP-9-induced mineralization. Altogether, our results point at the BMPR2-JNK signaling axis as a key pathway regulating inflammation-induced EndMT and contributing to calcification.

Introduction

Vascular calcification is a prevalent feature in cardiovascular diseases associated with elevated risk of mortality. The underlying cellular and molecular mechanisms involved have been an area of intense study in recent decades. Mineralization of blood vessels occurs through a coordinated crosstalk between different cell types and cytokines, but ultimately relies on osteoblast lineage cells for synthesis and mineralization of the calcified matrix 1. We and others have shown that endothelial cells (ECs) can function as an additional source of osteogenic progenitors in vascular calcification 2, 3, 4. ECs can undergo a process known as endothelial‐to‐mesenchymal transition (EndMT), which involves loss of endothelial features and acquisition of a fibroblast‐like phenotype, eventually leading to cells with osteogenic potential. EndMT is modulated by different extracellular growth factors [transforming growth factor beta (TGF‐β) family ligands, inflammatory cytokines, fibroblast growth factors] and conditions (mechanical stress, hypoxia) (reviewed in ref 5). During the onset and progression of calcified plaques in the aorta, both inflammatory cytokines [such as tumor necrosis factor alpha (TNF‐α) and interleukin‐1 beta (IL‐1β) 6] and TGF‐β family ligands [including the bone morphogenetic proteins (BMPs) 7, 8) coincide in the affected area. Whereas BMP‐2 and BMP‐4 mainly have a paracrine function, BMP‐9 and BMP‐6 are found in the systemic circulation 9. BMP‐9 regulates vascular homeostasis by controlling proliferation 10, angiogenesis 11, permeability 12 and monocyte recruitment 13. Interestingly, BMP‐9 induces heterotopic ossification when expressed ectopically using adenovirus 14. Importantly, administration of BMP antagonists prevents experimental atherosclerosis in preclinical animal models 15, 16, highlighting the relevance of this pathway in the context of atherosclerosis.

TGF‐β signaling is initiated with the oligomerization of receptor complexes at the cell membrane upon ligand binding. BMPs signal via complexes consisting of a type I receptor named activin receptor‐like kinase (ALK)1/2/3/6, and one of three possible type II receptors: BMPR2, activin type II receptor A, or activin type II receptor B (ACVR2A, ACVR2B) 17. Receptor activation induces the phosphorylation of SMAD1/5/8, which form heteromeric complexes with SMAD4 and translocate into the nucleus to regulate gene expression. In osteogenic cells, BMPs induce the transcription of genes related to osteoblast differentiation and activation 18. In addition to this so‐called canonical SMAD signaling pathway, TGF‐β family ligands modulate the activity of different mitogen‐activated protein (MAP) kinases, including ERK (extracellular signal‐regulated kinase), p38, and JNK (c‐Jun N‐terminal protein kinase) in a cell‐specific and context‐dependent fashion 19, 20. Furthermore, how BMP signaling is fine‐tuned by inflammation in EndMT‐derived cells and cells with osteogenic potential is yet to be determined. Whereas some have suggested that inflammatory stimuli inhibit BMP signaling and osteoblast differentiation and activation 21, 22, others have shown that inflammation is needed for osteogenesis 23, 24.

In this study, we investigated the interplay between inflammation and BMP signaling in ECs. We found that TNF‐α and IL‐1β induce EndMT in primary ECs, which become prone to undergo osteogenic differentiation in response to BMP‐9. We identified BMPR2 downregulation as a key event in this process, which leads to decreased JNK activation, thereby enhancing BMP‐9‐induced mineralization. Finally, we supported our findings in a preclinical animal model of atherogenesis, as well as patient‐derived tissues from atherosclerotic donors. Our results provide a better understanding of the molecular mechanisms underlying EndMT and vascular calcification, and may have broad implications for other human inflammatory pathologies involving BMP signaling.

Materials and methods

Cell culture and reagents

Human aortic endothelial cells (HAoECs, CC‐2535, Batch number 0000316662), pulmonary aortic endothelial cells (PAECs, CC‐2530), and coronary microvascular endothelial cells (cMVECs, CC‐7030) were purchased from Lonza (Walkerville, MD, USA). Human skin microvascular endothelial cells (HMECs) have been described elsewhere 25. All cells were cultured in complete EBM‐2 medium (Lonza) on 1% w/v gelatin‐coated wells. Endothelial colony‐forming cells (ECFCs) were isolated as previously described 26 and cultured in complete EBM‐2 medium. All experiments were performed with cells grown to near‐confluency between passage 6 and 8. 2H‐11 cells are derived from endothelial cells isolated from the lymph nodes of adult C3H/HeJ mice transformed using SV40 27, 28. These cells were grown on 1% w/v gelatin‐coated dishes (Merck, Darmstadt, Germany) in DMEM medium (Invitrogen, Breda, The Netherlands) supplemented with 4.5 g/l d‐glucose, 110 mg/l sodium pyruvate, non‐essential amino acids (all from Invitrogen), 10% (v/v) heat‐inactivated fetal bovine serum (FBS; Sigma‐Aldrich Chemie, Steinheim, Germany), 0.5% (v/v) antibiotic/antimyotic solution (Invitrogen), and 2 mm l‐glutamine (Invitrogen).

EndMT assays

HAoECs were seeded at confluence in 12‐well plates (for qPCR) or Lab‐Tek II chamber slides (for immunofluorescence) in EBM2 complete medium containing 2% FBS. Next day, the cells were stimulated with the indicated ligands for 24 h in EBM2 complete medium containing 10% FBS, and endothelial and mesenchymal‐specific markers were analyzed by RT‐qPCR or immunofluorescent labeling.

Osteoblast differentiation assays

Formation of calcium and phosphate deposits was analyzed by Alizarin Red staining (ARS) of cells seeded in 48‐well plates at near‐confluency. HAoECs were pretreated with TNF‐α (10 ng/ml) or TGF‐β3 (5 ng/ml) for 4 days in EBM2 containing 10% FBS. Next, the medium was replaced by osteogenic medium (DMEM containing 10% FBS, 10−8 mol/l dexamethasone, 0.2 mmol/l ascorbic acid, and 10 mmol/l β‐glycerolphosphate) in the presence of BMP‐9 (10 ng/ml) for 14 days. 2H‐11 cells were incubated with the aforementioned osteogenic medium containing both TNF‐α (10 ng/ml) and BMP‐9 (10 ng/ml) for 14 days. The medium was refreshed every 4 days. Afterwards, cells were washed twice with PBS and fixed with 3.7% formaldehyde for 5 min, washed twice with distilled water, and ARS was performed as previously described 29. Precipitates from three independent assays were dissolved using 10% cetylpyridinium chloride and absorbance was measured at 570 nm. Representative pictures were obtained using a Leica DMIL LED microscope with 10× magnification.

Reverse transcription–quantitative PCR

Total RNA was extracted using NucleoSpin RNA II (MACHEREY‐NAGEL, Duren, Germany), 500 ng of RNA were reverse transcribed using RevertAid First Strand cDNA Synthesis Kits (Fermentas/Thermo Fisher Scientific, Freemont, CA, USA) and quantitative PCR (qPCR) was performed using SYBR Green (Bio‐Rad, Veenendaal, The Netherlands) and a Bio‐Rad CFX Connect. The primers used are listed in the supplementary material, Supplementary materials and methods. GAPDH was used for normalization.

Immunofluorescent labeling of cultured cells

HAoECs grown on coverslips were fixed with 4% formaldehyde for 30 min at room temperature, washed with glycine for 5 min, permeabilized with 0.2% Triton X‐100, and blocked in PBS containing 5% BSA for 1 h. The cells were incubated overnight at 4°C with primary antibody in blocking solution with gentle shaking. Next day, the cells were washed five times in washing buffer (PBS containing 0.05% Tween‐20 and 1% BSA) and incubated with secondary antibody (Alexa‐Fluor FITC goat anti‐mouse IgG, Alexa‐Fluor 555 anti‐rabbit IgG; Invitrogen; 1:200) or with phalloidin‐488 (1:100) in PBS with 0.5% BSA for 1 h. Finally, the cells were washed five times in washing buffer and mounted in Prolong Gold containing DAPI (Invitrogen). Preparations were imaged in a Leica SP5 confocal scanning laser microscope. A representative picture from each staining is shown (n = 3). The antibodies used are described in the supplementary material, Supplementary materials and methods.

Western blotting

The cells were seeded into 12‐well plates and cultured until they reached confluence. Next, the cells were stimulated as indicated, washed with cold PBS, and lysed in 2× sample buffer as previously described 30. The protocol details and the antibodies used are provided in the supplementary material, Supplementary materials and methods.

Lentiviral production and transduction

Lentiviral vectors were produced in HEK293T cells as previously described 29. Lentiviral vectors expressing specific shRNAs were obtained from Sigma (MISSION® shRNA; see supplementary materials and methods). Lentiviral vectors to overexpress mMKP‐1 have been described elsewhere 31.

Vein graft procedure

Vein grafts were performed as reported elsewhere 32. Details may be found in the supplementary material, Supplementary materials and methods.

Immunostaining of tissues

Immunofluorescent and immunohistochemical staining on human and murine aortic tissues was performed as previously described 3 (see the supplementary material, Supplementary materials and methods for details). Human aortic sections (4 μm) were prepared from each donor and classified according to the revised classification of the American Heart Association (AHA) 33, 34 by two independent observers with no knowledge of the donor characteristics. Sample collection and handling were performed in accordance with the guidelines of the Medical and Ethical Committee in Leiden, The Netherlands and the code of conduct of the Dutch Federation of Biomedical Scientific Societies.

Transfections, luciferase assays, and DNA constructs

For luciferase reporter assays, the cells were seeded in 24‐well plates and transfected with DharmaFECT Duo (Thermo Fisher Scientific) following the recommendations of the manufacturer. The cells were harvested and lysed 48 h after transfection. Luciferase activity was measured using the luciferase reporter assay system from Promega (Leiden, The Netherlands) by a Perkin Elmer luminometer Victor3 1420. Each transfection mixture was equalized with empty vector when necessary and experiments were performed in triplicate. The BRE‐Luc reporter is reported elsewhere 35. The expression vector encoding the constitutively active fusion protein MKK7–JNK3 and its characterization have been described previously 36.

Iodination and ligand affinity labeling of cell surface receptors

Iodination of BMP‐9 was performed according to the chloramine T method and cells were subsequently affinity labeled with the radioactive ligand as previously described 37. Cell lysates were immunoprecipitated with antibodies against ALK1, ALK2, BMPR2, and ACVR2A. The generation and characterization of these antibodies have been published previously 38, 39, 40. Image quantification was performed by densitometry using ImageJ analysis.

Statistical analysis

All experiments were performed at least three times; results are shown as mean ± SD. Student's t‐test was used for statistical analysis, and +,* p < 0.05, ++,** p < 0.01, +++,*** p < 0.001 were considered significant. Representative data are shown in the figures.

Results

TNF‐α and IL‐1β induce EndMT in HAoECs, thereby enhancing BMP‐induced mineralization

To determine the mechanisms by which ECs contribute to the generation of atherosclerotic plaques, we used primary human coronary aortic endothelial cells (HAoECs) stimulated with the indicated pro‐inflammatory cytokines and members of the TGF‐β superfamily of growth factors. After 24 h, TNF‐α and IL‐1β induced a decrease in the endothelial markers CDH5 (encoding for VE‐cadherin) and PECAM1, and an upregulation of the mesenchymal genes CDH2 (encoding for N‐cadherin) and FN1 (fibronectin) (Figure 1A). Accordingly, these two cytokines upregulated the expression of SNAI1 and TWIST1 (supplementary material, Figure S1). This correlated with loss of the typical endothelial cobblestone‐like phenotype, seen by cytoskeletal staining with phalloidin, and the downregulation of VE‐cadherin, PECAM‐1, and TIE2, and increase in N‐cadherin, αSMA, and vimentin (Figure 1B). Furthermore, pretreatment with TNF‐α sensitized HAoECs to become osteoblast‐like cells in response to BMP‐9, more potently than TGF‐β, as shown by ARS (Figure 1C) and enhanced levels of the osteogenic genes COL1A1 (collagen 1 alpha 1), RUNX2, and OSX (osterix/transcription factor Sp7) (Figure 1D). TNF‐α alone induced BMP7, and the combination of BMP‐9 and TNF‐α upregulated osteogenic BMP2 and BMP4 (supplementary material, Figure S2). These results show that the pro‐inflammatory cytokine TNF‐α induces EndMT in HAoECs and that EndMT‐derived cells exhibit an enhanced osteogenic response to BMP‐9.

Figure 1

The pro‐inflammatory cytokine TNF‐α induces EndMT in HAoECs, priming them for BMP‐9‐induced osteogenic differentiation. (A) Gene expression analysis of CDH5, PECAM1, FN1, and CDH2 in HAoECs stimulated for 24 h with the indicated cytokines and growth factors. CO: control; Tα: TNF‐α (10 ng/ml); IL: IL‐1β (10 ng/ml); LP: LPS (10 ng/ml); Tβ: TGF‐β3 (5 ng/ml); AA: activin A (50 ng/ml); B6: BMP‐6 (50 ng/ml); B9: BMP‐9 (10 ng/ml). Statistical significant difference with respect to control cells (*, **, ***, n.s non significant). (B) Fluorescence staining of HAoECs stimulated with TNF‐α, BMP‐9 or TGF‐β as above: PHA (phalloidin), VECAD (VE‐cadherin), NCAD (N‐cadherin), αSMA, PECAM1, VIM (vimentin), and TIE2. DAPI (blue) was used as a nuclear counterstain. (C) Alizarin Red staining and corresponding quantification on HAoECs incubated with BMP‐9 for 21 days under osteogenic conditions after induction of EndMT with TNF‐α or TGF‐β. GM: growth medium; OM: osteogenic medium. Quantification is shown on the right side. Statistical significant difference with respect to untreated (**) or BMP‐9 (+++) treated cells. (D) qPCR of the indicated osteogenic genes in HAoECs subjected to osteogenic differentiation. Statistical significant difference with respect to untreated (**, ***) or BMP‐9 (+, ++, +++) treated cells.

BMPR2 downregulation is necessary and sufficient for TNF‐α to induce EndMT in HAoECs

BMP‐9 signaling in ECs is mediated by two possible type I receptors, ALK1 and ALK2, and three type II receptors, BMPR2, ACVR2A, and ACVR2B 41. We investigated whether the expression of the BMP type II receptors is modified by stimuli inducing EndMT, thereby enhancing the osteogenic response of BMP‐9. The two potent inducers of EndMT, TNF‐α and IL‐1β, triggered a clear downregulation of BMPR2, whereas ACVR2A and ACVR2B were not affected (Figure 2A). This effect correlated with a decreased mRNA level of BMPR2 (supplementary material, Figure S3) and was not exclusive to HAoECs, as endothelial colony‐forming cells (ECFCs) and, to a lesser extent, human skin microvascular ECs (HMECs), displayed a similar response (supplementary material, Figure S4). Noteworthy, incubation with higher concentrations of TNF‐α led to reduced expression of all three BMP type II receptors (supplementary material, Figure S5).

Figure 2

TNF‐α‐induced downregulation of BMPR2 is necessary for HAoECs to undergo EndMT. (A) Western blot in HAoECs where EndMT has been induced for 24 h (see Figure 1A,B). (B) Western blot in HAoECs transduced with control (pLK) lentivirus or two shRNA constructs (#1 and #2) against BMPR2 (upper panel) or with lentivirus encoding a full‐length version of BMPR2 (BMPR2 FL) and treated with TNF‐α for 24 h (lower panel). (C–E) EndMT analysis by (C) qPCR and (D, E) immunofluorescent staining in HAoECs knocked down for BMPR2 or overexpressing BMPR2 FL, respectively. Statistical significant difference with respect to control or TNF‐α‐treated cells (*, **).

We next investigated whether BMPR2 downregulation is necessary for TNF‐α to induce EndMT in HAoECs, using knockdown and overexpression experiments (Figure 2B). Knockdown of BMPR2 in HAoECs using two different shRNA constructs resulted in reduced CDH5 and PECAM1, and increased CDH2 and FN1 mRNA levels and loss of endothelial morphology. In contrast, ectopic overexpression of a full‐length BMPR2 construct (BMPR2 FL) in HAoECs using lentiviral particles prior to TNF‐α stimulation prevented TNF‐α‐induced EndMT, shown by gene expression analysis (Figure 2C) and immunofluorescence (Figure 2D,E). Altogether, our results show that downregulation of BMPR2 is required for TNF‐α to induce EndMT in HAoECs, and that BMPR2 overexpression can partially prevent EndMT.

BMPR2 is downregulated in ApoE3Leiden mice in response to a high‐fat diet

We next aimed to validate our findings in vivo using an animal model of atherosclerosis. The contribution of EndMT to vascular calcification and atherosclerosis has been investigated in different animal models, including LDLR−/−, ApoE‐deficient, and InsAkita mice 4, 42, 43. The ApoE3Leiden mice constitute a valuable model of diet‐induced atherosclerosis with a human‐like lipoprotein profile when fed a high‐fat diet 30, 39. Furthermore, this model is currently considered as one of the models of native atherosclerosis closest to human pathology. Although some aspects of this model have already been characterized (e.g. a strong inflammatory infiltrate potentiates plaque development), the contribution of EndMT to plaque formation is not known. Therefore, we performed immunofluorescent co‐staining for PECAM‐1 and αSMA on sections from control animals (7 days normal diet vein grafts) and high hypercholesterolemic diet (high‐fat diet, HFD)‐induced atherosclerosis animals (after 7, 14, and 28 days). As previously reported 32, the luminal endothelial layer of the aorta of HFD mice downregulated PECAM‐1 after 7 days, in contrast to control animals (Figure 3A, low magnification). This was accompanied by evident neointimal thickening. Interestingly, we observed PECAM‐1 and αSMA double‐positive cells migrating from the luminal layer into the neointima area (Figure 3A, high magnification). This phenomenon was visible from day 7 to day 28 in HFD animals, with a progressive increase of αSMA in the neointima. Consistent with our in vitro findings, BMPR2 was markedly reduced in the aorta at the onset of atherosclerosis, as early as 7 days after administration of HFD (Figure 2B).

Figure 3

BMPR2 is downregulated in the aorta of ApoE3Leiden mice developing atherosclerotic disease. Aortic sections from ApoE3Leiden mice fed with a normal (Chow) or a high‐fat diet (HFD) for 7, 14 or 28 days and stained for PECAM‐1 and αSMA (A) or BMPR2 and PECAM‐1 (B). A representative picture per condition is shown, including zoomed areas indicated in dashed lines. White arrows indicate PECAM‐1/αSMA double‐positive cells, suggestive of EndMT.

Downregulation of BMPR2 does not compromise BMP‐induced canonical SMAD1/5 signaling

To investigate the mechanisms by which loss of BMPR2 induced by TNF‐α enhanced the osteogenic differentiation of ECs in response to BMP‐9, we characterized an easy‐to‐culture EC line, 2H‐11 murine ECs. First, we demonstrated that BMP‐2, BMP‐6, BMP‐7, and BMP‐9 induce cell calcification in 2H‐11 cells (supplementary material, Figure S6A), which can be blocked by co‐incubation with a BMP type I receptor kinase inhibitor (supplementary material, Figure S6B). Moreover, similar to primary HAoECs, knockdown of Bmpr2 (but not Acvr2A or Acvr2B) enhanced the osteogenic differentiation of 2H‐11 cells induced by BMP‐9 (supplementary material, Figures S7 and S8). On the contrary, stable overexpression of BMPR2 FL partially prevented the osteogenic effect of BMP‐9 (supplementary material, Figure S9). Since one of the main determinants of osteogenic differentiation is the induction of BMP canonical SMAD1/5 signal transduction, we first analyzed whether this was affected by TNF‐α. Although TNF‐α pretreatment resulted in a dose‐dependent increase in osteogenic differentiation of BMP‐9‐induced 2H‐11 cells (Figure 4A), the canonical BMP‐9‐induced transcriptional response (as measured using the BRE‐LUC reporter construct, Figure 4B) was not increased by TNF‐α. Using a previously reported protocol for iodination and ligand affinity labeling of cell surface receptors, we studied the high affinity receptors for BMP‐9 in 2H‐11 cells. As Figure 4C shows, while BMP‐9 mainly forms a receptor complex including BMPR2 in control conditions, upon BMPR2 knockdown, BMP‐9 recruits ACVR2A in a receptor signaling complex, together with ALK1/2 (supplementary material, Figure S10). To investigate whether this affects the canonical signaling response to BMP‐9 in 2H‐11 cells, we performed western blotting in stably knocked down 2H‐11 cells for Bmpr2, Acvr2A or Acvr2B. As expected, single knockdown of any of the type II receptors failed to compromise the BMP‐9‐induced p‐SMAD1/5 response. Interestingly, pretreatment with TNF‐α blocked p‐SMAD1/5 activation exclusively in cells knocked down for ACVR2A (Figure 4D). These results suggest that loss of BMPR2 induced by TNF‐α leads to enhanced recruitment of ACVR2A in the signaling receptor complex to balance downstream canonical signaling in ECs.

Figure 4

TNF‐α‐induced loss of BMPR2 does not compromise BMP‐9 canonical signaling in ECs. (A) Alizarin Red staining and quantification (lower figure) in 2H‐11 cells incubated with BMP‐9 (10 ng/ml) and increasing concentrations of TNF‐α. GM: growth medium. Statistical significant difference with respect to control or BMP‐9/TNF‐α‐treated cells (**, *** or non significant). (B) BMP‐9‐induced canonical transcriptional response in 2H‐11 cells stimulated with BMP‐9 (10 ng/ml) upon pre‐incubation with TNF‐α. Fold induction of untreated control cells is shown. (C) Ligand–receptor interaction assay in 2H‐11 cells stably knocked down for BMPR2 or a control vector (pLK0.1). Quantification is shown below. (D) Western blot of p‐SMAD1/5 and total SMAD1 in 2H‐11 cells stably knocked down for Bmpr2, Acvr2A or Acvr2B and treated with BMP‐9 (10 ng/ml) and TNF‐α (10 ng/ml).

Downregulation of BMPR2 decreases JNK pathway activation to fine‐tune EC mineralization

Since the enhanced osteogenic response to BMP‐9 of ECs lacking BMPR2 was not related to BMP‐9 canonical signaling, we tested the effect of chemical inhibitors of non‐canonical signaling on BMP‐9‐induced osteogenic differentiation. We used SP600125 to inhibit JNK, UO120 against ERK, SB203580 for p38 kinase, and PDTC (ammonium pyrrolidinedithiocarbamate) to block nuclear factor kappa B (NF‐κB) signaling. Among all the inhibitors used, only SP600125 enhanced the osteogenic differentiation of 2H‐11 cells (Figure 5A). We confirmed this using a mutant version of the phosphatase MKP‐1 (mMKP‐1) 31. 2H‐11 cells transduced with mMKP‐1 showed decreased phosphorylation of c‐Jun (p‐c‐Jun), a downstream target of JNK, in response to BMP‐9 (supplementary material, Figure S11). Next, we investigated p‐c‐Jun in 2H‐11 cells stably knocked down for BMPR2 or the other type II receptors. Downregulation of BMPR2, but not ACVR2A or ACVR2B, by two different shRNA constructs decreased p‐c‐Jun in response to BMP‐9 (Figure 5B). Furthermore, chemical inhibition of JNK in shBMPR2 2H‐11 cells failed to enhance the calcifying effect of BMP‐9, unlike control (pLK0.1), shACVR2A, and shACVR2B stable cells (Figure 5C). This might suggest that BMPR2 functionally interacts with JNK to activate JNK signaling, as previously suggested 44, 45. Accordingly, endogenous JNK was detected in GST pull‐down experiments performed with GST‐BMPR2 (supplementary material, Figure S12). Finally, to determine the contribution of JNK signaling to the osteogenic differentiation of 2H‐11 cells, we restored JNK signaling in cells with stable knockdown of BMPR2 by overexpressing the MKK7‐JNK3 fusion protein, which induces constitutive activation of JNK 36. Ectopic overexpression of MKK7–JNK3 increased the levels of p‐c‐Jun (supplementary material, Figure S13). Interestingly, this correlated with partial inhibition of BMP‐9‐induced mineralization (Figure 5D). Taken together, these data confirm that downregulation of JNK activity in ECs favors their osteogenic differentiation in response to BMPs.

Figure 5

Loss of BMPR2 decreases c‐Jun activation, thereby enhancing BMP‐9‐induced osteogenic differentiation of ECs. (A) Alizarin Red staining and quantification on 2H‐11 cells co‐treated for 14 days with BMP‐9 (10 ng/ml) and chemical inhibitors targeting the JNK (SP600125, 5 μmol/l), ERK (UO120, 10 μmol/l), p38 (SB203580, 10 μmol/l), and NF‐κB (PDTC, 50 μmol/l) signaling pathways, incubated under osteogenic conditions. Fold induction of DMSO‐treated control cells is shown. (B) Western blot of p‐c‐Jun and total c‐Jun in 2H‐11 cells stably knocked down for Bmpr2, Acvr2A or Acvr2B and stimulated with BMP‐9 (10 ng/ml) for 45 min. Quantification is shown below. (C) ARS staining and quantification on 2H‐11 cells stably knocked down for Bmpr2 (shRNA #1), Acvr2A (shRNA #1) or Acvr2B (shRNA #1) and stimulated for 14 days with BMP‐9 (10 ng/ml) or BMP‐9 and SP600125 (5 μmol/l) in osteogenic conditions. OM: osteogenic medium. (D) ARS staining and quantification in 2H‐11 cells stably knocked down for Bmpr2 and transfected with MKK7‐JNK3 or an empty vector (pcDNA3), incubated with BMP‐9 (10 ng/ml) under osteogenic conditions. Fold induction of pcDNA‐transfected untreated cells is shown. OM: osteogenic medium. Statistical significant difference with respect to control or BMP‐9 treated cells (*, **, ***).

ECs forming the vasa vasora in advanced human atherosclerosis exhibit reduced BMPR2 expression and JNK activation

A unique characteristic of human atherosclerosis is the progressive ingrowth of microcapillaries from the adventitia towards the intima layer in the aorta. We have previously shown that ECs lining such capillaries (known as vasa vasora) increase the expression of the transcription factor SLUG in the early stages of atherosclerosis 3, suggesting that these cells may undergo EndMT and thereby contribute to a population of cells with osteogenic potential. To further extend these studies, we performed immunofluorescent labeling on sections of human aorta corresponding to different stages of the atherosclerosis spectrum 34. We observed an accumulation of αSMA‐positive cells and loss of luminal PECAM‐1 in microcapillaries from sections corresponding to fibrotic calcified plaque (advanced) (Figure 6A). Notably, we found double‐positive PECAM‐1/αSMA cells around the luminal endothelium, which is suggestive of EndMT. Furthermore, whereas BMPR2 was very potently expressed in ECs from capillaries of normal and early fibroatheroma (early) sections (Figure 6B), ECs in advanced lesions were very poorly stained for BMPR2 and PECAM‐1. Finally, we analyzed the nuclear levels of p‐SMAD1/5 and p‐c‐Jun in vasa vasorum ECs (Figure 6C,D). As Figure 6E shows, p‐SMAD1/5 was significantly augmented in the ECs of early lesions and remained elevated in advanced lesions. On the contrary, p‐c‐Jun staining peaked in early stages and dramatically dropped in advanced lesions, in agreement with the loss of BMPR2. Altogether, these results support the mechanisms that we have identified in vitro, and propose BMPR2 as a novel biomarker and druggable target for human atherosclerosis.

Figure 6

BMPR2 is downregulated in ECs of advanced lesions of human atherosclerosis. Immunofluorescent staining on aortic sections of control, early fibroatheroma (early), and fibrotic calcified plaque (FCP) donors. A representative picture per condition is shown. Zoomed images focus on microcapillaries in intimal and medial regions underneath the fibrotic core. (A) Staining for PECAM‐1 (green), αSMA (red), and DAPI nuclear counterstain (blue). White arrows indicate PECAM‐1/αSMA double‐positive cells. (B) Staining for PECAM‐1 (green), BMPR2 (red), and DAPI (blue). Immunohistochemical staining for p‐SMAD1/5 (C) and p‐c‐Jun (D) on the aforementioned sections. Black arrows indicate positively stained cells. Low magnification (10×) and zoomed images (40×) are shown. (E) Quantification of C and D based on the numbers of p‐SMAD1/5 or p‐c‐Jun‐positive EC nuclei in one vessel/total nuclei in that vessel. A minimum of 55 vessels from three independent donors were considered per condition. Statistical significant difference between analyzed groups (**, ***).

Discussion

The identification of patients at increased risk of acute coronary events associated with vascular calcification, who may benefit from intensified preventative measures and for monitoring therapeutic efficacy, is a major ongoing challenge 46. Recent publications have suggested that ECs directly contribute to vascular calcification through EndMT 2, 3, 4, 42. We investigated the interplay between inflammatory cytokines and BMPs, a subfamily of the TGF‐β family, in ECs. We found that the pro‐inflammatory cytokine TNF‐α induces EndMT in HAoECs, thereby enhancing their differentiation into osteoblast‐like cells in response to BMP‐9. TNF‐α and IL‐1β induced EndMT in HAoECs, which results in BMPR2 downregulation. BMPR2 loss is sufficient and necessary for TNF‐α to trigger EndMT in HAoECs. In addition, BMPR2 downregulation coincides with neointima formation and a decrease of luminal PECAM1‐positive cells in ApoE3Leiden mice subjected to a high‐fat diet. Moreover, we showed that in the absence of BMPR2, BMP‐9 induces the formation of a receptor complex involving ACVR2A to induce downstream canonical signaling. Interestingly, we identified that BMPR2 downregulation leads to decreased JNK activation in ECs, as measured by phosphorylation of its downstream target c‐Jun, and we showed that JNK deactivation facilitates BMP‐induced mineralization in ECs. Finally, we showed that adventitial microvessels (vasa vasora) in advanced atherosclerotic disease exhibit loss of luminal PECAM1 and BMPR2 and accumulation of PECAM1/αSMA double‐positive cells in layers underneath, suggestive of EndMT. Congruently, whereas p‐SMAD1/5 increases in vasa vasora ECs at early stages of atherosclerosis and remains sustained in advanced lesions, p‐c‐Jun activation drops dramatically in vessels in fibrotic calcified plaque lesions. These results point to a key role of BMPR2 as an integrator of TGF‐β/BMP and inflammatory signaling in ECs, determining EndMT and subsequent calcification.

Endothelial cells with reduced BMPR2 expression displayed increased expression of inflammatory markers such as ICAM and VCAM 47. Interestingly, siRNA‐mediated knockdown of BMPR2 in human umbilical vein endothelial cells (HUVECs) led to reduced p‐SMAD1/5 in response to BMP‐4, although the authors did not test whether this effect was specific for BMPR2. BMP‐9, unlike BMP‐4, shows a high affinity for ALK1 in ECs 37 and BMPR2 48, which may result in a different contribution of BMPR2 in BMP‐9 or BMP‐4‐induced receptor complexes. Furthermore, ALK1 (and not ALK2) was recently shown to mediate LDL uptake in ECs 49, and exposure of ECs to an inflammatory cocktail lowered their LDL uptake capacity 50. Whether BMPR2 deficiency alters the ALK1–LDL interaction is an interesting area of study.

BMPR2 deficiency leads to pulmonary arterial hypertension (PAH), where heterozygous germline mutations in the BMPR2 gene are found in more than 70% of patients with hereditary PAH and in 20% of patients with idiopathic PAH 51, 52. Several publications have indicated EndMT as a causative factor for the vascular remodeling in the pulmonary arteries of PAH patients 53, 54. Furthermore, PAH patients display a higher tendency to develop calcified lesions within the pulmonary arteries 55, suggesting that BMPR2 may have a protective role by inhibiting calcification in cells with osteogenic potential. In this sense, osteoblast‐specific knockout of Bmpr2 led to increased bone mass in transgenic mice 56. Importantly, BMP‐9 has been recently proposed as a therapeutic option for PAH 12. Nevertheless, it remains to be determined whether systemic administration of BMP‐9 may lead to calcification of the arteries under inflammatory conditions, such as those considered risk factors for atherosclerotic disease (renal disease, diabetes, hypertension or hyperlipidemia).

In summary, we have shown that aortic ECs undergo EndMT in response to inflammatory stimuli, which favor the calcification of ECs. Dissection of the underlying mechanisms of such enhanced osteogenic potential suggests that non‐canonical signal transduction fine‐tunes BMP‐SMAD1/5 signaling, thereby enhancing the osteogenic differentiation of ECs (summarized in the supplementay material, Figure S14). A hallmark of this process is the downregulation of BMPR2. Therefore, monitoring BMPR2 as a biomarker of calcification or developing drugs specifically increasing BMPR2 expression to prevent EndMT in the arteries constitutes an area of interest for future research.

activin receptor‐like kinase

alizarin red solution

alpha smooth muscle actin

bone morphogenetic protein

BMP receptor

early fibroatheroma

endothelial‐to‐mesenchymal transition

fibrotic calcified plaque

osteogenic medium

pulmonary arterial hypertension

homolog of the Drosophila protein, mothers against decapentaplegic (MAD) and C. elegans protein SMA

vascular endothelial cadherin

Author contributions statement

GSD, PtD, JL, MJG, and MDV designed the study. GSD, AGVA, VP, STGJ, and MG collected the data. GSD, AGVA, VP, STGJ, and MG analyzed the data. GSD, PtD, JL, MJG, HVD, and MDV interpreted the data. GSD, PtD, JL, MJG, HVD, and MDV searched the literature. GSD, AGVA, VP, and STGJ generated the figures. GSD, PtD, MJG, JL, HVD, and MDV wrote the manuscript.

Supplementary materials and methods

Supplementary figure legends

Figure S1. TNF‐α and IL‐1β induce the upregulation of EndMT factors in HAoECs

Figure S2. Long‐term effect of TNF‐α and TGF‐β in HAoECs

Figure S3. TNF‐α and IL‐1β induce the downregulation of BMPR2 in HAoECs

Figure S4. TNF‐α induces the upregulation of BMPR2 in a cell type‐specific manner

Figure S5. TNF‐α downregulates BMPR2 in a dose‐dependent manner

Figure S6. BMP receptor activation is required to induce cell mineralization in 2H‐11 endothelial cells

Figure S7. Bmpr2 knockdown enhances BMP‐9‐induced mineralization in 2H‐11 cells

Figure S8. Knockdown of Acvr2A or Acvr2B does not affect BMP‐9‐induced mineralization in 2H‐11 cells

Figure S9. Bmpr2 overexpression partially prevents BMP‐9‐induced mineralization in 2H‐11 cells

Figure S10. Knockdown of Bmpr2 does not compromise BMP‐9 binding to ALK1 or ALK2

Figure S11. Inhibition of c‐Jun phosphorylation enhances BMP‐9‐induced mineralization in 2H‐11 cells

Figure S12. In vitro protein interaction BMPR2–JNK

Figure S13. MKK7–JNK3 overexpression restores p‐c‐Jun in 2H‐11 shBMPR2 cells

Figure S14. Graphical summary

Supplementary Material and Methods

Click here for additional data file.

Supplementary figure legends

Click here for additional data file.

Figure S1. TNF‐α and IL‐1β induce the up‐regulation of EndMT factors in HAoECs. Gene expression analysis of the genes SNAI1 (encoding SNAIL), SNAI2 (encoding SLUG), TWIST1, ZEB1 and ZEB2 in HAoECs stimulated for 24 h with the indicated cytokines and growth factors. CO: control; Tα: TNF‐α (10 ng/ml); IL: IL‐1β (10 ng/ml); LP: LPS (10 ng/ml); Tβ: TGF‐β3 (5 ng/ml); AA: activin A (50 ng/ml); B6: BMP‐6 (50 ng/ml); B9: BMP‐9 (10 ng/ml).

Click here for additional data file.

Figure S2. Long term effect of TNF‐α and TGF‐β in HAoECs. (A) Gene expression analysis of FN (encoding Fibronectin), CDH2 (encoding N‐CAHDERIN), CDH5 (encoding VE‐CADHERIN) and PECAM1 in HAoECs stimulated for 4 days with TNF‐α (Tα: 10 ng/ml) or TGF‐β3 (Tβ: 5 ng/ml); followed by 24 h of BMP‐9 (B9: 10 ng/ml). (B) qPCR gene expression analysis of the BMP ligand encoding genes BMP2, BMP4 and BMP7, in HAoECs treated as indicated in (A).

Click here for additional data file.

Figure S3. TNF‐α and IL‐1β induce the down regulation of Gene expression analysis of the BMP type II receptors BMPR2, ACVR2A and ACVR2B in HAoECs incubated for 24 h with the indicated growth factors in the presence of 10% of serum.

Click here for additional data file.

Figure S4: TNF‐α induces the up‐regulation of BMPR2 in a cell type specific manner. Western blot for BMPR2 (long and short exposures) in HAoEC, human pulmonary aortic ECs (PAEC), human endothelial colony forming cells (ECFC), human coronary microvascular EC (cMVEC) and human skin microvascular ECs (HMEC) treated for 24 h with TNF‐α (10 ng/ml) in medium containing 10% serum.

Click here for additional data file.

Figure S5. TNF‐α down regulates BMPR2 in a dose dependent manner. Western blot in HAoECs treated for 24 h with increasing concentrations of TNF‐α in medium containing 10% serum. CO: Control.

Click here for additional data file.

Figure S6. BMP receptor activation is required to induce cell mineralization in 2H‐11 endothelial cells. (A) Alizarin Red staining (ARS) of 2H‐11 cells stimulated with either BMP‐2, BMP‐6 or BMP‐7 (50 ng/ml) or BMP‐9 (10 ng/ml) for 14 days under osteogenic culture conditions (OM). Quantification is shown below as fold induction of OM control cells. (B) ARS of 2H‐11 cells stimulated for 14 days with BMP‐6 (50 ng/ml) and/or the BMP type I receptor kinase inhibitor LDN‐193189 (120 nM). Quantification is shown below as fold induction of OM control cells.

Click here for additional data file.

Figure S7. (A) Alizarin Red staining (ARS) of 2H‐11 cells stably transduced with two independent shRNA constructs targeting BMPR2 (#1 and #2) or an empty vector (pLK0.1) and incubated with BMP‐9 (10 ng/ml) for 14 days under osteogenic culture conditions (OM) or regular growth medium (GM). Quantification is shown below as fold induction of pLK0.1 stable cells in OM. (B) qPCR analysis of BMPR2, ACVR2A and ACVR2B in 2H‐11 cells knocked down for BMPR2.

Click here for additional data file.

Figure S8. Knock‐down of ACVR2A or ACVR2B does not affect BMP‐9 induced mineralization in 2H‐11 cells. (A) ARS of 2H‐11 cells stably transduced with two shRNA constructs targeting Acvr2A or Acvr2B (#1 and #2) or a control vector (pLK0.1) and incubated with BMP‐9 (10 ng/ml) for 14 days under osteogenic culture conditions (OM) or regular growth medium (GM). Quantification is shown below as fold induction of pLK0.1 stable cells in OM. (B) qPCR analysis of Bmpr2, Acvr2A and Acv2 in 2H‐11 cells knocked down for AcvrA and Acvr2B.

Click here for additional data file.

Figure S9. BMPR2 over‐expression partially prevents BMP‐9 induced mineralization in 2H‐11 cells. ARS of 2H‐11 cells stably over‐expressing a BMPR2 full length construct (BMPR2 FL) or an empty vector (EMPTY, pLV) and stimulated with BMP‐9 (10 ng/ml) and/or TNF‐α (10 ng/ml) for 14 days under osteogenic conditions. Quantification is shown below as fold induction of pLV stable cells in OM.

Click here for additional data file.

Figure S10. Knock‐down of Quantification by densitometry corresponding to a ligand‐receptor interaction assay performed in 2H‐11 stably infected with a control (pLK0.1) or BMPR2 knock‐down (shBMPR2) lentivirus. ALK1‐ALK2 intensity is shown. IP: Immunoprecipitation.

Click here for additional data file.

Figure S11. Inhibition of c‐Jun phosphorylation enhances BMP‐9 induced mineralization in 2H‐11 cells. (A) Western blot of 2H‐11 cells transduced with lentivirus encoding for a c‐Jun‐specific mutant version of MKP1 (mMKP1) or an empty vector and stimulated with BMP‐9 (10 ng/ml). (B) ARS of 2H‐11 cells infected with mMKP1 and stimulated with BMP‐9 (10 ng/ml) under osteogenic culture conditions (OM). Calcium deposits were solubilized and measured by absorbance.

Click here for additional data file.

Figure S12. JNK interacts with BMPR2 in GST‐BMPR2 pull down assay on whole cell lysate of HAoECs. Endogenous JNK interacts in vitro with GST‐BMPR2 FL, whereas GAPDH is only detected in the input.

Click here for additional data file.

Figure S13. MKK7‐JNK3 over expression restores p‐c‐Jun in 2H‐11 shBMPR2 cells. Western blot of 2H‐11 cells stably knocked‐down for BMPR2 and transfected with a MKK7‐JNK3 encoding construct or an empty vector (pcDNA3). Cells were serum starved for 16 h and stimulated for 45 min with BMP‐9 (10 ng/ml).

Click here for additional data file.

Figure S14. Graphical summary. In the presence of BMP‐9, a heterotetrameric BMP membrane receptor complex is formed consisting of ALK1/2 and BMPR2 in ECs. This induces the downstream activation of canonical SMAD1/5 and non‐canonical JNK signaling, leading to osteogenic differentiation and calcium deposition. Upon stimulation with TNF‐α, ECs undergo EndMT and down‐regulate BMPR2. BMP‐9 now interacts with a receptor complex consisting of ALK1/2 and ACVR2A in ECs. This induces the phosphorylation of SMAD1/5, but does not activate p‐c‐Jun potently. As p‐c‐Jun acts as a negative regulator of EC calcification, EndMT‐derived cells exhibit a higher osteogenic activity in response to BMP‐9.

Click here for additional data file.