ALK5 and ALK1 Play Antagonistic Roles in Transforming Growth Factor β-Induced Podosome Formation in Aortic Endothelial Cells

Abstract

Transforming growth factor β (TGF-β) and related cytokines play a central role in the vascular system. In vitro, TGF-β induces aortic endothelial cells to assemble subcellular actin-rich structures specialized for matrix degradation called podosomes. To explore further this TGF-β-specific response and determine in which context podosomes form, ALK5 and ALK1 TGF-β receptor signaling pathways were investigated in bovine aortic endothelial cells. We report that TGF-β drives podosome formation through ALK5 and the downstream effectors Smad2 and Smad3. Concurrent TGF-β-induced ALK1 signaling mitigates ALK5 responses through Smad1. ALK1 signaling induced by BMP9 also antagonizes TGF-β-induced podosome formation, but this occurs through both Smad1 and Smad5. Whereas ALK1 neutralization brings ALK5 signals to full potency for TGF-β-induced podosome formation, ALK1 depletion leads to cell disturbances not compatible with podosome assembly. Thus, ALK1 possesses passive and active modalities. Altogether, our results reveal specific features of ALK1 and ALK5 signaling with potential clinical implications.

INTRODUCTION

The cellular actin cytoskeleton plays a fundamental role in tissue remodeling directed by matrix-degrading proteases. Specialized cellular F-actin-based structures involved in matrix remodeling are known as podosomes (1). They appear as dynamic cylinder-like structures standing out perpendicular to the cell ventral membrane and consist of a core of actin filaments surrounded by an adhesive ring of integrins and adaptor proteins such as vinculin. This spatial organization of proteins, the presence of cortactin, and the podosomal marker Tks5 serve to distinguish podosomes from other adhesion complexes. Podosomes recruit transmembrane metalloprotease (MMP) MT1-MMP and secreted MMPs to break down extracellular matrix (ECM) proteins. In aortic endothelial cells (ECs), podosomes arise in response to transforming growth factor β (TGF-β) signals. In this model, and in contrast to myelomonocytic cells, podosome induction requires protein synthesis (2) and therefore involves transcription factor activation. Another specific feature of this model is that podosomes always self-organize in ring-shaped superstructures (podosome rosettes). By concentrating MMP activities, the device may strip the basement membrane, allowing the replacement of ECM components by matrix-producing cells. Matrix-degrading podosome rosettes have been visualized in the native endothelium of aortic explants exposed to TGF-β (3). Thus, the characterization and physiological relevance of endothelial podosomes have been established, but the signaling pathways involved in their formation remain underexplored.

TGF-β is a multifunctional cytokine that signals through heteromeric complexes of transmembrane serine/threonine kinase receptors, usually comprising two type II receptors (TβRII) and two type I receptors (TRI). TβRII is constitutively active and phosphorylates TRI on Ser/Thr residues in the GS domain in response to ligand binding. In turn, TRI propagates signals by phosphorylating Smad intracellular effectors at their C-terminal SSXS motifs (4, 5), activating them to transduce the signal to the nucleus and stimulate transcription of target genes. In parallel, non-Smad pathways are initiated, including those that modulate Smad signaling (6, 7). In most cases, ALK5 is the TRI receptor conveying TGF-β signals (TβRI) through phosphorylation of Smad2/3. However, ECs also express another TRI, ALK1, which requires ALK5 for signaling and regulates bone morphogenetic Smads: Smad1/5 (8). ALK5 is not the only factor necessary for the recruitment of ALK1 into a TGF-β receptor complex: kinase activity of ALK5 is also required for maximal ALK1 activation (8). Furthermore, ALK1 can antagonize ALK5/Smad2/3 signaling at the level of Smads (9, 10). The cross talk between ALK1 and ALK5 signals provides a delicate TGF-β balance to fine-tune EC functions.

Functionally, TGF-β/ALK5 signaling via Smad2/3 leads to inhibition of cell migration and proliferation, whereas TGF-β/ALK1 signaling via Smad1/5 promotes these events (9). As mentioned above, ALK1 modulates ALK5 signaling. The combined effects promote the activation phase of angiogenesis, during which vascular permeability increases, the basement membrane is degraded, and ECs proliferate and invade the stroma (11, 12). In a second step, ALK1 signaling is downregulated and ALK5-mediated pathways regulate the resolution phase of angiogenesis, in which ECs stop proliferating, differentiate, and restore the basal lamina. However, such a scenario was questioned when it was found that ALK1, in complex with BMPRII (or ActRIIA), was identified as the TRI for bone morphogenetic proteins (BMP) BMP9 and BMP10 (13–15). ALK1 triggering through these cytokines represses EC activation, and a novel picture emerges in which BMP9 acts as a quiescence factor (13). BMP9 is a circulating factor predominantly produced in the liver, whereas BMP10 expression is restricted to the developing or postnatal heart (13). However, recent studies show ALK1 can deliver proangiogenic signals in some situations (16).

The functionality of ALK1 has also been addressed by studying human diseases. Heterozygous loss of functional mutations of ALK1 in humans causes a rare adult-onset vascular dysplasia known as hereditary hemorrhagic telangiectasia2 (HHT2; Osler-Rendu-Weber syndrome). This autosomal dominant disorder is characterized by dilated leaky capillaries (telangiectases), hemorrhages, and arteriovenous malformations (17). Haploinsufficiency is the major mechanism for HHT2, and in most cases studied, the ALK1 mutation results in defective trafficking by retention of the mutated ALK1 protein in the endoplasmic reticulum and subsequent loss of cell surface expression (18, 19). Mouse models with targeted disruption of the ALK1 gene have been generated in which heterozygous ALK1^+/− mice develop a phenotype over time, with characteristics similar to those of HHT2 patients (10, 20–22).

Previous work with bovine arterial ECs (BAE cells) has established that both Smad2/3 and Smad1/5 are rapidly phosphorylated in response to TGF-β, indicating that TGF-β signals are transduced through both ALK5 and ALK1 in this model (2). As TGF-β has not been reported to induce podosome formation in non-ECs, we hypothesized that ALK1 could be more specifically involved in the process of podosome formation in ECs. In this scenario, podosome induction would enable cells to invade tissues, consistent with the reported predominance of ALK1 signaling during the activation phase of angiogenesis (11, 12). Results presented herein establish that although TGF-β stimulates ALK5 and ALK1, ALK5 signaling to Smad2/3 is the receptor triggering podosome formation and ALK1 mitigates this signal through Smad1. Thus, ALK5 and ALK1 play antagonistic roles in TGF-β-induced podosome formation in aortic ECs. In addition, a novel feature of ALK1 became apparent in the course of the experiments. ALK1 depletion but not ALK1 neutralization disturbs cell responses, so ALK1 possesses both passive and active modalities.

(This work is part of the doctoral theses of Filipa Curado and Isabel Egaña, University of Bordeaux, Bordeaux, France.)

MATERIALS AND METHODS

Cells and cell stimulation.

BAE cells (Lonza) were maintained in complete endothelial cell growth medium (EGM-MV; Promocell). Active TGF-β concentrations were around 10 pg/ml in the culture medium, as calculated from enzyme-linked immunosorbent assay (ELISA) titration of the serum (Ready-Set-Go, human/mouse TGF-β1, 2nd generation; eBiosciences). The BMP9 concentration was around 0.2 ng/ml. Cells were cultured at 37°C in a 5% CO₂ humidified atmosphere and used between passages three and six. Podosome formation assays were performed by double staining for F-actin/cortactin or F-actin/vinculin over a 20-h period in complete medium, as the process requires the presence of serum (2). For Western blot analysis of Smad phosphorylation, cells were starved overnight before stimulation (30 min). TGF-β was used at 5 ng/ml for assays requiring maximal stimulation and at 1 ng/ml or lower concentrations for regular assays.

Mice.

ALK1^+/− mice on a C57BL/6J background originally described by D. Marchuk (56) were obtained from Franck Lebrin (College de France, Paris, France) with the permission of Paul Oh (University of Florida, FL). Genotypes were confirmed by PCR. Littermates were used as wild-type (WT) controls. All experiments were carried out in accordance with the National Code of Ethics on Animal Experimentation (French Ministry of Agriculture and Fisheries, Animal Care and Use Certificate number A3312043) and approved by the Committee of Ethics of Bordeaux (authorization certificate A50120219). For ex vivo analysis of the endothelial responses, aortic vessel segments were derived from 6- to 8-week-old C57BL/6J mice (Charles River) that were anesthetized by intraperinoneal (i.p.) administration of a mixture of ketamine (100 mg/kg of body weight; Merial) and xylazine (10 mg/kg; Bayer) in 100 μl in a sterile saline solution. Examination of native tissues (ex vivo) was performed with 2-year-old ALK1^+/− mice and littermates that had received intracardiac (i.c.) injection of a 4% paraformaldehyde solution. Fixation of the samples was continued by a 1-h fixation step in vitro, and the tissue samples were processed for immunofluorescence as described previously (3).

Reagents.

Recombinant human TGF-β, BMP9, ALK1ecd, ALK6ecd, and BMP9 neutralizing antibodies (MAB3209) were obtained from R&D Systems and used at the concentrations indicated below. Antibodies against the following proteins were obtained as follows: cortactin (catalog no. 05-180), PSmad2/3 (catalog no. 3849), PSmad1 (catalog no. 06-702), and Smad3 (catalog no. 04-1035) were from Millipore; Smad5 (catalog no. ab40771) was from Abcam; PSmad1/5 (catalog no. 9511), Smad1 (catalog no. 9743), Smad2 (catalog no. 5339), Psmad2 (catalog no. 3108), and PSmad3 (catalog no. 9520) were from Cell Signaling Technologies; and VE-cadherin was from eBioscience. Vinculin (catalog no. V9264) was from Sigma. Alexa Fluor 546-phalloidin, Alexa Fluor 488- or 647-labeled secondary antibodies, and gelatin-coupled Oregon-488 were purchased from Invitrogen. SB431542 (used at 10 μg/ml) and LDN-193189 (used in this study at 200 nM, similar results were obtained at 100 nM) were purchased from Tocris.

EC loading of oligonucleotides.

BAE cells were loaded with ALK1- or ALK5-derived oligonucleotides using influx pinocytotic cell-loading reagent as previously described (8).

siRNA, miR-155, and transfection.

Small interfering RNA (siRNA; Qiagen) transfection into BAE cells was performed by two rounds of transfection with double-stranded RNA (25 nM) (2). Three 21-bp duplexes of siRNA were used for ALK1 (target gene; bov1ALK1, 5′-AGTTCGTCAACCACTACTGCT-3′; bov2ALK1, 5′-AAGAGCAACCTGCAGTGCTGC-3′; and bov3ALK1, 5′-AAGAGCAGATCCGAACCGACT-3′). Individually, each siRNA duplex was found to selectively reduce ALK1 expression as assessed by quantitative reverse transcription-PCR (qRT-PCR) analysis (data not shown). When used in combination, ALK1 expression was further reduced, but under these conditions, ALK5 expression was also affected. These findings are consistent with the fact that HHT2 patients with defective ALK1 expression show reduced ALK5 expression (23). To examine the consequences of ALK1 depletion under conditions in which ALK5 levels were maintained and which mimic ALK1 haploinsufficiency in ALK1^+/− mice, we chose to reduce ALK1 expression by transfection of a single siRNA duplex (bov1ALK1). The sequence selected had been validated in several studies (24, 25).

The other RNA target sequences, siALK5 (bov1ALK5, 5′-AACATATTGCTGCAACCAGGA-3′; bov2ALK5, 5′-AAGGCATGATTCGGCCACGGA-3′; and bov3ALK5, 5′-AACTCAGTCAGCAGGAAGGCA-3′), siRNA Smad1 (5′-AACACTGGTGCTCTATTGTCT-3′), siRNA Smad2 (5′-ATAGATCAGTGGGATACAACA-3′), siRNA Smad3 (5′-AACCTGAAGATCTTCAACAAC-3′) and siRNA Smad5 (5′-AAAGCCTTGAGCAGTCCAGGA-3′), were previously published (25, 26). Similar results were obtained using other siRNA duplexes designed against the corresponding bovine cDNA using online algorithms (Dharmacon and/or Applied Biosystems). Similar results were obtained using at least two distinct siRNA for each Smad protein (data not shown). Bta-miR-155 mature sequence (UUAAUGCUAAUCGUGAUAGGGGU; miRBase accession number MIMAT0009241) was purchased from Qiagen. AllStars negative-control siRNA (Qiagen) was used as a control in all gene-silencing experiments. Fluorescence-activated cell sorting experiments using Alexa Fluor 488-siRNA (Qiagen) showed 95 to 98% transfection efficiency.

Expression constructs and transfection.

pcDNA3-Myc, pcDNA3-HA-ALK1 WT, pcDNA3-HA-ALK5 WT, pcDNA3-HA-CA-ALK1 (Q201D), pcDNA3-HA-CA-ALK5 (T204D), pcDNA3-HA-ALK5-3A (L45 loop 3 substitutions in the L45 loop), and pcDNA3-Myc-Smad1 were kindly provided by Peter ten Dijke (Leiden University Medical Center, Leiden, The Netherlands) and pcDNA3-GFP-Smad2 and pcDNA3-GFP-Smad3 by Caroline Hill (CRUK, London, United Kingdom). Cells were transfected using Nucleofector technology (Neon; Invitrogen) according to the manufacturer's guidelines (4 μg for a 35-mm dish), and 24 h after transfection, the cells were stimulated with TGF-β as indicated below and analyzed.

Western blot analysis.

Cells were collected in reducing Laemmli sample buffer, and then the samples were boiled and subjected to SDS-PAGE. Proteins were transferred from gels to Immobilon polyvinylidene difluoride membranes (Millipore). Proteins were detected by chemiluminescence (Invitrogen) using horseradish peroxidase-coupled secondary antibodies (Dako). The amounts of proteins detected by Western blotting were determined by scanning the autoradiograph (densitometry), followed by processing of the data with ImageJ (27). The complete area of respective images was set as 100%. All the values were normalized against that for tubulin, which was used as a loading control.

Immunofluorescence staining and matrix degradation assay.

Subconfluent cells grown on glass coverslips were prepared for immunofluorescence as previously described (2). The coverslips were washed in water and mounted on microscope slides with ProLong Gold Antifade containing (or not containing) 4′,6-diamidino-2-phenylindole (DAPI; Life Technologies). For the matrix degradation assay, BAE cells were seeded on coverslips coated with green fluorescent Oregon Green 488 gelatin and stimulated as indicated below. Cells were fixed and samples were processed for immunofluorescence to visualize podosome rosettes and matrix degradation. Quantification of degradation areas on fluorescence-labeled gelatin was performed for at least 10 fields (10× objective lens) for each coverslip. The areas of degradation were quantified by using ImageJ software. Degraded areas were thresholded and measured by the Analyze Particles function. The total degradation area (expressed in μm²) was then normalized for the number of cells (degradation index) in respective fields. In most cases, control values were arbitrarily taken as 100%.

Microscopy and image analysis.

Cells and aortic vessel segments were analyzed by confocal imaging using a Zeiss LSM 510 inverted laser-scanning fluorescence microscope equipped with acquisition software (LSM 510 acquisition software; Zeiss) and a 63×, numerical aperture (NA) 1.4 oil immersion objective. Quadruple-color imaging using DAPI-, Alexa Fluor 488-, or Alexa Fluor 647-labeled secondary antibodies and Alexa Fluor 546-phalloidin was obtained using selective laser excitation at 350 nm, 488 nm, 633 nm, and 543 nm, respectively. Each channel was imaged sequentially using the multitrack recording module before merging. Fluorescent images were processed with ImageJ.

To quantify the translocation of Smad proteins to the nucleus, two optical sections were acquired for each field. The first section was chosen as the 0 plan (best focus for the actin − representative for the cytoplasmic pool of Smad proteins) and the second slice was acquired 1 μm higher than the 0 plan (representative for the nuclear pool). Five fields of each coverslip (50 to 60 cells) were acquired. Images were analyzed with the ImageJ program. The ratio between the nuclear and cytoplasmic pool of the Smad proteins was calculated as follows. SUM slice projection was performed to obtain the nuclear and cytoplasmic signal in one image. DAPI staining was used to create the mask for the nuclear pool. The borders of the nuclei were automatically drawn using Threshold, Binary and Analyze Particles function. For cytoplasmic pool (signal around the nucleus), the binary image of nuclei was dilated 15 iterations and a mask was created using the Analyze Particles functions. To subtract the nuclei from the cytoplasmic mask, the Image Calculator function was used. Finally, nuclear and cytoplasmic masks were used to measure the amount of Smad proteins in the respective compartments. The mean of each field was calculated, and the means ± standard deviations (SDs) is presented. Duplicate coverslips were analyzed per condition.

Quantification of staining intensity of junctional proteins.

Intensities of junctional proteins were measured using ImageJ. Circular regions of interest (ROIs) 5 μm in diameter were drawn at junctional areas and mean fluorescence intensities were measured. 50 ROIs were measured in 5 fields under each condition. Intensities were corrected for background fluorescence and statistically analyzed using GraphPad Prism6 software (GraphPad Software, Inc., San Diego, CA).

Transmission electron microscopy (TEM).

Tissues were fixed in situ by intracardiac perfusion of 4% paraformaldehyde and 0.2% glutaraldehyde in 0.1 M phosphate buffer (PB; 0,2 M NaH₂PO₄, 0.2 M Na₂HPO₄). After dissection, mouse aortas were postfixed with 1% osmium tetroxide in PB 0.1 M for 2 h at room temperature (RT), followed by exposure to 0.5% uranyl acetate. Samples were dehydrated through a graded series of ethanol solutions to absolute ethanol, followed by exposure to propylene oxide. Specimens were embedded in a mixture of propylene oxide and epoxy resin (Epon 812; Delta Microscopy) for 2 h and then in 100% resin overnight at RT. The resin was polymerized at 60°C for 48 h. The samples were sectioned using a diamond knife on an ultramicrotome (UC7; Leica Microsystems). Ultrathin sections (70 nm) were picked up on copper grids and then stained with uranyl acetate and lead citrate. The grids were examined with a transmission electron microscope (H7650; Hitachi) at 80 kV equipped with an Orion 11Mpixel camera (Roper).

Statistics.

Statistical analysis was performed on GraphPad Prism 6. Data are presented as means ± SDs of three independent experiments. Significance within groups was assessed by using one-way analysis of variance (ANOVA) using Dunnett's multiple-comparison test, as well as the Bonferroni multiple-comparison test between selected pairs. Differences were considered to be statistically significant at a P of <0.05.

RESULTS

Inhibition of ALK5 or ALK1 signaling affects TGF-β-mediated podosome induction.

Both TRI ALK5 and ALK1 are expressed in BAE cells (8). To explore their respective roles in podosome formation, we first examined the outcome of their stimulation by their ligands. Podosomes were visualized by double staining of F-actin and cortactin and authenticated by their matrix proteolysis function in a gelatin degradation assay in which round patches of degraded matrix occur underneath cells cultured on fluorophore-conjugated gelatin coating (2). Within 20 h of treatment, TGF-β (which triggers either ALK5 or ALK5/ALK1 activation [8]) stimulated podosome assembly and associated gelatin degradation (Fig. 1A), whereas BMP9 (which triggers ALK1 but not ALK5 activation) was completely ineffective (data not shown). TGF-β-induced podosome formation is known to be sensitive to the pharmacological ALK5 receptor inhibitor SB431542 (2). Since SB431542 inhibits both ALK5 and ALK5/ALK1 responses (8), specific siRNAs were used to discriminate TRI signals individually. siRNA efficiency was assessed by Western blotting (Fig. 1B). siRNAs targeting ALK5 were found to be inhibitory, consistent with the effect of SB431542 (8). The knockdown of ALK1 abolished podosome formation and the ability of ECs to degrade the matrix (Fig. 1C). Inhibitory effects were also obtained in experiments performed with antisense oligonucleotides against ALK1 (8) (Fig. 1C). These experiments indicated that (i) the podosome response is, like most TGF-β responses, TRI dependent and that (ii) both ALK5 and ALK1 signaling pathways are regulated in the process of TGF-β-induced podosome formation.

FIG 1.

Both ALK1 and ALK5 are regulated when cells assemble podosome rosettes in response to TGF-β treatment. (A) Cells were seeded on fluorescent gelatin-coated coverslips and either left untreated or stimulated with TGF-β (5 ng/ml) for 20 h. Samples were fixed and then processed for the detection of F-actin (red), cortactin (white), and gelatin (green) by immunofluorescence to highlight podosome rosettes. Scale bar: 20 μm (10 μm for zoomed images). Podosome rosettes in the boxed regions are shown at higher magnification in the bottom row. (B) Cell lysates from BAE cells transfected with the indicated siRNAs (mix, combination of 3 siRNAs; one, single duplex siRNA) were analyzed by Western blotting with the indicated antibodies. Data shown are representative of at least three independent experiments. (C) BAE cells transfected with siRNA designed against ALK1 or ALK5, or loaded with sense (SO) or antisense (ASO) oligonucleotides, were stimulated for 20 h with TGF-β and then processed as for panel A to score podosome rosettes and assess gelatin degradation. Data are presented as percentage of the control TGF-β response arbitrarily taken as 100%. ****, P < 0.0001 for comparison to siCT in degradation area measurements; ####, P < 0.0001 for comparison to siCT in podososome counts. (D) BAE cells were either left untreated or stimulated with increasing concentrations of TGF-β for 20 h. Cells were next fixed and stained, and those showing podosome rosettes were scored. **, P < 0.01; ****, P < 0.0001 fo comparison to untreated cells. (E) BAE cells were left untreated or stimulated with increasing concentrations of TGF-β for 30 min. (F) Kinetics of Smad phosphorylation upon stimulation with TGF-β (5 ng/ml). (G) Same as in panel F with cells stimulated with BMP9 (0.5 ng/ml). (H and I) BAEc were either left untreated or treated with SB431542 for 1 h and then left untreated or stimulated with 0.5 ng/ml of TGF-β (H) or 0.05 ng/ml of BMP9 (I) for 30 min. (J and K) BAE cells were transfected with the indicated siRNA and then left untreated or stimulated with 0.5 ng/ml of TGF-β (J) or 0.05 ng/ml of BMP9 (K) for 30 min. (E to K) Cells lysates were prepared from BAE cells serum starved overnight, and all samples were subjected to Western blotting and analyzed with the indicated antibodies. Tubulin was used as a loading control.*, nonspecific band.

To correlate podosome formation and TGF-β signaling, Smad immediate downstream nuclear effectors of TRI were analyzed. In a concentration-dependent manner, TGF-β stimulated the formation of podosome rosettes (scored 20 h after the onset of stimulation) (Fig. 1D) and the phosphorylation of Smads (Fig. 1E). There was no difference in kinetics between TGF-β phosphorylation of Smad1/5 versus Smad2 or Smad3 (Fig. 1E). BMP9 is known to induce Smad1/5 phosphorylation at lower concentrations than TGF-β (28). In BAE cells and at the concentrations used, there was no significant difference in Smad1/5 phosphorylation kinetics between TGF-β and BMP9 (Fig. 1F and G). Both Smad1/5 and Smad2/3 responses were reduced in the presence of the inhibitor SB431542 (Fig. 1H), consistent with the fact that ALK1 signaling requires ALK5 kinase activity (8). ALK5 silencing produced similar effects, whereas ALK1 depletion reduced the phosphorylation levels of Smad1/5 only (Fig. 1J). These results suggested that TGF-β-induced ALK1 signaling could be dissociated from ALK5 signaling. In contrast, when ALK1 signaling was stimulated by BMP9, Smad1/5 phosphorylation occurred, an event that appeared insensitive to either SB431542-mediated pharmacological inhibition or siRNA-mediated ALK5 depletion (Fig. 1I and K). ALK1 depletion efficiently reduced Smad1/5 phosphorylation in response to BMP9 (Fig. 1K). These data indicated that both ALK1 and ALK5 TRI receptors are functional in BAE cells and suggest that in combination with the cognate TRII, TGF-β uses the ALK1/ALK5 complex for signaling (8), whereas BMP9 utilizes ALK1/ALK1 homodimers to transduce its signal.

ALK5 mediates podosome formation, whereas ALK1 inhibits this signal.

The above-described resuts showed that both ALK1 and ALK5 signaling pathways are regulated upon TGF-β stimulation and that depletion of either receptor prevents podosome formation. We went on to explore the consequences of triggering ALK5 and ALK1 by exposing cells to TGF-β plus BMP9. Synergy between the two ligands has been seen in which the combined action of the two proteins improved the endothelial response to angiogenic stimuli (29). In the BAE cell model, BMP9 produced a dose-dependent inhibition of the TGF-β-induced podosome response (Fig. 2A), even though Smad2/3 and Smad1/5 became efficiently phosphorylated (Fig. 2B). Additive effects were seen on Smad1/5 phosphorylation (Fig. 2B and C). Thus, TGF-β stimulated podosome formation, whereas BMP9 reduced this response, while Smad1/5 was found to be phosphorylated in both situations.

FIG 2.

ALK5 stimulates podosome formation, whereas ALK1 signaling is inhibitory. (A) BAE cells were either left untreated or stimulated with the cytokines indicated for 20 h. Cells were fixed and then subjected to immunofluorescence staining for F-actin/cortactin to score cells showing podosome rosettes. **, P < 0.01; ***, P < 0.001; ****, P < 0.0001 for comparison to CT in each group. (B) Cells were serum starved overnight and then either left untreated of stimulated with the indicated cytokines (TGF-β, 1 ng/ml, and BMP9, 0.1 ng/ml) for 30 min. Whole-cell lysates were prepared and subjected to immunoblot analysis for pSmad and total Smad protein content. (C) Densitometric analysis of pSmad and total Smad proteins from the Western blot shown in panel B presenting the ratio of pSmad to tubulin (not shown). **, P < 0.01; ***, P < 0.001; ****, P < 0.0001 for comparison to CT in each group. (D) Cells were either left untreated or exposed to the indicated treatments (TGF-β, 5 ng/ml) for 20 h. Cells were fixed and then subjected to immunofluorescence as in panel A to score cells showing podosome rosettes. *, P < 0.05; ****, P < 0.0001 for comparison to TGF-β-stimulated cells. (E and F) BAE cells were seeded on fluorescent gelatin-coated coverslips, starved overnight, left untreated or treated with LDN-193189 for 1 h, and then left untreated or stimulated with TGF-β (5 ng/ml) or BMP9 (0.5 ng/ml) for 20 h. Samples were processed for immunofluorescence as in panel A to score cells showing podosome rosettes (E) and to calculate the matrix degradation index (F). **, P < 0.01; ****, P < 0.0001 for comparison to untreated CT. (G) BAE cells were serum starved overnight, then left untreated or treated with LDN-193189 for 1 h, and then left untreated or stimulated with TGF-β (5 ng/ml) or BMP9 (0.5 ng/ml) for 30 min. Samples were analyzed by Western blotting with the indicated antibodies. Tubulin was used as a loading control. *, nonspecific band. (H) Cells were transfected with the indicated constructs and then stimulated with TGF-β (5 ng/ml) on the next day for 20 h. Cells were fixed and then subjected to immunofluorescence staining for F-actin/vinculin to score those showing podosome rosettes. Results are presented as percentage of cells showing podosome rosettes. *, P < 0.05; **, P < 0.01; ****, P < 0.0001 for comparison to CT. (I) Cells were transfected with CA-ALK5 were seeded on fluorescent gelatin-coated coverslips and then processed as for panel H to visualize F-actin (red) and vinculin (green), score cells showing podosome rosettes, and analyze the matrix. Bars: 10 μm (zoomed image, 5 μm). #, P < 0.05; ##, P < 0.01; ###, P < 0.001; ####, P < 0.0001.

To examine the impact of BMP9 signaling on the podosome response, we made use of a recombinant ALK1-Fc fusion protein (ALK1ecd) which acts as a decoy ALK1 receptor (18, 19). ALK1 is unable to directly bind TGF-β and forms a high-affinity complex with TGF-β only in the presence of TβRII (30). In contrast, BMP9 directly binds ALK1, so ALK1ecd can be used to trap BMP9 contained in the serum. When ALK1ecd was added in combination with TGF-β, the podosome response was enhanced compared to that with TGF-β alone (Fig. 2D). ALK6ecd, a chimeric receptor of BMPR1B used as a negative control, was ineffective. The experiment carried out in the presence of neutralizing BMP9 antibodies produced a stimulatory effect similar to that obtained with ALK1ecd (Fig. 2D). The increased podosome response observed in the presence of ALK1ecd or BMP9 antibodies could result from the neutralization of serum-derived BMP9-ALK1 signaling or from the increased availability of membrane ALK1 receptor for TGF-β ligands.

To address this issue, the pharmacological inhibitor LDN-193189 was used to block ALK1 kinase activity independently of whether it was activated by TGF-β or BMP9 (31). Although the podosome response did not significantly change in response to TGF-β, the inhibitory effect of BMP9 was clearly suppressed (Fig. 2E). Matrix degradation was increased in cells where ALK1 was inhibited (Fig. 2F). Smad1/5 phosphorylation was efficiently reduced in the presence of the inhibitor, whereas Smad2/3 phosphorylation remained unaffected (Fig. 2G). Collectively, these results indicated that ALK1 signaling was not required for the podosome response.

To explore further the role of ALK1 and ALK5, we examined the consequence of increasing their expression levels by means of full-length TRI transient transfection. Whereas ALK5 overexpression stimulated TGF-β-induced podosome formation, ALK1 overexpression had the opposite effect (Fig. 2H). Expression of the kinase-impaired ALK5-3A construct, which is unable to phosphorylate Smad proteins (32), prevented podosome formation in response to TGF-β stimulation (Fig. 2H). To differentiate signals emanating from ALK5 from those induced by ALK1/ALK5, we used a constitutively active ALK5 mutant (CA-ALK5, which carries a T204D mutation), exploring whether this mutant could mimic TGF-β responses. CA-ALK5 expression bypassed the need for TGF-β for podosome induction (Fig. 2H and I) and matrix degradation (data not shown). When the constitutively active ALK1 mutant (CA-ALK1 with a Q201D mutation) was expressed, podosomes were not formed. In addition, expression of CA-ALK1 prevented TGF-β-induced podosome assembly (Fig. 2H). These results confirmed ALK1 inhibitory signaling on TGF-β-driven podosome assembly.

Collectively, these experiments showed that ALK5 is necessary and sufficient to induce the podosome program. Surprisingly, although ALK1 signaling reduced the TGF-β response, ALK1 depletion did not stimulate podosome formation but prevented it. This suggested that removing ALK1 proteins also produces an inhibitory effect on podosome rosette formation.

ALK1 depletion induces an activated phenotype not compatible with podosome assembly.

To better understand the role played by ALK1, we explored further the consequences of ALK1 depletion on BAE cell phenotype and behavior. Cells with or without ALK1 depletion were doubly stained for F-actin and the junctional marker VE-cadherin. ALK1 knockdown affected the basal state. Control BAE cells had the characteristic endothelial cobblestone shape at confluence. Upon ALK1 depletion, although no major alteration could be detected on the overall cytoskeletal organization, cells showed increased F-actin staining. Moreover, they underwent morphological changes, became unable to form proper cell-cell junction, and lost contact inhibition (Fig. 3A). The overall staining intensity of VE-cadherin did not change (Fig. 3A), but the staining appeared diffuse and delocalized from cell-cell junctions in ALK1-depleted cells compared to controls (Fig. 3A and B). Thus, depletion of ALK1 induced an activated phenotype, indicating that unligated ALK1 signals to cells. These effects were not mimicked by BMP9-neutralizing antibodies or ALK1ecd (Fig. 3C). However, they were observed to some extent when the conditioned medium of ALK1-depleted cells was added to naive ECs (Fig. 3D), indicating that a soluble factor was involved. The defect shown in ALK1-silenced cells was not observed in ALK5 knockdown cells (Fig. 3B). In sharp contrast, cells with ALK5 depletion showed a very regular pattern and seemed to align in the same direction (Fig. 3A).

FIG 3.

ALK1 depletion induces an activated phenotype not compatible with podosome assembly. (A) Cells were transfected with the indicated siRNA and seeded at confluence on glass coverslips. On the next day, cells were fixed and then subjected to immunofluorescence staining for VE-cadherin and F-actin (CT, top set; ALK1, middle set; ALK5, bottom set). Red hot panels show the fluorescence intensity of VE-cadherin, indicating its expression level from low (red/orange) to high (white). Bars: 20 μm. The boxed regions are shown at higher magnification in the bottom row. (B) Quantification of staining intensity of junctional proteins in cells shown in panel A using ImageJ. ****, P < 0.0001. (C) BAE cells seeded at confluence on glass coverslips and exposed to anti-BMP9 antibodies (2 μg/ml) or ALK1ecd (1 μg/ml). On the next day, cells were fixed and then subjected to immunofluorescence staining for VE-cadherin and F-actin. Isotype antibody control incubations were negative (data not shown). Intensities of junctional proteins were measured using ImageJ. (D) Conditioned medium (CM) was collected from siRNA-transfected samples (A and B) before fixation and used to stimulate confluent BAE cells seeded on glass coverslips. Cells were fixed 20 h after stimulation and processed for immunofluorescence analysis with VE-cadherin and F-actin. Intensities of junctional proteins were measured using ImageJ. *, P < 0.05.

To establish the physiological relevance of these observations, we examined the consequence of ALK1 insufficiency in the endothelium of the ALK1^+/− mouse aorta. Tissues were fixed in situ through intracardiac injection of paraformaldehyde in anesthetized animals (3). Next, vessel segments were harvested, cut along their long axis, and stained for F-actin, cortactin, and VE-cadherin, while nuclei were highlighted with DAPI stain. The specimens were then subjected to confocal microscopy, with ECs facing upward for “en face” viewing. VE-cadherin staining was detected in the endothelial layer, clearly delimiting the cellular boundaries and revealing the integrity of the endothelium in the control specimen. In contrast, VE-cadherin staining appeared diffuse and delocalized from the intercellular junctions in ALK1^+/− aortic samples (Fig. 4A and B). Comparing the endothelium ultrastructure of wild-type and ALK1^+/− samples by transmission electron microscopy (TEM) revealed striking differences. In controls, ECs were as expected, tightly associated and presenting intact junctions. In the ALK1^+/− tissue samples, ECs presented an unusual appearance, with numerous extensions protruding into the vessel lumen, and gaps were visible between adjacent ECs in some places (Fig. 4C). To examine the ability of ECs to assemble podosomes in situ, living segments were prepared (omitting the i.c. step) and exposed to TGF-β for 20 h ex vivo. Podosomes were detected in samples explanted from wild-type animals but not in those prepared from ALK1^+/− mice (Fig. 4D). Interestingly, whereas TGF-β induced the mobilization of cortical cortactin away from cell-cell junctions in control cells, this did not occur for ALK1^+/− ECs (Fig. 4D). These results confirmed that within their native environment, ALK1-deficient ECs fail to assemble podosome rosettes in response to TGF-β.

FIG 4.

The endothelium of ALK1^+/− mice shows defective intercellular junctions, and the ECs do not assemble podosome rosettes in response to TGF-β stimulation. (A) After fixation in situ, aortas were excised from WT and ALK1^+/− mice and the samples were subjected to immunofluorescence staining using VE-cadherin and F-actin. Red Hot LUT shows the fluorescence intensity of VE-cadherin. Bars: 20 μm. The boxed regions are shown at higher magnification in the bottom row. (B) Quantification of staining intensity of junctional proteins in cells shown in panel A using ImageJ. ***, P < 0.001; ****, P < 0.0001. (C) Aortas were fixed in situ by i.c. perfusion and excised from WT and ALK1^+/− mice, and the samples were prepared for TEM. The images show aorta cross sections of the endothelium. Bars: 0.5 μm. (D) Aortas were excised from anesthetized WT or ALK1^+/− mice, and the samples were exposed to TGF-β for 20 h, fixed, and subjected to immunofluorescence staining with F-actin and cortactin, and the endothelium was observed by en face viewing. Podosome rosettes are detected in TGF-β-treated WT (arrow in zoomed image) but not ALK^+/− samples. Bars: 20 μm.

TGF-β regulates Smad1, Smad2, and Smad3 but not Smad5.

Next, we explored the molecular mechanisms underlying the interplay between ALK5 and ALK1 receptors on podosome formation on downstream effectors. This investigation had the potential to unravel how TGF-β and BMP9 could deliver distinct signals to ALK1 in BAE cells. Immunofluorescence experiments were carried out to examine R-Smad translocation to the nucleus as a readout of their regulation in response to these ligands. Whereas TGF-β efficiently elicited Smad1, Smad2, and Smad3 relocalization from the cytosol to the nucleus, Smad5 remained cytosolic (Fig. 5A and B). In contrast, BMP9 stimulated both Smad1 and Smad5 nuclear translocation (Fig. 5A and B). In Smad1-depleted cells, Smad5 did translocate in response to BMP9 but not to TGF-β (Fig. 5C), suggesting that ALK1 regulates Smad1 but not Smad5 in response to TGF-β. In ALK1-depleted cells, Smad1 translocation did not occur in response to either BMP9 or TGF-β, and the Smad5 relocalization that occurred in BMP9-treated control cells became impaired (data not shown). The analysis of Smad phosphorylation supported the nuclear translocation data. TGF-β elicited no signal when Smad1-depleted cell lysates were probed for Smad1/5 phosphorylation, confirming that Smad5 was not phosphorylated upon TGF-β–ALK1 triggering (Fig. 5D). Selective Smad phosphorylation was further verified by using an electrophoretic shift assay that we set up. Whereas both BMP9 and TGF-β induced a shift in the electrophoretic mobility of Smad1, BMP9 but not TGF-β induced a shift in the electrophoretic mobility of Smad5 (Fig. 5E). Similar results were obtained for Smad5-depleted cells (Fig. 5F). However, Smad5 shifted in mobility only in response to BMP9, and not TGF-β, in Smad1-depleted cells (Fig. 5F). Smad1 depletion did not affect Smad2 phosphorylation (Fig. 5D). Smad1 depletion produced effects similar to those induced by ALK1 depletion at the level of R-Smads (Fig. 1I and K and Fig. 5D), thus confirming that ALK1 regulates Smad1 but not Smad5 in response to TGF-β in BAE cells.

FIG 5.

TGF-β regulates Smad1 but not Smad5 in BAE cells. (A) Cells were seeded on glass coverslips. On the next day, cells were either left untreated or stimulated with TGF-β (5 ng/ml) or BMP9 (0.5 ng/ml) for 30 min, then fixed, and stained for the indicated Smad for confocal laser scanning microscopy analysis. (B) The amount of nuclear Smad versus perinuclear Smad was quantitated as described in Materials and Methods. *, P < 0.05; ***, P < 0.001; ****, P < 0.0001 for comparison to CT in each group. (C) Cells were transfected with the indicated siRNA and seeded on glass coverslips. On the next day, cells were either left untreated or stimulated with TGF-β or BMP9 at the concentration indicated for 30 min, then fixed, and stained for the indicated Smad for confocal laser scanning microscopy analysis. (D) Cells were transfected with control siRNA (siCT) or siRNA against Smad1 and then serum starved overnight. On the next day, cells were stimulated with increasing concentrations of TGF-β for 30 min. Total cell lysates were prepared and subjected to immunoblot analysis with the indicated antibodies. Tubulin was used as a loading control. *, nonspecific band. (E) Cells were seeded on culture dishes. On the next day, cells were either left untreated or stimulated with TGF-β (5 ng/ml) or BMP9 (0.5 ng/ml) for 30 min. Whole-cell lysates were prepared, run on a 7% SDS-PAGE gel, and analyzed for Smad1 or Smad5 electrophoretic mobility by immunoblotting with the corresponding antibodies. (F) Cells were transfected with the indicated siRNA, serum starved overnight, and then left untreated or stimulated with TGF-β (5 ng/ml) or BMP9 (0.5 ng/ml) for 30 min. Samples were analyzed by Western blotting as described for panel E. Tubulin was used as a loading control.

TGF-β-mediated regulation of Smad2 and Smad3 promotes podosome formation, whereas TGF-β-mediated regulation of Smad1 reduces it.

Next, the siRNA strategy was employed to address the role of each R-Smad on podosome formation by knocking down Smad proteins individually. Knockdown efficiency was verified by Western blot analysis (Fig. 6A). Smad5 depletion slightly affected Smad1 expression. Likewise, Smad2 depletion affected Smad3 expression. Podosome formation was assessed under each condition. The single depletion of each Smad protein was found to induce a contrasting effect on podosome formation. Depletion of both Smad1 and Smad5 was found to stimulate TGF-β-induced podosome assembly (Fig. 6B). Whereas depletion of Smad1 promoted the formation of well-formed circular podosome rosettes, depletion of Smad5 resulted in disorganized rings (Fig. 6C). In sharp contrast, Smad3 depletion abolished podosome formation (Fig. 6B). Depletion of Smad2 had a weaker inhibitory effect on the process than that of Smad3 (Fig. 6B). The changes in the podosome response were reflected in the extent of matrix degradation (Fig. 6D). In BAE cells transfected with the plasmid encoding CA-ALK5, podosome formation was induced in the absence of TGF-β (Fig. 2H). In CA-ALK5-expressing cells, Smad3 depletion and, to a lesser extent, Smad2 depletion prevented podosome formation (Fig. 6E). The same experiment could not be done for Smad1, as cells with Smad1 depletion did not survive CA-ALK5 transfection. These data indicated that ALK5 is sufficient to induce podosome formation without any requirement for ALK1. In the opposite approach, Smad proteins were overexpressed individually. Podosome formation was induced in cells transfected with either Smad2- or Smad3-encoding plasmid in the absence of TGF-β (Fig. 6F). Adding the cytokine further enhanced podosome induction (Fig. 6F). Although the transfected cells presented abnormal podosome arrangements with loose clusters of podosomes (Fig. 6H), matrix degradation occurred (Fig. 6I). Conversely, overexpressing Smad1 did not enhance but reduced TGF-β-induced podosome formation (Fig. 6G). These results indicated that although Smad1 is regulated by TGF-β, its activation opposes the process of podosome formation regulated by Smad2 and Smad3.

FIG 6.

Smad depletion or Smad overexpression affects TGF-β-induced podosome formation. (A) Cells were transfected with siRNA designed against Smad1, Smad2, Smad3, or Smad5 and then serum starved. On the next day, cells were either left untreated or stimulated with TGF-β (1 ng/ml) for 30 min. Total cell lysates were prepared and subjected to immunoblot analysis with the indicated antibodies. Tubulin was used as a loading control. (B) Cells were transfected with the indicated siRNA and seeded on glass coverslips. On the next day, cells were stimulated with TGF-β (1 ng/ml) for 20 h, then fixed, and subjected to immunofluorescence staining for F-actin and cortactin to score those showing podosomes. *, P < 0.05; ****, P < 0.0001 for comparison to siCT; ####, P < 0.0001 for comparison to siCT stimulated with TGF-β. (C) Representative images of BAE cells transfected with the indicated siRNA and stimulated with TGF-β (1 ng/ml) for 20 h. (D) Cells were transfected with the indicated siRNA and seeded on fluorescent gelatin-coated coverslips. On the next day, cells were stimulated with TGF-β (1 ng/ml) for 20 h and then fixed, and matrix-degrading activity was assessed. *, P < 0.05 for comparison to siCT; ##, P < 0.01; ###, P < 0.001 for comparison to siCT stimulated with TGF-β. (E) BAE cells were transfected with the indicated siRNA and then transfected with the CA-ALK5-encoding plasmid. On the next day, cells were fixed and stained with F-actin and cortactin, and podosome rosettes were scored. Data show results for one experiment representative of three independent ones. (F) BAE cells were transfected with the indicated constructs. On the next day, cells were either left untreated or stimulated with TGF-β (5 ng/ml) for 20 h. The samples were then fixed and stained with F-actin and cortactin, and podosome rosettes were scored. **, P < 0.01; ***, P < 0.001 for comparison to CT. (G) The experiment was carried out as for panel F using a Smad1-encoding construct. **, P < 0.01; ***, P < 0.001. (H and I) Cells were transfected with the indicated construct. On the next day, cells were fixed and subjected to immunofluorescence analysis for F-actin (red), cortactin (green), and gelatin degradation. Representative images of transfected cells displaying podosomes or degradation areas (H) and histograms of the matrix degradation index (I) are shown. *, P < 0.05; **, P < 0.01 for comparison to CT.

Altogether, these results indicated that TGF-β activates Smad2 and Smad3 through ALK5 and regulates Smad1 through ALK1. ALK1 signals mitigate ALK5 signaling in the process of TGF-β-induced podosome formation. Depletion of ALK1 perturbs endothelial cell responses as cells become unable to form podosomes in response to TGF-β.

Physiological relevance of the TGF-β–BMP9 interplay.

To provide physiological relevance to these findings, we explored the effect of neutralizing BMP9 signals on podosome formation in the living endothelium. We reasoned that under normal physiological conditions, the action of TGF-β could be counterbalanced by that of BMP9. Live aortic vessel segments were cultured ex vivo in the presence of exogenously added recombinant TGF-β with or without BMP9 neutralizing antibodies, under conditions similar to those used for cultured BAE cells. Aortic explants were then fixed, stained for F-actin and cortactin, and prepared for en face viewing. In this setting, there was a trend for increasing TGF-β-induced podosome rosette assembly upon exposure of the specimens to BMP9 antibodies (5.03 ± 0.16 [anti-BMP9] versus 3.8 ± 0.05 [control]). Thus, neutralization of ALK1 ligands enhances TGF-β-induced podosome formation in the aortic endothelium.

A physiological situation in which BMP9 signals may be impaired occurs in pathological situations associated with miR-155 overexpression, a condition that may occur in cardiovascular diseases (33, 34). miR-155 directly targets Smad5 (35) and is therefore expected to affect BMP9 signals. The TGF-β-induced podosome response was thus explored with BAE cells that had been transfected with miR-155. Figure 7A shows that both Smad1 and Smad5 expression were reduced in these cells. In agreement with these regulations, TGF-β-induced Smad1/5 phosphorylation was impaired. Under these conditions, ECs became hyperresponsive to TGF-β stimulation for podosome rosette formation (Fig. 7B) and matrix degradation (Fig. 7C), suggesting that they were deprived of BMP9 regulatory signals. In addition, miR-155-transfected cells displayed podosome clusters resembling those induced by Smad2 or Smad3 overexpression (Fig. 6H) even in the absence of TGF-β stimulation (Fig. 7D). Thus, miR-155 appears to be a novel player in the signaling pathways controlling podosome formation in ECs.

FIG 7.

miR-155 expression increases podosome formation and associated matrix degradation. (A) Cells were transfected with control miR or miR-155 and then serum starved overnight. On the next day, cells were either left untreated or stimulated for 30 min with TGF-β (1 ng/ml), and cell extracts were prepared and subjected to immunoblot analysis with the indicated antibodies. Tubulin was used as a loading control. (B) BAE cells transfected with control miR or miR-155 were seeded on glass coated coverslips. On the next day, cells were either left untreated or stimulated with TGF-β (1 ng/ml) for 20 h. After fixation, cells were stained for cortactin and F-actin and podosome rosettes were scored. (C) BAE cells transfected with control miR or miR-155 were seeded on gelatin-coated glass coverslips. On the next day, cells were either left untreated or stimulated with TGF-β (1 ng/ml) for 20 h. After fixation, matrix-degrading activity was assessed. *, P < 0.05; **, P < 0.01. (D) Representative images of miR-155-transfected cells. Cells transfected with miR-155, seeded on fluorescent coated gelatin, and stimulated on the following day with TGF-β (1 ng/ml) for 20 h are shown.

DISCUSSION

The proteomic composition of a cell influences the cellular response to TGF-β signaling (36). Aortic ECs respond to this cytokine by assembling podosome rosettes that may be devoted to ECM remodeling. Although the context in which these structures come into play remains unknown, those associated with local TGF-β activation, such as vessel injury (37) or vessel diameter enlargement (38), might be physiological situations favoring their emergence. More generally, factors that stimulate TGF-β activation, such as angiotensin II, mechanical stress, endothelin-1, high glucose, steroids, and reactive oxygen species, are likely to promote podosome rosette assembly in aortic ECs (37). Deciphering of the signaling pathways involved in their formation holds the key to understanding their role in pathophysiology.

The results presented herein show that TGF-β-induced ALK5 signaling provides the decisive signal for podosome rosette assembly. ALK1-mediated signaling induced by either TGF-β or BMP9 mitigates this ALK5 response. Such modulation has been observed for other ALK5-dependent processes, such as ALK5 target gene regulation (8, 15, 39). However, using specific PSmad1 antibodies, Smad1/Smad5 electrophoretic mobility shift analysis, and selective Smad depletion, we found that ALK1 signaling differed between TGF-β and BMP9. TGF-β induced Smad1 phosphorylation, but this occurred without any detectable effect on Smad5. The use of pharmacological inhibitors confirmed that Smad1 phosphorylation occurred through ALK1. This situation contrasts with the ALK1 signaling triggered by BMP9, in which both Smad1 and Smad5 were phosphorylated. As a consequence, BMP9 exerts a more potent inhibitory effect on TGF-β-induced podosome rosette formation. Thus, the results show for the first time that ALK1 triggering by TGF-β does not elicit a full ALK1 response. TGF-β signals through TβRII in association with ALK1 and ALK5, whereas BMP9 signals through BMPRII (or ActRIIA) in association with ALK1 only. It is likely that the difference observed in signaling is due to the composition of the receptor complexes. A higher expression level of Smad1 than of Smad5 in BAE cells may also contribute to distinct phosphorylation efficiencies for the two proteins. The unique combination of R-Smad regulated by TGF-β provides a situation where the formation of mixed Smad complexes (for instance, Smad1/2 or Smad1/3) may to some extent compromise Smad2/3 activation (40). In this view, depleting Smad1 would promote the formation of Smad2/3 complexes and enhance the TGF-β signal.

Neutralization of BMP9 (naturally present in the serum) with functional antibodies or ALK1-Fc chimera increased the proportion of podosome forming cells both in vitro and ex vivo. However, ALK1 depletion did not induce the same effects as the inhibition of ALK1 signaling. At variance with ligand deprivation, the knockdown of ALK1 promoted the emergence of an activated phenotype. In ALK1-depleted cells, intercellular junctions were disturbed as detected by the mislocalization of VE-cadherin. In addition, the basal activity of some noncanonical pathways was found to be elevated (F. Curado, unpublished data), and these events are likely to contribute to disorganized signaling in these cells. By trapping the cytokine contained in the medium, the quiescence signal provided by BMP9 was suppressed (13, 14), but this did not alter the cell phenotype, as the cell junction defect was not observed. It should be pointed out that the two phenotypes are the results of profoundly different perturbations of ALK1. In the case of neutralizing antibodies or decoy receptor addition, the intracellular domain is free to interact with intracellular proteins, of which the cytoplasmic immunophilin FKBP12 (41), the regulatory subunit of casein kinase 2 β (42), and liver X receptor β (43) have been identified. The entire receptor is missing following the knockdown strategy, and ALK1 intracellular binding partners are released in the cytoplasm. Since ALK1 deficiency was reported to result in elevated vascular endothelial growth factor (VEGF) expression, this factor may be causing the observed defect in cell-cell junctions (24).

The observations made on the native endothelium in vivo confirmed that disturbances in cell-cell junctions are a prominent defect in ALK1-deficient mice. Mutations in the ACVRL1 encoding ALK1 are responsible for HHT2, a disease characterized by telangiectasias (44). These consist of clusters of abnormally dilated thin walled vessels with reduced mural cell coverage that are typically found in the skin and mucocutaneous tissues. The reduction in pericyte numbers appears at least in part attributable to defective vessel maturation. TEM studies have revealed the presence of gaps between microvascular ECs (45). It is conceivable that the small venules of the skin and mucous membranes, which lack other supportive elements, depend heavily upon the perivascular bed for their integrity. Our studies show that the same defects occur in the aortic endothelium: the integrity of large vessels is also disturbed in the disease condition. Such abnormalities do not lead to gross hemorrhage or other detectable manifestations in large vessels, which are better protected by perivascular elements than capillaries. However, it is conceivable that the defect affects interendothelial cell communication and thus cellular functions. The physiological processes involving podosome-forming ECs are coming under intense scrutiny. Podosome formation is likely to be impaired in HHT2 patients, and this defect may also contribute to the pathogenicity of the disease.

Impairment of BMP9 signaling is not restricted to genetic disorders. miR-155 is dysregulated in many pathological states, including vascular diseases (33, 34). Our studies show that expression of miR-155 promotes podosome rosette formation in BAE cells. It is thus possible that increased expression of the miR-155 gene is causally involved in vascular diseases, at least in part, through the upregulation of podosome formation. This situation is reminiscent of that described for vascular smooth muscle cells, in which alterations in miR-143 and miR-145 expression result in the formation of podosomes (46, 47). In addition, Smad5 signals to cells in the basal state. Smads constantly shuttle between the cytosol and nucleus, even in the absence of TGF-β/BMP9 (48). By reducing Smad5 expression, miR-155 is likely to alter cell homeostasis by suppressing a regulatory signal. Lastly, Smad5 is not the only gene in the TGF-β-induced responses targeted by this microRNA. miR-155 was also shown to influence endothelial to mesenchymal transition in ECs (49). Overexpression of miR-155 is known to counteract RhoA function, and this may influence podosome formation (50).

Thus, podosome assembly is critically dependent on Smad3 and is dependent to a lesser extent on Smad2 expression. Structurally, Smad2 and Smad3 are very similar. In vitro, most of the TGF-β responses can be mediated by either Smad. However, Smad2 and Smad3 have distinct patterns of gene activation (51). In addition, animal models have demonstrated the distinct biological functions of these two Smad proteins. Targeted deletion of Smad2 in mice results in early embryonic death, whereas Smad3 knockout mice are viable but die prematurely due to defects in immune function (52, 53). Our knowledge of the specificities of these Smad proteins in their function is limited, so our understanding of their role in podosome rosette formation remains only partial. As the Smad pathway is the primary signaling pathway for most gene responses and because podosome assembly requires protein synthesis (2), it is likely that Smad-dependent transcription controls the expression of key proteins of the podosome response. For instance, MT1-MMP expression is upregulated by TGF-β and contributes to proper podosome rosette assembly (2).

However, neither Smad2 nor Smad3 overexpression was able to induce proper podosome rosette formation. These results highlight the need for both Smad and non-Smad pathways for podosome rosette assembly. The noncanonical pathways are dependent on the type of cells (36). Cdc42 is a master regulator of podosome formation (2, 54), and its regulation occurs through the guanine exchange factor Fgd1 (55). Although Fgd1 is rapidly activated in TGF-β-treated cells (55), podosomes are detected after several hours and protein synthesis is involved, pointing to the contribution of Smad in the process. BMP9 affects podosome formation by altering the pattern of activated Smads. Further work will explore whether BMP9 regulatory functions are also involved at the level of the noncanonical pathways.

In conclusion, our results show that TGF-β stimulates both ALK5 and ALK1 signaling. ALK5 is the TRI driving podosome formation, whereas ALK1 fine-tunes ALK5 signals. Moreover, TGF-β-induced ALK1 stimulation does not trigger the full BMP Smad response, as observed with BMP9. Finally, ALK1 depletion does not produce the same effects as ALK1 neutralization. Future studies will explore a possible link between ALK1 deficiency and the process of endothelial-to-mesenchymal transition. Our results shed light on the functions of ALK1 based upon in vitro studies in ECs and phenotypic analysis of ALK1^+/− heterozygous mice, thus providing novel insights into the pathophysiology of HHT2 disease.