BMPR2 is required for zebrafish CVP sprouting angiogenesis and human EC filopodia formation
Sprouting angiogenesis from the zebrafish axial vein is dependent on BMP and independent of VEGF-A (50). During embryogenesis, zebrafish express two BMP type II receptors, bmpr2a and bmpr2b that could mediate this effect (50, 51). To assess this possibility, we depleted Bmpr2b by injections of an antisense morpholino oligo (MO) into the vascular reporter transgenic line Tg(kdrl:GFP)s843 which substantially reduced bmpr2b spliced transcript levels at 8 hours post fertilization (hpf) (Fig. 1 ̶ figure supplement 1). By 25 hpf, bmpr2b morphants had significantly fewer sprouts protruding from the CVP (Fig. 1a, b). This specific phenotype was also present at 26 hpf and resulted in a deficient plexus formation at 32 hpf (Fig. 1 ̶ figure supplement 2) suggesting it was not simply caused by a developmental delay since other hallmarks of development were not affected. In accordance with Wakayama et al. (43, 50),we show that BMP signaling via bmpr2b regulates zebrafish CVP sprouting, which is an indispensable mechanism for vascular plexus formation in this region.
To next address whether BMPR2 deficiency impairs the typical migratory phenotype of human ECs, we took advantage of a previously established immortalized human umbilical vein EC (HUVEC) cell-line depleted by CRISPR/Cas9 for BMPR2 expression (BMPR2+/-), resulting in BMPR2 haploinsufficiency (53, 54) (Fig. 1 ̶ figure supplement 3). To complement the EC line, we used primary HUVECs, treated with self-delivering siRNA to efficiently knock down (KD) BMPR2 (Fig. 1 ̶ figure supplement 3). To mimic typical TC-related sprouting characteristics in 2D in-vitro (i.e. polarization, filopodia formation and migration), we used seeding inserts. Upon removal, they leave behind a cell-free gap towards which ECs can migrate by polarizing and forming a leading edge in the presence of pro-angiogenic cues delivered by the EC-activation medium (Fig. 1c). In this approach, the first row of “leader cells” adopts a TC-like behavior, notably recapitulating filopodia formation at the leading edge. F-actin staining of leader cells revealed that BMPR2+/- ECs display significantly less filopodia per cell edge when compared to their wild-type (WT) counterparts (Fig. 1d, upper). This finding was observed also at higher magnification by Scanning Electron Microscopy (SEM) (Fig. 1d, lower). In line with this, siRNA-mediated KD of BMPR2 in HUVECs also resulted in a significantly reduced number of filopodia or even a filopodia-free morphology at the leading edge (Fig. 1e). However, in some cells, a few isolated filopodia remained upon BMPR2 depletion. High throughput analysis using FiloQuant (55) did not detect significant differences in the length of the few remaining filopodia neither in siBMPR2-treated HUVECs nor in BMPR2+/- cells when compared to controls (Fig. 1 - figure supplement 4). Together this shows that immortalized ECs and primary HUVECs both depend on BMPR2 for filopodia formation of leader cells migrating in 2D. We next aimed to rescue the filopodial phenotype of BMPR2+/− ECs by re-expressing BMPR2 as a GFP-tagged fusion protein. BMPR2-GFP overexpression effectively rescued filopodia deficiency in BMPR2+/− ECs when compared to controls transfected with GFP only (Fig. 1f). Moreover, excessive filopodia formation was observed when BMPR2-GFP was ectopically expressed in primary HUVECs (Fig. 1 - figure supplement 5). We furthermore found BMPR2 foci within filopodia but also at their tips and close to sites where filopodia originate (herein referred to as filopodia base) (Fig. 1f). The same result was observed and quantified for ectopic BMPR2-GFP expression in the parental EC line (BMPR2wt) (Fig. 1- figure supplement 6). This shows that in both EC model systems, ectopic expression of BMPR2 positively correlates with greater number of filopodia. Interestingly, the transfection of different amounts of BMPR2-GFP plasmid positively correlated with increased appearance of so called “migration tracks” (aka. cellular remnants). This phenomenon of cells migrating in-vitro occurs during rear-end retraction and indicates strong adherence of cells to their substratum due to enrichment in FAs or disturbed FA disassembly mechanisms (56) (Fig. 1- figure supplement 7). To investigate the role of BMP-induced SMAD signaling in BMPR2-mediated EC filopodia formation, we overexpressed BMPR2 under concomitant inhibition of pan type I receptor kinase activity, required for SMAD activation, using the small molecule inhibitor LDN-193189 (LDN) (57) (Fig. 1- figure supplement 8). BMP type I receptor inhibition led to strong reduction in basal SMAD phosphorylation when cells were cultured in EC activation media supplemented with LDN-193189 (Fig. 1- figure supplement 8, left). BMPR2-overexpressing ECs showed significantly increased filopodia formation even in the presence of LDN (Fig. 1- figure supplement 8, right). This suggests that BMPR2-dependent filopodia formation does not directly rely on the type I receptor kinase activity, which is required for canonical SMAD signaling, and that BMPR2-dependent filopodia formation might rely on non-canonical BMP signaling.
Our previous investigation on BMPR2 subcellular localization (Fig. 1f) suggested its membrane presence along all regions of the filopodia. We next investigated further on BMPR2 subcellular localization, particularly its membrane mobility in proximity to cell edges of living cells. Membrane mobility of distinct BMPRs can report on their respective involvement in SMAD vs. non-SMAD signaling complexes (58). For this we performed single particle tracking (SPT) microscopy using quantum-dot (QDot)-labelled HA-tagged BMPR2 in living cells (58, 59). After successful validation of BMPR2 labelling efficacy (Fig. 1- figure supplement 9), SPT analysis for single QDot-labelled BMPR2 molecules (Fig. 1g, left) revealed several BMPR2 foci at the cellular periphery (F-actin labelled by LifeActGFP), close to filopodia base. Confinement of mobile BMPR2 upon addition of ligand indicates BMPR2 in complex with BMP and BMP type I receptors, forming active signaling complexes termed BMP-induced signaling complexes (BISCs) (58, 60). We could previously show, that BISC activation is associated with non-canonical signaling responses (58). To prove that the observed mobility behavior of BMPR2 at filopodial bases is indeed indicative for such BISC formation and in line with our previously published data (58), we induced confinement of mobile, i.e. ligand-unbound BMPR2 receptors in starved cells, upon pulsed stimulation with BMP-2 (Fig. 1g, right). BMP2 stimulation led to a reduction of BMPR2 lateral mobility at the filopodia base as shown by shift of individual receptor displacements (Fig. 1g, right). This result is indicative for a confinement of BMPR2 induced by BMP2 and enrollment of BMPR2 into BISC signaling complexes at those sites. Together these data show that BMPR2 is located in all filopodia regions, although with different lateral mobilities, revealing a particular involvement of BMPR2 in BMPR signaling complexes at the filopodia base related to non-canonical signaling.
In sum, data from Fig. 1 show that BMPR2 expression levels correlate with ECs ability to form filopodia at their leading edge using two independent cellular and one organismal model system. Furthermore, BMPR2 is present throughout the filopodial membrane but adopts a primarily confined lateral mobility at the base of the filopodia upon BMP stimulation, indicative for BISC formation and non-canonical signaling emanating at those sites.
Loss of BMPR2 alters EC migration dynamics, tube-formation and focal adhesion morphology accompanied by defects in spatial actomyosin organization
While BMP-2 and other BMP family members facilitate EC motility (13), their role as classical angiogenic guidance cues is still under debate, despite data showing chemotactic responses of ECs towards BMP gradients (24, 30). An interesting hypothesis is that instead of directly inducing EC polarity and migration like typical endothelial guidance cues, BMPs promote only the migratory velocity of cells that already display an established front-to-rear end polarity. To investigate on this further, we performed dynamic analysis of gap closure migration assays under EC activating medium conditions. We found that BMPR2+/− cells migrate with reduced efficacy into the cell-free gap when compared to controls (Fig. 2a). However, more revealing was dynamic cell migration component analysis on image stacks of wildtype (green) and BMPR2+/− (magenta) ECs simultaneously migrating over 16 hrs within the same gap-field-of-view (Fig. 2b, upper left). Particle image velocimetry analysis applied to these data (Fig. 2b, lower left) shows that movements of BMPR2+/− cells appear in larger clusters, with individual cells within the cluster adopting the same directionality, creating a swirl-like displacement pattern (Fig. 2b, inlet I). On the other hand, WT cells appeared to form only small clusters and displayed lesser cluster dependency in their directionality of movement (Fig. 2b, inlet II). Subsequent migration component analysis, in which all trajectories were decomposed into directional (velocity) and random (diffusivity) contributions to cell migration (61), confirmed reduced migratory diffusivity of BMPR2+/− ECs with reduced forward displacements of cells over time (Fig. 2b, right). Taken together, these dynamic migration data suggest an increased collective cell migration in 2D for ECs lacking BMPR2. This observation may be a consequence of increased cell-to-cell contacts during movement which reduces the overall persistence in migratory directionality and speed of single cells to a level impeding with efficient individual cell forward movement (Fig. 2b, right). Conversely, WT cells migrated with lesser interdependency displaying sustained speed and pronounced forward directionality. This becomes particularly clear when seeding both cell types (BMPR2wt green, BMPR2+/− magenta) in equal amounts as a dual color mosaic within the same gap-field-of-view (supplementary Movie 2). Here WT cells (green) are preceding BMPR2 deficient cells (magenta) during gap closure suggesting an impaired leader-cell phenotype of BMPR2 deficient cells. These data are further underlined by confocal analysis of filamentous actin (F-actin) stained ECs at the migrating front showing that indeed BMPR2+/− cells display increased cell-to-cell contact sites and lack of free space between individual leader cells (Fig. 2 - figure supplement 1). Exacerbated collective cell migration of ECs may be a functional consequence of changes in the cell-to-cell adhesion forces (62) facilitating cellular contractility (63, 64). The Rho-ROCK-myosin II pathway acts upstream of actomyosin mediated cell-contractility (65, 66). Investigating the relative localization of active, i.e. phosphorylated myosin light chain (pMLC) to cortical actin bundles at the cells leading edge revealed altered co-localization of pMLC and cortical F-actin in BMPR2+/− cells compared to controls (Fig. 2c). For this, we measured relative intensity profiles from the proximal leading edge towards a more distal side of the cell. In WT ECs, typical cortical F-actin bundles are seen at the cell periphery and the actomyosin rich zone (green) located more segregated inwards within the region of the lamellum (Fig. 2c, left). In contrast, pronounced co-localization of F-actin bundles with pMLC containing actomyosin was found at the very periphery of BMPR2+/− cells, having less distance from the leading edge cell membrane (Fig. 2c, left). This is indicating a change in spatial organization of the contractile actomyosin cytoskeleton in the absence of BMPR2 expression in ECs (Fig. 2c, right). To next gain more information on the adhesive properties of BMPR2+/− cells, we investigated the organization of vinculin-rich FAs, which displayed decreased circularity in case of BMPR2 loss (Fig. 2 - figure supplement 2). However, it appeared in 2D only when cells were seeded on stiff glass substrate but not on softer polydimethylsiloxane (Fig. 2 - figure supplement 2). From this we concluded, that the role of BMPR2 in modulating EC mechanical behavior during angiogenic processes should be further studied in physiologically more relevant 3D models, also providing a more complex ECM environment.
Albeit not recapitulating essential aspects of angiogenic sprouting, tube formation is still used to analyze the angiogenic potential of cells and stimuli. Computational models suggest that in the absence of TC phenotype, ECs still form blood vessel-like structures, however displaying abnormal morphology (67). We thus speculated that tube capillary network formation may be affected upon BMPR2 deficiency. Accordingly, we performed mosaic tube formation assays by seeding equal amounts of wildtype (green) and BMPR2+/− cells (magenta) together on soft Matrigel™. We opted for this mosaic approach to reveal if BMPR2 promotes certain occupancy of ECs within the tube-network due to different migratory and adhesive properties as suggested by the findings before. At early time points after seeding, BMPR2-deficient ECs (magenta) preceded WT ECs in early cell-to-cell contact events (1.5–4 hrs). These are required for initiation of fusion and first immature capillary tube-network formation. Only at later stages of tube formation (6 hrs), WT ECs (green) also participated in the tube network (Fig. 2- figure supplement 3) particularly when tube-sizes increased and network matured. This suggests that in a softer 3D ECM environment, BMPR2 deficiency affects angiogenesis by increasing cell-to-cell adhesion e.g. during tube network formation.
BMPR2 regulates spatial organization of pulling forces in nascent sprouts revealed by a 3D in-vitro model combined with traction force microscopy
Since our 2D findings suggested altered actomyosin organization at the leading edge of migrating BMPR2+/− leader cells, we decided to further investigate the mechanical consequences of BMPR2 deficiency, particular on TC pulling forces during 3D angiogenic sprouting. For this, we applied traction force microscopy (TFM) with sprouting ECs originating from spheroids in a 3D fibrin-based hydrogel with a fibrin concentration of 2.5mg/ml (3D TFM). Wildtype or BMPR2+/− EC spheroids were embedded in fibrin gel forming higher ordered ECM network (68) (Fig. 2- figure supplement 4). Fibrin hydrogels have superb optical properties for imaging, mimic the in-vivo microenvironment of immediate wound healing after blood coagulation and recapitulate major aspects of the transitional ECM in regenerative processes which was shown to provide a beneficial environment for sprouting angiogenesis (69). The embedding of fluorescent fiducial markers within the gels (Fig. 2- figure supplement 4), allowed subsequently the tracking of ECM displacements generated by the pulling forces of nascent sprouts upon experimentally induced cell relaxation (Rock inhibitor (Y-27632) treatment). To our knowledge, the combination of using fibrin hydrogels with 3D TFM on EC spheroids was not performed before (Fig. 2d). Rho kinase inhibition led to cellular relaxation and sprout collapse with similar kinetics compared to Cytochalasin-D treatment (supplementary Movie 3). We first set out to investigate ECM displacements by sprouts emanating from spheroids which were generated by either BMPR2wt or BMPR2+/− ECs (Fig. 2d). With this approach we found for both BMPR2wt and BMPR2+/− sprouts a typical peak of displacements around sprout tips, where pulling by the TC is expected to occur prominently. The observed displacements were in the range of 1–3 µm and in accordance with previous data gained from WT ECs sprouting in a different but related technical setup (41). The data shown in Fig. 2d suggested a different displacement pattern of BMPR2 deficient sprouts as compared to WT controls. However, we observed a high level of data variability in 3D TFM experiments depending on the morphology of the sprout and the level of sprout matureness at the timepoint of measurement as well as the overall sprout density. We were therefore concerned about the reproducibility of those data and performed > 80 experimental replicates per condition to address this assay-dependent variability and to be able to average the data. By analyzing the average 3D TFM displacements along the normalized distance of sprouts of various length, we found that the averaged 3D TFM displacements do not differ between BMPR2wt ECs and BMPR2+/− ECs (Fig. 2e and Fig. 2- figure supplement 5). This result suggests that a loss of BMPR2 does not generally interfere with the ECs ability to pull and interact with the ECM, and that TC pulling is also possible in the absence of BMPR2 expression by the TC.
BMPR2 is required for efficient sprout outgrowth and determines tip-cell position in EC spheroids
Our previous 2D characterization of BMPR2-deficient ECs suggested impaired phenotypical features of leader cells including lack of filopodia formation, impaired forward movement during cell migration and altered actomyosin organization at the leading edge. While our 3D TFM experiments did not allow to identify significantly altered TC tractions upon BMPR2 depletion, the 3D spheroid sprouting assay in fibrin gels still proved beneficial to better understand the role of BMPR2 in sprouting angiogenesis, particularly the need for BMPR2 expression by ECs who aim to acquire the TC position in a developing sprout. Therefore, we first quantified and compared the sprouting areas of BMPR2wt and BMPR2+/− spheroids for a period of 60 hrs. Spheroid sprouting assay showed a reduced outgrowth area for BMPR2+/− cells when compared to the sprouting area of WT controls (Fig. 3a and supplementary Movie 4) suggesting a lack of efficient sprout elongation over time. Kinetic analysis of sprouting area progression showed that WT cells increase in elongation speed at about half of the experimental time when compared to their BMPR2 deficient counterparts (Fig. 3b). This finding was also recapitulated when analyzing the average distance of TCs from their sprout origins (Fig. 3b). Moreover, after investigating TC morphology more thoroughly, we found that BMPR2 deficient TCs harbor less protrusive structures in 3D in agreement with our 2D data (Fig. 3c). Analysis of TC morphology, as performed by measuring their “solidity” (70) revealed that BMPR2 deficient TCs adopt a conical shape with a “bullet-like” morphological appearance. During sprouting angiogenesis, it is known that SCs dynamically compete for the TC position over the course of sprouting (71). Why this “tug-of-war” of SCs and TCs for the relative positions within a sprout is required for efficient sprout outgrowth is not completely understood (71). Eventually, this is decisive for sprouts to perform efficient sprouting angiogenesis in a complex 3D environment. Thus, we next investigated whether BMPR2-deficient cells that are devoid of filopodia would still acquire TC position when seeded as a 1:1 mosaic with WT ECs. Strikingly, after 36 hrs of sprouting, sprouts from mosaic spheroids (BMPR2wt, green; BMPR2+/−, magenta) were significantly devoid of BMPR2+/− cells (magenta) in TC position (Fig. 3d). Dynamic quantification of sprout composition at 16 hrs, 36 hrs and 60 hrs confirmed that TC position is almost systematically occupied by BMPR2wt ECs (Fig. 3e and supplementary Movie 5).
Together our data show that BMPR2 expression is required for efficient sprouting angiogenesis of EC spheroids in 3D fibrin matrix. While ECs lacking BMPR2 expression would still form some sprouts, the overall number of sprouts, their protrusiveness and their sprouting area are largely reduced indicating lack of efficient sprout outgrowth in the absence of BMPR2 expression. However, during the entire sprouting process the TC position is occupied by BMPR2 expressing TCs which cannot be replaced by a BMPR2 deficient SC when seeded as a mosaic approach. Thus, BMPR2 expression may be required for dynamic SC to TC exchange at the sprout front. We confirmed our experimental setup by inverting the dyes used to label the cells, with BMPR2wt ECs now labeled in magenta and BMPR2+/− cells now labeled in green, to exclude any effects of the dyes on the sprouting behavior and phenotypes observed (Fig. 3- figure supplement 2). For nearly all investigated sprouts, we again found BMPR2 expressing cells in the TC position (Fig. 3d and Fig. 3- supplementary Fig. 2, right), underlining the necessity for BMPR2 expression to acquire TCs characteristics in this experimental setting.
BMPR2 governs filopodia formation via CDC42 regulation and interacts with the CDC42 effector and actomyosin regulator BORG5
The phenotypical and functional characterization presented in Figs. 1–3 unambiguously demonstrates the indispensability of BMPR2 for EC filopodia formation, EC migration and the spatial regulation of actomyosin contractility. These findings also underscore the pivotal role of BMPR2 expression facilitating ECs to acquire TC position during angiogenic sprouting. However, the underlying molecular mechanism through which BMPR2 acts as a central regulator, upstream of these biochemical and biomechanical processes, remain less elucidated. Filopodia formation in TCs critically depends on the activity of the small Rho GTPase CDC42, as established by previous studies (72–75). Consistent with these observations, the EC-specific deletion of CDC42 has been shown to impede filopodia formation and lead to aberrant sprouting angiogenesis (44), a phenotype recapitulated in our BMPR2+/− cell model. This prompted us to hypothesize, that BMPR2 might play a role in regulating CDC42 activity in ECs. To investigate this hypothesis, we first examined if expression of constitutively active (c.a.) CDC42-GFP fusion protein could rescue filopodia loss in BMPR2+/− ECs. The introduction of c.a.-CDC42-GFP successfully induced robust filopodia formation in BMPR2 deficient cells that would otherwise show the typical blunted membrane appearance devoid of filopodial protrusions (GFP-only control) (Fig. 4a). The mechanism by which BMPs regulate Rho GTPase signaling including CDC42 activation is not entirely understood. Some data from cell types other than ECs suggest that SMAD signaling may transcriptionally regulate Rho-guanine nucleotide exchange factors (GEFs) (76). However, our own data (Fig. 1- figure supplement 8) suggested that SMAD signaling may not be involved. Moreover, previous kinetic studies by others analyzing Rho GTPases activation by BMPs argue for a quick activation within seconds-to-minutes, suggesting no need for transcriptional activity of the BMP pathway (49, 77–79). To exclude general transcriptional changes in CDC42 expression, we assessed CDC42 protein levels, which showed no change between BMPR2+/− ECs compared to controls (Fig. 4- figure supplement 1). Thus, we speculated that BMPR2 dependent regulatory mechanisms independent of gene transcription are acting upstream of CDC42 activity. Along these lines, we previously demonstrated that BMPR2 is a regulator of PI3K activity, leading to BMP-induced phosphatidylinositol-3,4,5-trisphosphate (PIP3) production at the cell’s leading edge in the minutes range (21). Membrane-bound PIP3 serves as a second messenger for Rho-GEF recruitment and subsequent CDC42 activation (80). Recently, it was found that the strongest GEFs facilitating endothelial CDC42 activity belong to the Pleckstrin Homology (PH) and RhoGEF Domain containing Family G (81) with PH-domains required to tether GEFs to PIP3. To confirm alterations in PI3K signaling in BMPR2-deficient HUVECs (siBMPR2) and in BMPR2+/− ECs, we assessed the phosphorylation of the PH-domain protein Akt, indicative for active PI3K-PIP3 signaling (82). When cultured in presence of EC activation media, we observed reduced Akt phosphorylation in siBMPR2-treated HUVECs and BMPR2+/− ECs compared to controls. Furthermore, Akt phosphorylation was further diminished upon treatment with LY294002, confirming impaired PI3K- PIP3 -PH-domain protein signaling under BMPR2 deficiency (Fig. 4b). Immunostaining revealed that CDC42, like BMPR2, localizes in close proximity to filopodial protrusions (Fig. 4c, magenta). Notably, overexpressed HA-tagged BMPR2 locates at the same foci as CDC42 (Fig. 4c, green), suggesting spatial proximity and a potential regulatory interplay between these two proteins. To substantiate the notion that BMPR2 is essential for regulating CDC42 activity, we utilized a CDC42 single-chain FRET biosensor, providing a readout of CDC42 activity in living cells (83) (d4D). Expression of the CDC42 biosensor in BMPR2wt and BMPR2 +/− cells revealed significantly reduced CDC42 sensor activity in BMPR2+/− ECs compared to BMPR2wt controls (Fig. 4d, lower). This biosensor is primarily anchored to the cell plasma membrane via a myristoyl group (Fig. 4d, lower right). Our data therefore indicate that the presence of BMPR2 in the EC plasma membrane positively correlates with CDC42 activation, possibly through a PI3K-PIP3 dependent mechanism. Additionally, both BMPR2 SPTM analysis and the localization of CDC42 and BMPR2 at the filopodia base imply a local interdependency among BMPR2-containing BISC localization, CDC42 activation and filopodia formation.
Previous proteomics approaches (84) aiming to identify novel BMPR2 interacting proteins suggested that the CDC42-effector protein1 (EP1), also known as binder of Rho GTPase 5 (BORG5), interacts with the cytosolic domain of BMPR2 (85). Interestingly, BORG5 was shown before to promote angiogenesis by regulating persistent directional EC migration (86), reminiscent of the migratory data presented here in Fig. 2b. Importantly, loss of endothelial BORG5 also results in disorganized spatial localization of actomyosin, mirroring the findings reported in our Fig. 2c (86). BORG5 binds active i.e. GTP-bound CDC42 (87), participating in regulation of the Septin cytoskeleton, which, in turn, governs actomyosin contractility (87). Due to these obvious overlaps of reported BORG5 features with our here reported data, we investigated whether there might be increased spatial proximity between BORG5 and BMPR2. For this, we co-expressed carboxy-terminally Myc-tagged BMPR2 and GFP-tagged BORG5 and used proximity ligation assay (PLA) to visualize their interaction in intact cells. PLA in HUVECs expressing these tagged proteins indeed revealed a close spatial proximity between BMPR2 and BORG5, primarily at the filopodia base, but also some at shaft, and tips (Fig. 4e, left). Conversely, no PLA signal (magenta) could be detected when only one interaction partner was expressed (Fig. 4e, right).
In summary, our mechanistic findings support a model in which BMPR2 governs filopodia formation, a crucial aspect of TC function and positioning during angiogenic sprouting, presumably through a PI3K- dependent pathway which involves activation of CDC42. Furthermore, in search for BMPR2-specific signaling proteins involved in transducing CDC42 activity towards spatial organization of F-actin, FAs and actomyosin we propose the CDC42 effector BORG5 as a new BMPR2 interacting (cytoskeletal) protein (see Fig. 5; graphical summary).