Summary

VE-cadherin-mediated cell–cell junction weakening increases paracellular permeability in response to both angiogenic and inflammatory stimuli. Although Semaphorin 3A has emerged as one of the few known anti-angiogenic factors to exhibit pro-permeability activity, little is known about how it triggers vascular leakage. Here we report that Semaphorin 3A induced VE-cadherin serine phosphorylation and internalisation, cell–cell junction destabilisation, and loss of barrier integrity in brain endothelial cells. In addition, high-grade glioma-isolated tumour-initiating cells were found to secrete Semaphorin 3A, which promoted brain endothelial monolayer permeability. From a mechanistic standpoint, Semaphorin 3A impinged upon the basal activity of the serine phosphatase PP2A and disrupted PP2A interaction with VE-cadherin, leading to cell–cell junction disorganization and increased permeability. Accordingly, both pharmacological inhibition and siRNA-based knockdown of PP2A mimicked Semaphorin 3A effects on VE-cadherin. Hence, local Semaphorin 3A production impacts on the PP2A/VE-cadherin equilibrium and contributes to elevated vascular permeability.

Introduction

Endothelial cells that mat the inner side of blood vessels tightly regulate vascular integrity and homeostasis. Cell–cell junction leakiness is believed to orchestrate the paracellular pathway, which allows plasma molecules and cells to pass between endothelial cells (Vestweber et al., 2009). Two cell–cell junction proteins, VE-cadherin and claudin-5, are essential for ensuring vascular integrity (Taddei et al., 2008). VE-cadherin belongs to a conserved family of calcium-dependent homophilic adhesion proteins, and is a fundamental constituent of adherens junctions. Indeed, VE-cadherin forms a molecular bond connecting the extracellular space, the plasma membrane and the intracellular cytoskeleton (Dejana, 2004). Its pivotal role in vascular integrity and angiogenesis is supported by the embryonic lethality observed in VE-cadherin knockout mice as a result of vascular insufficiency (Carmeliet et al., 1999). Conversely, vascular permeability is circumvented by VE-cadherin adhesion during the course of adult neovascularisation or in response to inflammation (Schulte et al., 2011; Crosby et al., 2005). Coordinated disruption of VE-cadherin adhesive function through dual tyrosine and serine phosphorylation, internalisation, and mechanical forces, converges on cell-cell junction weakening and vascular leakage (Eliceiri et al., 1999; Gavard, 2009; Gavard and Gutkind, 2006; Murakami et al., 2008; Stockton et al., 2004). Vascular permeability is frequently elevated under diverse pathological conditions including cancer, and serves notably as a diagnostic and prognostic parameter in brain tumours (Sorensen et al., 2009). Indeed, remodelling of VE-cadherin-based junctions has been observed within the tumour microenvironment, ultimately corroborating with tumour growth, vascularisation and spreading (Le Guelte et al., 2011; Dwyer et al., 2011; Lin et al., 2007; Weis et al., 2004).

Recently, the semaphorin guidance molecules have been proposed to regulate endothelial migration and tumour angiogenesis by providing both repulsive and attractive signals (Serini et al., 2009). Endothelial type A/D Plexins (Plx), and Neuropilins (NRP-1 and NRP-2), which are co-opted by vascular endothelial growth factor (VEGF), serve as transducer and binding receptors, respectively. Notably, class III semaphorins including, Semaphorins 3A to G, share the unique ability to hamper vascular formation and tumour vascularisation (Gaur et al., 2009). Although PlxA1 and NRP-1 likely convey Semaphorin 3A (S3A) anti-angiogenic actions in endothelial cells, the downstream intracellular signalling mechanisms remain hypothetical. It was proposed that S3A stimulation might shift NRP-1 association with VEGF-R2 to the Plx pool (Acevedo et al., 2008; Guttmann-Raviv et al., 2007). Under these conditions, S3A attenuates VEGF-R2-based angiogenesis, and yet, does not subvert all VEGF signalling function. Accordingly, S3A and VEGF cooperate to induce vascular permeability and enhance VE-cadherin tyrosine phosphorylation (Acevedo et al., 2008). Consistently, S3A overexpression reduces both pericyte coverage of tumour vessels and metastasis (Casazza et al., 2011). A role for S3A in tumour vasculature normalisation has also been reported together with its anti-angiogenic activity (Maione et al., 2009). Thus, exactly how S3A coordinates endothelial barrier function remains obscure.

In this study, we co-cultured human cerebral microvascular endothelial cells that phenocopy the blood brain barrier, with patient-derived brain tumour stem-like cell lines (Galan-Moya et al., 2011). We found that these glioma stem-like cells (GSCs) induced profound remodelling of the endothelial monolayer. This essentially resulted from GSC-produced S3A and required the PlxA1 transducing receptor. Both recombinant and GSC-derived S3A promoted VE-cadherin serine phosphorylation and internalisation, which strictly correlated with cell–cell junction destabilisation and elevated endothelial permeability. Furthermore, while the serine/threonine phosphatase PP2A was found constitutively bound to VE-cadherin in quiescent endothelial cells, stimulation with S3A transiently disrupted this complex, thereby allowing VE-cadherin phosphorylation. Hence, our data suggest that perturbation of the finely tuned PP2A/VE-cadherin balance might support tumour vessel abnormalities, such as heightened vascular permeability.

Results

Semaphorin 3A promotes loss of brain endothelial cell monolayer integrity

Besides its canonical repulsive activity in neurons, Semaphorin 3A (S3A) acts as anti-angiogenic factor with pro-permeability activity in vessels (Acevedo et al., 2008; Casazza et al., 2011). Indeed, S3A local injections in mice significantly increased vascular permeability by 4-fold, as assessed by Evans blue extravasation (Fig. 1A). When administered in vitro to human cerebral microvascular endothelial cells, recombinant S3A dose-dependently increased passage of fluorescent dextran through monolayers (Fig. 1B). Of note, a plateau was reached at 100 ng/ml; a concentration found below toxicity (supplementary material Fig. S1). Confocal microscopy analysis of brain endothelial junctions revealed that upon S3A treatment, the adherens junction molecule VE-cadherin, as well as its cytoplasmic partners, β- and p120-catenins, evolved from straight lines to a more punctuated pattern at the cell–cell interface (Fig. 1C). We then aimed to study VE-cadherin serine phosphorylation and internalisation upon S3A stimulation, as it has been previously observed that these events play a key role in controlling endothelial barrier function (Gavard and Gutkind, 2006). Indeed, VE-cadherin-containing cell–cell junction reorganization correlated with rapid, as early as 5 minutes, and transient phosphorylation on serine (Fig. 1D). This was accompanied by a diminution of VE-cadherin exposed at the cell surface, as evaluated by surface biotinylation assay (Fig. 1E). Furthermore, S3A induced VE-cadherin internalisation into compartments (Fig. 1F), that partially colocalised with both Early Endosomal Antigen 1 and adaptin α, suggesting that internal VE-cadherin accumulated in endocytic coated-vesicles (supplementary material Fig. S2). Consistent with the actions elicited by recombinant S3A, conditioned medium collected from S3A–GFP-expressing HEK-293T heightened both VE-cadherin internalisation and brain endothelial cell permeability (Fig. 1G,H,I). Altogether, our data suggest that recombinant and secreted S3A disrupt VE-cadherin-mediated junctional organisation.

Fig. 1.

Semaphorin 3A promotes loss of brain endothelial cell monolayer integrity. (A) Mice were injected with S3A (50 ng/ml) or PBS as a control, in duplicate, in each flank. The images show the Evans blue extravasation at the site of injection, 1h post-treatment. The graph shows the mean of Evans blue extravasation (n = 8). (B) Three-day-old confluent human brain endothelial cells (ECs) were stimulated with Semaphorin 3A (S3A) at the indicated doses for 1 h. FITC-dextran permeability is expressed as fold increase with respect to untreated cells. (C) Confocal microscopy analysis of VE-cadherin (VEC), β-catenin (β-cat) and p120-catenin (p120-cat) staining in untreated (ctrl) and S3A (100 ng/ml, 30 min)-stimulated ECs. Scale bar: 10 µm. (D,E) ECs were exposed to S3A for the indicated times and levels of phospho-S665 (p) and cell-surface (surf.)-exposed VE-cadherin (VEC) were tested by western blotting. Total VEC was used as a loading control. (F) Confocal analysis of internalised VE-cadherin staining after an acid wash (iVEC, green) in control (ctrl) and S3A (100 ng/ml, 15 min)-stimulated ECs. Nuclei are stained blue (DAPI). Scale bar: 10 µm. The number of cells exhibiting vesicular staining was scored in ctrl and S3A-treated ECs for the indicated times and expressed as the percentage of the total cells; n = 300. (G) Myc-GFP (GFP)- or S3A-GFP (S3A)-transfected HEK-293T were analysed for GFP expression by western blotting. (H,I) ECs were stimulated with conditioned medium (CM) prepared from GFP- and S3A-transfected HEK-293T. VEC internalisation (iVEC) (H) and FITC-dextran permeability (I) were monitored.

Fig. 1.

Semaphorin 3A promotes loss of brain endothelial cell monolayer integrity. (A) Mice were injected with S3A (50 ng/ml) or PBS as a control, in duplicate, in each flank. The images show the Evans blue extravasation at the site of injection, 1h post-treatment. The graph shows the mean of Evans blue extravasation (n = 8). (B) Three-day-old confluent human brain endothelial cells (ECs) were stimulated with Semaphorin 3A (S3A) at the indicated doses for 1 h. FITC-dextran permeability is expressed as fold increase with respect to untreated cells. (C) Confocal microscopy analysis of VE-cadherin (VEC), β-catenin (β-cat) and p120-catenin (p120-cat) staining in untreated (ctrl) and S3A (100 ng/ml, 30 min)-stimulated ECs. Scale bar: 10 µm. (D,E) ECs were exposed to S3A for the indicated times and levels of phospho-S665 (p) and cell-surface (surf.)-exposed VE-cadherin (VEC) were tested by western blotting. Total VEC was used as a loading control. (F) Confocal analysis of internalised VE-cadherin staining after an acid wash (iVEC, green) in control (ctrl) and S3A (100 ng/ml, 15 min)-stimulated ECs. Nuclei are stained blue (DAPI). Scale bar: 10 µm. The number of cells exhibiting vesicular staining was scored in ctrl and S3A-treated ECs for the indicated times and expressed as the percentage of the total cells; n = 300. (G) Myc-GFP (GFP)- or S3A-GFP (S3A)-transfected HEK-293T were analysed for GFP expression by western blotting. (H,I) ECs were stimulated with conditioned medium (CM) prepared from GFP- and S3A-transfected HEK-293T. VEC internalisation (iVEC) (H) and FITC-dextran permeability (I) were monitored.

Glioma stem-like cell-secreted Semaphorin 3A remodels brain endothelial cell monolayers

Because S3A may target different cell types and be produced within the tumour microenvironment (Acevedo et al., 2008; Catalano et al., 2006; Maione et al., 2009; Maione et al., 2012), we analysed its expression pattern in glioma cell lines and glioma stem-like cells (GSCs), isolated from high-grade brain tumours (Patru et al., 2010), by confocal microscopy, RT-PCR and immunoblotting (Fig. 2A–D). S3A mRNA and processed protein were detected in four patient-derived GSCs grown as neurospheres (Fig. 2A–C). Lower levels of S3A messenger were detected in several human glioma lines and in normal human foetal glial cells, when compared to GSC (Fig. 2D). Furthermore, S3A receptors, Neuropilins (NRP), and transducers, Plexins (Plx) were present on both endothelial cells and GSCs (Fig. 2E), suggesting that GSC-derived S3A might function in both an autocrine and paracrine fashion. Of note, levels of S3A and its PlxA1 and NRP-1 receptors could be efficiently decreased by RNA interference (Fig. 2F,G,H).

Fig. 2.

Expression of Semaphorin 3A and its receptors in endothelial, glioma and glioma stem-like cells. (A) Semaphorin 3A (S3A, green) expression was analysed by confocal microscopy in high grade glioma stem-like cells grown as neurospheres (GSCs#1–#4). Nuclei were counterstained with DAPI (blue). Scale bar: 10 µm. (B–D) S3A expression was checked by RT-PCR (B,D) and by western blotting (C) in GSCs#1–#4, in glioblastoma cell lines (U87, U138, U251, U373 and LN229), SVGp12 foetal glial lines and Jurkat T lymphocytes (Jrk). β-actin and H2O were used as positive and negative controls, respectively in PCR experiments. Tubulin was used as a protein loading control. (E) Plexin (Plx) A1, A2, A3 and A4 expression was measured by RT-PCR in hCMEC/D3, HUVEC and GSC#1, along with neuropilins (NRP) 1 and 2. (F) S3A expression was evaluated by western blotting in GSC#1 transfected with non-silencing (sic) and three different siRNA sequences targeting S3A. β-catenin (β-cat) was used as a loading control. (G) PlxA1 expression was explored by RT-PCR in non-silencing (sic) and plxA1 siRNA transfected hCMEC/D3. β-actin was used as housekeeping gene. (H) NRP-1 expression was evaluated by western blotting in hCMEC/D3 transfected with non-silencing (sic) and three different siRNA sequences targeting NRP1. Tubulin was used as a loading control.

Fig. 2.

Expression of Semaphorin 3A and its receptors in endothelial, glioma and glioma stem-like cells. (A) Semaphorin 3A (S3A, green) expression was analysed by confocal microscopy in high grade glioma stem-like cells grown as neurospheres (GSCs#1–#4). Nuclei were counterstained with DAPI (blue). Scale bar: 10 µm. (B–D) S3A expression was checked by RT-PCR (B,D) and by western blotting (C) in GSCs#1–#4, in glioblastoma cell lines (U87, U138, U251, U373 and LN229), SVGp12 foetal glial lines and Jurkat T lymphocytes (Jrk). β-actin and H2O were used as positive and negative controls, respectively in PCR experiments. Tubulin was used as a protein loading control. (E) Plexin (Plx) A1, A2, A3 and A4 expression was measured by RT-PCR in hCMEC/D3, HUVEC and GSC#1, along with neuropilins (NRP) 1 and 2. (F) S3A expression was evaluated by western blotting in GSC#1 transfected with non-silencing (sic) and three different siRNA sequences targeting S3A. β-catenin (β-cat) was used as a loading control. (G) PlxA1 expression was explored by RT-PCR in non-silencing (sic) and plxA1 siRNA transfected hCMEC/D3. β-actin was used as housekeeping gene. (H) NRP-1 expression was evaluated by western blotting in hCMEC/D3 transfected with non-silencing (sic) and three different siRNA sequences targeting NRP1. Tubulin was used as a loading control.

To test whether GSC-produced S3A was functional and able to exert its well-known repulsive activity on endothelial cells (Guttmann-Raviv et al., 2007), we used a previously designed co-culture model (Galan-Moya et al., 2011). Briefly, the behaviour of S3A-secreting GSCs, grown as neurospheres, on top of confluent quiescent human brain endothelial monolayers, was monitored. GSCs promoted significant remodelling of endothelial monolayers, as evidenced by the appearance of asymmetric empty areas surrounding the neurospheres (Fig. 3A). This was achieved even when blocking matrix degradation or actomyosin contractility (supplementary material Fig. S3). The stemness marker Sox2 (Galan-Moya et al., 2011) identified GSCs as the repellent cells, while VE-cadherin labelling demonstrated that endothelial cells had been moved back (Fig. 3A). Furthermore, S3A-expressing HEK-293T mimicked GSC properties, as they also induced the formation of cell-free gaps within the endothelial mat under co-culture conditions (Fig. 3B). Again, this phenotype did not result from cell detachment or death (supplementary material Fig. S1). Interestingly, Sox2/S3A-positive Tubulin βIII-negative cells were observed at the interface with endothelial cells, suggesting that S3A-producing GSCs did not undergo differentiation (Fig. 3C; supplementary material Fig. S4). Thus, these data suggest that GSCs repulse brain endothelial monolayers likely via repulsive interactions between the two cell types. Conversely, treatment of GSCs with S3A blocking antibody or functional S3A-targeting siRNA dramatically reduced GSC-mediated endothelial repulsion (Fig. 3D–F). Likewise, this effect was lost in PlxA1- or NRP-1-siRNA-transfected endothelial cells (Fig. 3D,F; supplementary material Fig. S3). Altogether, our experiments suggest a functional link between GSC-secreted S3A and endothelial-expressed PlxA1/NRP-1 in co-culture.

Fig. 3.

Glioma stem-like cell-secreted Semaphorin 3A remodels brain endothelial cell monolayers. (A) Three-day-old confluent human brain endothelial cells (ECs) were cultured alone or with glioblastoma stem-like cells (GSCs#1–#4) for 18 h under deprivation conditions. Upper panel: representative phase-contrast images were coloured using Photoshop (ECs in green, GSCs in orange). Repulsion areas are in grey and indicated by black arrows. Scale bar: 60 µm. Lower panel: VE-cadherin (VEC, green) and Sox2 (red) were analysed by confocal microscopy. Scale bar: 10 µm. (B) ECs were cultured with myc–GFP (GFP) and Semaphorin 3A–GFP (S3A)-transfected HEK-293T for 2 days. Representative phase-contrast images were coloured using Photoshop (ECs in green, HEK-293T in blue, repulsion areas in grey). Scale bar: 60 µm. Inserted numbers are the percentage of HEK-293T that repelled EC monolayers; n = 100. (C) Co-culture assays, as described in A, were fixed and processed for immunostaining for Sox2 (red) and VEC or S3A (green). Nuclei were counterstained with DAPI. Scale bars: 10 µm. (D–F) Co-culture assays were performed as described in A. GSCs were either pre-treated for 1 h with pre-immune (ctrl; 2 µg/ml, 1 h) and S3A (N15; 2 µg/ml, 1 h) antibodies (ab) or transfected with 50 nM non-silencing sequence (sic) and S3A siRNA (S3Asi) 3 days earlier. ECs received 50 nM of sic and plexin A1 siRNA (plxsi) 3 days prior exposure to GSCs. Graphs represent the percentage of the total number of GSC neurospheres (NS) that repelled EC monolayers; n = 100.

Fig. 3.

Glioma stem-like cell-secreted Semaphorin 3A remodels brain endothelial cell monolayers. (A) Three-day-old confluent human brain endothelial cells (ECs) were cultured alone or with glioblastoma stem-like cells (GSCs#1–#4) for 18 h under deprivation conditions. Upper panel: representative phase-contrast images were coloured using Photoshop (ECs in green, GSCs in orange). Repulsion areas are in grey and indicated by black arrows. Scale bar: 60 µm. Lower panel: VE-cadherin (VEC, green) and Sox2 (red) were analysed by confocal microscopy. Scale bar: 10 µm. (B) ECs were cultured with myc–GFP (GFP) and Semaphorin 3A–GFP (S3A)-transfected HEK-293T for 2 days. Representative phase-contrast images were coloured using Photoshop (ECs in green, HEK-293T in blue, repulsion areas in grey). Scale bar: 60 µm. Inserted numbers are the percentage of HEK-293T that repelled EC monolayers; n = 100. (C) Co-culture assays, as described in A, were fixed and processed for immunostaining for Sox2 (red) and VEC or S3A (green). Nuclei were counterstained with DAPI. Scale bars: 10 µm. (D–F) Co-culture assays were performed as described in A. GSCs were either pre-treated for 1 h with pre-immune (ctrl; 2 µg/ml, 1 h) and S3A (N15; 2 µg/ml, 1 h) antibodies (ab) or transfected with 50 nM non-silencing sequence (sic) and S3A siRNA (S3Asi) 3 days earlier. ECs received 50 nM of sic and plexin A1 siRNA (plxsi) 3 days prior exposure to GSCs. Graphs represent the percentage of the total number of GSC neurospheres (NS) that repelled EC monolayers; n = 100.

Glioma stem-like cell-secreted Semaphorin 3A jeopardizes brain endothelial cell monolayer integrity

Because S3A has also been reported to exhibit pro-permeability activity (Acevedo et al., 2008), we next wondered whether GSC-produced S3A could alter endothelial cell barrier properties. To this end, endothelial cells were cultured with GSC-conditioned medium in order to recapitulate paracrine signalling. Treatment of endothelial cells with GSC-conditioned medium significantly enhanced VE-cadherin internalisation, to a level comparable to that induced by recombinant S3A protein (Fig. 4A). In sharp contrast, reducing S3A availability by either blocking antibody or siRNA nearly completely abolished VE-cadherin internalisation driven by GSC-conditioned medium (Fig. 4A,B). Similarly, introduction of PlxA1 siRNA into endothelial cells also reduced VE-cadherin internalisation triggered by exposure to GSC conditioned medium (Fig. 4B). In line with these findings, GSC conditioned media promoted brain endothelial cell permeability and VE-cadherin serine phosphorylation, in an S3A-dependent fashion (Fig. 4C,D). Finally, siRNA-mediated knockdown of VE-cadherin enhanced endothelial cell permeability in quiescent cells, while having no impact on GSC-governed monolayer repulsion, a phenomenon most likely already at its maximum (supplementary material Fig. S5). In short, S3A secreted by GSCs appears to play a central role in VE-cadherin endothelial cell–cell junction reorganization and elevated permeability.

Fig. 4.

Semaphorin 3A secreted by glioma stem-like cells jeopardizes brain endothelial cell monolayer integrity. (A,B) Three-day-old confluent human brain endothelial cells (ECs) were exposed for 15 min to glioblastoma stem-like cell conditioned medium (GSC CM) pre-incubated with pre-immune (ctrl, 2 µg/ml, 1 h) and S3A antibodies (ab, N15, 2 µg/ml, 1 h) and to CM prepared from non-silencing (sic)- and S3A siRNA (si)-transfected GSCs. Alternatively, sic- and plexin siRNA (plxsi)-transfected ECs were incubated with GSC CM. VE-cadherin antibody uptake was performed for each condition. Number of cells exhibiting vesicular staining was quantified and expressed as a percentage of the total number of cells; n = 300. (C) FITC-dextran permeability was measured in hCMEC/D3 treated for 1 h with CM prepared from GSCs#1–#4 transfected with sic and S3Asi. (D) ECs were exposed for 5 min to GSC CM pre-incubated with ctrl- and S3A-blocking antibodies, as defined in B. CM from sic- and S3Asi-transfected GSCs were similarly applied. Phospho-VE-cadherin (p-VEC) and total VEC expression levels were analysed by western blotting.

Fig. 4.

Semaphorin 3A secreted by glioma stem-like cells jeopardizes brain endothelial cell monolayer integrity. (A,B) Three-day-old confluent human brain endothelial cells (ECs) were exposed for 15 min to glioblastoma stem-like cell conditioned medium (GSC CM) pre-incubated with pre-immune (ctrl, 2 µg/ml, 1 h) and S3A antibodies (ab, N15, 2 µg/ml, 1 h) and to CM prepared from non-silencing (sic)- and S3A siRNA (si)-transfected GSCs. Alternatively, sic- and plexin siRNA (plxsi)-transfected ECs were incubated with GSC CM. VE-cadherin antibody uptake was performed for each condition. Number of cells exhibiting vesicular staining was quantified and expressed as a percentage of the total number of cells; n = 300. (C) FITC-dextran permeability was measured in hCMEC/D3 treated for 1 h with CM prepared from GSCs#1–#4 transfected with sic and S3Asi. (D) ECs were exposed for 5 min to GSC CM pre-incubated with ctrl- and S3A-blocking antibodies, as defined in B. CM from sic- and S3Asi-transfected GSCs were similarly applied. Phospho-VE-cadherin (p-VEC) and total VEC expression levels were analysed by western blotting.

Inhibition of PP2A alters brain endothelial cell monolayer integrity

To tackle the molecular pathways initiated by S3A, endothelial cell lysates were prepared following treatment with recombinant S3A in the absence of additional external factors, such as VEGF or FGF. This triggered the tyrosine phosphorylation of two proteins, the non-receptor tyrosine kinase Src and the holoenzyme, protein phosphatase 2A (PP2A) (Fig. 5A). Concomitantly, the catalytic domain of PP2A (PP2A-C) bound to Src following S3A treatment (Fig. 5A; supplementary material Fig. S6). Interestingly, serine/threonine phosphatase activity of PP2A was rapidly and transiently reduced upon S3A in quiescent mature endothelial monolayers (Fig. 5B).

Fig. 5.

Semaphorin 3A operates through Src in brain endothelial cell monolayer remodelling. (A) Three-day-old confluent human brain endothelial cells (ECs) were stimulated with 100 ng/ml Semaphorin 3A (S3A) for the indicated times. Phosphorylation (p) levels of Src (Y416) and PP2A-C (Y307) were analysed by western blotting. Src and PP2A-C served as loading controls. Alternatively, PP2A-C immunoprecipitated fractions (IP) were blotted for Src and PP2A. Src and PP2A-C intensity ratios are indicated below. (B) PP2A activity was measured in ECs stimulated with S3A for the indicated times. (C,D) ECs transfected with non silencing (sic) and Src targeting (srcsi) duplexes were stimulated with S3A for 5 minutes and analysed for PP2A activity (C) and p-VE-cadherin (p-VEC), Src and VEC by western blotting (D). (E–G) ECs were pre-treated with DMSO vehicle (ctrl) and SU6656 (Srcinh, 1 µM) prior S3A stimulation, and processed for p-VEC and VEC western blotting (E) FITC-dextran passage (F) and VEC antibody uptake (G).

Fig. 5.

Semaphorin 3A operates through Src in brain endothelial cell monolayer remodelling. (A) Three-day-old confluent human brain endothelial cells (ECs) were stimulated with 100 ng/ml Semaphorin 3A (S3A) for the indicated times. Phosphorylation (p) levels of Src (Y416) and PP2A-C (Y307) were analysed by western blotting. Src and PP2A-C served as loading controls. Alternatively, PP2A-C immunoprecipitated fractions (IP) were blotted for Src and PP2A. Src and PP2A-C intensity ratios are indicated below. (B) PP2A activity was measured in ECs stimulated with S3A for the indicated times. (C,D) ECs transfected with non silencing (sic) and Src targeting (srcsi) duplexes were stimulated with S3A for 5 minutes and analysed for PP2A activity (C) and p-VE-cadherin (p-VEC), Src and VEC by western blotting (D). (E–G) ECs were pre-treated with DMSO vehicle (ctrl) and SU6656 (Srcinh, 1 µM) prior S3A stimulation, and processed for p-VEC and VEC western blotting (E) FITC-dextran passage (F) and VEC antibody uptake (G).

Src is instrumental in governing permeability processes and PP2A is one of its ubiquitous targets (Eliceiri et al., 1999; Gavard and Gutkind, 2006; Sontag, 2001; Yokoyama and Miller, 2001). Moreover, PP2A has been shown to play a paramount role in E-cadherin stability and tight junctions in epithelial cells (Nunbhakdi-Craig et al., 2002; Suzuki and Takahashi, 2006). We therefore focused our attention on the molecular mechanisms by which S3A modulates this Src/PP2A network. First, treatment of endothelial cells with Src siRNA blocked S3A-mediated PP2A inactivation (Fig. 5C), suggesting that Src conveys S3A signalling to PP2A in endothelial cells. This is in agreement with a role for Src as an upstream negative regulator of PP2A. Src siRNA also impaired VE-cadherin serine phosphorylation (Fig. 5D). In addition, S3A-mediated increased permeability, as well as VE-cadherin internalisation and serine phosphorylation, were significantly reduced when endothelial cells were pre-treated with a Src inhibitor (Fig. 5E–G). In sharp contrast, pharmacological inhibition of PP2A with calyculin A enhanced vascular leakage in vivo and endothelial permeability in vitro, together with both serine phosphorylation and internalisation of VE-cadherin (Fig. 6A–D).

Fig. 6.

Inhibition of PP2A by Set alters brain endothelial cell monolayer integrity. (A) Mice received intra-dermal injection of DMSO vehicle (ctrl) or PP2A inhibitor (Caly, 1 µM and 100 nM), in duplicate in each flank. The images show the Evans blue extravasation at the site of injection, 1h post-treatment. The graph shows the mean of Evans blue extravasation (n = 6). (B–D) Three day-old confluent human brain endothelial cells (ECs) treated with DMSO vehicle (ctrl) and Caly (10 nM, unless specified) were processed for FITC-dextran passage (B), phospho-VE-cadherin (p-VEC) and VEC western blotting (C), and VEC antibody uptake (D). (E) FLAG immunoprecipitations (IP) were performed in HEK-293T transfected with FLAG–Set and PP2A-Cα, as indicated. FLAG and PP2A-C western blots are shown for IP and total cell lysates (TCL). Additionally, S3A-stimulated ECs were processed for PP2A-C immunoprecipitations (IP). Anti-Set and anti-PP2A-C were monitored by western blotting. (F–H) Non silencing (sic)- and Set-targeting (Setsi)-transfected ECs stimulated with S3A were processed for PP2A activity (F), p-PP2A-C, PP2A-C and Set western blotting (G) and total VEC staining (H). (I–L) Sic- and Setsi-transfected ECs exposed to medium conditioned by glioblastoma stem-like cells (GSC CM) were processed for p-VEC and VEC western blotting (I), VEC antibody uptake (J), FITC-dextran passage (K) and EC monolayer repulsion assays in co-culture (L).

Fig. 6.

Inhibition of PP2A by Set alters brain endothelial cell monolayer integrity. (A) Mice received intra-dermal injection of DMSO vehicle (ctrl) or PP2A inhibitor (Caly, 1 µM and 100 nM), in duplicate in each flank. The images show the Evans blue extravasation at the site of injection, 1h post-treatment. The graph shows the mean of Evans blue extravasation (n = 6). (B–D) Three day-old confluent human brain endothelial cells (ECs) treated with DMSO vehicle (ctrl) and Caly (10 nM, unless specified) were processed for FITC-dextran passage (B), phospho-VE-cadherin (p-VEC) and VEC western blotting (C), and VEC antibody uptake (D). (E) FLAG immunoprecipitations (IP) were performed in HEK-293T transfected with FLAG–Set and PP2A-Cα, as indicated. FLAG and PP2A-C western blots are shown for IP and total cell lysates (TCL). Additionally, S3A-stimulated ECs were processed for PP2A-C immunoprecipitations (IP). Anti-Set and anti-PP2A-C were monitored by western blotting. (F–H) Non silencing (sic)- and Set-targeting (Setsi)-transfected ECs stimulated with S3A were processed for PP2A activity (F), p-PP2A-C, PP2A-C and Set western blotting (G) and total VEC staining (H). (I–L) Sic- and Setsi-transfected ECs exposed to medium conditioned by glioblastoma stem-like cells (GSC CM) were processed for p-VEC and VEC western blotting (I), VEC antibody uptake (J), FITC-dextran passage (K) and EC monolayer repulsion assays in co-culture (L).

Last, we investigated the role of a potent and specific natural inhibitor of PP2A, I2PP2A, also known as Set (Samanta et al., 2009; Switzer et al., 2011; Westermarck and Hahn, 2008). Although they were found to constitutively interact when overexpressed in HEK-293T cells, Set interaction with PP2A-C was increased in brain endothelial cells after S3A stimulation (Fig. 6E; supplementary material Fig. S6). By reducing Set levels with siRNA, we found that it was required to promote PP2A inactivation and phosphorylation downstream of S3A (Fig. 6F,G). This enhanced PP2A serine/threonine phosphatase activity in Set-depleted endothelial cells withheld S3A-produced VE-cadherin disorganization (Fig. 6H), while hindering phosphorylation and internalisation of VE-cadherin, as well as permeability elicited by GSC conditioned medium (Fig. 6I–K). This also led to a reduced repulsive action of GSCs on endothelial cells (Fig. 6L). Altogether, our data support the idea that S3A depends on both Src and Set to abolish PP2A serine/threonine phosphatase activity, subsequently triggering VE-cadherin serine phosphorylation and endocytosis. Herein our data demonstrate that PP2A might operate at the crossroad between Set, Src and VE-cadherin in order to maintain the endothelial barrier.

PP2A activity is required to maintain brain endothelial cell monolayer integrity

To further examine the specific role of PP2A in endothelial cell barrier function, siRNA targeting the two PP2A-C isoforms, α and β, were used. Their respective efficiency and specificity were confirmed at the RNA level using specific primers for α or β isoforms, and, at the protein level using polyclonal antibodies targeting the two PP2A-C isoforms (Fig. 7A,B). PP2A-C knockdown by siRNA led to drastic disorganization of the endothelial monolayers, in combination with augmented serine phosphorylation and internalisation of VE-cadherin in non-stimulated cells (Fig. 7B–E). As a consequence, endothelial cell permeability was increased (Fig. 7F). Ectopic mapping experiments revealed that PP2A-C interacts with the C-terminal portion of the VE-cadherin intracellular domain, which includes the β-catenin binding site (Fig. 7G). By contrast, no binding with the juxtamembrane domain was detected. Moreover, a series of immunoprecipitations unveiled a basal interaction of endogenous VE-cadherin with PP2A-C in quiescent human endothelial cell monolayers (Fig. 7H; supplementary material Fig. S6). This binding was rapidly and transiently lost upon S3A stimulation (Fig. 7H). Furthermore, Src but not Set expression was required for S3A-mediated PP2A/VE-cadherin dissociation, as siRNA-mediated silencing of Src, but not Set, blocked PP2A release from VE-cadherin upon S3A treatment (Fig. 7I). Finally, we found that the PP2A/VE-cadherin interaction was also diminished in brain endothelial cells exposed to GSC-conditioned medium, but rescued with S3A-depleted GSC-conditioned medium (Fig. 7J,K). This last piece of data highlights the central role of PP2A in modulating VE-cadherin fate. In summary, PP2A might contribute to endothelial cell barrier homeostasis by maintaining low levels of VE-cadherin phosphorylation on the S665 residue in quiescent cells. Our data suggest that this equilibrium is perturbed by S3A-based Src activation and/or Set downregulation, as directed by local production from GSCs (Fig. 7L).

Fig. 7.

PP2A activity is required to maintain brain endothelial cell monolayer integrity. (A–F) Non-silencing (sic)-, PP2A-Cα- and PP2A-Cβ siRNA (si)-transfected human brain endothelial cells (ECs) were analysed after 3 days at confluence for PP2A-Cα and PP2A-Cβ expression by RT-PCR (A), levels of phospho-VE-cadherin (p-VEC), VEC and PP2A-Cα/β, by western blotting (B), total VE-cadherin staining (C), VEC internalisation before acid wash [iVEC (green) D; nuclei were counterstained with DAPI (blue)] and FITC-dextran passage (F). Scale bar: 10 µm (C,D). (G) Myc and FLAG immunoprecipitations (IP) were performed on HEK-293T transfected with FLAG–PP2A-Cα, myc–VEC-intracellular domain (ICD) and myc–VEC-ICD, comprising amino acids 621–784, 621–728 or 703–784, as indicated. Anti-myc and anti-FLAG western blots of IP and total cell lysates (TCL) are shown. (H) ECs were stimulated with 100 ng/ml Semaphorin 3A (S3A) for the indicated times. Anti-VEC and anti-PP2A-C IP were blotted for PP2A-C and VEC. (I) Sic-, Setsi- and Srcsi-transfected ECs stimulated with S3A (5 min) were processed for PP2A-C IPs. VEC and PP2A-C were analysed by western blotting. (J,K) VEC and PP2A-C levels were monitored, by western blotting, in PP2A-C IP from ECs exposed to glioblastoma stem-like cell conditioned medium (GSC CM) for 5 min (J). Alternatively, CMs were prepared from sic- and S3Asi-transfected GSC#1 (K). (L) GSC-secreted S3A can signal locally through endothelial Neuropilin and PlexinA1. This triggers PP2A phosphorylation on Y307 and inactivation by a two-pronged mechanism, involving both Src and Set. Ultimately, VE-cadherin and PP2A interaction is lost, resulting in elevation of endothelial permeability. In this model, PP2A might protect VE-cadherin from S665 phosphorylation and internalisation in quiescent cells.

Fig. 7.

PP2A activity is required to maintain brain endothelial cell monolayer integrity. (A–F) Non-silencing (sic)-, PP2A-Cα- and PP2A-Cβ siRNA (si)-transfected human brain endothelial cells (ECs) were analysed after 3 days at confluence for PP2A-Cα and PP2A-Cβ expression by RT-PCR (A), levels of phospho-VE-cadherin (p-VEC), VEC and PP2A-Cα/β, by western blotting (B), total VE-cadherin staining (C), VEC internalisation before acid wash [iVEC (green) D; nuclei were counterstained with DAPI (blue)] and FITC-dextran passage (F). Scale bar: 10 µm (C,D). (G) Myc and FLAG immunoprecipitations (IP) were performed on HEK-293T transfected with FLAG–PP2A-Cα, myc–VEC-intracellular domain (ICD) and myc–VEC-ICD, comprising amino acids 621–784, 621–728 or 703–784, as indicated. Anti-myc and anti-FLAG western blots of IP and total cell lysates (TCL) are shown. (H) ECs were stimulated with 100 ng/ml Semaphorin 3A (S3A) for the indicated times. Anti-VEC and anti-PP2A-C IP were blotted for PP2A-C and VEC. (I) Sic-, Setsi- and Srcsi-transfected ECs stimulated with S3A (5 min) were processed for PP2A-C IPs. VEC and PP2A-C were analysed by western blotting. (J,K) VEC and PP2A-C levels were monitored, by western blotting, in PP2A-C IP from ECs exposed to glioblastoma stem-like cell conditioned medium (GSC CM) for 5 min (J). Alternatively, CMs were prepared from sic- and S3Asi-transfected GSC#1 (K). (L) GSC-secreted S3A can signal locally through endothelial Neuropilin and PlexinA1. This triggers PP2A phosphorylation on Y307 and inactivation by a two-pronged mechanism, involving both Src and Set. Ultimately, VE-cadherin and PP2A interaction is lost, resulting in elevation of endothelial permeability. In this model, PP2A might protect VE-cadherin from S665 phosphorylation and internalisation in quiescent cells.

Discussion

In summary, we present evidence that GSC-produced S3A destabilises the brain endothelial barrier by interrupting the continuous protective effect of PP2A on VE-cadherin. Our data indeed suggests that PP2A secures mature endothelial junctions against vascular integrity disruption by preventing both serine phosphorylation and internalisation of VE-cadherin. Conversely, S3A-mediated transient inhibition of PP2A activity weakens cell–cell adhesion. Of note, the endothelial specific tyrosine phosphatase VE-PTP is required for endothelial cell–cell junction maintenance (Nawroth et al., 2002; Nottebaum et al., 2008; Broermann et al., 2011), further suggesting that an intricate interplay exists between tyrosine and serine/threonine phosphatases to double-lock the endothelial barrier.

How S3A modulates the endothelial signalling remains rather complex. In vivo studies have previously illuminated the role of NRP-1 in S3A biological responses (Gu et al., 2003). However, NRP-1 might relay multiple signals as it has been described to modulate the behaviour of other class III semaphorins, VEGF and also integrins (Gu et al., 2003; Serini et al., 2003). In a model in which NRP-1 acts as a dual receptor for VEGF and semaphorins, VEGF might compete with S3A for NRP-1 binding and therefore impact on GSC-mediated endothelial monolayer reshaping. On the other hand, S3A could also function through a PlxA1/NRP-1 signalling complex independently of VEGF, whereby PlxA1 acts as the signal transducer (Acevedo et al., 2008; Guttmann-Raviv et al., 2007). In our settings, S3A, NRP-1 and PlxA1 participate in GSC-directed endothelial repulsion and cell–cell junction opening, in the absence of exogenous VEGF. In contrast to Acevedo et al. who reported Src-independent modulation of barrier properties in peripheral endothelium, our findings suggest that, at least in our model, there is a Src addiction in brain endothelial cells (Acevedo et al., 2008). Indeed, we found that Src was critically required for S3A-mediated inhibition of PP2A and subsequent internalisation and serine phosphorylation of VE-cadherin culminating in increased endothelial permeability. Although plexin-driven signalling pathways in endothelial cells are poorly understood, Src family kinases, Focal Adhesion Kinase and PI3K have been proposed to take part in S3A actions in other cell systems (Serini et al., 2009). Herein, we report that S3A negatively modulates PP2A activity in human brain endothelial cells, which is an essential step for cell–cell junction destabilisation. Collectively, our data suggest that PlxA1/NRP-1 elicit a direct signalling pathway involving Src, Set and PP2A, which ultimately modulates VE-cadherin status (Fig. 7L). The exact mechanism by which Set interferes with PP2A activity in the context of endothelial plasticity remains elusive. However, membrane targeting and molecular bridges have all been proposed to modulate Set action (Switzer et al., 2011; ten Klooster et al., 2007). For example, active Rac targets Set to the plasma membrane where it then inhibits PP2A (ten Klooster et al., 2007). In line with this report, VE-cadherin serine phosphorylation and internalisation require the Rac/p21-activated kinase pathway (Gavard and Gutkind, 2006). Thus, in the future it will be interesting to test whether Rac could control PP2A activity through Set in S3A-exposed endothelial cells.

Besides its pro-permeability functions, S3A also emerges as a potent anti-angiogenic factor. In fact, S3A expression is frequently downregulated during the course of cancer progression, especially during the angiogenic switch, in pancreatic, cervical and skin tumours (Maione et al., 2009). In addition, S3A re-expression correlates with lower blood vessel density and reduced tumour growth and dissemination (Casazza et al., 2011; Chakraborty et al., 2012; Maione et al., 2009; Maione et al., 2012). Likewise, S3A silencing allows re-vascularization in ischemic tissues (Joyal et al., 2011). S3A also influences tumour cell migration and spreading in glioma (Bagci et al., 2009), therefore suggesting both autocrine and paracrine actions of S3A in brain tumours. Thus, these studies converge on the idea that restoring S3A expression might be of therapeutic interest to slow down tumour progression and aggressiveness. However, several studies have identified cancer stem-like cells in brain tumour cultures as a heterogeneous subpopulation of tumour cells, which express neural stemness markers, such as Nestin and Sox2, can self-renew and retain the ability to initiate and form a tumour mimicking the original (Galli et al., 2004; Patru et al., 2010; Singh et al., 2004; Yuan et al., 2004). In our study, we proposed that S3A is preferentially expressed by a small subpopulation of brain tumour cells (GSCs) that have endothelial progenitor properties and closely interact with endothelial cells (Calabrese et al., 2007; Ricci-Vitiani et al., 2010; Wang et al., 2010). Depending on their microenvironment and culture conditions, GSCs can lose their stemness properties and engage a pseudo-differentiation process. Remarkably, glioma vasculature comprises a variable number of cells emanating from the tumour mass, as evidenced by their karyotype and specific genetic aberrations (Ricci-Vitiani et al., 2010; Wang et al., 2010). Keeping with this, GSCs are found close to capillaries in brain tumours, where endothelial cells are believed to sustain GSC self-renewal (Batchelor et al., 2007; Calabrese et al., 2007; Galan-Moya et al., 2011). Although GSCs harbour endothelial features and can ultimately participate in tumour neovascularisation, the underlying molecular mechanisms remain obscure. Although rather speculative, our data suggest a specific and confined action of GSC-secreted S3A, which, when coupled to its localised pro-permeability function, might remodel the endothelial wall to allow progenitor recruitment and insertion into active sites of angiogenesis. One can thus imagine that this mechanism could contribute, at least partially, to tumour endothelial aberrations observed in human glioma, including increased vascular permeability and tumour cell integration.

Materials and Methods

Cell culture, transfections and siRNA

Immortalised human cerebral microvascular endothelial cells (hCMEC/D3) were maintained in Endothelial Basal Medium-2 (EBM-2, Lonza), containing 5% foetal bovine serum (FBS Serum Gold, PAA Laboratories), 1% penicillin/streptomycin (P/S), HEPES, and chemically defined lipid concentrate (Invitrogen), hydrocortisone (1.4 µM), acid ascorbic (5 µg/ml) and basic fibroblast growth factor (1 ng/ml; Sigma) as described by Weksler et al. (Weksler et al., 2005). Previously characterized long-term human glioblastoma-stem-like cells [namely GSC #1–#4 (Galan-Moya et al., 2011; Patru et al., 2010; Thirant et al., 2012)] were cultivated in Dulbecco's modified Eagle's DMEM:F-12 medium (1:1) supplemented with N2, B27 and G5 (Invitrogen) and maintained as spheroids. Human umbilical vein endothelial cells [HUVEC, clone Eahy.926 (Edgell et al., 1983)], human embryonic kidney cells (HEK-293T, ATCC), U87, U138, U373, U251 and LN229 human glioblastoma cell lines (ATCC), and human foetal glial cell line SVGp12 (ATCC) were grown in DMEM, 10% FBS, 1% P/S (Invitrogen); Jurkat lymphocyte cells (ATCC) in RPMI, 10% FBS, 1% P/S (Invitrogen). DNA and siRNA were transfected using Turbofect (Fermentas) and Lipofectamine RNAiMax (Invitrogen), respectively. Non-silencing (Low GC Duplex) and pre-designed Stealth siRNA sequences targeting human Plexin A1 (HSS108150, HSS108149, HSS182381), Semaphorin 3A (HSS115876, HSS115877, HSS115878), Neuropilin 1 (HSS113022, HSS189594, HSS189595) and Src (HSS186080, HSS186081, HSS186082) were purchased from Invitrogen, universal negative control and pre-designed Mission siRNA targeting human PP2A-Cα (HA01985008), PP2A-Cβ (HA01985010) and Set (Hs0200327365, Hs0200327366) were from Sigma.

DNA constructs

Human intracellular domains of VE-cadherin corresponding to amino acids 621–784, 621–728 or 703–784 were cloned between NheI and NotI in frame with myc tag in pCEFL-GFP plasmid. Human full-length PP2A-Cα and Set in pCMV-Sport6 plasmids were purchased from Open Biosystems (ThermoFisher) and subcloned between BamHI and NotI in 3×FLAG plasmid. Semaphorin 3A-GFP construct was described previously (Piaton et al., 2011).

Miles permeability assays

All animal studies were carried out using protocols approved by the French Veterinary Department, in compliance with the European Guide for the Care and Use of Laboratory Animals. C57BL/6 mice, 6- to 8-weeks old, were obtained from Charles River Laboratories. Permeability assays were performed as described in Acevedo et al. (Acevedo et al., 2008). Briefly, Evans Blue dye (1% in PBS, Sigma) was administered through the retro-orbital venous sinus, 5 min prior vehicle, mouse Semaphorin 3A (50 ng/ml), and Calyculin A (1 µM and 100 nM) intra-dermal injections in equal volume (25 µl). One hour later, mice were sacrificed and skin samples were excised using dermal biopsy punches (diameter 0.8 cm, Kai medical), and photographed. Dye extraction was achieved by incubating skin specimens in formamide (56°C, 3 days) and measured (595 nM, Thermo Electron Corporation). Background dye extravasation was subtracted from non-injected sites.

Reagents and antibodies

Recombinant human and mouse Semaphorin 3A-Fc chimera were from R&D, Src family kinase inhibitor su6656 from Merck, and the PP2A phosphatase inhibitor calyculin A from Invitrogen. The following primary antibodies were used: VE-cadherin, Tubulin, Semaphorin 3A (N-15 and H-130 clones), Neuropilin 1, Src, PP2A-Cα/β and pre-immune serum (Santa Cruz), pY416-Src, AKT, Src and myc (Cell Signaling), Sox2 and VE-cadherin BV6 clone (Millipore), β-catenin, Set and FLAG (Sigma), p120 catenin (BD), pY307-PP2A-Cα/β (Epitomics), GFP (Roche) and pS665-VE-cadherin (Gavard and Gutkind, 2006). Alexa-Fluor-680-, Alexa-Fluor-488- and Alexa-Fluor-546-conjugated donkey anti-mouse, anti-rabbit and anti-goat antibodies were from Invitrogen.

RT-PCR

RNA was extracted using the RNeasy Mini Kit as per the manufacturer's directions (Qiagen). Equal amounts of RNA were reverse transcribed using the Superscript III RT kit (Invitrogen) and the resulting cDNA was used to amplify S3A, PlxA1-A4 and NRP-1/2 genes by PCR using gene-specific primers for human Semaphorin 3A (Fwd: 5′-TGGACATCATCCTGAGGACA-3′; Rev: 5′-CTCTGTCCTGATTGGGTGGT-3′), human PlexinA1 (Fwd: 5′-GACAGACATCCACGAGCTGA-3′; Rev: 5′-TCAGCGACTTCTCCACATTG-3′), human PlexinA2 (Fwd: 5′-CGTGTTTGACATCCACAAGG-3′; Rev: 5′-CTCCACCCAGCTCTTGTAGC-3′), human PlexinA3 (Fwd: 5′-AAACTGCTCTACGCCAAGGA-3′; Rev: 5′-TGGCGGTACTTGGTGACATA-3′), human PlexinA4 (Fwd: 5′-CAGAGGACCCAGAGTTGAGC-3′; Rev: 5′-AAGGGTCAATGTTCGAGGTG-3′), human Neuropilin 1 (Fwd: 5′-CCACAGTGGAACAGGTGATG-3′; Rev: 5′-GCACGTGATTGTCATGTTCC-3′), human Neuropilin2 (Fwd: 5′-ATACCACACCAAGGCTGGAG-3′; Rev: 5′-ACCACCTAGTCCGGGAGAGT-3′), PP2A-Cα (Fwd: 5′-GAATCCAACGTGCAAGAGGT-3′; Rev: 5′-CGTTCACGGTAACGAACCTT-3′), PP2Aβ (Fwd: 5′-CCCTGGATCGTTTACAGGAA-3′; Rev: 5′-TGGGTGCACTGAAAATGGTA-3′) in the presence of Taq DNA polymerase (Invitrogen). Human β-actin (Fwd: 5′-AGCACTGTGTTGGCGTACAG-3′; Rev: 5′-GGACTTCGAGCAAGAGATGG-3′) was also amplified as a control for input. PCR products were separated by electrophoresis on SYBR green-containing agarose gels (Invitrogen).

Endothelial repulsion assays

Three-day-old confluent hCMEC/D3 monolayers were placed either alone or in co-culture with GSC for 18 h (Galan-Moya et al., 2011). Cells were then fixed and processed for morphology analysis by phase and immunofluorescence microscopy. GSC neurospheres that repelled monolayers were counted as positive and expressed as a ratio over the total number of seeded neurospheres.

Internalisation, immunofluorescence and permeability assays

VE-cadherin internalisation was tracked using the anti-VE-cadherin BV6 clone antibody uptake assay as described previously (Gavard and Gutkind, 2006) (supplementary material Fig. S3). Briefly, sub-confluent cultures were first incubated with BV6 antibody (1:200 dilution, 4°C, 1 h) to label surface-exposed VE-cadherin. Samples were then washed with ice-cold PBS and further incubated for 5 to 30 min with the indicated media at 37°C. After extensive PBS washes, cells were either fixed for staining or subjected to acid wash before fixation (Glycine 0.1 M, pH 2, 15 min, RT). Cells were counted as positive when more than five VE-cadherin dots were visualised in cells. Immunofluorescence assays were performed as described previously (Gavard and Gutkind, 2006). Specifically, cells were fixed in PBS–paraformaldehyde 4% for 15 min, permeabilised in PBS–Triton X-100 (0.5%) for 5 min and further blocked in PBS–BSA 3% for 30 min. Primary and secondary antibodies were diluted in PBS–BSA 3% at 1:100 an 1:400 dilution, respectively, for 1 h. Samples were then stained with DAPI (1:10,000 dilution in PBS, 5 min, Invitrogen) and mounted in Fluoromount (Sigma). Slides were analysed on a SP5 confocal microscope (Institut Andre Lwoff, Villejuif).

FITC-dextran transcellular passage was estimated on 3-day-old hCMEC/D3 monolayers cultured on 3 µm pore collagen-coated PTFE membranes (Corning) as described previously (Gavard and Gutkind, 2006) and analysed with a fluorescence plate reader (Fusion, Packard Bioscience). Data were normalised to untreated samples and expressed as the mean on three independent experiments.

PP2A activity

PP2A activity was measured by phosphate release using the Human/Mouse/Rat Active PP2A Activity Assay (DuoSet IC, R&D).

Immunoprecipitation, surface biotinylation and western blotting

Proteins lysates were collected as described previously (Galan-Moya et al., 2011). Equal amounts of proteins (MicroBCA Protein Assay Kit, Thermo Scientific) were pre-cleared with Protein G agarose (Roche) and further processed for immunoprecipitation using 1 µg of antibodies. Biotinylation assays were performed using the EZ-Link Sulfo-NHS-Biotin and biotinylation kits, as per the manufacturer's instructions (Thermo Scientific). For western blots, proteins were separated on 4–20% polyacrylamide Nupage gels (Invitrogen) and transferred onto PVDF membranes (Thermo Scientific). Membranes were scanned using the Odyssey Infrared imaging system (LiCor).

Statistical analysis

Graphs are shown as a mean value + s.e.m. from at least three independent experiments. Confocal pictures and western-blot scans are representative of at least three independent experiments. siRNA experiments were performed with at least two independent sequences. Student's t-tests were performed for statistical analyses.

Acknowledgements

We are grateful to P. O. Couraud (Institut Cochin, Paris, France) for sharing the human brain endothelial cell line, and C. Lubetzki and C. Ros for Semaphorin 3A plasmid (Pitie-Salpetriere Hospital, Paris, France).

Funding

This work was supported by the Ligue Nationale contre le Cancer [grant number RS12/7595-8 to J.G.]; Fondation ARC pour la Recherche sur le Cancer [grant number SFI20101201140 to J.G. and JR/AD/MDV-A09/4 to E.M.G.M.]; Association des NeuroOncologues d'Expression Française [grant number 074391 to J.G.]; Fondation pour la Recherche Medicale [grant number DPC20111122987 to J.G.]; Agence Nationale pour la Recherche sur le SIDA et les hepatites virales [grant number NM/no1380/CSS3/AO 2011-2 to J.D.]; Agence Nationale pour la Recherche [grant number ANRJCJC2010/130601 to N.B.]; and a Marie Curie International Reintegration Grant [grant number 239126 to J.G.].

Supplementary material available online at Supplementary Information

References

Acevedo
L. M.
,
Barillas
S.
,
Weis
S. M.
,
Göthert
J. R.
,
Cheresh
D. A.
(
2008
).
Semaphorin 3A suppresses VEGF-mediated angiogenesis yet acts as a vascular permeability factor.
Blood
111
,
2674
2680
.
Bagci
T.
,
Wu
J. K.
,
Pfannl
R.
,
Ilag
L. L.
,
Jay
D. G.
(
2009
).
Autocrine semaphorin 3A signaling promotes glioblastoma dispersal.
Oncogene
28
,
3537
3550
.
Batchelor
T. T.
,
Sorensen
A. G.
,
di Tomaso
E.
,
Zhang
W. T.
,
Duda
D. G.
,
Cohen
K. S.
,
Kozak
K. R.
,
Cahill
D. P.
,
Chen
P. J.
,
Zhu
M.
et al.  (
2007
).
AZD2171, a pan-VEGF receptor tyrosine kinase inhibitor, normalizes tumor vasculature and alleviates edema in glioblastoma patients.
Cancer Cell
11
,
83
95
.
Broermann
A.
,
Winderlich
M.
,
Block
H.
,
Frye
M.
,
Rossaint
J.
,
Zarbock
A.
,
Cagna
G.
,
Linnepe
R.
,
Schulte
D.
,
Nottebaum
A. F.
et al.  (
2011
).
Dissociation of VE-PTP from VE-cadherin is required for leukocyte extravasation and for VEGF-induced vascular permeability in vivo.
J. Exp. Med.
208
,
2393
2401
.
Calabrese
C.
,
Poppleton
H.
,
Kocak
M.
,
Hogg
T. L.
,
Fuller
C.
,
Hamner
B.
,
Oh
E. Y.
,
Gaber
M. W.
,
Finklestein
D.
,
Allen
M.
et al.  (
2007
).
A perivascular niche for brain tumor stem cells.
Cancer Cell
11
,
69
82
.
Carmeliet
P.
,
Lampugnani
M. G.
,
Moons
L.
,
Breviario
F.
,
Compernolle
V.
,
Bono
F.
,
Balconi
G.
,
Spagnuolo
R.
,
Oosthuyse
B.
,
Dewerchin
M.
et al.  (
1999
).
Targeted deficiency or cytosolic truncation of the VE-cadherin gene in mice impairs VEGF-mediated endothelial survival and angiogenesis.
Cell
98
,
147
157
.
Casazza
A.
,
Fu
X.
,
Johansson
I.
,
Capparuccia
L.
,
Andersson
F.
,
Giustacchini
A.
,
Squadrito
M. L.
,
Venneri
M. A.
,
Mazzone
M.
,
Larsson
E.
et al.  (
2011
).
Systemic and targeted delivery of semaphorin 3A inhibits tumor angiogenesis and progression in mouse tumor models.
Arterioscler. Thromb. Vasc. Biol.
31
,
741
749
.
Catalano
A.
,
Caprari
P.
,
Moretti
S.
,
Faronato
M.
,
Tamagnone
L.
,
Procopio
A.
(
2006
).
Semaphorin-3A is expressed by tumor cells and alters T-cell signal transduction and function.
Blood
107
,
3321
3329
.
Chakraborty
G.
,
Kumar
S.
,
Mishra
R.
,
Patil
T. V.
,
Kundu
G. C.
(
2012
).
Semaphorin 3A suppresses tumor growth and metastasis in mice melanoma model.
PLoS ONE
7
,
e33633
.
Crosby
C. V.
,
Fleming
P. A.
,
Argraves
W. S.
,
Corada
M.
,
Zanetta
L.
,
Dejana
E.
,
Drake
C. J.
(
2005
).
VE-cadherin is not required for the formation of nascent blood vessels but acts to prevent their disassembly.
Blood
105
,
2771
2776
.
Dejana
E.
(
2004
).
Endothelial cell-cell junctions: happy together.
Nat. Rev. Mol. Cell Biol.
5
,
261
270
.
Dwyer
J.
,
Le Guelte
A.
,
Galan Moya
E. M.
,
Sumbal
M.
,
Carlotti
A.
,
Douguet
L.
,
Gutkind
J. S.
,
Grange
P. A.
,
Dupin
N.
,
Gavard
J.
(
2011
).
Remodeling of VE-cadherin junctions by the human herpes virus 8 G-protein coupled receptor.
Oncogene
30
,
190
200
.
Edgell
C. J.
,
McDonald
C. C.
,
Graham
J. B.
(
1983
).
Permanent cell line expressing human factor VIII-related antigen established by hybridization.
Proc. Natl. Acad. Sci. USA
80
,
3734
3737
.
Eliceiri
B. P.
,
Paul
R.
,
Schwartzberg
P. L.
,
Hood
J. D.
,
Leng
J.
,
Cheresh
D. A.
(
1999
).
Selective requirement for Src kinases during VEGF-induced angiogenesis and vascular permeability.
Mol. Cell
4
,
915
924
.
Galan–Moya
E. M.
,
Le Guelte
A.
,
Lima Fernandes
E.
,
Thirant
C.
,
Dwyer
J.
,
Bidere
N.
,
Couraud
P. O.
,
Scott
M. G.
,
Junier
M. P.
,
Chneiweiss
H.
et al.  (
2011
).
Secreted factors from brain endothelial cells maintain glioblastoma stem-like cell expansion through the mTOR pathway.
EMBO Rep.
12
,
470
476
.
Galli
R.
,
Binda
E.
,
Orfanelli
U.
,
Cipelletti
B.
,
Gritti
A.
,
De Vitis
S.
,
Fiocco
R.
,
Foroni
C.
,
Dimeco
F.
,
Vescovi
A.
(
2004
).
Isolation and characterization of tumorigenic, stem-like neural precursors from human glioblastoma.
Cancer Res.
64
,
7011
7021
.
Gaur
P.
,
Bielenberg
D. R.
,
Samuel
S.
,
Bose
D.
,
Zhou
Y.
,
Gray
M. J.
,
Dallas
N. A.
,
Fan
F.
,
Xia
L.
,
Lu
J.
et al.  (
2009
).
Role of class 3 semaphorins and their receptors in tumor growth and angiogenesis.
Clin. Cancer Res.
15
,
6763
6770
.
Gavard
J.
(
2009
).
Breaking the VE-cadherin bonds.
FEBS Lett.
583
,
1
6
.
Gavard
J.
,
Gutkind
J. S.
(
2006
).
VEGF controls endothelial-cell permeability by promoting the beta-arrestin-dependent endocytosis of VE-cadherin.
Nat. Cell Biol.
8
,
1223
1234
.
Gu
C.
,
Rodriguez
E. R.
,
Reimert
D. V.
,
Shu
T.
,
Fritzsch
B.
,
Richards
L. J.
,
Kolodkin
A. L.
,
Ginty
D. D.
(
2003
).
Neuropilin-1 conveys semaphorin and VEGF signaling during neural and cardiovascular development.
Dev. Cell
5
,
45
57
.
Guttmann–Raviv
N.
,
Shraga–Heled
N.
,
Varshavsky
A.
,
Guimaraes–Sternberg
C.
,
Kessler
O.
,
Neufeld
G.
(
2007
).
Semaphorin-3A and semaphorin-3F work together to repel endothelial cells and to inhibit their survival by induction of apoptosis.
J. Biol. Chem.
282
,
26294
26305
.
Joyal
J. S.
,
Sitaras
N.
,
Binet
F.
,
Rivera
J. C.
,
Stahl
A.
,
Zaniolo
K.
,
Shao
Z.
,
Polosa
A.
,
Zhu
T.
,
Hamel
D.
et al.  (
2011
).
Ischemic neurons prevent vascular regeneration of neural tissue by secreting semaphorin 3A.
Blood
117
,
6024
6035
.
Le Guelte
A.
,
Dwyer
J.
,
Gavard
J.
(
2011
).
Jumping the barrier: VE-cadherin, VEGF and other angiogenic modifiers in cancer.
Biol. Cell
103
,
593
605
.
Lin
M. I.
,
Yu
J.
,
Murata
T.
,
Sessa
W. C.
(
2007
).
Caveolin-1-deficient mice have increased tumor microvascular permeability, angiogenesis, and growth.
Cancer Res.
67
,
2849
2856
.
Maione
F.
,
Molla
F.
,
Meda
C.
,
Latini
R.
,
Zentilin
L.
,
Giacca
M.
,
Seano
G.
,
Serini
G.
,
Bussolino
F.
,
Giraudo
E.
(
2009
).
Semaphorin 3A is an endogenous angiogenesis inhibitor that blocks tumor growth and normalizes tumor vasculature in transgenic mouse models.
J. Clin. Invest.
119
,
3356
3372
.
Maione
F.
,
Capano
S.
,
Regano
D.
,
Zentilin
L.
,
Giacca
M.
,
Casanovas
O.
,
Bussolino
F.
,
Serini
G.
,
Giraudo
E.
(
2012
).
Semaphorin 3A overcomes cancer hypoxia and metastatic dissemination induced by antiangiogenic treatment in mice.
J. Clin. Invest.
122
,
1832
1848
.
Murakami
M.
,
Nguyen
L. T.
,
Zhuang
Z. W.
,
Moodie
K. L.
,
Carmeliet
P.
,
Stan
R. V.
,
Simons
M.
(
2008
).
The FGF system has a key role in regulating vascular integrity.
J. Clin. Invest.
118
,
3355
3366
.
Nawroth
R.
,
Poell
G.
,
Ranft
A.
,
Kloep
S.
,
Samulowitz
U.
,
Fachinger
G.
,
Golding
M.
,
Shima
D. T.
,
Deutsch
U.
,
Vestweber
D.
(
2002
).
VE-PTP and VE-cadherin ectodomains interact to facilitate regulation of phosphorylation and cell contacts.
EMBO J.
21
,
4885
4895
.
Nottebaum
A. F.
,
Cagna
G.
,
Winderlich
M.
,
Gamp
A. C.
,
Linnepe
R.
,
Polaschegg
C.
,
Filippova
K.
,
Lyck
R.
,
Engelhardt
B.
,
Kamenyeva
O.
et al.  (
2008
).
VE-PTP maintains the endothelial barrier via plakoglobin and becomes dissociated from VE-cadherin by leukocytes and by VEGF.
J. Exp. Med.
205
,
2929
2945
.
Nunbhakdi–Craig
V.
,
Machleidt
T.
,
Ogris
E.
,
Bellotto
D.
,
White
C. L.
 3rd
,
Sontag
E.
(
2002
).
Protein phosphatase 2A associates with and regulates atypical PKC and the epithelial tight junction complex.
J. Cell Biol.
158
,
967
978
.
Patru
C.
,
Romao
L.
,
Varlet
P.
,
Coulombel
L.
,
Raponi
E.
,
Cadusseau
J.
,
Renault–Mihara
F.
,
Thirant
C.
,
Leonard
N.
,
Berhneim
A.
et al.  (
2010
).
CD133, CD15/SSEA-1, CD34 or side populations do not resume tumor-initiating properties of long-term cultured cancer stem cells from human malignant glio-neuronal tumors.
BMC Cancer
10
,
66
.
Piaton
G.
,
Aigrot
M. S.
,
Williams
A.
,
Moyon
S.
,
Tepavcevic
V.
,
Moutkine
I.
,
Gras
J.
,
Matho
K. S.
,
Schmitt
A.
,
Soellner
H.
et al.  (
2011
).
Class 3 semaphorins influence oligodendrocyte precursor recruitment and remyelination in adult central nervous system.
Brain
134
,
1156
1167
.
Ricci–Vitiani
L.
,
Pallini
R.
,
Biffoni
M.
,
Todaro
M.
,
Invernici
G.
,
Cenci
T.
,
Maira
G.
,
Parati
E. A.
,
Stassi
G.
,
Larocca
L. M.
et al.  (
2010
).
Tumour vascularization via endothelial differentiation of glioblastoma stem-like cells.
Nature
468
,
824
828
.
Samanta
A. K.
,
Chakraborty
S. N.
,
Wang
Y.
,
Kantarjian
H.
,
Sun
X.
,
Hood
J.
,
Perrotti
D.
,
Arlinghaus
R. B.
(
2009
).
Jak2 inhibition deactivates Lyn kinase through the SET-PP2A-SHP1 pathway, causing apoptosis in drug-resistant cells from chronic myelogenous leukemia patients.
Oncogene
28
,
1669
1681
.
Schulte
D.
,
Küppers
V.
,
Dartsch
N.
,
Broermann
A.
,
Li
H.
,
Zarbock
A.
,
Kamenyeva
O.
,
Kiefer
F.
,
Khandoga
A.
,
Massberg
S.
et al.  (
2011
).
Stabilizing the VE-cadherin-catenin complex blocks leukocyte extravasation and vascular permeability.
EMBO J.
30
,
4157
4170
.
Serini
G.
,
Valdembri
D.
,
Zanivan
S.
,
Morterra
G.
,
Burkhardt
C.
,
Caccavari
F.
,
Zammataro
L.
,
Primo
L.
,
Tamagnone
L.
,
Logan
M.
et al.  (
2003
).
Class 3 semaphorins control vascular morphogenesis by inhibiting integrin function.
Nature
424
,
391
397
.
Serini
G.
,
Maione
F.
,
Giraudo
E.
,
Bussolino
F.
(
2009
).
Semaphorins and tumor angiogenesis.
Angiogenesis
12
,
187
193
.
Singh
S. K.
,
Hawkins
C.
,
Clarke
I. D.
,
Squire
J. A.
,
Bayani
J.
,
Hide
T.
,
Henkelman
R. M.
,
Cusimano
M. D.
,
Dirks
P. B.
(
2004
).
Identification of human brain tumour initiating cells.
Nature
432
,
396
401
.
Sontag
E.
(
2001
).
Protein phosphatase 2A: the Trojan Horse of cellular signaling.
Cell. Signal.
13
,
7
16
.
Sorensen
A. G.
,
Batchelor
T. T.
,
Zhang
W. T.
,
Chen
P. J.
,
Yeo
P.
,
Wang
M.
,
Jennings
D.
,
Wen
P. Y.
,
Lahdenranta
J.
,
Ancukiewicz
M.
et al.  (
2009
).
A “vascular normalization index” as potential mechanistic biomarker to predict survival after a single dose of cediranib in recurrent glioblastoma patients.
Cancer Res.
69
,
5296
5300
.
Stockton
R. A.
,
Schaefer
E.
,
Schwartz
M. A.
(
2004
).
p21-activated kinase regulates endothelial permeability through modulation of contractility.
J. Biol. Chem.
279
,
46621
46630
.
Suzuki
K.
,
Takahashi
K.
(
2006
).
Induction of E-cadherin endocytosis by loss of protein phosphatase 2A expression in human breast cancers.
Biochem. Biophys. Res. Commun.
349
,
255
260
.
Switzer
C. H.
,
Cheng
R. Y.
,
Vitek
T. M.
,
Christensen
D. J.
,
Wink
D. A.
,
Vitek
M. P.
(
2011
).
Targeting SET/I(2)PP2A oncoprotein functions as a multi-pathway strategy for cancer therapy.
Oncogene
30
,
2504
2513
.
Taddei
A.
,
Giampietro
C.
,
Conti
A.
,
Orsenigo
F.
,
Breviario
F.
,
Pirazzoli
V.
,
Potente
M.
,
Daly
C.
,
Dimmeler
S.
,
Dejana
E.
(
2008
).
Endothelial adherens junctions control tight junctions by VE-cadherin-mediated upregulation of claudin-5.
Nat. Cell Biol.
10
,
923
934
.
ten Klooster
J. P.
,
Leeuwen
I.
,
Scheres
N.
,
Anthony
E. C.
,
Hordijk
P. L.
(
2007
).
Rac1-induced cell migration requires membrane recruitment of the nuclear oncogene SET.
EMBO J.
26
,
336
345
.
Thirant
C.
,
Galan–Moya
E. M.
,
Dubois
L. G.
,
Pinte
S.
,
Chafey
P.
,
Broussard
C.
,
Varlet
P.
,
Devaux
B.
,
Soncin
F.
,
Gavard
J.
et al.  (
2012
).
Differential proteomic analysis of human glioblastoma and neural stem cells reveals HDGF as a novel angiogenic secreted factor.
Stem Cells
30
,
845
853
.
Vestweber
D.
,
Winderlich
M.
,
Cagna
G.
,
Nottebaum
A. F.
(
2009
).
Cell adhesion dynamics at endothelial junctions: VE-cadherin as a major player.
Trends Cell Biol.
19
,
8
15
.
Wang
R.
,
Chadalavada
K.
,
Wilshire
J.
,
Kowalik
U.
,
Hovinga
K. E.
,
Geber
A.
,
Fligelman
B.
,
Leversha
M.
,
Brennan
C.
,
Tabar
V.
(
2010
).
Glioblastoma stem-like cells give rise to tumour endothelium.
Nature
468
,
829
833
.
Weis
S.
,
Shintani
S.
,
Weber
A.
,
Kirchmair
R.
,
Wood
M.
,
Cravens
A.
,
McSharry
H.
,
Iwakura
A.
,
Yoon
Y. S.
,
Himes
N.
et al.  (
2004
).
Src blockade stabilizes a Flk/cadherin complex, reducing edema and tissue injury following myocardial infarction.
J. Clin. Invest.
113
,
885
894
.
Weksler
B. B.
,
Subileau
E. A.
,
Perrière
N.
,
Charneau
P.
,
Holloway
K.
,
Leveque
M.
,
Tricoire–Leignel
H.
,
Nicotra
A.
,
Bourdoulous
S.
,
Turowski
P.
et al.  (
2005
).
Blood-brain barrier-specific properties of a human adult brain endothelial cell line.
FASEB J.
19
,
1872
1874
.
Westermarck
J.
,
Hahn
W. C.
(
2008
).
Multiple pathways regulated by the tumor suppressor PP2A in transformation.
Trends Mol. Med.
14
,
152
160
.
Yokoyama
N.
,
Miller
W. T.
(
2001
).
Protein phosphatase 2A interacts with the Src kinase substrate p130(CAS).
Oncogene
20
,
6057
6065
.
Yuan
X.
,
Curtin
J.
,
Xiong
Y.
,
Liu
G.
,
Waschsmann–Hogiu
S.
,
Farkas
D. L.
,
Black
K. L.
,
Yu
J. S.
(
2004
).
Isolation of cancer stem cells from adult glioblastoma multiforme.
Oncogene
23
,
9392
9400
.

Supplementary information