Neuropilin-1 regulates platelet-derived growth factor receptor signalling in mesenchymal stem cells

MSCs (mesenchymal stem cells) in bone marrow and perivascular niches throughout the body are reservoirs of multipotent cells that can differentiate along mesenchymal lineages, including smooth muscle, and undergo endothelial transdifferentiation in response to VEGF (vascular endothelial growth factor) [1–3]. We have previously shown that multipotent human MSCs express NRP-1 (neuropilin-1) and PDGFRs [PDGF (platelet-derived growth factor) receptors] α and β, but not VEGFRs (VEGF receptors), and that both PDGFs and VEGF-A stimulate PDGFRs thereby regulating proliferation, migration and smooth muscle-specific cytoskeleton [2,4].

NRP-1 is a type I transmembrane glycoprotein that regulates vascular and neural development and acts as a co-receptor for VEGFRs and plexins [5–11]. NRP-1-deficient or -overexpressing mice display severe abnormalities in nervous and cardiovascular systems [12,13], whereas NRP-1-null zebrafish have loss of circulation via angiogenic vessels [14]. The large extracellular region of NRP-1 comprises two CUB (complement binding) domains (designated a1a2), two coagulation factor V/VIII homology domains (designated b1b2) and a MAM (meprin, A5 antigen, receptor tyrosine phosphatase μ) domain (designated c). The last three C-terminal residues of NRP-1 form a PDZ-binding motif that influences NRP-1-mediated angiogenesis [15,16]. NRP-1 is highly expressed by numerous tumour cell lines, and enhances tumour survival, growth and vascularization in vivo [17–19].

In vascular endothelial cells, NRP-1 and the VEGFR2 co-cluster, but do not interact directly in the absence of VEGF-A₁₆₅ [20,21]. NRP-1 b1b2 domains can bind the basic C-terminal tail of the heparan-sulfate-binding growth factor VEGF-A₁₆₅, which bridges extracellularly between VEGFR2 and NRP-1, generating a complex with enhanced VEGFR2 signalling that can induce angiogenic sprouting [7,22–26]. Cytoplasmic domains also contribute to VEGFR2–NRP-1 receptor complexes, since inhibiting VEGFR phosphorylation or deleting the PDZ domain of NRP-1 reduces this association [27]. In tumour cells that lack expression of VEGFR2, NRP-1 supports VEGF-mediated endothelial cell migration through PI3K (phosphoinositide 3-kinase)/Akt signalling, implying the existence of other receptors for NRP-1-mediated VEGF function [28,29]. Indeed, NRP-1 associates with heparan-sulfate-binding growth factors bFGF (basic fibroblast growth factor) and HGF (hepatocyte growth factor) [30], and can regulate HGF-induced c-met phosphorylation [31]. PDGF-B also influences vascular smooth muscle cell motility by up-regulating and associating with NRP-1 [32].

The PDGFR and VEGFR tyrosine kinases, and their growth-factor ligands, are closely related structurally and evolutionarily [33,34]. PDGFs induce receptor-specific activation, with PDGF-AA stimulating only PDGFRαα, whereas PDGF-BB stimulates all PDGFR dimers αα, ββ and αβ [35]. PDGF-CC binds to PDGFRs αα and αβ [35], whereas PDGF-AB mainly signals through PDGFRαβ [36]. In early embryonic development, PDGFRα and its major ligand PDGF-A are co-expressed from the two-cell stage, and PDGF-A-stimulated PDGFRα signalling is critical for differentiation of ES (embryonic stem) cells into mesenchymal, neural crest, cranial and myogenic cells, and for epithelial–mesenchymal transformation [37–39]. PDGF-A knockout is embryonic lethal, PDGFRα-null mice die during embryonic development, and mice null for PDGF-C die perinatally [34,40]. PDGFRs are also essential regulators of vessel-wall development [41] and remodelling following injury [42], with PDGF-B a major mitogenic and chemotactic ligand for smooth muscle cells and their mesenchymal precursors. NRP-1 expression also identifies vascular precursors in ES cells [43].

It was recently shown that bone marrow cells are recruited to sites of neovascularization through NRP-1 [44]. In the present study, using MSCs lacking VEGFRs, we show that NRP-1 co-localization with phosphorylated PDGFRs regulates their signalling in a ligand-specific manner, and has an indispensable role in PDGFRα-induced migration and MSC network assembly. This novel receptor cross-talk may thus control the recruitment of MSCs in vascular remodelling.

Human MSCs from normal bone marrow of 20- and 26-year-old females and 18-, 22- and 24-year-old males (obtained from Lonza), were cultured on 0.1% gelatine (Sigma–Aldrich) and maintained and characterized as described previously [45]. For each analysis, MSCs were analysed at passage 4. HUVECs (human umbilical vein endothelial cells) from 35- and 29-year-old females (Cascade Biologics) were maintained as described previously [45]. All growth factors were obtained from R&D Systems and VEGFR2 tyrosine kinase inhibitor V was supplied by Merck.

For single-colour flow cytometry, MSCs (4×10⁶ cells/ml) were incubated with either PE (phycoerythrin)-conjugated anti-human NRP-1 (FAB3870P), VEGFR2 (FAB357P) or control anti-IgG₁ (IC002P) (R&D Systems) antibodies, then processed as described previously [2].

MSCs were cultured on round glass coverslips in 24-well culture dishes, previously coated with 0.1% gelatin overnight at 4 °C, or a thin-layer of growth-factor-reduced Matrigel™ (BD Biosciences) incubated at 37 °C for 30 min. Cells were fixed with 4% (w/v) paraformaldehyde for 20 min, incubated in 0.2 M glycine for 20 min, then permeabilized using 0.5% Triton X-100 in PBS for 4 min. After blocking in 2% fish-skin gelatin in PBS (Sigma–Aldrich), pairs of primary antibodies in blocking solution (2% fish-skin gelatin) were incubated overnight at 4 °C. Primary antibodies were all obtained from Santa Cruz Biotechnology: anti-human NRP-1 (sc-5541), NRP-1 (sc-7239), p-PDGFRα-Tyr⁷⁵⁴ (where p- indicates phosphorylated) (sc-12911), p-PDGFRα-Tyr⁷²⁰ (sc-12910), PDGFRα (sc-338), p-PDGFRβ-Tyr¹⁰²¹ (sc-12909-R), p-PDGFRβ-Tyr⁷⁵¹ (sc-21902-R), p-Flk-1-Tyr¹¹⁷⁵ (sc-101819) and PDGFRβ (sc-339). Cells were then incubated with appropriate Alexa Fluor® 488 and Alexa Fluor® 555 fluorophores (Invitrogen) in blocking solution for 2 h at room temperature (20 °C) and coverslips were mounted on to glass slides with ProLong Gold antifade reagent with DAPI (4′,6-diamidino-2-phenylindole; Invitrogen). Images were collected with a Nikon C1 confocal microscope using a TE2000 PSF inverted microscope, utilizing 60×/NA (numerical aperture) 1.40 Plan Apo or 20×/NA 0.50 Plan Fluor objectives and 3×confocal zoom. Different sample images detecting the same antibodies were acquired under constant acquisition settings. Images were processed using Nikon EZ-C1 FreeViewer v3.3 software. For co-localization analysis, images were processed using ImageJ software and a co-localization analysis plugin. For each analysed image, similar best-fit lower threshold values were determined to reduce the signal background of the corresponding red and green channels, then particle sizes for the red and green channels were set at a minimum of 1 pixel and maximum of 1000 pixels, then co-localization between channels was determined and represented by a yellow image.

MSC migration was determined using a modified Boyden chamber assay as described previously [2]. The number of migratory MSCs on the membrane underside (cells/field using a 10×/NA 0.3 UPlan F1 objective) were determined using an Olympus BX51 widefield microscope. Images were captured with a CoolSNAP camera system and processed using MetaMorph imaging v5.0 software. To determine proliferation, MSCs (2000 cells/well) in growth medium were seeded into 96-well plates coated with 0.1% gelatin and incubated with or without PDGF ligands at 37 °C, with growth medium and ligands exchanged every 24 h. At the end of each time point, a CyQuant cell proliferation assay kit (Invitrogen) was used to detect MSC proliferation as described previously.

MSCs (5×10⁵ cells) together with 3 μg of siRNAs were transfected by electroporation using a human Nucleofector kit (Amaxa), then cultured for 20 h in growth medium at 37 °C in a humidified atmosphere of 5% CO₂ in air. Two different validated siRNAs, functionally tested to provide ≥70% target gene knockdown for NRP-1 were obtained from (i) Qiagen (S102663213) and (ii) Ambion (4390824) and a scrambled siRNA control was also obtained from Qiagen. The targeting specificity and efficiency following individual siRNA knockdowns was evaluated using primers and RT (reverse transcription)–PCR analysis, as described previously [2,45].

MSC lysates were isolated as described previously [45], then 100 μg of lysate was incubated with antibodies against human NRP-1 (sc-7239), PDGFRα (sc-338) or PDGFRβ (sc-339) (Santa Cruz Biotechnology) overnight at 4 °C. Immune complexes were isolated by incubation with 10% (w/v) protein A–Sepharose for 2 h, followed by immunoblot analysis as described previously [45], using antibodies against human PDGFRα (sc-338), PDGFRα-Tyr⁷⁵⁴ (sc-12911), PDGFRβ (sc-339), PDGFRβ-Tyr¹⁰²¹ (sc-12909) or NRP-1 (MAB38701) (R&D Systems). For quantification, the densities of bands were determined using Gene Tools software (Syngene), and normalization to the corresponding loading control.

A cell-based human p-PDGFRβ-Tyr⁷⁵¹ ELISA kit (R&D Systems), was used to measure p-PDGFRβ-Tyr⁷⁵¹ and modified to measure p-PDGFRα-Tyr⁷⁴² utilizing an anti-p-PDGFRα-Tyr⁷⁴² antibody (AF2114) (R&D Systems). MSCs (10000 cells/well) in serum-free medium were seeded on to 0.1% gelatin, stimulated with fresh serum-free medium containing a specific growth factor, then immediately analysed according to the manufacturer's protocol. The p-PDGFR fluorescence at 600 nm in each well was normalized to the total PDGFR fluorescence at 450 nm, and the means of triplicate readings were determined.

Round glass coverslips were coated with a thin layer of growth-factor-reduced Matrigel™ (BD Biosciences), allowed to set, and then seeded with MSCs (2×10⁴) in 0.5% serum growth medium and incubated at 37 °C. For quantification of network formation, the average number of branch points/field after 24 h was determined. Each assay was performed in duplicate, with the number of branch points/field counted from at least six random fields per well.

Briefly, fertilized White Leghorn chicken eggs were incubated at 38 °C for 5 days. Under aseptic conditions in a laminar flow cabinet, a small window at the top of the shell was carefully excised and the CAM blood vessels exposed. MSCs (2×10⁴ cells) were seeded on to a Matrigel™-coated coverslip and incubated for 45 min at 37 °C to allow adherence. Coverslips were implanted MSCs face down on to a highly vascularized area of CAM, the shell opening sealed, and the MSCs incubated in ova at 38 °C for 24 h. Afterwards, coverslips were carefully removed from the CAM, washed in PBS and then processed for immunofluorescence microscopy.

In all quantification experiments, results are expressed as the means±S.D. Statistical differences between sets of data were determined by using a paired Student's t test with SigmaPlot 8.0 software, with P<0.05 considered significant.

We have previously shown that multipotential MSCs which express PDGFRs α and β, but no VEGFRs, on their cell surface, also expressed NRP-1 [2,4]. Using flow cytometry, we confirmed that MSCs express NRP-1 on their cell surface, but not VEGFR-2 (Figure 1A).

Co-immunoprecipitation experiments were conducted to examine whether NRP-1 associates with PDGFRα and/or PDGFRβ (Figure 1B). NRP-1 co-immunoprecipitated with PDGFRα, and vice versa, predominantly in the presence of its ligands PDGF-AA, PDGF-BB or VEGF-A₁₆₅ [Figure 1B (i and iii)]. PDGFRβ and NRP-1 co-immunoprecipitated in the presence of PDGF-BB or VEGF-A₁₆₅ [Figure 1B (ii and iv)]. These data demonstrate ligand regulation of the association of PDGFRs α and β with NRP-1.

To estimate the percentage of PDGFRs in a particular cell lysate which interact with NRP-1, co-immunoprecipitation analysis of total PDGFRs [Figure 1C (i and ii)] and phosphorylated PDGFRs [Figure 1C (iii and iv)] was evaluated. Using unstimulated control MSCs, co-immunoprecipitation analysis demonstrated ~5.0±0.7% total PDGFRα or PDGFRβ associated with NRP-1 [Figure 1C (i and ii)], but MSC exposure to PDGF-AA resulted in 67±8% total PDGFRα being associated with NRP-1, while exposure to PDGF-BB induced 36±7% total PDGFRβ to associate with NRP-1 [Figures 1C (i and ii) and 1D (i)]. Similarly, co-immunoprecipitation analysis of PDGFRα phosphorylated at Tyr⁷⁵⁴ or PDGFRβ at Tyr¹⁰²¹ [Figure 1C (iii and iv)], demonstrated that unstimulated MSCs displayed ~5.0±0.7% phosphorylated PDGFRs associated with NRP-1, whereas exposure to PDGF-AA or PDGF-BB induced 63±6% PDGFRα phosphorylated at Tyr⁷⁵⁴ and 31±6% PDGFRβ phosphorylated at Tyr¹⁰²¹ respectively to associate with NRP-1 [Figures 1C (iii and iv) and 1D (ii)]. These results are based on the proportion of receptor association within the immunoprecipitates. The values are likely to be indicative of total amounts of associated receptors, although the immunoprecipitation of each receptor from a cell lysate may not be 100% efficient. Because the estimated proportions of total and phosphorylated PDGFRs which co-immunoprecipitated with NRP-1 in a particular cell lysate are comparable [see Figure 1D (i and ii)], the data suggest that virtually all of the PDGFRs which associate with NRP-1 are phosphorylated.

To further demonstrate that PDGF ligand stimulation induces NRP-1 to associate with PDGFRs, we examined the cellular distribution of NRP-1 and phosphorylated PDGFRs by immunofluorescence microscopy.

Analysis of unstimulated control MSCs demonstrated that PDGFRα-Tyr⁷⁵⁴ and PDGFRβ-Tyr¹⁰²¹ immunoreactivity predominantly localized around perinuclear regions, but was also detected at low levels peripherally, whereas NRP-1 immunoreactivity had a wider cellular distribution which in some cases extended towards the cell surface (Figures 2A and 2C). In contrast, MSCs exposed to PDGF-AA or PDGF-BB showed widespread cellular PDGFRα-Tyr⁷⁵⁴ (Figure 2B), or PDGFRβ-Tyr¹⁰²¹ (Figure 2D) immunoreactivity respectively. Co-localization analysis of unstimulated MSCs demonstrated minimal co-localization between NRP-1 and PDGFRα-Tyr⁷⁵⁴ and PDGFRβ-Tyr¹⁰²¹; however, MSCs exposed to PDGF-AA or PDGF-BB resulted in a significant increase (P<0.001, compared with unstimulated controls) in co-localization between NRP-1 and PDGFRα-Tyr⁷⁵⁴ or PDGFRβ-Tyr¹⁰²¹ respectively (Figure 2E), with PDGFRα-Tyr⁷⁵⁴ consistently producing the highest level of co-localization.

We also examined the cellular distribution of NRP-1 and total PDGFRs, using pan-PDGFR antibodies. In both unstimulated control and ligand-stimulated MSCs, while total PDGFRα immunoreactivity predominantly localized to perinuclear regions, the total PDGFRβ and NRP-1 immunoreactivity had a wider cellular distribution (Supplementary Figures S1A–S1D at http://www.BiochemJ.org/bj/427/bj4270029add.htm). Co-localization analysis demonstrated a low level of co-localization between NRP-1 and total PDGFRs in unstimulated MSCs, but a significant increase (P<0.001, compared with unstimulated controls) on exposure to PDGF ligands (Supplementary Figure S1E), similar to the co-localization determined between NRP-1 and phosphorylated PDGFRs (Figure 2E), but at a relatively lower level.

To compare the distribution of ligand-stimulated NRP-1/PDGFR in MSCs with ligand-induced NRP-1/VEGFR2 co-localization in endothelial cells (HUVECs), we examined VEGF-A₁₆₅-induced NRP-1/PDGFRα-Tyr⁷⁵⁴ and NRP-1/PDGFRβ-Tyr¹⁰²¹ co-localization within MSCs, with VEGF-A₁₆₅-induced NRP-1/VEGFR2-Tyr¹¹⁷⁵ co-localization within HUVECs. Although HUVECs generally displayed a wider NRP-1 distribution than MSCs, unstimulated HUVECs and MSCs both demonstrated minimal NRP-1/VEGFR2 and NRP-1/PDGFR co-localization respectively (Figures 3A, 3C and 3E). In contrast, VEGF-A₁₆₅ stimulation significantly increased (P<0.001, compared with unstimulated controls) the co-localization of NRP-1 with VEGFR2-Tyr¹¹⁷⁵ in HUVECs (Figure 3B), as well as NRP-1 with PDGFRα-Tyr⁷⁵⁴ (Figure 3D) and PDGFRβ-Tyr¹⁰²¹ (Figure 3F) in MSCs.

Thus in MSCs, PDGF and VEGF-A₁₆₅ ligands induce co-localization of NRP-1 with phosphorylated PDGFRs, similar to VEGF-A₁₆₅-induced NRP-1 co-localization with p-VEGFR2 in HUVECs.

Having established that NRP-1 can associate and co-localize with phosphorylated PDGFRα and PDGFRβ, we investigated whether NRP-1 regulates PDGFR signalling. Following NRP-1 knockdown using two different siRNAs, NRP-1 protein expression was virtually ablated (Figure 4A), whereas RT–PCR analysis of their targeting specificity demonstrated they did not affect PDGFR transcripts (Figure 4B). We therefore utilized NRP-1 knockdown during the present study, first examining PDGFRα phosphorylation levels in serum-free conditions, using an ELISA for PDGFRα-Tyr⁷⁴². NRP-1 knockdown had little impact on basal levels of unstimulated PDGFRα phosphorylation (Figure 4C). However, exposure to PDGF-AA strongly stimulated PDGFRα phosphorylation, with NRP-1 knockdown dramatically reducing this phosphorylation to near-basal levels (Figure 4C). Although PDGF-CC and VEGF-A₁₆₅ stimulated lower levels of PDGFRα phosphorylation, NRP-1 knockdown also reduced their phosphorylation to near-basal levels (Figure 4C). Thus NRP-1 markedly regulates PDGFRα signalling when stimulated by these growth factors. However, NRP-1 knockdown did not inhibit PDGF-AB-stimulated PDGFRα phosphorylation. The differential effects of NRP-1 knockdown on PDGF-CC- and PDGF-AB-induced PDGFRα phosphorylation reflect these growth-factor-binding specificities; PDGF-CC can bind the PDGFRαα homodimer and PDGFRαβ heterodimer [35,46], whereas PDGF-AB mainly binds PDGFRαβ [36]. These results thus imply ligand-induced NRP-1-dependent PDGFRα homodimer signalling.

We also investigated whether NRP-1 regulates PDGFRβ signalling in serum-free conditions using an ELISA for PDGFRβ-Tyr⁷⁵¹. Control scrambled and target NRP-1 siRNA knockdowns resulted in comparable basal levels of unstimulated PDGFRβ phosphorylation (Figure 4D). While exposure to PDGF-BB strongly stimulated PDGFRβ phosphorylation, NRP-1 knockdown only partially reduced this phosphorylation (Figure 4D). However, whereas VEGF-A₁₆₅ stimulated lower levels of PDGFRβ phosphorylation, NRP-1 knockdown effectively inhibited this phosphorylation. Thus NRP-1 also regulates ligand-induced PDGFRβ phosphorylation. Since NRP-1 knockdown only decreased PDGF-BB-induced PDGFRβ phosphorylation by ~44±5%, we examined whether NRP-1 knockdown primarily affects PDGF-BB-induced PDGFRαβ phosphorylation, using PDGF-AB and PDGF-CC which bind the PDGFRαβ heterodimer, but not a PDGFRββ homodimer [35,36,46]. While both of these ligands induced PDGFRβ phosphorylation, indicating heterodimer stimulation, NRP-1 knockdown had no inhibitory effect in either case (Figure 4D). These results thus imply that NRP-1 influences ligand-induced PDGFRβ homodimer signalling.

Having established that NRP-1 plays a prominent role in regulating ligand-induced PDGFR signalling, the functional importance of this receptor cross-talk was investigated. We have previously demonstrated that VEGF-A- or PDGF-induced PDGFR signalling stimulates migration of MSCs [2]. In the present study we examined, in serum-free conditions, whether NRP-1 regulates PDGFR-mediated MSC migration. In the absence of ligand, control scrambled and target NRP-1 siRNA knockdowns resulted in comparable basal levels of unstimulated MSC migration (Figures 5A and 5B), similar to NRP-1 effects on PDGFR phosphorylation (Figures 4C and 4D). Exposure to PDGF-AA (PDGFRα homodimer mediated) increased MSC migration, but was inhibited to virtually basal levels by NRP-1 knockdown (Figures 5A and 5B). Likewise, NRP-1 knockdown also decreased PDGF-BB- and VEGF-A₁₆₅-induced MSC migration, by ~38±6% and ~56±8% respectively (Figures 5A and 5B). Control scrambled and NRP-1 knockdowns, in the presence or absence of a VEGFR2 tyrosine kinase inhibitor, produced a similar level of VEGF-A₁₆₅-induced MSC migration (Figure 5B), indicating that a VEGFR2–NRP-1 complex was not contributing to VEGF-A₁₆₅-stimulated MSC migration. NRP-1 had a minimal effect on PDGF-CC- or PDGF-AB-induced MSC migration, confirming that these ligands probably stimulate migration through PDGFRαβ, but independently of NRP-1 (results not shown).

We also investigated the effects of NRP-1 knockdown on serum-stimulated MSC proliferation, in the absence or presence of supplementary PDGFR ligands. In serum growth medium alone, scrambled knockdown control MSCs proliferated up to 5 days, which was increased by supplementary PDGF-AA (Figure 5C) or PDGF-BB (Figure 5D). In comparison, at each timepoint NRP-1 knockdown significantly inhibited serum-stimulated MSC proliferation (Figures 5C and 5D). Moreover, NRP-1 knockdown decreased serum- and PDGF-ligand-supplemented proliferation to comparable levels (Figures 5C and 5D), indicating that NRP-1 knockdown was inhibiting PDGF-ligand-stimulated MSC proliferation. Similar results were obtained following NRP-1 knockdown of serum-stimulated MSC proliferation, in the absence or presence of supplementary VEGF-A₁₆₅ (results not shown).

These results highlight the crucial contribution of NRP-1 to ligand-induced PDGFR-mediated MSC migration and proliferation.

MSCs are critical contributors to neovascularization [47], and NRP-1 and PDGFRs play essential roles in this process [11,41]. Having established that NRP-1 plays a crucial role in regulating PDGFR-mediated MSC phosphorylation, migration and proliferation, we went on to examine the function of NRP-1 in regulating the assembly of MSC network formation in a Matrigel™ culture model over 24 h.

We first evaluated the distribution of NRP-1 and p-PDGFRs during MSC network assembly by determining the immunolocalization of NRP-1 and PDGFRα-Tyr⁷⁵⁴ or PDGFRβ-Tyr¹⁰²¹. During the initial stages of network assembly, MSCs exhibited intense widespread NRP-1 immunoreactivity (Figure 6). After 2 h seeding on to Matrigel™, high levels of PDGFRα-Tyr⁷⁵⁴ immunoreactivity were predominantly localized around the cell surface, where it conspicuously co-localized with NRP-1 (Figure 6A). In comparison, PDGFRβ-Tyr¹⁰²¹ was distributed throughout the cell, including the cell surface, but co-localization with NRP-1 was at a lower level (Figure 6B). After 6 h, PDGFRα-Tyr⁷⁵⁴ and NRP-1 co-localization had a wider cellular distribution (Figure 6C), whereas co-localization of PDGFRβ-Tyr¹⁰²¹ with NRP-1 became more prominent, especially at the cell surface (Figure 6D). By 24 h, MSCs had assembled to form abundant capillary-like network structures, which displayed high levels of co-localized NRP-1 with PDGFRα-Tyr⁷⁵⁴ and PDGFRβ-Tyr¹⁰²¹ (Figures 6E and 6F). To further substantiate these data, NRP-1 was also shown to co-localize with PDGFRα-Tyr⁷²⁰ and PDGFRβ-Tyr⁷⁵¹ in MSC networks (Supplementary Figures S2A and S2B at http://www.BiochemJ.org/bj/427/bj4270029add.htm).

We confirmed that VEGFR2-Tyr¹¹⁷⁵ was not expressed in MSC networks (Supplementary Figure S3A at http://www.BiochemJ.org/bj/427/bj4270029add.htm). However, HUVECs in Matrigel™ readily formed capillary-like network structures, as expected, which not only displayed prominent co-localization of NRP-1 with VEGFR2-Tyr¹¹⁷⁵ (Supplementary Figure S3B), but also with PDGFRα-Tyr⁷⁵⁴ and PDGFRβ-Tyr¹⁰²¹ (Supplementary Figures S3C and S3D), suggesting that PDGFRs may also influence NRP-1 function in endothelial cells.

Having demonstrated that NRP-1 regulates PDGFR signalling and that both receptors are abundantly co-localized within MSC network structures, we went on to examine MSC network assembly following NRP-1 knockdown (Figure 7). Control scrambled siRNAs resulted in MSCs forming widespread capillary-like network structures within 24 h, containing pronounced NRP-1, PDGFRα-Tyr⁷⁵⁴ and PDGFRβ-Tyr¹⁰²¹ immunoreactivity and co-localization (Figures 7A, 7C and 7E), comparable with untransfected MSCs (see Figures 6E and 6F). However, after 24 h, NRP-1 knockdown MSCs produced distinctly disorganized structures (Figures 7B, 7D and 7F), containing significantly fewer branch points compared with control MSCs (Figure 7G). As expected, there was a dramatic reduction in NRP-1 immunoreactivity, confirming the efficiency of the knockdown, concurrent with a distinct decrease in both PDGFRα-Tyr⁷⁵⁴ and PDGFRβ-Tyr¹⁰²¹ immunoreactivity (Figures 7B and 7D).

These results indicate that NRP-1 regulation of PDGFR signalling plays a crucial role in directing MSC network assembly.

To further demonstrate the importance of NRP-1 during MSC network formation, we examined the effects of NRP-1 knockdown utilizing an in vivo angiogenesis model system: the CAM of the developing chicken embryo [48]. Control scrambled or NRP-1 knockdown MSCs were seeded on to Matrigel™, then implanted in contact with a highly vascularized area of CAM for 24 h. As a control, identically prepared MSCs were also cultured in vitro. Following intimate association with the underlying CAM blood vessel microenvironment for 24 h, control MSCs formed widespread capillary-like network structures, containing abundant NRP-1, PDGFRα-Tyr⁷⁵⁴ and PDGFRβ-Tyr¹⁰²¹ immunoreactivity (Figures 8A, 8C, 8E and 8F). Both PDGFRα-Tyr⁷⁵⁴ and PDGFRβ-Tyr¹⁰²¹ displayed a high level of co-localization with NRP-1 in these in vivo assembled networks (Figures 8E and 8F), similar to the control in-vitro-cultured MSCs (results not shown) as previously demonstrated (see Figures 7A and 7C). In striking contrast, however, after 24 h of CAM exposure, NRP-1 knockdown resulted in widespread clusters of MSCs maintaining a rounded cellular morphology, which only exhibited trace levels of NRP-1, PDGFRα-Tyr⁷⁵⁴ or PDGFRβ-Tyr¹⁰²¹ immunoreactivity (Figures 8B and 8D). In comparison, the control in-vitro-cultured NRP-1 knockdown MSCs produced highly disorganized network assemblies (results not shown), as previously demonstrated (see Figures 7B and 7D).

Thus NRP-1 is critical for PDGFR signalling and the in vivo assembly of MSC network structures within the CAM vasculature microenvironment.

MSCs, which offer immense potential for cell-based tissue regeneration, have the capability to differentiate along vascular cell lineages [1–3]. Previously, we have shown that multipotent human MSCs express NRP-1 and PDGFRs, but not VEGFRs, and that PDGFs regulate MSC proliferation and migration, and the smooth muscle-specific cytoskeleton [2,4]. PDGFRα is an essential regulator of mesenchymal tissue formation in early embryonic development [38], and both PDGFRs contribute to vessel-wall development and remodelling following injury [42]. The essential contribution of NRP-1 to vascular development and neovascularization is also well documented [11] and, although its mechanisms of action remain incompletely understood, it is thought to regulate cell-surface-receptor clustering and signalling in a ligand-dependent manner. Our discovery that NRP-1 regulates the phosphorylation and signalling responses of PDGFRs, especially PDGFRα, sheds important light on fundamental cellular mechanisms of tissue development and neovascularization.

NRP-1 co-immunoprecipitated and co-localized with p-PDGFRs, and this association was significantly increased in the presence of growth-factor ligands, indicating that the PDGFR cross-talk with NRP-1 that we have identified may occur through a receptor-bridging mechanism. Indeed, in-vitro-binding studies indicate that PDGFRα and NRP-1 do not interact directly, but PDGF ligands, PDGF-AA, PDGF-BB and VEGF-A₁₆₅, all bind NRP-1 (Supplementary Figure S4 at http://www.BiochemJ.org/bj/427/bj4270029add.htm). PDGF-AA-mediated PDGFRα responses were particularly dependent upon NRP-1, implying that NRP-1 may be indispensable for PDGFRα function in tissue development and remodelling. PDGFRβ dependence on NRP-1 was also significant, so NRP-1 must regulate PDGFRβ-dependent smooth muscle cell migration, proliferation and differentiation during vessel-wall maturation and repair.

While NRP-1 is a transmembrane protein, immunofluorescence analysis demonstrated that the majority of NRP-1 in permeabilized MSCs and HUVECs was localized intracellularly. Exposure to VEGF-A₁₆₅ has been shown to promote NRP-1 on the surface of HUVECs to internalize, with immunofluorescence analysis of the permeablized HUVECs demonstrating NRP-1 predominantly localized around perinuclear regions [49]. Thus a similar mechanism resulting in rapid ligand-induced NRP-1 internalization may occur in MSCs.

MSCs readily formed extensive networks in Matrigel™ and CAM assays, highlighting their potential to contribute to blood-vessel formation. Co-localization of NRP-1 with phosphorylated PDGFRs occurred prominently in these networks, and the essential role for NRP-1/PDGFR cross-talk in network formation was confirmed by knockdown of NRP-1 which caused dramatically reduced PDGFR phosphorylation and grossly disrupted network formation. Prominent pericellular co-localization of NRP-1 with PDGFRα during early network formation suggests that this relationship is particularly important in initiating cellular changes leading to network formation. In MSCs, VEGF-A also induced NRP-1/PDGFR co-localization, similar to VEGF-A-induced NRP-1 co-localization with VEGFR2 in HUVECs. Since in HUVECs, NRP-1 co-localized with both VEGFR2 and PDGFRs in response to VEGF-A, NRP-1/PDGFR cross-talk is likely to contribute to endothelial functions mediated by VEGF-A. Our MSCs do not express VEGFRs, so their response to PDGFs and VEGF-A ligands is channelled through PDGFRs, but in endothelial and other cells expressing both VEGFRs and PDGFRs, the relative abundance of each receptor, local ligand concentrations and receptor affinities may combine to modify NRP-1-dependent receptor signals.

The essential contribution of PDGFRα to the formation of embryonic mesoderm and mesenchymal tissues is well documented [34], and we have demonstrated a high PDGFRα/PDGFRβ ratio in our MSCs [4]. PDGFRα-null mice die at around E10 (where E is embryonic day) due to vascular and other defects [40], whereas conditional null mice highlight that both PDGFRs are essential for early yolk sac vascular development [41]. The NRP-1-knockout mouse is also embryonic lethal, with major yolk sac and embryonic vascular defects, dying between E10 and E12.5 [50]. Thus the functional cross-talk between NRP-1 and both PDGFRs, especially PDGFRα, that we have identified, suggests a fundamental developmental relationship between these receptors.

In summary, in the present study we have shown ligand-dependent cross-talk between NRP-1 and phosphorylated PDGFRs that controls receptor signalling, migration, network formation and proliferation of MSCs. We have thus identified NRP-1 as an essential co-receptor for PDGFR signalling, which may critically contribute to the formation of blood vessels and other mesenchymal tissues. This mechanism may be exploited in the application of MSCs in tissue regeneration.

CAM

chorioallantoic membrane

DAPI

4′,6-diamidino-2-phenylindole

ES

embryonic stem

HGF

hepatocyte growth factor

HUVEC

human umbilical vein endothelial cell

MSC

mesenchymal stem cell

NA

numerical aperture

NRP-1

neuropilin-1

p-

phosphorylated-

PDGF

platelet-derived growth factor

PDGFR

PDGF receptor

RT

reverse transcription

siRNA

small interfering RNA

VEGF

vascular endothelial growth factor

VEGFR

VEGF receptor

Stephen Ball contributed to experimental design, conducted MSC experiments and contributed to the writing of the manuscript. Christopher Bayley conducted the molecular-binding studies shown in Supplementary Figure S4. Adrian Shuttleworth contributed to experimental design and the writing of the manuscript. Cay Kielty directed the experiments and contributed to the writing of the manuscript.

We thank Ms Jemima Whyte (University of Manchester, Manchester, U.K.) for her contribution towards developing the CAM assay, and Ms Maybo Chiu for recombinant VEGF-A used in the binding assays.

This work was funded by the Medical Research Council (UK) [grant number G0700712]. C.M.K. holds a Royal Society Wolfson Research Merit award.