Ena/VASP proteins mediate repulsion from ephrin ligands

Ena/VASP family proteins inhibit cell motility by modulating the architecture of the actin cytoskeleton (Bear et al., 2002). Capping of the barbed ends of non-productive actin filaments prevents continued polymerisation and maintains a short, highly-branched lamellipodial actin network, an arrangement more suitable for productive protrusion (Pollard and Borisy, 2003). The presence of Ena/VASP proteins at the leading edge of cells antagonises capping proteins at barbed ends of actin filaments, but crucially enables actin polymerisation to continue, generating long, unbranched actin filaments (Bear et al., 2002; Barzik et al., 2005). However, these Ena/VASP-modulated filaments are prone to buckling and thus destabilise lamellipodia, leading to increased membrane ruffling and reduced cell motility (Bear et al., 2000; Bear et al., 2002).

The Ena/VASP family consists of VASP (Ddx4; Mouse Genome Informatics), Mena (Enah; MGI) and Evl in vertebrates (Krause et al., 2003), each of which can function interchangeably, at least with respect to their roles in the regulation of fibroblast and Listeria motility (Geese et al., 2002; Laurent et al., 1999; Loureiro et al., 2002). In addition to these roles, this family has been implicated downstream of several repulsive guidance cue receptors, including Robo, Dlar and Sema6D (Bashaw et al., 2000; Toyofuku et al., 2004; Wills et al., 1999), but as yet there is no evidence suggesting a link downstream of Eph receptors.

Eph receptors constitute the largest family of receptor tyrosine kinases found in mammalian genomes and fall into two distinct classes on the basis of sequence homology (Pasquale, 2005): EphA receptors typically bind GPI-anchored ephrin-A ligands, whereas EphBs bind the transmembrane ephrin-Bs (Gale et al., 1996), although examples of promiscuity between these groupings exist (Gale et al., 1996; Himanen et al., 2004). Since both Eph receptors and ephrins, are expressed at the cell surface, cell-cell contact leads to Eph receptor activation (Davis et al., 1994), the most characterised outcome of which is cell-cell repulsion or growth cone collapse (Drescher et al., 1995; Wang and Anderson, 1997; Xu et al., 1999). For example, the polarised expression of ephrin-Bs in the posterior halves of the somitic sclerotome is important in restricting the migration of neural crest cells, through cell-cell repulsion, to the anterior halves (Krull et al., 1997; Wang and Anderson, 1997). Indeed, neural crest cells avoid stripes of substrate-bound ephrin-B1 and ephrin-B2 in vitro and the application of soluble ephrin-B1 interferes with Eph receptor-ephrin interactions, disrupting segmented migration through the somites in trunk explant assays (Krull et al., 1997; Wang and Anderson, 1997).

Although a number of downstream signalling pathways have been implicated in the generation of ephrin-induced repulsion and cell-cell separation, the exact mechanisms remain poorly characterised. Interfering with various signalling molecules downstream of EphAs partially blocks ephrin-A-induced growth cone collapse, suggesting that several parallel signalling pathways contribute to repulsion from ephrin-As. Src family kinases, which are required for full collapse responses (Knoll and Drescher, 2004), phosphorylate Ephexin, a Rho family GTPase GEF, downstream of EphA activation to shift its specificity towards Rho (Sahin et al., 2005; Shamah et al., 2001). Increasing the ratio of Rho and Rac/Cdc42 activity enhances ROCK-induced contractility and opposes protrusion (Kozma et al., 1997; Shamah et al., 2001; Wahl et al., 2000). In addition, downregulation of R-Ras activity may reduce integrin-mediated adhesion during growth cone collapse (Dail et al., 2006; Zou et al., 1999).

High affinity Eph receptor-ephrin contacts must also be attenuated to allow cell-cell separation during repulsive responses. Two mechanisms have been demonstrated to facilitate this: metalloproteases cleave ephrin-As to facilitate the withdrawal of neuronal processes (Hattori et al., 2000), whereas Rac-dependent endocytosis of EphB-ephrin-B complexes appears to terminate these cell-cell contacts (Marston et al., 2003; Zimmer et al., 2003). The Rac-dependent internalisation of Eph receptors temporally correlates with ephrin-induced cell-cell separation (Marston et al., 2003) and growth cone collapse (Cowan et al., 2005; Jurney et al., 2002).

In this study we examined neural crest cell responses to both substrate-bound and soluble ephrin-B2. These treatments destabilised neural crest cell lamellipodia and may therefore contribute to the generation of repulsion, downstream of Eph receptor activation. Next we investigated whether the Ena/VASP family contributes to repulsion from substrate-bound ephrin ligands by comparing the repulsive responses of VASP^–/– mouse embryonic fibroblasts (MEFs) and MV^D7 cells to wild-type MEFs; MV^D7 cells are derived from VASP^–/–Mena^–/– mice and lack detectible VASP, Mena and Evl (Bear et al., 2000). Both VASP^–/– MEFs and MV^D7 cells displayed reduced repulsion from substrate-bound ephrin-A5 and ephrin-B2. Additionally, we found that VASP and Mena co-accumulated with activated Eph receptors at sites of plasma membrane contact between EphB4- and ephrin-B2-expressing cells and that this localisation is necessary for internalisation of Eph receptor-ephrin complexes. Taken together, our data suggest that the Ena/VASP family plays a functional role downstream of Eph receptor signalling.

In order to investigate the dynamic behaviour of neural crest cells at boundaries demarcated by ephrin ligands, we examined these cells using timelapse video microscopy as they migrated over stripes of substrate-bound ephrin-B2-Fc. Stripes of ephrin-B2-Fc were adsorbed to glass coverslips using specially designed silicon matrices (Vielmetter et al., 1991); areas of ephrin-B2-Fc adsorption were labelled by the incorporation of Texas Red-dextran in the ephrin-B2-Fc stripes. Neural tube explants were plated perpendicularly to these patterned coverslips, such that, when neural crest cells delaminated from neural tube explants, they encountered the stripes of substrate-bound ephrin-B2-Fc.

Neural crest cells avoided the substrate-bound ephrin-B2-Fc, resulting in a striped pattern of cells emerging from neural tube explants, analogously to their restricted migration through the anterior halves of the somites in vivo (Fig. 1A). This behaviour was confined to the periphery of the neural crest cell sheet; presumably contact-inhibition takes precedence over responses to ephrin-B2-Fc close to the neural tube, where neural crest cells are at a higher cell density. By contrast, neural crest cells emerging onto negative control stripes, from which ephrin-B2-Fc was omitted, failed to respond repulsively (Fig. 1B), revealing this behaviour to be specific to the presence of ephrin-B2-Fc and not a non-specific effect of adsorbing stripes of protein to glass coverslips.

Static images revealed that the majority of neural crest cells distal to the neural tube avoided the ephrin-B2-Fc stripes. However, timelapse movies showed that neural crest cells do extend lamellipodia over ephrin-B2-Fc stripes, suggesting that contact with ephrin-B2-Fc was not absolutely repulsive (Fig. 1D). Nevertheless, analysis of these movies revealed that dynamic interactions with ephrin-B2-Fc stripes generally caused a rapid loss of lamellipodia in the neural crest cells (Fig. 1C; supplementary material, Movie 1); changes to the structure of lamellipodia were visible as soon as 1 minute post-contact, with collapse occurring as rapidly as within 4 minutes. Neural crest cells were observed to internalise phase-bright vesicles, particularly after ephrin-B2-Fc-contact (Fig. 1E).

Neural crest cells appeared to exhibit a `love-hate' relationship with ephrin-B2-Fc, with some cells preferring to migrate along the edges of the ephrin-B2-Fc stripes. This was not an artefact of generating striped coverslips, since this behaviour was not observed on negative control stripes (Fig. 1B) and was therefore specific to the presence of ephrin-B2-Fc. Neural crest cells underwent repeated cycles of protrusion onto, followed by retraction from, the ephrin-B2-Fc stripes as they advanced away from the neural tube explant (Fig. 1D; supplementary material, Movie 1). Retraction was associated with the rearwards collapse of protrusions as membrane ruffles (Fig. 1D,E). This behaviour appears to be crucial to the loss of neural crest cell lamellipodia, and thus the avoidance of stripes of substrate-bound ephrin-B2-Fc by neural crest cells.

To investigate how ephrin-B2-Fc altered the dynamics of neural crest cell lamellipodia, cells distal to the neural tube explant were examined using timelapse video microscopy for 30 minutes prior to stimulation and then for 45-60 minutes in the presence of 2 μg/ml soluble, pre-clustered ephrin-B2-Fc. Treatment with ephrin-B2-Fc failed to elicit a synchronised membrane retraction or cell contraction event, as observed for retinal ganglion cell (RGC) growth cones (Meima et al., 1997); however, neural crest cells appeared to cycle between lamellipodial protrusion and collapse more frequently in the presence of ephrin-B2-Fc (supplementary material, Movie 2). The loss of lamellipodial protrusions was once again characterised by the formation of membrane ruffles, which were associated with the internalisation of phase-bright vesicles (Fig. 2A,B; supplementary material, Movie 2).

The dynamics of neural crest cell lamellipodia were analysed by kymography before and after stimulation (Bear et al., 2002; Hinz et al., 1999). In kymograph analysis a single pixel line is taken from each frame in a timelapse series of images through the leading edge of the cell, along the central axis of protrusion and retraction. These single pixel lines are then placed side by side to assemble the kymograph, such that the y axis represents displacement of the leading edge, while the x axis corresponds to the timecourse of the timelapse movie. Since neural crest cells often repolarise, timelapse movies were divided up into sections after each repolarisation for each cell (a new axis of protrusion and retraction was then assigned for each section); these `mini' kymographs were assembled for each section and combined to produce a single kymograph covering the entire movie. Lamellipodial protrusion results in a positive gradient and retraction in a negative one, thus kymograph peaks represent transitions between protrusion and retraction.

Prior to stimulation with soluble, pre-clustered ephrin-B2-Fc, kymograph peaks were smooth and infrequent (15.5±1.2 peaks/hour); after stimulation with 2 μg/ml pre-clustered ephrin-B2-Fc, peaks became more jagged and their frequency increased to 25.2±1.1 peaks/hour (Fig. 2C,D; nine cells processed from three independent experiments). Therefore, both substrate-bound and soluble, pre-clustered ephrin-B2-Fc appeared to destabilise neural crest cell lamellipodia; this may facilitate the withdrawal of cell processes from sites of cell-ephrin-B2-Fc contact by membrane ruffling.

Since Ena/VASP proteins are believed to destabilise fibroblast lamellipodia (Bear et al., 2002) and are known to contribute to repulsion downstream of guidance cue receptors such as Robo and Dlar (Bashaw et al., 2000; Wills et al., 1999), we next examined whether Ena/VASP proteins could mediate cellular repulsion from ephrins.

Since VASP^–/–Mena^–/– mice display defects in closure of the neural tube (Menzies et al., 2004), we used embryonic fibroblast cell lines generated from this and other knockout mice to examine the contribution of Ena/VASP proteins to repulsion from substrate-bound ephrins. Wild-type MEFs express functional Eph receptors that become tyrosine phosphorylated in response to treatment with 1 μg/ml pre-clustered ephrin-A5-Fc, as shown by immunoprecipitation of EphA4 and subsequent western blotting of samples for EphA4 and phospho-tyrosine (Fig. 3A). Indeed, when 6×10³ wild-type MEFs were plated onto ephrin-A5-Fc, ephrin-B2-Fc or negative control (human Fc) stripes for 18 hours, these cells avoided both types of ephrin, but were distributed randomly over negative control stripes (Fig. 3B-E). These data show that wild-type MEFs express endogenous Eph receptors that are activated in response to soluble, pre-clustered ephrins (Fig. 3A, Fig. 4E) and mediate repulsion from substrate-bound ephrin-A5-Fc and ephrin-B2-Fc (Fig. 3B-D). We next investigated whether the degree of repulsion might be reduced in the absence of Ena/VASP proteins.

Wild-type MEFs, VASP^–/– MEFs (Hauser et al., 1999) and Ena/VASP-deficient MV^D7 cells (Bear et al., 2000) were stimulated for 10 minutes with 1 μg/ml pre-clustered ephrin-A5-Fc, ephrin-B2-Fc or Fc (negative control) as affinity probes. Following fixation, coverslips were stained for the goat anti-human IgG antibodies, used to pre-cluster these affinity probes, to detect clusters of ephrin-bound Eph receptors. Each cell type bound both ephrin affinity probes, suggesting that they can respond to both ephrin-A (Fig. 4A-D) and ephrin-B ligands (data not shown); cells failed to bind pre-clustered Fc, demonstrating specificity of ephrin-Eph receptor interactions (Fig. 4B). Furthermore, each cell line was able to activate EphA4 in response to 1 μg/ml pre-clustered ephrin-A5-Fc (Fig. 4E), revealing that the loss of Ena/VASP proteins does not interfere with receptor activation.

To test the contribution of Ena/VASP proteins to repulsion from substrate-bound ephrin, 6×10³VASP^–/– MEFs and MV^D7 cells were plated onto ephrin-A5-Fc, ephrin-B2-Fc and negative control (human Fc) stripes for 18 hours and their behaviour compared to that of wild-type MEFs. In order to quantify this assay the percentage of cells avoiding ephrin or equivalent stripes in negative controls was calculated; the repulsive response was defined as the difference between the percentage of cells avoiding ephrin and the percentage of cells avoiding the equivalent (red) stripe in negative controls (see arrow in Fig. 3E), i.e. that specific to the presence of ephrin. This was then normalised according to the wild-type MEF response for each experiment. VASP^–/– MEFs exhibited a moderate, but significant decrease in repulsion from both substrate-bound ephrin-A5-Fc and ephrin-B2-Fc to 87±2% and 75±6% of the wild-type MEF repulsive response, respectively (Fig. 4F). The repulsive response was further reduced in MV^D7 cells to 34±27% and 56±17% of the wild-type MEF response to ephrin-A5-Fc and ephrin-B2-Fc, respectively (Fig. 4F). Both VASP^–/– MEFs and MV^D7 cells displayed a slightly higher than expected avoidance of negative control stripes (114±5% and 108±6% of the wild-type MEF avoidance on negative control stripes, respectively), however the percentage of cells avoiding ephrin was not similarly increased, indicating a reduction in the repulsive response.

Thus, it seems that Ena/VASP proteins play a functional role downstream of Eph receptors and are necessary for normal levels of repulsion from substrate-bound ephrins. The fact that the level of repulsion was further decreased in MV^D7 cells suggests that Mena acts partially redundantly with VASP, similarly to its role in regulation of fibroblast and Listeria motility (Geese et al., 2002; Laurent et al., 1999; Loureiro et al., 2002). Fibroblast cell lines are not thought to express the other vertebrate Ena/VASP protein, Evl (Lambrechts et al., 2000), and so Evl is unlikely to contribute to repulsion in fibroblast-derived cell types. Despite the absence of VASP, Mena and Evl in MV^D7 cells, these cells can still partially respond to ephrin ligands, suggesting that Ena/VASP-dependent mechanisms are not exclusively responsible for the generation of repulsion in this assay.

Physiological interactions between Eph receptors and ephrins occur at the surface of neighbouring, touching cells (Davis et al., 1994; Marston et al., 2003; Zimmer et al., 2003). One approach to studying the events following Eph receptor-ephrin contact at the cell surface is to microinject a line of confluent, starved Swiss 3T3 fibroblasts with an Eph receptor expression construct, and to express ephrin in an immediately adjacent line of cells (Marston et al., 2003). Injection of 100 μg/ml pCIneo-EphB4 into cells adjacent to cells injected with 200 μg/ml pRK5-ephrin-B2 typically resulted in the formation of membrane protrusions and ruffles by EphB4-expressing cells at sites of Eph receptor activation, followed by the internalisation of activated Eph receptor/ephrin-B2 complexes into EphB4-expressing cells, which is believed to facilitate cell-cell separation, after expression for approximately 3.5 hours (Fig. 5A,D,E) (Marston et al., 2003).

In order to address whether Ena/VASP proteins were suitably localised to exert an influence on repulsion from ephrin, cells were fixed after 2-3 hours of EphB4/ephrin-B2 expression and immunostained for activated Eph receptors (phospho-tyrosine/phospho-Eph; p-Y/p-Eph) and Ena/VASP proteins (VASP/Mena). The rabbit anti-p-Eph antibody recognises the phosphorylation of two conserved juxtamembrane tyrosine residues required for the full activation of Eph receptors (Binns et al., 2000; Marston et al., 2003; Shamah et al., 2001). At both 2 and 3 hours post-microinjection, small protrusions were visible at sites of ephrin-B2-induced Eph receptor activation (Fig. 5D-G). VASP and Mena accumulated at the EphB4/ephrin-B2 interface [relative to levels at the periphery of uninjected cells (Fig. 5B,C) and the rear of EphB4-expressing cells], at the edges of the small protrusions that extended towards the ephrin-B2-expressing cells, adjacent to activated Eph receptor staining (Fig. 5F,G). Some co-localisation between activated Eph receptors and VASP and Mena was also detected along these interfaces (Fig. 5F,G). VASP and Mena were not observed at stress fibres, since these cells were starved prior to microinjection and therefore lack these structures. At 3 hours, the internalisation of activated Eph receptors had begun, but VASP and Mena remained at the EphB4/ephrin-B2 interface, rather than with activated Eph receptor-containing vesicles (Fig. 5F,G). Thus, VASP and Mena are suitably positioned to modify the architecture of the actin-based protrusions that result from Eph receptor activation. We next investigated whether this localisation was required for the Eph receptor internalisation response of EphB4-expressing cells adjacent to cell surface-tethered ephrin-B2.

We wished to examine the more physiologically relevant repulsive response of EphB4-expressing cells to cell surface-tethered ephrin-B2 in the absence of Ena/VASP proteins. Since MV^D7 cells failed to form suitable monolayers to perform neighbouring cell injections and multiple siRNAs would be required to ensure knockdown of Ena/VASP proteins by RNA-mediated interference, we instead co-expressed EphB4 and FPPPP-mito to sequester Ena/VASP proteins away from the cell periphery (Bear et al., 2000). The FPPPP-mito construct consists of the four FPPPP motifs from ActA, attached to a mitochondrial localisation signal and an enhanced green fluorescent protein (EGFP) tag. FPPPP-mito localises to mitochondria, where it recruits Ena/VASP family members via their EVH1 (Ena/VASP homology1) domains (Bear et al., 2000; Niebuhr et al., 1997). The substitution of phenylalanine (F) in FPPPP motifs with alanine (A) disrupts binding of EVH1 domains (Niebuhr et al., 1997); hence, the expression of pcDNA3-APPPP-mito can be used as a negative control (Bear et al., 2000).

Confluent, starved Swiss 3T3 fibroblasts were co-injected with 100 μg/ml pCIneo-EphB4 and either 100 μg/ml pcDNA3-EGFP-FPPPP-mito or 100 μg/ml pcDNA3-EGFP-APPPP-mito. A line of adjacent cells was subsequently injected with 200 μg/ml pRK5-ephrin-B2 and 100 μg/ml biotin-dextran (injection marker). Cells were left to express for 3.5 hours prior to fixation, then stained using rabbit anti-p-Eph to detect activated Eph receptors. FPPPP/APPPP-mito and EphB4-co-expressing cells were identified through EGFP fluorescence (green) and increased levels of p-Eph staining (red), respectively, compared with non-injected neighbouring cells (Fig. 6A-D). EphB4- and FPPPP/APPPP-mito-co-expressing Swiss 3T3 fibroblasts that contacted ephrin-B2-injected cells were scored for internalisation of p-Eph and the formation of large, invasive, p-Eph-positive protrusions, both normal consequences of Eph receptor signalling in this assay (Marston et al., 2003). VASP co-localised with FPPPP-mito but not APPPP-mito, revealing that FPPPP-mito was specifically able to ectopically localise Ena/VASP proteins to the mitochondria (data not shown).

When APPPP-mito was co-expressed with EphB4, the typical response was the internalisation of p-Eph at sites of ephrin-B2 contact by EphB4-expressing cells (69±3% of cells, n=5; Fig. 6A,B). This response was comparable to the levels of internalisation observed on expression of EphB4 alone (64±4% of cells, n=4; data not shown). By contrast, the presence of FPPPP-mito dramatically reduced internalisation, with p-Eph instead accumulating at sites of EphB4-ephrin-B2 contact (28±4% of cells, n=5; Fig. 6C,D). This corresponded to a 58% reduction compared to cells co-expressing APPPP-mito and EphB4 (Fig. 6E).

Small protrusions were observed along the EphB4/ephrin-B2 boundary (Fig. 6A-D) and in some cells large, invasive protrusions were also expanded. However, in contrast to the large differences in internalisation, the percentages of cells with large, invasive protrusions at sites of EphB4-ephrin-B2 contact were comparable whether FPPPP-mito (32±3% of cells, n=5) or APPPP-mito (33±4% of cells, n=5) were co-expressed with EphB4 (Fig. 6E). In some EphB4-expressing cells that failed to internalise p-Eph, tubule-like structures extended intracellularly from the EphB4/ephrin-B2 interface (Fig. 6D).

Thus, Ena/VASP proteins are necessary for endocytosis of Eph receptors, since the internalisation of activated Eph receptors was significantly reduced when these proteins were sequestered away from the periphery at the mitochondria. This result is consistent with the localisation of VASP and Mena at the tips of lamellipodial protrusions formed downstream of EphB4 activation in Swiss 3T3 cells, and a requirement for Ena/VASP proteins in generating repulsion from stripes of substrate-bound ephrin. It is conceivable that VASP and Mena destabilise these protrusions to enable withdrawal of cell processes and the internalisation of ephrin-bound Eph receptors at sites of EphB4-ephrin-B2 contact and in this way contribute to cell-cell repulsion.

Trunk neural crest cells undergo segmented migrations in vivo that are at least partially controlled by Eph receptor signalling (Krull et al., 1997). Consistent with this in vivo behaviour, we observe that rat trunk neural crest cells avoid stripes of substrate-bound ephrin-B2-Fc, a repulsive response characterised by the rearward collapse of lamellipodia and membrane ruffling. Furthermore, soluble, pre-clustered ephrin-B2-Fc reduces the persistence of neural crest cell lamellipodia, with cycles of protrusion and retraction occurring with greater frequency in the presence of ephrin-B2-Fc. We hypothesised that Ena/VASP proteins might contribute to the destabilisation of lamellipodia and facilitate repulsion at sites of Eph receptor activation. We show that VASP^–/– MEFs display defective repulsion from stripes of both ephrin-A5-Fc and ephrin-B2-Fc compared to wild-type MEFs, a defect further enhanced in Ena/VASP-deficient MV^D7 cells. Additionally, we show the sequestration of Ena/VASP proteins away from the periphery to mitochondria inhibits internalisation of activated Eph receptors at sites of EphB4-ephrin-B2 contact. Taken together, our data show that Ena/VASP proteins are involved in ephrin-induced Eph receptor signalling events.

Given the localisation of VASP and Mena at the edges of ruffling protrusions and the inhibition of activated Eph receptor internalisation at EphB4/ephrin-B2 interfaces in the presence of FPPPP-mito, it is tempting to speculate that Ena/VASP proteins might facilitate the generation of long, unbranched actin filaments to destabilise membrane protrusions, similar to Ena/VASP-dependent negative regulation of fibroblast motility (Bear et al., 2000; Bear et al., 2002). However, involvement in the regulation of actin filaments at adhesion complexes or during the internalisation process itself could also underlie these observations (Hoffman et al., 2006; Han et al., 2002; Coppolino et al., 2001).

During ephrin-induced RGC growth cone collapse, Rho and ROCK-dependent contractility facilitates axonal retraction (Meima et al., 1997; Wahl et al., 2000). Inhibition of Rho or ROCK using C3 or Y27632, respectively, fails to block ephrin-A-induced loss of growth cone lamellipodia (Harbott and Nobes, 2005); therefore, Ena/VASP-mediated lamellipodia destabilisation may function in parallel to contribute to this loss of lamellipodia. Intriguingly, interfering with Abl activity, a protein known to regulate and interact with Ena/VASP proteins, prevents this loss of lamellipodia in RGC neurons in response to ephrin-As (Harbott and Nobes, 2005). Although the GEF Ephexin may activate Rho downstream of EphA receptor activation (Shamah et al., 2001; Sahin et al., 2005), the activation of this GTPase is less closely associated with EphB signals (for a review, see Noren and Pasquale, 2004). For example, ephrin-B ligands fail to elicit the same levels of RGC axonal retraction as ephrin-As (Meima et al., 1997). Therefore, destabilisation of lamellipodia through Ena/VASP activity may be particularly important downstream of EphB receptors.

It is becoming increasingly apparent that Rac-induced protrusion is required for repulsive events: Rac and dynamin activity are necessary for cell-cell separation at sites of EphB4-ephrin-B2 contact (Marston et al., 2003) and Rac is required for ephrin-induced growth cone collapse (Jurney et al., 2002; Cowan et al., 2005). Furthermore, Rac and Cdc42 activity are associated with actin polymerisation-dependent dendritic spine formation, downstream of activated EphB receptors (Irie and Yamaguchi, 2002; Penzes et al., 2003). Rac, dynamin and Ena/VASP proteins are each required during Fcγ receptor-mediated phagocytosis (Coppolino et al., 2001; Gold et al., 1999; Marston et al., 2003) and internalisation of activated Eph receptors at sites of EphB4-ephrin-B2 contact (Marston et al., 2003). It seems unlikely that Ena/VASP proteins instigate the protrusions made by EphB4-expressing cells, as there was no difference in the proportion of cells with large, invasive protrusions on co-expression of FPPPP-mito or APPPP-mito.

Since Ena/VASP proteins can bind the barbed ends of actin filaments via their EVH2 domains (Bachmann et al., 1999; Bear et al., 2002), de novo actin polymerisation may contribute to their accumulation at sites of EphB4-ephrin-B2 contact. However, numerous Ena/VASP ligands have been identified that bind Ena/VASP proteins via EVH1 domain-FPPPP-containing peptide interactions (Krause et al., 2004; Lafuente et al., 2004; Moeller et al., 2004; Niebuhr et al., 1997). Notably, during random fibroblast motility, the regulation of Ena/VASP localisation is partially controlled through an Ena/VASP-binding scaffold protein, Lamellipodin (Krause et al., 2004). Moreover, the adaptor protein Fyb/SLAP operates downstream of T cell and Fcγ receptors to recruit Ena/VASP proteins (Coppolino et al., 2001; Krause et al., 2000), hence a similar mechanism could be employed downstream of Eph receptors.

Thus, these studies reveal a novel and important role for the Ena/VASP family downstream of Eph receptors. Although the exact links between these families of proteins remain to be characterised, repulsion from ephrin and the internalisation of activated Eph receptors at sites of ephrin contact are clearly perturbed in the absence of Ena/VASP proteins. This work therefore reveals a potential Ena/VASP-dependent mechanism that contrasts with, and may function in parallel to, Rho- and ROCK-induced contractility during repulsive responses to ephrin ligands, downstream of Eph receptors.

Trunk neural tubes, from which neural crest cells were about to emerge, were dissected from E11 Sprague-Dawley rats (Erickson et al., 1989). Neural tube explants were transferred to coverslips pre-coated with 50 μg/ml fibronectin or patterned with substrate-bound ephrin stripes (see below). Both fibronectin-coated and striped coverslips were pre-coated with 1 mg/ml poly-L-lysine. Explants were cultured for 18-24 hours at 37°C, 5% CO₂ in either fully defined medium, described by Woodhoo et al. (Woodhoo et al., 2004), but with 10.1 ng/ml triiodothyronine and 400 μg/ml thyroxine, or in B27-supplemented medium, which contained DMEM and Hams F12 medium in a 1:1 ratio, supplemented with 1× B27 growth factor supplement (Gibco), 100 IU/ml penicillin and 100 μg/ml streptomycin (Gibco), 100 ng/ml insulin-like growth factor (IGF1; R&D), 10 ng/ml NDFβ3 (Neu differentiation factor β3, also known as Heregulin-β1; R&D), and 3 ng/ml basic fibroblast growth factor (bFGF; R&D). Neural crest cells were stimulated with 2 μg/ml pre-clustered ephrin-B2-Fc (R&D); ephrin-B2-Fc was pre-clustered (via its human Fc domain) at 37°C, 10% CO₂ in a 1:10 ratio (w/w) for 30 minutes in pre-conditioned neural crest cell medium, using goat anti-human IgG antibodies (Jackson).

Serum-starved, confluent Swiss 3T3 fibroblasts for microinjection of plasmids were prepared as in Marston et al. (Marston et al., 2003). Immortalised wild-type and VASP^–/– MEFs were derived from VASP^–/– mice (Hauser et al., 1999) and cultured in DMEM supplemented with 15% FCS and penicillin/streptomycin at 10% CO₂, 37°C. Ena/VASP-deficient cells (MV^D7 cells), a clonal cell line derived from E9.5 MEFs obtained from Mena^–/–VASP^–/– mice, which also lacks detectible Evl expression, were cultured as described by Bear et al. (Bear et al., 2000). To detect Eph receptor expression, wild-type MEF, VASP^–/– MEFs and MV^D7 cells were treated with 1 μg/ml pre-clustered ephrin-A5-Fc (R&D) or ephrin-B2-Fc for 10 minutes. Ephrin-bound Eph receptors were then stained via the goat anti-human IgG antibodies used to pre-cluster, using rabbit anti-human-FITC (Jackson).

Two variations upon published protocols were used for the stripe assay (Wang and Anderson, 1997; Krull et al., 1997; Eickholt et al., 1999). Neural crest cell stripes were patterned onto coverslips pre-coated with 1 mg/ml poly-L-lysine. 50 μl of PBS containing 200 μg/ml fibronectin, 100 μg/ml goat anti-human IgG and 2 μM Texas Red-dextran was injected into silicon matrices and incubated for 1 hour at 37°C to form the first set of stripes. Next, 50 μl of 20 μg/ml ephrin-B2-Fc in PBS were injected for 1 hour at 37°C; ephrin-B2-Fc is thus captured by pre-adsorbed goat anti-human IgG. 100 μl of 2 mg/ml BSA in PBS was used to prevent further adsorption of proteins for 30 minutes at 37°C. Incorporation of Texas-Red-dextran enables visualisation of the stripe containing ephrin. After removal of the coverslip from the matrix, 100 μl of 50 μg/ml fibronectin solution was applied to the entire coverslip for 1 hour at 37°C. Coverslips were finally blocked in 100 μl of 2 mg/ml BSA in PBS for 30 minutes at 37°C and left in 500 μl neural crest cell medium for 30 minutes before plating of neural tube explants. PBS was injected or applied to wash coverslips between each step. This protocol was adapted from Krull et al. (Krull et al., 1997) and Wang and Anderson (Wang and Anderson, 1997). For negative control stripes, the ephrin-B2-Fc injection was replaced with an injection of PBS.

MEFs and MV^D7 cells were plated to stripes patterned onto acid-washed borosilicate glass coverslips [protocol adapted from Eickholt et al. (Eickholt et al., 1999)]. 75 μl of 10 μg/ml pre-clustered ephrin-Fc (90 nM) was injected for 45 minutes at 37°C, to form the first set of stripes; ephrin-Fc dimers were pre-clustered in a 1:2 molar ratio with rabbit anti-human IgG antibodies for 30 minutes at 37°C. 100 μl of 2 mg/ml BSA was then used to block this first set of stripes for 15 minutes at 37°C. 100 μl pre-clustered human Fc solution (18 nM) was then applied to the whole coverslip for 45 minutes at 37°C to form the second set of stripes; human Fc dimers were pre-clustered in a 1:2 molar ratio, using goat anti-human IgG antibodies. Lastly, 100 μl of 2 mg/ml BSA in PBS was applied for 15 minutes at 37°C. PBS washes were again performed between each step and coverslips were left for 30 minutes in the appropriate media at 37°C, 10% CO₂ prior to plating of cells. In order to make negative control stripes, 75 μl of 90 nM pre-clustered Fc was injected instead of pre-clustered ephrin-Fc to form the first set of stripes. Since anti-human IgG antibodies raised in different species were used to pre-cluster the recombinant proteins in the first and second sets of stripes, Cy3-labelled anti-rabbit IgG (Jackson) was used to specifically recognise the first set of stripes. A cell is considered to avoid a particular stripe if greater than 50% of its spread area lies off of that stripe. The percentage of cells avoiding the first set of stripes (stained red) was quantified for ephrin and negative control (human Fc) stripes, with 100-150 cells counted per coverslip. The difference between these two values revealed the repulsive response specific to ephrin, and was expressed as a percentage of the wild-type repulsive response. Numerical data is presented as mean ± s.e.m.

Cell fixation and staining was performed as described by Marston et al. (Marston et al., 2003), except that cells expressing EGFP-tagged proteins were not treated with sodium borohydride to preserve EGFP fluorescence. Actin filaments were stained using 0.1 μg/ml FITC-labelled phalloidin (Molecular Probes); biotin-dextran injection marker was detected using 5 μg/ml Alexa Fluor 350-streptavidin (Molecular Probes). The following primary antibodies were used in immunostaining protocols: goat-anti-EphB4 (R&D; 1:100), mouse anti-Mena (1:10), rabbit anti-phospho-Eph (Marston et al., 2003) (1:500), mouse anti-phospho-tyrosine (Sigma; clone PT66; 1:100) and rabbit anti-VASP (Alexis Biochemicals; 1:700). Fluorescently labelled secondary antibodies (Jackson; 1:200) were used to detect primary antibodies. All dilutions were made in PBS.

Phase timelapse movies of neural crest cells before and after stimulation with pre-clustered ephrin-B2-Fc were taken using Openlab software (Improvision); kymographs were generated using the `volume slicing' tool in Openlab. Neural crest cells were selected for this analysis provided that they remained at the edge of the neural crest cell sheet and did not divide or escape the field of view during the course of the timelapse movie.

Cell lines were washed twice in Tris-buffered saline (TBS) and lysed in 500 μl immunoprecipitation buffer, as described by Harbott and Nobes (Harbott and Nobes, 2005). Samples were immunoprecipitated using mouse anti-EphA4 (Zymed) and analysed by SDS-PAGE and western blotting for phospho-tyrosine (Sigma; clone PT66; 1:5000) and EphA4 (Zymed; 1:5000), to determine the levels of EphA4 activation and loading, respectively. These primary antibodies were then detected using HRP-conjugated secondary antibodies (Jackson).

A line of quiescent, confluent, starved Swiss 3T3 fibroblasts was microinjected with 100 μg/ml pCIneo-EphB4 at 8.5% CO₂, 37°C. The cells immediately adjacent were then injected with 200 μg/ml pRK5-ephrin-B2 and 100 μg/ml biotin-dextran injection marker. Neighbouring cells subsequently express Eph receptor or ephrin, resulting in receptor-ligand contact, binding and subsequent activation and downstream signalling (Marston et al., 2003). Fixation at various timepoints and subsequent staining enabled the nature of these responses to be determined. pRK5-ephrin-B2-injected cells were detected via staining for biotin-dextran; EphB4 expression was detected via staining for activated Eph receptors. Neither FPPPP-mito, nor APPPP-mito significantly suppressed EphB4 expression. Coverslips were examined `blind' to eliminate bias and cells co-expressing FPPPP/APPPP-mito and EphB4 (showing EGFP and p-Eph staining), adjacent to ephrin-B2-expressing cells (showing biotin-dextran staining), were scored for internalisation of p-Eph and the formation of large, invasive, p-Eph-positive protrusions.

The following constructs were received as gifts: pCIneo-EphB4 (from D. Wilkinson, NIMR, London), ephrin-B2 cDNA (from R. Adams, CRUK, London). Ephrin-B2 and FPPPP/APPPP-mito were sub-cloned into pRK5 and pcDNA3 vectors, respectively. The FPPPP-mito and APPPP-mito constructs have been described previously (Bear et al., 2000).

Phase-contrast timelapse movies were taken on an inverted Zeiss microscope equipped with a heated stage and controlled CO₂ environment (37°C, 8.5% CO₂), using Openlab software (Improvision); images were taken every 20 seconds. Immunofluorescent images of fixed cells were taken either using an upright Zeiss microscope or on a Leica confocal microscope.

We are grateful to past and present members of the Nobes and Martin Labs for advice and discussion, particularly Paul Martin. We thank the MRC for providing an Infrastructure Award to establish the School of Medical Sciences Cell Imaging Facility and Mark Jepson and Alan Leard for their assistance. This work was funded by the Medical Research Council; C.D.N. is an MRC Senior Research Fellow. I.R.E. is an MRC-funded student.