Par3 controls neural crest migration by promoting microtubule catastrophe during contact inhibition of locomotion

Cell migration underlies numerous events during development and in adult life, controlling such diverse processes as morphogenesis, angiogenesis, axon pathfinding, wound healing, immune surveillance and metastasis. During migration, cells must determine the direction in which to migrate and reorient their cytoskeleton so as to establish protrusions at the leading edge and retraction at the trailing edge. This cell polarisation can be induced by cell-cell contact during contact inhibition of locomotion (CIL), the phenomenon by which a cell, upon contact with another cell, collapses its protrusions, reorientates and migrates in another direction (Abercrombie and Heaysman, 1953) (for reviews, see Batson et al., 2013; Stramer et al., 2013). Indeed, it has been shown that CIL plays a major role in controlling the directional migration of neural crest (NC) cells (Carmona-Fontaine et al., 2008; Matthews et al., 2008; Theveneau et al., 2010; Theveneau et al., 2013), a mesenchymal cell population with a migratory behaviour that has been likened to cancer metastasis (Theveneau and Mayor, 2011).

Polarisation by CIL is a complex process that involves several steps: cell contact, collapse of cell protrusions, formation of new protrusions away from the contact, cell separation and migration (Mayor and Carmona-Fontaine, 2010; Stramer et al., 2010). Some molecular regulators of CIL have been identified in recent years, such as small GTPases, planar cell polarity (PCP) signalling, N-cadherin and Eph-ephrins (Carmona-Fontaine et al., 2008; Kadir et al., 2011; Matthews et al., 2008; Theveneau et al., 2010), but how these molecules can control the cytoskeletal rearrangements required for a CIL response remains unknown.

Cell-cell contacts play a major role in the collective directional migration of NC in Xenopus, zebrafish and chick embryos (Carmona-Fontaine et al., 2008; Matthews et al., 2008), presumably because cell contact is a key step in CIL. Recent evidence indicates that, during cell contact between NC cells, a transient cell adhesion complex is formed (Theveneau et al., 2010; Theveneau et al., 2013). The formation of a cell-cell adhesion complex is unexpected, as NC cells are mesenchymal cells with reduced cell-cell adhesion (Theveneau and Mayor, 2012). In epithelial cells, the adhesion complex identifies and reinforces the boundary between apical and basolateral membranes to maintain cell polarity and tissue structure. It also serves as a signalling focal point, orchestrating various signalling events via the recruitment of regulatory factors such as the cell polarity protein Par3 (also known as Pard3) (Gao et al., 2002; Izumi et al., 1998; Kuchinke et al., 1998; Suzuki et al., 2001). However, the composition of a cell adhesion complex in mesenchymal cells, such as the NC, is poorly characterised and, more importantly, its downstream effectors remain elusive.

Par3 promotes the maturation of the cell adhesion complex and has been linked to the regulation of microtubule dynamics and Rac1 activity through interaction with the Rac-GEF Tiam1 (Chen et al., 2013; Chen and Macara, 2005; Mack et al., 2012; Mertens et al., 2005; Mishima et al., 2002; Xue et al., 2013). In epithelial cells, Par3 associates with cell-cell adhesion complexes and appears to be important for establishing apicobasal polarity; however, whether Par3 controls PCP during CIL remains to be investigated. Par3 also appears to have roles in controlling microtubule dynamics during periods of cell motility (Du et al., 2010; Nishimura et al., 2005; Pegtel et al., 2007; Schmoranzer et al., 2009; Shin et al., 2007). For example, in wound healing assays, Par3 has been found to play an important role at the leading edge of migratory cells, controlling centrosome polarisation, microtubule orientation and microtubule dynamics (Du et al., 2010; Schmoranzer et al., 2009; Shin et al., 2007). Thus, as Par3 controls microtubule dynamics, cell migration and polarity and is localised at the cell-cell junction of epithelial cells, it represents a possible candidate to participate in CIL of mesenchymal cells.

To investigate a possible role of Par3 in CIL during collective cell migration, we turned to NC cells. We found that Par3 is required for NC migration by controlling the CIL response of these cells. Par3 is localised at cell-cell contacts between NC cells, where it interacts with the Rac-GEF Trio. We show that Par3 exerts its effect by promoting microtubule catastrophe at the cell contact and that Par3 inhibition can be rescued by concurrent inhibition of Trio. Thus, we propose that CIL requires of local destabilisation of microtubules at the cell-cell contacts, which is controlled in a Par3/Trio-dependent manner.

To investigate the role of Par3 in NC migration, we performed a loss-of-function experiment using an antisense morpholino (MO) against Par3 (Par3MO), which efficiently decreases the level of Par3 protein in Xenopus embryos (Fig. 1A). Injection of Par3MO did not affect NC induction, as analysed by in situ hybridisation against Snail2, FoxD3, Sox9 or Sox10, all of which are known markers for pre-migratory NC (Fig. 1B,C). However, Par3MO injection had a dramatic effect on NC migration, as analysed by in situ hybridisation against Twist, which marks migrating NC (Fig. 1D,E,G). Importantly, this effect can be rescued by co-injection of zebrafish par3 mRNA, which does not contain the sequence targeted by the MO against Xenopus Par3 (Fig. 1F-H), demonstrating specificity for the Par3MO. The requirement of Par3 is cell-autonomous, as grafts of Par3-depleted cells into normal host show a clear defect in NC migration (supplementary material Fig. S1).

To further assess the necessity for Par3 in NC migration, NC explants were cultured on fibronectin and observed by time-lapse imaging. Control explants tended to disperse after a few hours of cell culture (Fig. 1I), as previously described (Alfandari et al., 2003), whereas explants injected with Par3MO failed to disperse (Fig. 1J; supplementary material Movie 1). We quantified cell dispersion by measuring the area of the triangles formed between the nuclei using Delaunay triangulation as previously described (Carmona-Fontaine et al., 2011). A dramatic increase in cell dispersion starts at ∼6 hours in the control explants, but this is very much reduced in Par3MO-injected explants (Fig. 1K). This reduced dispersion is not due to an effect on cell motility, as control and Par3MO-injected cells exhibited similar speeds and persistence during the migration of individual cells (Fig. 1L,M). These results demonstrate that Par3 is not required for cell motility but is required for NC dispersion.

Our results show an effect of Par3MO on Xenopus NC migration in vivo and in vitro. Next, we turned to zebrafish embryos to perform live imaging of NC migration, as it is easier to obtain high-resolution images in zebrafish than in Xenopus for quantitative analysis in vivo. First, we confirmed that a similar NC phenotype was achieved using a previously characterised zebrafish Par3MO (Wei et al., 2004). Indeed, injection of Par3MO leads to a clear inhibition of migration in trunk (Fig. 2B-D) and cephalic (Fig. 2B,E,F) NC cells, without affecting NC induction (Fig. 2A). To analyse cell behaviour in vivo, a new transgenic zebrafish line that expresses membrane-GFP and nuclear-RFP under the NC-specific Sox10 promoter was developed. Similar to the in vitro observation using Xenopus NC, Par3MO reduced NC dispersion in vivo (Fig. 2G-I; supplementary material Movie 2).

One way in which Par3 could affect dispersion is through controlling cell-cell adhesion, as has been shown in other cell types (Chen and Macara, 2005; Mertens et al., 2005; Mishima et al., 2002). We analysed the effect of Par3MO on the level and localisation of various cell adhesion proteins. No effect was found on the localisation of β-catenin or p120-catenin (also known as δ1 catenin) at the cell-cell junction in Par3MO-injected cells compared with control Xenopus NC cells (Fig. 3A-F), nor in the level or localisation of N-cadherin between control or Par3MO-injected cells in zebrafish embryos (Fig. 3G-L). Furthermore, we performed a cell-sorting assay to evaluate whether Par3MO influenced cell-cell adhesion (Fig. 3M). When control and N-cadherin morphant cells are mixed they sort out, indicating differential cell adhesion (Fig. 3P) (Friedlander et al., 1989). However, when control and Par3MO-injected cells are combined, a mixed cell population results with no difference between control and Par3 morphant cells (Fig. 3N,O). Together, our results did not support a role for Par3 in regulating cell adhesion between NC cells, and an alternative mechanism for the effect of Par3 inhibition on NC migration and dispersion needed to be explored.

An alternative way in which Par3 could affect NC dispersion is through controlling CIL, as CIL promotes dispersion by repolarising the cells away from each other upon cell contact (Mayor and Carmona-Fontaine, 2010). We performed three different assays that have been used previously to analyse CIL (Abercrombie and Heaysman, 1953; Carmona-Fontaine et al., 2008; Theveneau et al., 2010). First, two NC explants confronted with each other do not overlap if they exhibit CIL (Fig. 4A). Par3MO-injected explants widely overlapped, as did pairs comprising one control and one Par3MO-injected explant, indicating a failure in CIL (Fig. 4A-D). Second, CIL dictates that two adjacent cells do not make protrusions on top of one another (Fig. 4E) (Abercrombie and Heaysman, 1953); however, Par3MO-injected NC cells extended protrusions on top of their neighbours (Fig. 4F,G; supplementary material Movie 3). Third, the hallmark of CIL is that collision between two single cells will cause them to change velocity if CIL occurs (Fig. 4H) (Abercrombie and Heaysman, 1953). However, colliding Par3MO-injected cells showed a significantly reduced change in velocity upon collision compared with control cells (Fig. 4I,J; supplementary material Movie 4). Together, these results indicate that Par3 is required for CIL between NC cells in Xenopus embryos. Next, we used the sox10:mGFPnRFP zebrafish line to analyse CIL in vivo. Collisions between NC cells were analysed after time-lapse imaging of migrating NC. ControlMO-injected NC cells exhibited the typical CIL response by changing the direction of migration after collision (Fig. 5A-C; supplementary material Movie 5). By contrast, Par3MO-injected NC cells failed to change the direction of migration after colliding with another NC cell (Fig. 5D-F; supplementary material Movie 6). These experiments confirm that Par3 is required for CIL between NC cells in vitro and in vivo.

To determine the precise role of Par3 during CIL, it was important to assess Par3 localisation. To analyse the dynamics of Par3 localisation, embryos were injected with Par3GFP mRNA and time-lapse imaging was performed on cultured NC. Par3GFP localised to cell junctions both in clusters of NC cells that were already in contact (Fig. 6A) and during collisions between individual cells (Fig. 6B; supplementary material Movie 7). To confirm that endogenous Par3 was expressed in Xenopus NC and localised to cell-cell contacts, membraneRFP (mbRFP)-labelled NC cells were transplanted into wild-type host embryos to identify the migrating NC and immunostaining against Par3 was performed. Our results showed that endogenous Par3 was present in the NC and colocalised with mbRFP at cell-cell contacts (Fig. 6C-F).

How does Par3 regulate CIL in NC cells? Par3 has been shown to control microtubule array polarisation in other cell types by affecting microtubule dynamics or centrosome positioning (Pegtel et al., 2007; Schmoranzer et al., 2009). Furthermore, microtubules are involved in CIL (Kadir et al., 2011; Stramer et al., 2010). Thus, we next investigated a possible interaction between Par3 and microtubules as a means of regulating CIL. However, Par3 inhibition had no effect on the entire microtubule array (Fig. 6G,H), nor on a stable microtubule subpopulation containing acetylated tubulin (Fig. 6I,J).

To analyse microtubule dynamics we labelled NC cells with EB3GFP and performing time-lapse imaging, which allows high-resolution visualisation of growing microtubules (supplementary material Movie 8). We tracked EB3GFP comets and compared the number of microtubules undergoing catastrophe (disappearing comets) with the total number of microtubules (comets) at the cell contact and at the free edge. In control cells, a greater percentage of microtubules experienced catastrophe at the contact (Fig. 7A) compared with the free edge (Fig. 7C). However, inhibiting Par3 abolished the increase in microtubule catastrophe at the contact (Fig. 7B,C; supplementary material Movie 9). This suggests that Par3 is required at the cell-cell contacts to promote microtubule catastrophe.

A caveat of using EB3GFP-labelled cells is that collapsing and pausing microtubules are not easy to distinguish or quantify manually. Therefore, we used the automatic tracking software plusTipTracker (Applegate et al., 2011) to perform an exhaustive analysis of microtubule dynamics. Importantly, it calculates which tracks belong to the same microtubule, which may be cycling between phases of growth, pause and catastrophe (Applegate et al., 2011). Movies of EB3GFP-labelled cells were analysed at cell-cell contacts during collisions of Xenopus NC cells. Microtubule tracks were classified as being slow or fast and as short-lived or long-lived based on the mean growth speed (26 μm/minute) and mean lifetime (20 seconds) of the data set. Par3 inhibition reduced the proportion of slow, short-lived microtubules in favour of slow, long-lived microtubules (Fig. 7D-F). This suggests that slow microtubules live longer in Par3MO cells, which implies that they must be undergoing catastrophe less frequently. Indeed, statistical analysis revealed that Par3MO injection decreased the rate of microtubule catastrophe (Fig. 7G), whereas no effect of Par3 inhibition could be found on other parameters routinely used for measuring microtubule dynamics, including growth rate, depolymerisation rate, rescue rate or pause duration (Fig. 7H-K). Overall, our analysis indicates that Par3 promotes microtubule catastrophe at the cell contact during CIL.

Par3 promotion of microtubule catastrophe at the cell-cell contact could control CIL by initiating a change in cell polarity. To test whether the effect of Par3 on CIL was dependent on microtubule catastrophe, we attempted to rescue the Par3 morphant phenotype using a mild concentration of nocodazole (13 nM), which has a partial effect on microtubule depolymerisation (supplementary material Fig. S2) and has no effect on NC dispersion (supplementary material Fig. S3). As described above, Par3MO blocks NC dispersion; however, treatment of Par3MO-injected cells with a mild concentration of nocodazole was able to rescue cell dispersion (Fig. 8A, compare black and blue curves; supplementary material Movie 10), in spite of a small reduction in the motility of Par3MO-injected cells (Fig. 8B,C, blue bars). On control cells, nocodazole inhibited CIL (not shown). Importantly, the explant confrontation assay showed that the increased overlap area induced by Par3MO (Fig. 8D,E,H) was reduced by nocodazole treatment (Fig. 8F,H, blue bar), indicating that CIL behaviour was restored. Furthermore, analysis of CIL in single cell collisions showed that the loss of CIL in Par3MO-injected cells (Fig. 8I,J) was rescued by nocodazole treatment (Fig. 8K; supplementary material Movie 11). This supports the notion that Par3 controls CIL by promoting microtubule collapse at the cell-cell contact.

It is known that Rac1 can influence cell polarity by stabilising microtubules (Pegtel et al., 2007; Wittmann et al., 2003). Thus, inhibition of Rac1 activity is a potential mechanism through which Par3 could affect microtubules. Indeed, a low concentration of the Rac1 inhibitor NSC23766 was able to rescue the inhibition of cell dispersion induced by Par3MO in NC explants (Fig. 8B,C, yellow curve) without affecting the speed or persistence of migration (Fig. 8B,C, yellow bars; supplementary material Movie 10). Importantly, NSC23766 was also able to rescue CIL between Par3MO-injected explants (Fig. 8G,H, yellow bar) and between Par3MO-injected colliding cells (Fig. 8L; supplementary material Movie 11).

As Rac1 inhibition can rescue CIL in Par3MO-injected NC cells, Par3 might inhibit Rac1 during CIL. In other cell types, Par3 interacts with Rac-GEFs to influence Rac1 activity (Pegtel et al., 2007). It has been shown that the Rac-GEF Trio is an important regulator of NC migration (Kashef et al., 2009) and is therefore a possible candidate to be regulated by Par3. Indeed, we found that Par3 and Trio co-immunoprecipitate (Fig. 9A) and colocalise at NC cell contacts (Fig. 9B-E). We next investigated whether inhibiting Trio could rescue Par3 morphant phenotypes. We developed an antisense MO against Trio (TrioMO) that efficiently reduced GFP fluorescence when co-injected with Trio-GFP (supplementary material Fig. S4); its NC migration phenotype can be rescued by human TRIO, which does not hybridise with TrioMO (supplementary material Fig. S5), demonstrating the specificity of TrioMO. We co-injected TrioMO with Par3MO to perform a double loss-of-function experiment. Trio inhibition rescued the Par3 morphant phenotype by promoting explant dispersion (Fig. 9F; supplementary material Fig. S6) without affecting the speed or persistence of migration (Fig. 9G,H; supplementary material Movie 12). Co-injection of TrioMO also rescued Par3MO-induced CIL deficiency, as analysed by the change in velocity after single cell collisions (Fig. 9I-K; supplementary material Movie 13). Finally, we analysed whether inhibition of NC migration by Par3MO in vivo could be rescued by inhibiting Trio. Co-injection of TrioMO with Par3MO efficiently restored NC migration in vivo (Fig. 9L-P). TrioMO rescue of the Par3 morphant phenotype demonstrates that Par3 inhibits Trio in NC cells.

As Rac1 has been shown to stabilise microtubules (Wittmann et al., 2003), Trio activation of Rac1 could promote microtubule stability. Thus, Par3 could promote microtubule catastrophe by inhibiting the Rac-GEF Trio. We analysed microtubule dynamics by performing time-lapse imaging of EB3GFP-labelled NC cells. Trio inhibition increased microtubule catastrophe in Par3MO-injected cells at the cell-cell contact to a level comparable to that of control cells (Fig. 10A-D). These results were supported by plusTipTracker analysis, which showed that inhibition of both Trio and Par3 increased the microtubule catastrophe rate to control levels (Fig. 10E). Finally, we measured Rac1 activity at the cell contact by fluorescence resonance energy transfer (FRET) analysis (Matthews et al., 2008). As shown previously (Theveneau et al., 2010), Rac1 activity is reduced at the cell contact in NC (Fig. 10F, control). However, inhibition of Par3 leads to a significant increase in Rac1 activity at the cell-cell interface, which is lost upon co-inhibition of Par3 and Trio (Fig. 10F,G). This supports the notion that Par3 is required at contacts between NC cells to inhibit Trio-dependent activation of Rac1, thus inhibiting Rac1-dependent microtubule stabilisation.

Altogether, our results demonstrate that Par3 controls cell protrusion collapse and reversal of cell polarity during CIL by promoting microtubule catastrophe via inhibition of the Rac-GEF Trio.

Effective NC migration is dependent on CIL. Here, we show that Par3 is required for NC migration, as it controls CIL both in vivo and in vitro. We demonstrate that Par3 controls CIL by inhibiting the Rac-GEF Trio to prevent Trio-mediated activation of Rac1 at cell-cell contacts. We propose the following model for the migration of NC cells (Fig. 11). In individually migrating cells, microtubule stability in the lamellipodia is promoted by Rac1, which is activated by Trio (Fig. 11B). Upon collision with another NC cell, Par3 is recruited to the cell-cell contact, sequestering Trio and preventing it from activating Rac1. This local inhibition of Rac1 activity triggers microtubule catastrophe at the cell-cell contact, resulting in a reversal of cell polarity (Fig. 11C,E), the formation of a new protrusion and a change in the direction of migration, which are typical of contact inhibition behaviour (Fig. 11D). In the absence of Par3, Trio and Rac1 remain active after cell contact, the polymerisation of microtubule continues and cell protrusions do not collapse upon cell contact (Fig. 11F).

Localised microtubule catastrophe leads to an asymmetry in stability across the microtubule array. Work by others has shown that this can result in a change in cellular polarity and migration away from the cell contact as part of a CIL response (Kadir et al., 2011; Stramer et al., 2010). Our results suggest Par3 as a likely candidate to link the cell-cell contact signal to the microtubule catastrophe response in this process.

The Wnt-PCP pathway and N-cadherin are both necessary to reduce Rac1 activity and enhance RhoA activity effectively at the cell-cell contact, as is required to promote CIL (Carmona-Fontaine et al., 2008; Theveneau et al., 2010). Here, we show that Par3 localises to contacts between NC cells, similar to previous findings for epithelial cells (Gao et al., 2002; Izumi et al., 1998; Kuchinke et al., 1998; Suzuki et al., 2001) and wounded fibroblast sheets (Schmoranzer et al., 2009). In contrast to both of these cell types, we find that Par3 neither controls centrosome position (not shown) nor modulates cell-cell adhesion in NC cells. Nonetheless, our results suggest that cell adhesion molecules, such as N-cadherin or cadherin 11, are likely to be regulators of Par3 localisation or activity at the cell contact. This hypothesis is further supported by the fact that N-cadherin-dependent contacts have been shown to affect microtubule organisation and cell polarity in other cell types (Camand et al., 2012; Chausovsky et al., 2000; Dupin et al., 2009). Our analysis of cell adhesion molecules and cell sorting does not support the notion that Par3 regulates cell adhesion in NC cells, as has been shown for other cell types (Xue et al., 2013); however, we cannot rule out the possibility that subtle changes in cell adhesion, undetected in our assays, are modulated by Par3 in NC cells. Nonetheless, our data clearly demonstrate a role for Par3 in CIL in NC cells, which could explain some of the phenotypes previously attributed to cell-cell adhesion in other cell types.

Here, we report a novel interaction between Par3 and the Rac-GEF Trio, which could be added to the previously characterised interaction between Par3 and the Rac-GEF Tiam1 at the leading edge and cell-cell contacts (Pegtel et al., 2007). In NC cells, the Rac-GEF Trio has been shown to interact with cadherin 11, which is localised to cell protrusions as well as cell-cell contacts (Kashef et al., 2009). We show that Trio inhibition compensates for a lack of Par3, suggesting that Par3 inhibits Rac-GEF Trio activity at cell-cell contacts during NC migration. RhoA is active at contacts between NC cells (Carmona-Fontaine et al., 2008) and ROCK phosphorylation of Par3 has been shown to disrupt formation of the Par3-aPKC-Par6 complex, through which Par3 often acts (Nakayama et al., 2008). However, ROCK phosphorylation of Par3 does not affect Par3-Tiam2 interactions (Nakayama et al., 2008), suggesting that Par3 is likely to regulate Rac-GEFs independently of its involvement in the Par complex, as has been shown for the regulation of Rac by Par3 in embryonic kidney cells (Mack et al., 2012).

Microtubules are important in the determination and maintenance of cell polarity. Rapid microtubule dynamics allows the microtubule array to respond to signals quickly and effectively. Par3 has been shown to stabilise microtubules and promote directional migration at the leading edge of single migrating cells (Pegtel et al., 2007) or during the establishment of neuronal polarity (Chen et al., 2013). However, again illustrating the context-dependent activity of Par3, our data show that Par3 inhibits Rac1 activation at the cell-cell contact to promote microtubule catastrophe. An asymmetry in stability across the microtubule array leads to a change in cell polarity and thus to a change in the direction of migration. In this model, Par3 simply induces a change in direction rather than dictating the direction in which the new protrusion should be extended. As such, it is interesting to note that microtubule depolymerisation, Rac1 inhibition and Trio inhibition - all of which presumably affect the entire cell - are each able to rescue the effect of Par3 inhibition. It is likely that other mechanisms also influence cell-cell contact-induced polarity and that Par3 could have a direct role in microtubule destabilisation. However, our results provide a molecular explanation for the hypothesis put forward by Kadir and colleagues (Kadir et al., 2011) that reaching a threshold of microtubule dynamics at the cell-cell contact is sufficient to induce CIL.

In this study we report a novel interaction between Par3 and the Rac-GEF Trio. We show that Par3 inhibition of Trio at cell-cell contacts promotes microtubule collapse, is required for CIL between NC cells, and is thus essential for the collective and directional migration of NC cells.

Xenopus embryos were obtained and staged as described previously (Newport and Kirschner, 1982; Nieuwkoop and Faber, 1994). Embryos were injected at the 8- to 16-cell stage as described (Aybar et al., 2003). Zebrafish strains were maintained and bred according to standard procedures (Westerfield, 2000), injected before division commenced and staged as described (Kimmel et al., 1995). We used two transgenic lines: sox10:egfp (Carney et al., 2006) and sox10:mGFPnRFP (C.L. and R. Mayor, unpublished). In the latter, the sox10(4.9) promoter (Carney et al., 2006) drives NC expression of membrane-bound GFP and nuclear localised RFP separated by the self-cleaving 2A sequence (Stewart et al., 2009).

MOs against Xenopus Par3 (Par3MO; 8 ng, 3′-TACTTCCACTGCCACTCGAAACCCT-5′) and Trio (TrioMO; 8 ng, 3′-TGCGCATAGCTACAGCTGAAAAAAA-5′) were designed by GeneTools. Zebrafish Par3MO has been reported previously [0.21 pmol (Wei et al., 2004)]. Equimolar concentrations of standard control MO (ControlMO) were used in both species. N-cadherinMO was used as described (Theveneau et al., 2010). Fluorescein-dextran (Invitrogen, D1821; 3 μg) or Rhodamine-dextran (Invitrogen, D1824; 5 μg) were used as tracers. Plasmids were linearised and mRNA transcribed as described (Harland and Weintraub, 1985). mRNA constructs injected were: membraneGFP (mbGFP), membraneRFP (mbRFP), EB3GFP (all 300 pg); Par3GFP, Par3RFP (both 500 pg); and p120cateninGFP (120 pg). TrioGFP was injected as DNA (300 pg).

Co-immunoprecipitation (Kashef et al., 2009) and western blots (Kuriyama and Mayor, 2009) were performed as described previously using the following antibodies: Par3 (Millipore, 0-330; 1:1000); p42/44 MAPK (Cell Signaling, 9102S; 1:1000); rabbit anti-HA (Abcam; 1:2000); chicken anti-GFP (Abcam; 1:3000); rabbit IgG HRP (ECL, NA934; 1:25,000); rabbit IgG AP (Dianova; 1:2000); chicken IgG AP (Abcam; 1:2000). Band intensity was measured using ImageJ (NIH). In situ hybridisation was performed as described (Harland, 1991). NC was labelled with digoxigenin-labelled RNA probes against Snail2 (Mayor et al., 1995) or Twist (Hopwood et al., 1989) in Xenopus and foxd3 in zebrafish (Odenthal and Nüsslein-Volhard, 1998). Cryosectioning was performed using standard procedures. Immunostaining was performed according to standard procedures with α-tubulin (DSHB, 12G10; 1:25), acetylated tubulin (Sigma, 611B1; 1:100), Par3 (1:200), β-catenin (Abcam, ab6302; 1:500), rabbit IgG Alexa 488 (Invitrogen, A11034; 1:100) and mouse IgG Alexa 488 (Invitrogen, A11017; 1:200) antibodies. If required, DAPI was applied with the secondary antibody (Sigma, D9542; 20 μg/ml).

Transplants were performed as described (De Calisto et al., 2005). Cranial NC was dissected using a standard technique (Alfandari et al., 2003). Fibronectin dishes were prepared as described (Theveneau et al., 2010) using 10 μg/ml or 50 μg/ml fibronectin (Sigma) for plastic or glass dishes, respectively. When required, nocodazole (Sigma, M1404) or NSC23766 (Tocris, 2161) was added to provide final concentrations of 13 nM and 20 μM, respectively. The cell-sorting assay was performed as described previously (Carmona-Fontaine et al., 2011). Fixed cells or sections were imaged using a Leica SPE confocal microscope. NC explants were observed by low-magnification time-lapse on a Leica DM5500 compound microscope, with frames collected every 5 minutes. High-magnification time-lapse on a spinning disk microscope was used to image individual cells, with frames collected every 2 seconds. In vivo time-lapse of zebrafish NC was performed on a spinning disk microscope, with frames taken every 5 minutes. Pixel intensity was measured using ImageJ.

Cells were tracked using the ImageJ Manual Tracking plugin. Track speed and persistence were determined using the ImageJ Chemotaxis Tool plugin. To measure dispersion, the ImageJ Delaunay Triangulation plugin was used. Several methods to study CIL have been developed (Scarpa et al., 2013); we used explant confrontation and cell collision assays as described by Carmona-Fontaine et al. (Carmona-Fontaine et al., 2008). EB3GFP movies were analysed automatically using plusTipTracker (Applegate et al., 2011). The following parameters were used: maximum gap length, 30 frames; minimum track length, 3 frames; search radius range, 5-25 pixels; maximum forward angle, 35°; maximum backward angle, 10°; maximum shrinkage factor, 1.5; fluctuation radius, 2 pixels. For manual analysis, 5×5 μm squares were selected at the front and back of a cell. The total number of EB3GFP⁺ tips that disappeared within the square during a period of 2 minutes was compared with the number that entered the square. Microtubules that were within the square in the final frame were excluded from analysis. All error bars indicate s.e.m. FRET for Rac1 was performed as previously described (Theveneau et al., 2010).

Comparison of percentages was performed using contingency tables as described previously (Taillard et al., 2008). Two data sets were considered significantly different (null hypothesis rejected) if T>3.841 (α=0.05), T>6.635 (α=0.01) or T>10.83 (α=0.001). Normality of data sets was tested using Kolmogorov-Smirnov’s test, d’Agostino and Pearson’s test and Shapiro-Wilk’s test using Prism4 (GraphPad). A data set was considered normal if found to be normal by all three tests. Data sets following a normal distribution were compared using Student’s t-test (two-tailed, unequal variances) or a one-way ANOVA with a Dunnett’s multiple comparisons post-test using Excel (Microsoft) or Prism4. Data sets that did not follow a normal distribution were compared using Mann-Whitney’s test or a non-parametric ANOVA (Kruskal-Wallis with Dunn’s multiple comparisons post-test) using Excel or Prism4. Cross-comparisons were performed only if the overall P-value of the ANOVA was less than 0.05.

We thank J. Leslie and B. Stramer for comments on the manuscript; and N. Ueno, I. Daar and A. Debant for the EB3, Par3 and Trio fusion proteins, respectively.