We have shown here that cell clusters exhibit radial polarity with large stable protrusions in the polarized outer cells and high level of Rac1 at the free edge and nonpolarized inner cells. This radial symmetry is broken upon addition of a chemoattractant that further stabilizes protrusions and increases Rac1 activity at the front, leading to directional migration of the cluster toward the source of the chemoattractant (P). This cell polarity is N-cadherin/CIL dependent and essential for efficient chemotaxis (P and 7Q). Inhibition of cell interactions leads to a loss of CIL resulting in a loose arrangement of cells with no difference between inner or outer cells, no stable cell polarity, and poor chemotaxis (Q). These data, alongside in vivo description of Xenopus
NC cells migration, suggest that these cells migrate as a cohesive cluster progressively breaking away as single cells, similar to the original description of cluster migration done by Trinkaus (Trinkaus, 1988
). The effect of cell dissociation on chemotaxis in vivo may be counterbalanced by inhibitory cues at the border of the migration route in order to maintain the cells in close proximity (R) or by a hypothetical attraction between the cells.
One of the big issues with collective movement is how the driving force is generated. While in small clusters like the Drosophila
border cells it is possible that a couple of front cells may pull the rest of the group, it is unlikely that a few front cells would be sufficient to achieve the same effect in large populations like the NC. In fact, studies on cell sheet migration have shown that the main driving force arises from cells inside the group while leading cells are mainly giving direction (Trepat et al., 2009
). However, despite the fact that we demonstrated the requirement of cell interactions, NC cells remain a mesenchymal population. Relative positions of a given cell and its direct neighbors are not fixed. Cells do exchange positions and gaps are constantly appearing in between the cells leading some inner cells to form protrusions and behaving as front cells for a while before colliding with the cells in front or next to them. These observations strongly indicate that the NC cells population should be seen as a relatively cohesive population progressively breaking up as a collection of small clusters of variable cell composition that are constantly splitting, colliding, and reassembling (R), rather than as a group with stable organization over time in which a wide group of inner cells would have to be pulled by a few front cells. Consequently, we think that NC migration cannot be directly compared with epithelial movements during wound healing or lateral line migration in terms of physical motion of the group.
Another aspect of collective migration that studies on lateral line have highlighted is the possibility that inner cells could act as a sink by trapping Sdf1 using Cxcr7 and therefore helping to shape the gradient itself (Dambly-Chaudiere et al., 2007; Valentin et al., 2007
). In this system, Sdf1 expression in the surrounding tissues is homogenous (David et al., 2002
), making necessary an additional system like the sink model to shape a gradient along which the cells can move. On the contrary, in Xenopus
, Sdf1 expression is progressively shifting ventrally and is constantly ahead of the NC cells position along the dorso-ventral axis. In addition, isolated small clusters, in which there are no inner cells, as they are all exposed to a free space and produce protrusions, migrate as efficiently as big groups. Moreover, transient contacts between single cells are sufficient to partially restore chemoattraction. All these observations indicate that a sink system similar to that described for the lateral line is unlikely to be required for NC cell migration. Although our results demonstrate a crucial role for cell interactions during NC directional migration, we can not exclude that other mechanisms, such as inner cell acting as sink for chemoattractant signals or a global detection of chemoattractants by the whole cluster, could also cooperate with CIL in vivo.
Different alternatives about how chemoattractant are generating directional migration have been proposed (Andrew and Insall, 2007; Iglesias and Devreotes, 2008
). Some argue that chemokines induce the formation of cell protrusions and use the protrusions as a physical markers of responding cells (Haas and Gilmour, 2006
), while others suggested that stabilization of cell protrusions formed independently of the chemotactic signaling could be sufficient to generate directional movement (Andrew and Insall, 2007
). A recent study clearly showed that the increase of protrusion stability correlates with an increase in cell persistence (Harms et al., 2005
), reinforcing the possibility that stabilizing pre-existing protrusions can lead to directional migration. Our data in NC cells support the notion that chemoattractants stabilize protrusions at the front of a cell cluster, creating an asymmetry and leading to directional migration of the cell group.
Our results on cell-contact-dependent polarity are consistent with our recent findings showing that activation of RhoA at regions of NC cell contact is essential for migration during CIL (Carmona-Fontaine et al., 2008
), in which cell protrusions are inhibited after cell-cell contact (Abercrombie and Heaysman, 1953
). Here we further show that cell contacts, CIL dependent and mediated by N-cadherin, are essential for NC chemotaxis and that the polarization of the small GTPases by cell contact is important for optimal response to a chemoattractant. Besides, we have shown here that N-cadherin is required for CIL at the cell contacts and that N-cadherin inhibition leads to an increase of Rac1 activity at the juxtamembrane domain probably due to a lack of RhoA activation downstream of the Wnt/PCP pathway. The precise mechanism of interaction between N-cadherin and Wnt/PCP during CIL remains to be investigated.
Our data also indicate that type I cadherin-mediated cell interactions are essential for proper collective migration of a highly mesenchymal and invasive cell population such as the neural crest. These data further support the idea proposed for cancer cells that transient epithelial-like cell interactions do not prevent mesenchymalization and migration (Yang and Weinberg, 2008
). Interestingly, tip-like contacts were described during chick NC cells migration (Kulesa and Fraser, 2000; Teddy and Kulesa, 2004
). We propose that such contacts could achieve the same effect on cell polarity that the pseudo epithelial-like interactions present in a migratory cohesive cell group. A possibility further supported by data showing that these cells exhibit CIL-like behaviors during migration (Kulesa and Fraser, 2000; Teddy and Kulesa, 2004
). We show here that even invasive mesenchymal cells can benefit from cell-cell interactions, and it would be interesting to address the role of cell contacts during collective phase of cancer cell migration.
The results presented here may lead to the reinterpretation of recent studies. For example, inhibition of N-cadherin blocks directional migration of cerebellar granule neurons (Rieger et al., 2009
) and LL (Kerstetter et al., 2004
) consistent with our conclusion that cell contacts are required for directional migration and suggest that these phenotypes may be due to a loss in chemotactic response, as we have demonstrated for NC cells.
Finally our results, alongside the influence of CIL, give a more complete view of how large populations of cells can achieve directional migration by integrating cell interactions and external cues. They indicate that invasive cells need to interact not only with their local environment, but also with each other in order to migrate efficiently, and may give new angles to better understand and tackle invasive issues.