Supracellular contraction at the rear of neural crest cell groups drives collective chemotaxis

See allHide authors and affiliations

Science  19 Oct 2018:
Vol. 362, Issue 6412, pp. 339-343
DOI: 10.1126/science.aau3301

Supracellular cable drives collective cell movement

Neural crest cells migrate far and wide through a vertebrate embryo during development. Shellard et al. used Xenopus and zebrafish embryos to study how these clumps of mesenchymal cells migrate (see the Perspective by Adameyko). Movement was powered by a supracellular actomyosin cable that contracted around the rear of the clump. Similar supracellular contractility at the front was inhibited by a chemotactic signal. The imbalance in forces caused cells to rearrange so that the whole clump would be propelled forward.

Science, this issue p. 339; see also p. 290


Collective cell chemotaxis, the directed migration of cell groups along gradients of soluble chemical cues, underlies various developmental and pathological processes. We use neural crest cells, a migratory embryonic stem cell population whose behavior has been likened to malignant invasion, to study collective chemotaxis in vivo. Studying Xenopus and zebrafish, we have shown that the neural crest exhibits a tensile actomyosin ring at the edge of the migratory cell group that contracts in a supracellular fashion. This contractility is polarized during collective cell chemotaxis: It is inhibited at the front but persists at the rear of the cell cluster. The differential contractility drives directed collective cell migration ex vivo and in vivo through the intercalation of rear cells. Thus, in neural crest cells, collective chemotaxis works by rear-wheel drive.

Directed migration orchestrates events in development, homeostasis, and disease (14). Most long-range directed migration in vivo occurs by chemotaxis (2, 49), in which cells follow gradients of soluble chemical cues. This has been best understood in individually migrating cells, for which several mechanisms have been proposed (1013), but collective migration has been less well studied.

In collective migration, leader cells have dynamic actin-based protrusions (Fig. 1A, darker red) (1, 6), form contacts with follower cells and with the extracellular matrix, and are responsive to chemotactic signals (3, 14, 15). We asked whether cells at the group’s rear (Fig. 1A, dotted rectangle) contribute to collective cell chemotaxis. To investigate the mechanism of collective chemotaxis ex vivo and in vivo, we studied Xenopus and zebrafish cranial neural crest cells, an embryonic cell population that undergoes collective cell migration (6, 16) in a manner similar to cancer cells (17), unlike neural crest cells of other species or those in the trunk, where less is known about the collectiveness (18). Although contact inhibition of locomotion and cluster confinement (19, 20) are needed for cephalic neural crest directional movement in Xenopus and zebrafish, they are not sufficient, as collective chemotaxis toward SDF1 is essential for long-range directed movement (6).

Fig. 1 Xenopus neural crest clusters exhibit a contractile actomyosin ring.

(A) A neural crest cluster with protrusions (red) at the edge undergoes chemotaxis to SDF1. SDF1 stabilizes the protrusions at the front (darker red) (7). Dotted rectangle, rear cells. (B) Immunofluorescence of a neural crest explant in the absence of SDF1. MLC, myosin light chain. Scale bar, 50 μm. (C to E) Immunofluorescence of a cell at the edge of a neural crest explant (C and E) and a corresponding diagram (D). N-Cad, N-cadherin; Memb, membrane. Scale bar, 10 μm. (F) Protein fluorescence levels (means ± SEM) along the actin cable. Position 0 μm represents the cell contact. a.u., arbitrary units. n = 8 cells. (G) Spontaneous contraction of the actomyosin cable. Green arrowheads, cell-cell contacts. Scale bar, 10 μm. (H) Actomyosin length (means ± SEM) measured over time. Contractions start at 0 s. n = 20 cells. (I) Multicellular contraction of the actomyosin cable. Scale bar, 10 μm. (J) Distribution of actomyosin contractility at different angles without (−SDF1) or with (+SDF1) an SDF1 gradient. n = 150 contractions. (K) Relative actomyosin length at the front (brown line) and rear (green line) of a cluster and the positions of the front (red line) and rear (blue line) of the cluster. r.u., relative units.

Imaging of fluorescently tagged actin and myosin in neural crest explants revealed the presence of a multicellular actomyosin ring localized at the periphery of the cell group, in both the absence and presence of an SDF1 gradient (Fig. 1B and fig. S1, A and B). Enrichment of N-cadherin near the actomyosin cable at the cell junction (Fig. 1, C to F, and fig. S1, C to E) suggests that this cable is supracellular. Premigratory neural crest and neural crest overexpressing E-cadherin, but not N-cadherin, have internalized myosin localization, rather than myosin at the cluster periphery (fig. S1, F to J), suggesting that the switch of cadherin expression during the epithelial-to-mesenchymal transition may be required for the formation of the actomyosin cable.

To determine whether the actomyosin cable is contractile, we performed laser photoablation of the structure, resulting in recoil of both the actomyosin cable and cell-cell junctions (fig. S2, A and B), followed by the cable’s reformation (fig. S2, C and D). To assess contractility, we measured actomyosin length, and we found frequent shortening (Fig. 1, G and H) independent of SDF1. These contractions were multicellular, as adjacent cells contracted synchronously (Fig. 1I and fig. S2E). A second ablation in a nearby cell after an initial ablation resulted in reduced actomyosin recoil (fig. S2, F and G), indicating that the tension of the cable is transmitted between cells. In epithelial cells, the presence of an actomyosin cable seems to inhibit protrusion formation (21), but this inhibition does not occur in mesenchymal neural crest cells (fig. S2, H and I).

Although exposure to SDF1 gradients did not affect the magnitude of actomyosin contractions (Fig. 1H), contractions occurred less frequently in front cells during collective chemotaxis without affecting cells at the rear (Fig. 1J and fig. S3A). A similar inhibition of front contractions was observed with the chemoattractant PDGF-A (22) (fig. S3B). Mechanistically, this contractility gradient is likely set up by SDF1 activation of Rac1 in front cells, which inhibits RhoA and myosin phosphorylation (fig. S4). Uniform SDF1, unlike the SDF1 gradient, did not inhibit contractility (fig. S5A), suggesting that the cluster responds to the chemotactic gradient instead of to absolute SDF1 levels. This was further supported by the observation that rear contractility (fig. S5B) and cluster speed (fig. S5C) were unchanged when clusters were closer to the chemoattractant source, where higher SDF1 levels should be present.

To explore the connection between the asymmetric actomyosin contraction and collective chemotaxis, we simultaneously measured the positions of front and rear cells of explants during migration, as well as the length of the actomyosin cable at the front and rear. Pulsatile contraction of the cable at the rear (Fig. 1K, green lines, and fig. S6A) coincided with the forward movement of the rear (Fig. 1K, blue lines, and fig. S6A). Both events immediately preceded the movement of the front of the cluster (Fig. 1K, red lines, and fig. S6A). A similar local contraction precedes a short forward movement in the absence of SDF1 (fig. S6, B and C), but with no long-range directed movement. Together, these results suggest that supracellular actomyosin contractility at the rear may drive collective cell chemotaxis.

We tested the role of rear contractility of the actomyosin ring in collective chemotaxis by performing laser ablation. Chemotaxis was impaired by ablation of the actomyosin ring in rear cells but not by equivalent ablations in front cells or by other control ablations (Fig. 2, A to C, and fig. S7), suggesting the necessity of a rear supracellular actomyosin cable for chemotaxis. To test the requirement of rear contractility, we used an optogenetic system (23, 24) to either increase (via the protein construct optoGEF-contract) or decrease (via the construct optoGEF-relax) contractility and myosin phosphorylation in the actomyosin cable upon illumination with low doses of blue light (fig. S8). No effect was observed on cell protrusions (fig. S9), focal adhesions (fig. S10), cell dispersion (fig. S11), or the phosphorylation of myosin located basally outside the cable (fig. S8K) upon illumination under the conditions of our assay. We first tested whether high contractility at the rear is necessary for collective chemotaxis by photoactivating optoGEF-relax at the rear of migrating clusters exposed to SDF1 (Fig. 2D). Inhibition of contractility in rear cells (Fig. 2D) impaired chemotaxis (Fig. 2, E and F). By contrast, inhibition of contractility in front cells failed to affect collective chemotaxis (fig. S12). To determine whether rear contractility is sufficient to drive collective cell migration, we activated contractility in rear cells in the absence of SDF1 (Fig. 2G). Whereas control neural crest cells did not exhibit directional migration, activated neural crest cells moved forward, away from the region of photoactivation (Fig. 2, H and I).

Fig. 2 Rear contractility is necessary and sufficient for collective chemotaxis of Xenopus neural crest.

(A) (Top) Examples of two neighboring cells with ablations (red arrowheads). Scale bar, 10 μm. (Bottom) Images of explants exposed to SDF1 gradients during ablations between the indicated times. Scale bar, 50 μm. (B) Position of the front of explants during chemotaxis (means ± SEM). The dashed line indicates when ablations begin. n = 6 to 8 clusters. (C) Chemotaxis index (means ± SEM) of clusters. n = 6 to 8 clusters. ***P ≤ 0.001 (two-tailed Student’s t test); ns, not significant. For (A) to (C): red, front actomyosin cable ablation; blue, rear actomyosin cable ablation. (D to O) Experimental setup for treated explants [(D), (G), (J), and (M)], representative cluster tracks [(E), (H), (K), and (N)], and the distance migrated (means ± SEM) over times as indicated in methods [(F), (I), (L), and (O)]. n = 10 to 23 clusters (F), n = 10 or 11 clusters (I), n = 14 to 18 clusters (L), and n = 11 or 12 clusters (O). ***P ≤ 0.001 (two-tailed Student’s t test). Scale bars, 40 μm [(E) and (K)]; 20 μm [(H) and (N)]. Green box, initial illumination area; cross, initial cluster position. The top of all pictures is the rear.

To test whether SDF1-dependent inhibition of contractility in front cells is required for collective chemotaxis, we activated contractility in front cells of migrating clusters exposed to SDF1 (Fig. 2J); this repressed chemotaxis (Fig. 2, K and L), suggesting that low front contractility is essential for collective chemotaxis. Lastly, we asked whether front inhibition of contractility by SDF1 was sufficient to generate directed migration. We inhibited front contractility in the absence of SDF1 (Fig. 2M), which resulted in directional migration (Fig. 2, N and O). These optogenetic treatments affected contractility (fig. S13) (23) and not cell motility (fig. S14). Together, these results suggest that collective migration requires greater contractility at the rear than at the front of the cell cluster.

To understand how rear cell contractility might drive directed collective cell migration, we implemented a cell-centered computational model of a cell group with contractile edge cells (methods; Fig. 3A; and fig. S15, A to C). Cells interact through a soft-core repulsion and midrange attraction; to model contractions, cells at the edge (either around the cluster or at the rear) periodically attract one another with additional force (Fig. 3A, red springs). Similar to ex vivo clusters, only simulations with rear but not uniform contractility were able to migrate forward (Fig. 3B, fig. S15D, and movie S1). Other migration parameters were comparable between in silico and ex vivo clusters (fig. S15, E and F). Unexpectedly, analysis of cell movements in silico revealed that rear cells in contractile regions were intercalated forward, into the cell group (Fig. 3C). As predicted by the model, we found an equivalent intercalation at the rear of neural crest clusters (Fig. 3D and fig. S15G). Furthermore, our simulations predicted that the effect of this local cell rearrangement is spread through the whole cell group such that when the cluster’s rear contracts, the rear cells are intercalated, triggering a wave of cell movement that propagates from the rear toward the front of the cluster (Fig. 3, E and F). A similar wave was observed ex vivo (Fig. 3, G and H), as predicted by the model. This suggests that rear cell intercalation after rear contractions pushes cells forward progressively over time. Averaging cell movement over time and subtracting cluster movement reveals an intracluster flow of cells in silico, whereby rear cell intercalation causes a drift forward through the middle of the group and cells at the front and sides move backward, replacing rear cells (Fig. 3I). This was then confirmed to occur ex vivo as well (Fig. 3J). We found a positive correlation between the speed of ex vivo and in silico clusters during collective migration and the amount of rear cell intercalation (Fig. 3, K and L), consistent with this mechanism’s driving cluster movement. Nonmigratory ex vivo and in silico clusters had low intercalation, and migratory clusters had comparable cluster speeds (Fig. 3L). We observed that contractions were normally accompanied by relaxation events (fig. S16A, green and red bars); however, we showed that ex vivo and in silico clusters were able to migrate directionally, independently of the level of rear relaxation (fig. S16, A and B, and movie S2). Altogether, these results suggest that rear contractility drives collective cell migration by inducing cell intercalation, which pushes the group forward.

Fig. 3 Modeling contractility-driven collective migration.

(A) Illustration of the computational model cluster. Yellow, edge cells; green, internal cells; red, contraction; horizontal line, distinction between front and rear, with rear outer cells contracting (red spring). (B) Directionality (means ± SEM) of clusters. n = 10 clusters. ***P ≤ 0.001 (two-tailed Student’s t test); ns, not significant. (C and D) Intercalation of a rear cell (purple) between two adjacent cells (orange) in silico (C) and ex vivo (D) during directional migration. Scale bars, 20 μm. (E to H) Wave of contraction. Heat maps indicate speed during migration in silico (E) and ex vivo (G). Scale bars, 40 μm. Graphs show speed profiles (means ± SEM) from clusters in silico (F) and ex vivo (H) at different times during directional migration. Position 0 μm represents the rear of the cluster; positions 200 and 170 μm [(F) and (H), respectively] represent the front of the cluster. n = 5 clusters. (I and J) Direction of intracluster cell movements shown from time-averaged cell tracks in silico (I) and particle image velocimetry ex vivo (J) after subtraction of cluster movement. n = 5 clusters. Scale bars, 40 μm. (K) Cluster speed and rear cell intercalation during migration. (L) Cluster speed (means ± SEM) and rear cell intercalation (means ± SEM) of clusters. Abl, laser ablation of the actomyosin ring in rear cells. n = 6 to 21 clusters. The top of all pictures is the rear.

Next, we analyzed whether this model of collective cell chemotaxis explains the in vivo migration of neural crest cells. As in ex vivo clusters, an actomyosin cable is present at the edge of the neural crest in both Xenopus (Fig. 4, A and B, and fig. S17, A and B) and zebrafish (fig. S18, A and B). Live imaging of the actomyosin cable shows that it is a contractile structure in vivo in both Xenopus (Fig. 4C and fig. S17C) and zebrafish (fig. S18, C and D) and contracts more often at the rear of the neural crest stream than at the front (fig. S17D). Rear contractility precedes forward movement of the cluster in vivo (Fig. 4D), as it does ex vivo. Less phosphomyosin was present at the front than at the rear at the beginning of migration (figs. S17, E to H, and S19). To identify whether individual neural crest cells flowed through clusters, as predicted from in silico and ex vivo results, we tracked live cells during migration. In both Xenopus and zebrafish, cells that were initially at the rear of the group were intercalated forward during migration (Fig. 4E and fig. S20, A and B). As with the ex vivo and in silico data, subtracting cluster movement to in vivo cell tracks revealed an intracluster flow (Fig. 4F and fig. S20C). This suggests that rear contractility is driving neural crest migration in vivo.

Fig. 4 Actomyosin drives collective chemotaxis in vivo in Xenopus.

(A and B) Immunofluorescence of the rear (A) and front (B) of the Xenopus neural crest stream. Dashed lines, cell-cell contacts between neural crest cells. Scale bar, 10 μm. (C) Contraction of the actomyosin cable of Xenopus neural crest in vivo. Green arrowheads, cell-cell contacts; dashed lines, cell edges. Scale bar, 10 μm. (D) Actomyosin length at the front (brown line) and rear (green line) of a Xenopus cluster in vivo and the positions of the front (red line) and rear (blue line) of the cluster. (E) Intercalation of a rear cell (purple) between two adjacent cells (orange) in vivo. Scale bar, 20 μm. (F) Tracks of rear neural crest cells in vivo after subtraction of the cluster movement. Gray dots, initial cell positions. Scale bar, 30 μm. (G to O) Experimental design of treated Xenopus embryos [(G), (J), and (M)], representative tracks of neural crest clusters [(H), (K), and (N)], and migration indexes (means ± SEM) [(I), (L), and (O)]. Green boxes, initial illumination area; crosses, starting position of the explant. n = 10 clusters. ***P ≤ 0.001 (two-tailed Student’s t test). Scale bars, 50 μm. (P) The model: Collective cell chemotaxis is driven by actomyosin contractility at the rear (red arrows). The top of all pictures is the rear.

To test whether rear contractility is required for neural crest migration in vivo, we grafted neural crest expressing optoGEF-contract or optoGEF-relax into wild-type Xenopus embryos. Activation of contractility at the front of the stream (Fig. 4, G to I, and movie S3) or inhibition at the rear (Fig. 4, J to L, and movie S4) impaired neural crest migration, indicating that greater contractility at the rear than at the front was necessary for migration in the embryo. Neural crest grafted into host embryos lacking SDF1 failed to migrate, but activation of contractility at the rear of such grafts rescued migration (Fig. 4, M to O, and movie S5), demonstrating that high actomyosin contractility at the rear can drive directed collective migration in vivo. We conclude that rear contractility, as produced by a supracellular actomyosin cable, can drive collective cell chemotaxis in vivo (Fig. 4P).

The theory of active gels shows how anisotropies in viscoelastic materials can generate rotating flows similar to the cellular flows described here (25, 26). In addition, physicists have proposed that cells can move by using tangential retrograde movement of their surfaces (27) and that this movement is more energetically efficient than other modes of swimming (28). However, only recently has such surface retrograde propulsion been described for the migration of single cells (29). Our work identifies an equivalent surface retrograde propulsion for collective cell migration, suggesting that the whole cluster behaves as a “supracell.”

It is likely that for in vivo collective chemotaxis, rear actomyosin contractility works together with protrusions at the front to drive migration. Notably, peripheral actomyosin has been similarly observed in the collective migration of other cell types, including cancer cells (30, 31), suggesting that other cell types may migrate under similar principles.

Supplementary Materials

Materials and Methods

Figs. S1 to S20

References (3247)

Movies S1 to S5

References and Notes

Acknowledgments: We thank G. Charras, L. Cramer, C. Stern, B. Stramer, and D. Wilkinson for critical reading of the manuscript and members of the Mayor laboratory for discussions. We thank M. Tada, M. Meyer, and E. Sahai for providing us with vectors and antibodies, and E. Scarpa for preliminary data. Funding: This study was supported by grants from the Medical Research Council (M010465 and J000655 to R.M.), the Biotechnology and Biological Sciences Research Council (M008517 to R.M.), the Wellcome Trust (102489/Z/13/Z Wellcome Trust Ph.D. fellowship to A.Sh.), the Spanish Ministry of Economy and Competitiveness/FEDER (BFU2015-65074-P to X.T.), the Generalitat de Catalunya and CERCA program (2014-SGR-927 to X.T.), the European Research Council (CoG-616480 to X.T.), and the European Commission (project H2020-FETPROACT-01-2016-731957 to X.T.) and by a Marie Curie fellowship (329968 to A.Sz.). Author contributions: R.M. conceptualized the study; A.Sh., A.Sz., X.T., and R.M. performed the methods; A.Sz. provided software; R.M. provided resources; A.Sh. and R.M. wrote the original draft of the manuscript; A.Sh., A.Sz., X.T., and R.M. reviewed and edited the manuscript; and R.M. supervised the study, provided project administration, and acquired funding. Competing interests: The authors declare no competing interests. Data and materials availability: All data are available in the main text or the supplementary materials.
View Abstract

Stay Connected to Science

Navigate This Article