Cxcl12a induces snail1b expression to initiate collective migration and sequential Fgf-dependent neuromast formation in the zebrafish posterior lateral line primordium

Interactions between cells determine morphogenesis, cell fate and cell behavior in developing organs. The simplicity and accessibility of the zebrafish posterior lateral line (pLL) system and our ability to manipulate and monitor its development in live embryos has led to its establishment as an extremely effective model system for studying self-organization of developing organ systems. The lateral line is a sensory system that allows fish to sense the pattern of water flow over the surface of their bodies (Coombs and Montgomery, 1999). It consists of sensory organs called neuromasts that are distributed in a stereotyped pattern over the body surface. Each neuromast has a cluster of sensory hair cells at its center. They are innervated by neurons of the lateral line ganglia, which carry sensory information from neuromasts back to the brain (Ghysen and Dambly-Chaudière, 2007).

Formation of the pLL system in zebrafish is pioneered by the pLL primordium, a group of about 140 cells that separates from the pLL placode, leaving behind cells that form the sensory neurons of the pLL ganglion (Kimmel et al., 1995). Whereas leading cells in the pLL primordium have a quasi-mesenchymal morphology, trailing cells are more epithelial; they have distinct apical-basal polarity and they reorganize to form nascent neuromasts (protoneuromasts) sequentially (Lecaudey et al., 2008; Nechiporuk and Raible, 2008). The primordium begins migration toward the tip of the tail at about 22 h post-fertilization (hpf). Cells incorporated into protoneuromasts in the migrating primordium are deposited as neuromasts, whereas cells that were not are deposited between neuromasts as inter-neuromast cells. Eventually, the primordium ends its migration about 24 h later after depositing five or six neuromasts and by resolving into two or three terminal neuromasts (Sarrazin et al., 2010).

Establishment of polarized Wnt and Fgf signaling systems coordinates morphogenesis and migration of the primordium. Wnt signaling dominates at the leading end and promotes its own activity (Aman and Piotrowski, 2008; Jiang et al., 2014). At the same time, activation of the Wnt/β-catenin signaling pathway drives expression of fgf3 and fgf10. However, leading cells, which produce Fgf, do not respond to these signals because Wnt activity simultaneously promotes expression of factors that inhibit activation of Fgf receptors (Aman and Piotrowski, 2008; Matsuda et al., 2013). Instead, the Fgfs activate Fgf signaling at the trailing end of the primordium, where Wnt signaling is weakest. There, Fgf signaling determines expression of the diffusible Wnt antagonist Dkk1b, which counteracts Wnt signaling to help establish stable Fgf-responsive signaling centers (Aman and Piotrowski, 2008). Once established, activation of the Fgf signaling pathway in trailing cells coordinates morphogenesis of protoneuromasts by simultaneously promoting the reorganization of cells into epithelial rosettes and by initiating expression of factors that help specify a hair cell progenitor at the center of each forming neuromast (Itoh and Chitnis, 2001; Lecaudey et al., 2008; Matsuda and Chitnis, 2010; Nechiporuk and Raible, 2008). Over time, the initially broad Wnt active domain shrinks closer to the leading edge and additional Fgf signaling centers form sequentially in its wake, each associated with formation of additional protoneuromasts (Dalle Nogare and Chitnis, 2017).

Polarized Wnt-Fgf signaling systems in the pLL primordium also coordinate collective migration of the primordium along the horizontal myoseptum on a path defined by the chemokine Cxcl12a (previously known as Sdf1a) (Sapede et al., 2005). Directed migration of the primordium in a caudal direction is determined by the polarized expression of two chemokine receptors, Cxcr4b and Cxcr7b (Ackr3b – Zebrafish Information Network) (Dambly-Chaudière et al., 2007; Haas and Gilmour, 2006; Valentin et al., 2007). Cells in the leading two thirds of the primordium express cxcr4b, which encodes a chemokine receptor that allows cells to interact with Cxcl12a and to respond via G protein-coupled signaling pathways with migratory behavior. The absence of Wnt signaling allows expression of cxcr7b in the trailing part of the primordium (Aman and Piotrowski, 2008). Cxcr7b interacts with Cxcl12a but its distinctive intracellular domain cannot engage the G protein-coupled system to determine migratory behavior (Balabanian et al., 2005; Graham, 2009). As a result, the migratory response of leading cells, coupled with unresponsive trailing cells, determines polarized response to Cxcl12a in the path of the primordium. In addition, Cxcl12a binds Cxcr7b with greater affinity than it does with Cxcr4b (Balabanian et al., 2005). Once bound, the complex is internalized and Cxcr7b-expressing cells degrade Cxcl12a. This creates a local gradient of Cxcl12a, which contributes to the robust polarized migratory response of leading primordium cells to Cxcl12a in the path (Donà et al., 2013; Venkiteswaran et al., 2013).

Although the framework described above provides a broad description of the early self-organization of the pLL primordium, the mechanisms that link the dynamics of Wnt-Fgf signaling to systematic changes in cell morphology, organization and migratory behavior remain poorly understood. The transcription factor Snail is well known for its role in coordinating expression of factors that promote epithelial mesenchymal transition (EMT) and migratory behavior (Barrallo-Gimeno and Nieto, 2005; Lamouille et al., 2014). An initial screen to identify factors expressed in the primordium that might promote mesenchymal morphology revealed that snail1b (snai1b – Zebrafish Information Network) is expressed in leading cells of the primordium. Its expression in the leading zone suggested that Wnt-dependent snail1b expression might determine the relatively mesenchymal morphology and migratory potential of leading cells. However, as this study describes, snail1b expression is not dependent on Wnt signaling, nor does interfering with its function, in any obvious way, prevent leading cells from acquiring a mesenchymal morphology. Instead, we discovered that snail1b expression is initiated by chemokines first encountered by leading primordium cells and its expression plays a crucial role in initiating collective migration of the pLL primordium. Furthermore, we found that migratory behavior of the primordium, more generally, has an essential role in initiating sequential formation of protoneuromasts and the progressive shrinking of the Wnt system.

snail1b is expressed in leading cells of the pLL primordium at 32 hpf (Fig. 1A, arrow) and in underlying cells of the horizontal myoseptum over which the primordium migrates. As a result, in some whole-mount in situ hybridization images of the pLL primordium, snail1b expression is also seen, slightly out of focus at a deeper plane, under the primordium (Fig. 1A, arrowhead; Fig. S1A-E). snail1b is expressed at the leading end of the pLL primordium at 20 hpf, just before onset of collective migration, and continues to be expressed during migration (Fig. S1A-E). Its expression in the pLL primordium is lost once its migration is complete at about 48 hpf (data not shown).

Like snail1b, Wnt-dependent expression of lef1 is restricted to a leading zone of the primordium at 32 hpf (Fig. 1A,B). However, in contrast to lef1, expression of which initially spans the entire primordium and gradually becomes restricted to a smaller leading zone (Matsuda et al., 2013; Valdivia et al., 2011), snail1b expression is always restricted to the leading end (Fig. S1A-E). Nevertheless, we investigated whether Wnt signaling plays some role in determining snail1b expression in leading cells by examining embryos treated with the glycogen synthase kinase 3 inhibitor BIO, which exaggerates Wnt/β-catenin signaling (Meijer et al., 2003). Whereas lef1 expression expanded into the trailing domain following exposure to BIO, snail1b expression was indistinguishable from controls (Fig. 1C,D). This suggested that Wnt signaling does not promote expression of snail1b. However, when the secreted Wnt antagonist Dkk1b was expressed in response to heat shock using a HS-dkk1b transgenic fish line (Stoick-Cooper et al., 2007), snail1b expression expanded into the trailing domain, whereas it resulted in the loss of lef1 expression in the primordium (Fig. 1E,F). No change in expression of snail1b and lef1 was observed in embryos not exposed to heat shock and in heat-shocked non-transgenic controls (Fig. 1G-J). These observations show that, rather than promoting its expression, some Wnt-dependent process prevents snail1b expression in the trailing zone.

As Fgfs expressed in response to Wnt signaling determine activation of Fgf receptors in the trailing zone, we investigated whether snail1b expression is prevented in this region by Fgf signaling. snail1b, usually confined to the leading region, expanded into the trailing domain in embryos treated with the Fgf signaling inhibitor SU5402 (Mohammadi et al., 1997), whereas Fgf-dependent pea3 (etv4 – Zebrafish Information Network) expression was lost (Aman and Piotrowski, 2008) (Fig. 1K-N). Similarly, expression of a dominant-negative form of the Fgf receptor under control of a heat shock promoter expanded expression of snail1b and suppressed pea3 expression (Fig. 1O,P) in HS-DN-fgfr1 transgenic embryos, whereas transgenic embryos not exposed to heat shock and heat-shocked non-transgenic siblings were not affected (Fig. 1Q-T). Together, these results confirm that Fgf signaling prevents snail1b expression in the trailing zone.

As Wnt signaling does not promote snail1b expression, we investigated whether Cxcl12a dependent-chemokine signaling determines snail1b expression in leading cells. Consistent with this possibility, snail1b expression was reduced or lost in embryos injected with cxcl12a morpholino (Fig. 2A,B). Cxcr7b, expressed by trailing cells, ensures a polarized response to chemokines by preventing trailing cells from responding to Cxcl12a and knockdown of cxcr7b allows cells at both ends of the primordium to respond to Cxcl12a (Valentin et al., 2007). Consistent with this change, snail1b expression was seen at both ends of the primordium in cxcr7b morphants (Fig. 2C, brackets). Furthermore, heat shock-induced expression of cxcl12a in HS:cxcl12a transgenic fish transiently promoted snail1b expression in the entire primordium (Fig. 2D), whereas no change was seen in non-heat-shocked HS:cxcl12a transgenic and heat-shocked non-transgenic embryos (Fig. 2E,F). Heat shock induced broad ectopic expression of cxcl12a, whereas no change was seen in non-heat-shocked transgenic and heat-shocked non-transgenic controls (Fig. 2G-I).

Two morpholinos were used to study snail1b function, a translation-blocking morpholino (TB) that was validated in a previous publication (Blanco et al., 2007) and a splice-blocking morpholino (SB) that was validated by visualizing aberrantly spliced fragments induced by its injection in embryos (Fig. S2A,B). Potential p53 (Tp53 – Zebrafish Information Network)-dependent off-target effects in all morphants (Eisen and Smith, 2008; Robu et al., 2007) were minimized by the use of a relatively low dosage of morpholino (0.5 ng TB and 0.75 ng SB) and with co-injection of p53 morpholino.

In embryos injected with control morpholino provided by Gene Tools (see Materials and Methods), the primordium deposited five or six neuromasts and completed migration to the tip of the tail where they dissociated to form terminal neuromasts by 52 hpf (Fig. 3A,H). By contrast, both TB and SB snail1b morpholinos reduced the number of deposited neuromasts to three or four and the primordium did not complete its migration by 52 hpf (Fig. 3B,C,H).

The specificity of the effects of TB morpholino was demonstrated by suppressing the morphant phenotype with co-injection of snail1b mRNA. Injection of relatively high levels of snail1b mRNA (200-250 pg) resulted in developmental defects including a shortened body axis, as described previously (data not shown) (Blanco et al., 2007). However, injection of 100 pg snail1b mRNA did not cause obvious morphological defects. This concentration was used to suppress the morphant phenotype. Whereas embryos injected with TB morpholino alone had only three or four deposited neuromasts, co-injection of the TB morpholino with 100 pg snail1b mRNA restored the number of deposited neuromasts to an average of five and the primordium completed its migration with formation of terminal neuromasts by 50 hpf (Fig. 3D,H).

ZO-1 (Tjp1a – Zebrafish Information Network), which is associated with tight junctions, was examined to determine whether the reduced number of deposited neuromasts was related to a delay in their sequential formation within the migrating primordium. ZO-1 was typically seen associated with the apical constrictions of two relatively mature trailing protoneuromasts and a third most recently formed leading protoneuromast in control primordia. Less prominent localization of ZO-1 was seen in a single relatively immature trailing protoneuromast in the primordia of both TB and SB morpholino-injected embryos (Fig. 3E-G). This confirmed that deposition of fewer neuromasts is related, at least in part, to a delay in their morphogenesis within the migrating primordium.

In addition to depositing fewer neuromasts, time-lapse movies showed that in morphants there was a delay in deposition of the first neuromast (L1) in the morphants; the primordium began its migration following a short delay, and migration was slower (Movie 1, Fig. 3I-K). Furthermore, whereas the control primordia had reached the tail tip, stopped migrating and formed terminal neuromasts by 50 hpf, morphant primordia were still actively migrating and had not formed terminal neuromasts (Movie 1) at 50 hpf. These problems were prevented by co-injecting snail1b TB morpholino with snail1b mRNA (Movie 2).

We tried to generate homozygous snail1b mutants to compare with snail1b morphants. F1 progeny from CRISPR/Cas9- and snail1b guide RNA-injected G0 fish were identified carrying insertions or deletions expected to result in loss of snail1b function (data not shown). However, heterozygous fish carrying these mutations did not produce either functional sperm or eggs and no homozygous mutant progeny could be obtained for analysis. Although their F1 progeny were infertile, key changes observed in snail1b morphants were also observed in pLL primordia of some CRISPR/Cas9- and snail1b guide RNA-injected G0 embryos (Fig. 3L-O and Fig. 4X-Z). At 28 hpf, in 20/50 G0 embryos, primordia on one side had migrated a shorter distance than on the other side (Fig. 3L,M), suggesting that in this subset of primordia CRISPR-mediated mutations had caused a delay in migration similar to that observed in snail1b morphant embryos. Two of these embryos subsequently died, but when the surviving embryos were re-examined at 50 hpf, all of them had fewer neuromasts on one side of the embryo and the primordia had not completed migration (Fig. 3N,O). The observation of such changes, only on one side and in a subset of injected embryos, is consistent with the expected mosaic distribution of cells with effective CRISPR-induced somatic mutations.

The delay in initiation of effective primordium migration in snail1b morphants could potentially result from changes in the primordium itself, whereby snail1b is expressed in leading cells, and/or from problems in the underlying snail1b-expressing horizontal myoseptum cells, which express cxcl12a to define the migratory path of the pLL primordium. To distinguish between these possibilities, we first examined the expression of cxcl12a at various time points and found no obvious change in the spatiotemporal pattern of cxcl12a expression along the horizontal myoseptum in snail1b morphants (Fig. S3A-F), suggesting no obvious problem in the differentiation of the cells that define the migratory path. Next, to confirm that the delay in initiating collective migration is related to a function of snail1b in the pLL primordium, rhodamine-dextran labeled cells from uninjected donors were transplanted into snail1b morpholino-injected embryos at the 1000-cell stage. Donor cells were placed halfway between the animal pole and the margin in host embryos, where they frequently give rise to ectoderm-derived primordium cells. Embryos were selected in which wild-type cells were only delivered to a primordium on one side of the morphant embryo and the primordium on the other side, composed only of morphant cells, served as a control. Viewing the embryo from the dorsal side allowed us to track migratory behavior of the primordium on the right and left of the embryo simultaneously (Fig. 4A-A″). When wild-type donor cells were successfully transplanted into snail1b morphant pLL primordia, they were also often located in the overlying skin, as both primordium and skin cells have neighboring ectodermal origins. However, it was relatively easy to identify morphant primordia with transplanted wild-type cells, as there were no wild-type cells in the underlying horizontal myoseptum cells (Fig. 4B-B″). This is consistent with the relatively distinct spatial separation of prospective primordium cells in the ectoderm, and prospective horizontal myoseptum in the mesoderm at the 1000-cell stage. Transplant experiments showed that 11/13 snail1b morphant primordia that had received transplanted wild-type cells migrated further than morphant pLL primordia on the contralateral side that had received no wild-type cells (Fig. 4A-A″, Movie 3). In one case, in which a snail1b morphant host primordium received donor wild-type cells but did not migrate faster than the contralateral morphant primordium, the wild-type cells were at the trailing end of the primordium, not at the leading end where snail1b is expressed (Movie 4). Together, these observations show that problems with migration in snail1b morphants have more to do with the function of snail1b at the leading end of the primordium and less to do with snail1b expression in underlying horizontal myoseptum cells, where it is also expressed.

Previous studies have demonstrated that slower migration of axial mesendoderm cells during gastrulation in snail1b morphants is related to de-repression of cdh1 (cadherin 1; also known as e-cadherin) (Blanco et al., 2007). Although there were no obvious changes in cdh1 expression in the snail1b morphant primordium (data not shown), there were reciprocal changes in the expression of transcripts encoding two other cell adhesion molecules, epithelial cell adhesion molecule (Epcam) and Cdh2 (cadherin 2; also known as N-cadherin). cdh2 and epcam are expressed throughout the primordium; however, cdh2 is most prominent in the leading two thirds of the primordium, whereas epcam is expressed in the trailing part in association with maturing protoneuromasts (Fig. 4C,D). However, in snail1b morphants cdh2 expression was reduced, especially in the leading region, whereas, in contrast, epcam expression became much more prominent in the leading region (Fig. 4E,F). Co-injection of 100 pg snail1b mRNA restored cdh2 expression and reduced epcam expression in the leading part of the primordium of snail1b morphants (Fig. 4G,H). Consistent with a role for snail1b in repressing epcam expression, injection of 100 pg snail1b mRNA on its own resulted in an overall reduction in epcam (Fig. 4J, Fig. S4B). By contrast, there was no obvious increase in cdh2 expression following ectopic snail1b expression (Fig. 4I, Fig. S4A). Reciprocal changes in cdh2 and epcam were not non-specifically related to slower migration as no obvious change in expression was seen when migratory behavior of the primordium was inhibited by exposure to Src kinase inhibitor PP1 (Hanke et al., 1996) (Fig. 4K-P).

To determine whether the expansion of epcam expression into the leading zone of the primordium is a crucial change that determines the delay in initiating collective migration we investigated whether partially knocking down epcam reduces this delay in snail1b morphants. At 28 hpf, the pLL primordium has typically migrated halfway along the length of the yolk extension in control embryos (Fig. 4Q). However, in snail1b morphants the primordium is still close to the otic vesicle because of the delay in initiating collective migration (Fig. 4R). However, in 26/30 snail1b morphants co-injected with 1.2 ng epcam morpholino, there did not appear to be any delay in migration (Fig. 4S), suggesting that expanded epcam expression is a significant determinant of delayed collective migration in snail1b morphants. It is important to note that although epcam knockdown prevented a migration problem associated with expanded epcam expression in snail1b morphants, injection of 1.2 ng epcam morpholino on its own did not significantly affect either pLL migration (Fig. 4T) or the pattern in which neuromasts were deposited (Fig. S5D-G). It did, however, result in aberrant morphogenesis of otoliths (Fig. S5A-C), recapitulating both the ear phenotype and absence of lateral line phenotype previously described in epcam mutants (Slanchev et al., 2009).

Consistent with expanded epcam expression having a crucial role in determining delayed primordium migration in snail1b morphant embryos, at 26 hpf the primordium had migrated a much shorter distance in embryos injected with 30 pg of epcam mRNA compared with uninjected controls (Fig. 4U-W).

As expanded epcam expression was a consistent feature of snail1b morphant primordia we investigated whether such expanded epcam expression was also observed in the subset of primordia that were characterized by slower migration in CRISPR/Cas9- and snail1b guide mRNA-injected G0 embryos. All surviving primordia with slower migration scored in the earlier section (Fig. 3L,M) were found to have expanded epcam expression (Fig. 4X-Z) supporting the interpretation that this is a key change that accompanies slower/delayed migration when snail1b function is lost.

As formation of protoneuromasts is dependent on Fgf signaling, we investigated whether the delay in their formation in snail1b morphants is associated with a delay in the progressive restriction of Wnt signaling to a smaller leading zone and to the stable establishment of the trailing Fgf signaling system. By 26 hpf, Wnt signaling-dependent expression of lef1 is restricted to a leading domain in control embryos, whereas in snail1b morphants, lef1 expression remains expanded and extends into the trailing zone (Fig. 5A,D). At 35 hpf, the lef1 expression domain in snail1b morphants remains larger than that in controls (Fig. 5B,E). It is nevertheless restricted to a leading zone by this stage. Conversely, although well-established Fgf signaling centers, demarcated by pea3 expression, were seen in control embryos at 26 hpf, pea3 expression was either absent or just being initiated at the trailing end of the primordium in snail1b TB morphants (Fig. 5G,J). The pea3 expression domain was, however, well established in snail1b morphants by 35 hpf (Fig. 5H,K).

The delay in the shrinking of the leading Wnt system was accompanied by a corresponding delay in the eventual loss of Wnt signaling. In control embryos, the lef1 expression domain progressively shrinks and is lost in most embryos (18/24) by 48 hpf. However, in snail1b morphants lef1 was still expressed in a relatively broad leading zone at 48 hpf (32/32) (Fig. 5C,F). Similarly, pea3 expression moves progressively closer to the leading end of the primordium and by 48 hpf in control embryos is associated with the entire primordium, which at this stage has stopped migrating and has resolved to form terminal neuromasts. In snail1b morphants, however, the primordium had not resolved to form terminal neuromasts and pea3 expression was still restricted to a broad trailing zone of the primordium at 48 hpf (Fig. 5I,L). One interpretation of the delay in Wnt and Fgf signaling dynamics in snail1b morphants is that the size of the Wnt system reflects its position along the migratory path and the delay in the shrinking of the Wnt system simply reflects the delay in initiating primordium migration. However, plotting the size of the leading Wnt system as a function of the distance travelled by the primordium, rather than as a function of time, showed that shrinking of the Wnt system is delayed, even when one considers the position of the primordium along its migratory path (Fig. 5Y).

The delay in restricting Wnt signaling and establishing a trailing Fgf signaling system was accompanied by a delay in establishment of polarized chemokine receptor expression. cxcr4b, typically restricted to the leading zone by 26 hpf in control embryos, remained relatively broad in snail1b TB morphants (Fig. 5M,P). Similarly, whereas there is normally robust expression of cxcr7b in a trailing zone of the pLL primordium in control embryos by 26 hpf, its expression was absent or just being initiated in snail1b TB morphant embryos (Fig. 5S,V). By 35 hpf, as in control embryos, polarized cxcr7b expression was established in the trailing part of the primordium in snail1b morphants (Fig. 5T,W). However, unlike controls, expression of cxcr4b remained broad and relatively unpolarized in snail1b morphants at 35 hpf and 48 hpf (Fig. 5N,Q,O,R). On the other hand, the delay in the demise of the Wnt system and the resolution of the primordium to form terminal neuromasts was accompanied by a failure to lose cxcr7b expression in the trailing zone at 48 hpf (Fig. 5U,X). This was accompanied by persistent migratory behavior of the morphant primordium at a time when the control primordium had terminated its migration and resolved to form terminal neuromasts (data not shown).

As shown previously in the context of morphogenesis and migration phenotypes, co-injection of snail1b mRNA in snail1b morphants prevented the delay in shrinking of the Wnt system and in the establishment of the trailing Fgf signaling system. Although 100 pg snail1b mRNA had no obvious effect on expression on its own, co-injection with snail1b TB morpholino restored timely restriction of lef1 expression to a leading zone and the establishment of pea3 and cxcr7b expression in the trailing zone by 26 hpf (Fig. S6A-H,M-P). Co-injection of snail1b mRNA in the morphants did not, however, prevent aberrant expression of cxcr4b in the trailing zone (Fig. S7I-L)

As expression of snail1b at the leading end of the primordium is determined by Cxcl12a-dependent chemokine signaling, we investigated whether loss of cxcl12a prevents snail1b expression and delays restriction of Wnt and establishment of the trailing Fgf signaling systems, as observed in snail1b morphants. Examination of lef1 showed that, whereas lef1 was restricted to the leading domain at 22 hpf in control embryos (Fig. 6A), lef1 expression remained broad in cxcl12a morphant primordia (Fig. 6B).

Furthermore, a small pea3 expression domain was established in cxcl12a morphants, without delay, at the trailing end of the prospective pLL primordium by 22 hpf (Fig. 6D,E). However, this pea3 expression domain was in the middle of a broad lef1 expression domain, which represents a pLL primordium and pLL ganglion that have failed to separate in the absence of effective collective migration (compare Fig. 6B and E). Although a small pea3 expression domain was established at the trailing end of the prospective pLL primordium in cxcl12a morphants, no additional Fgf signaling centers, marked by progressive expansion of pea3 expression, were established.

Although there were similar delays in shrinking of the Wnt system in snail1b morphants and cxcl12a morphants, in which snail1b expression in the primordium is indirectly lost, there were important differences. Whereas progressive restriction of the Wnt system was delayed in snail1b morphants, it was permanently blocked in cxcl12a morphants, in which caudal migration is not just delayed but permanently lost. This suggested that the severity of problems associated with the failure to progressively restrict the domain of Wnt activity correlated with the delay or complete failure of caudal migration, not specifically with loss of snail1b function. To determine whether loss of migratory behavior causes a delay in restriction of Wnt signaling and sequential formation of Fgf signaling-dependent protoneuromasts, we examined lef1 and pea3 expression in embryos in which collective migration was inhibited with a Src kinase inhibitor, PP1. Exposure of embryos to 30 µM PP1 at 18 hpf stalled migration of the primordium. As in snail1b and cxcl12a morphants, restriction of lef1 expression to the leading domain and establishment of expanded pea3 expression in the trailing domain was delayed (Fig. 6C,F). Nevertheless, in this context, chemokine-dependent expression of snail1b in leading cells was not affected (Fig. 6G-I), suggesting that the similar phenotype seen in cxcl12a morphants was not specifically dependent on either cxcl12a or chemokine signaling-dependent expression of snail1b. Instead, these observations suggest that the delay in sequential formation of Fgf signaling-dependent protoneuromasts is likely to be related, at least in part, to the delay in initiation of effective collective migration of the primordium.

Fgf signaling in the trailing part of the primordium determines the reorganization of sequential groups of cells as epithelial rosettes to form protoneuromasts, whereas Wnt signaling is thought to maintain the relatively mesenchymal morphology of leading cells, where it dominates (Aman and Piotrowski, 2008; Lecaudey et al., 2008; Nechiporuk and Raible, 2008). However, how Wnt signaling determines mesenchymal morphology was not clear. Our discovery that snail1b is expressed in leading cells led us to speculate that its expression might be determined by Wnt signaling and that, as a factor known for its role in promoting EMT (Barrallo-Gimeno and Nieto, 2005; Dang et al., 2011; Lamouille et al., 2014), it might determine mesenchymal morphology of leading cells. However, our analysis revealed that Wnt signaling does not determine snail1b expression in the leading zone. Instead, its expression in the leading zone is promoted by chemokine signals first encountered by leading cells of the primordium (Fig. 7A), whereas Fgf signaling prevents snail1b expression in trailing cells (Fig. 7B). The effect of knocking down snail1b function was also not what we expected. It did not compromise the ability of leading cells to adopt a mesenchymal morphology. Instead, sequential morphogenesis of epithelial rosettes associated with formation of protoneuromasts in the trailing domain was delayed in snail1b morphants (Fig. 7C). Our analysis revealed, however, that the delay in protoneuromast formation is not related to a specific role of Snail1b in morphogenesis of epithelial rosettes. Instead, it is indirectly related to the role of Snail1b in helping to initiate effective collective migration of the primordium, a role consistent with its previously described role in promoting cell movement (Barrallo-Gimeno and Nieto, 2005). Interestingly, we found that other manipulations that prevent effective primordium migration also cause a similar delay in sequential formation of Fgf signaling centers, associated protoneuromasts and shrinking of the leading Wnt system. These observations, together, reveal an unexpected role for collective migration of the primordium in kick-starting sequential formation of additional protoneuromasts. Finally, we showed that problems in initiating collective migration in the primordium may, at least in part, be related to aberrant expansion of epcam into the leading zone and/or the loss of cdh2 expression from the leading zone of the primordium following knockdown of snail1b function.

Horizontal myoseptum cells, located caudal to the recently formed primordium, secrete Cxcl12a and determine a migratory response in Cxcr4b-expressing cells at the caudal end of the primordium, to initiate and bias the initial direction of primordium migration (Dambly-Chaudière et al., 2007). This study now shows that, in addition to providing a directionally biased guidance cue to leading cells, Cxcl12a initiates snail1b expression in cells at the caudal end of the primordium. This helps to kick-start effective collective migration of the entire primordium. Though the initial exposure of the primordium to Cxcl12a is asymmetric, once the primordium is migrating over the horizontal myoseptum, along the path defined by cxcl12a expression, it is exposed at both its leading and trailing ends to Cxcl12. At this stage, establishment of Fgf signaling in the trailing end of the primordium helps to restrict snail1b expression to the leading end of the primordium, as loss of Fgf signaling allows snail1b expression to expand into the trailing zone.

The failure to initiate effective collective migration of the pLL primordium in snail1b morphants has additional consequences related to sequential formation of protoneuromasts and associated changes in Wnt and Fgf signaling in the primordium. snail1b morphants are associated with a delay in formation of an expanding Fgf signaling system at the trailing end of the pLL primordium and in the accompanying restriction of the initially broad Wnt signaling system to a progressively smaller leading zone in the primordium. Interestingly, inhibition of collective migration by other manipulations, including knockdown of cxcl12a or exposure to the Src kinase inhibitor PP1, caused a similar delay in progressive restriction of the Wnt system to a smaller leading zone and in establishment of additional Fgf signaling centers associated with formation of protoneuromasts. Although it remains unclear how sequential formation of additional protoneuromasts is related to collective migration, one simple possibility is that migratory behavior alters the shape of the primordium. In the absence of collective migration all the cells in the primordium tend to coalesce to form one or two relatively large unstable epithelial rosettes. Initiation of collective migration alters the shape of the primordium by stretching the initially large ellipsoid collection into a longer column of cells. In the context of self-organizing mechanisms that might impose a minimum distance between sequentially formed epithelial rosettes, stretching the primordium into a longer column following the effective initiation of collective migration may be all it takes to initiate formation of additional protoneuromasts. This hypothesis will be explored in future studies.

Understanding of the dynamics of Wnt and Fgf signaling in the pLL primordium has relied on examining genes like lef1 and pea3, expression of which is determined by Wnt and Fgf signaling, respectively. The discovery that snail1b expression is determined by cxcl12a-mediated chemokine signaling now provides an avenue for examining the pattern of chemokine signaling in the primordium. Previous studies suggested that polarized expression of chemokine-responsive Cxcr4b and chemokine-unresponsive Cxcr7b at leading and trailing ends of the primordium, respectively, determine a polarized response of the primordium to Cxcl12a. Furthermore, binding, internalization and degradation of Cxcl12a bound to the Cxcr7b receptor contributes to generation of a local gradient of chemokine availability, which facilitates unidirectional collective migration of the primordium. Consistent with this suggestion, chemokine signaling-dependent expression of snail1b is normally only observed in leading cells of the primordium. However, when cxcr7b is knocked down, the response to Cxcl12a is no longer polarized and snail1b expression is observed at both ends of the primordium.

A significant challenge we faced in this study is that we were unable to breed heterozygous fish with snail1b CRISPR/Cas9-mediated mutations. Consequently, we have had to rely on morphant phenotypes for our conclusions. There has been considerable debate about the strength of conclusions that are based on morphant studies (Eisen and Smith, 2008; Stainier et al., 2017). Nevertheless, there are many reasons to be confident of key conclusions made in this study. First, similar phenotypes were observed with independent splice-blocking and translation-blocking snail1b morpholinos. Second, delays in initiating migration and in sequential formation of protoneuromasts were suppressed or reversed by co-injecting morpholinos with low levels of snail1b mRNA. Similarly, changes in gene expression, which account for the observed deficits in migration and formation of protoneuromasts in the snail1b morphants, were reversed by co-injection of snail1b mRNA. Third, complementary changes in the pattern of gene expression were induced by loss-of-function and gain-of-function experiments. For example, snail1b knockdown resulted in epcam expanding into the leading zone, where snail1b is normally expressed, and ectopic expression of snail1b mRNA resulted in suppression of epcam expression. Together, these observations provide strong support to conclusions based on the analysis of snail1b morphants.

In conclusion, this study has provided a fresh perspective on events associated with formation of the pLL primordium, initiation of polarized caudal migration and sequential protoneuromast formation within the migrating primordium. In addition to initiating polarized migratory behavior of leading cells, the chemokine Cxcl12a triggers snail1b expression in cells at the leading end of the pLL primordium, which helps to initiate collective migration of the pLL primordium (Fig. 7A). Furthermore, prior to the initiation of effective collective migration and separation of the rostral pLL ganglion cells from caudal pLL primordium, there is broad Wnt signaling activity across the entire pLL placode. As a boundary between a rostral pLL ganglion and the caudal pLL primordium compartment forms within the pLL placode, an Fgf signaling center and the first protoneuromast is established at this boundary at the trailing end of the prospective pLL primordium (Fig. 7A). Initiation of collective migration then has an important impact on the subsequent dynamics of Wnt and Fgf signaling in the pLL primordium: as collective migration is initiated and the primordium moves away, leaving the pLL ganglion cells in its wake, additional Fgf signaling centers and protoneuromasts form in the pLL primordium. Furthermore, Fgf signaling in the trailing zone helps to restrict snail1b expression to the leading zone (Fig. 7B,C). Establishment of additional Fgf-dependent protoneuromasts is accompanied by progressive shrinking of the initially broad system of Wnt activity. In the absence of effective collective migration, additional protoneuromast do not form and the Wnt system remains broad and does not shrink in a timely manner (Fig. 7C, inset). We expect future studies to help define more clearly how Snail1b determines the timely initiation of collective migration and how this, in turn, facilitates sequential formation of protoneuromasts.

Danio rerio were maintained under standard conditions. Embryos were staged according to Kimmel et al. (1995). Tg[cldnb:lynGFP] (Haas and Gilmour, 2006), Tg[hsDKK1b] (Stoick-Cooper et al., 2007), Tg[hsp70l:dnfgfr1-EGFP]pd1 (Lee et al., 2005 ) and Tg[hsp:cxcl12a] (Knaut et al., 2005) were previously described. Heat shock was carried out at 37.5°C for 40 min. Embryos were incubated with 10 μm SU5402 (Tocris) or 5 μm BIO (Calbiochem) from 26 hpf until fixation, or with 30 μm PP1 (Sigma) from 18 hpf until fixation at 22 hpf. Drugs were made up in DMSO, and control embryos were incubated for similar periods in DMSO alone. Experiments were carried out in accordance with National Institute of Child Health and Human Development Animal Care and Use Committee Protocol Number 15-039.

All MOs were obtained from Gene Tools LLC. MOs used were (5′-3′): snail1b-MO1, TGCAAATATCCCCTCGAAACTTCAG (Blanco et al., 2007); snail1b-SB, CCGGTCTGGGAAGAGATCACTTAA; cxcl12a-MO1, TTGAGATCCATGTTTGCAGTGTGAA (Boldajipour et al., 2008); cxcr7b-MO2, TCATTCACGTTCACACTCATCTTGC (Dambly-Chaudière et al., 2007); epcam-MO, ACTAAAACCTTCATTGTGAGCGAGA (Villablanca et al., 2006). All the MOs were injected at the single-cell stage with co-injection of 1.5 ng of p53-MO (Robu et al., 2007). Standard control MO (CCTCTTACCTCAGTTACAAATTTATA; Gene Tools LLC) was injected at the same concentration as the test. Statistical significance was determined by unpaired Student's t-test.

Embryo were fixed in 4% paraformaldehyde (PFA) overnight and dehydrated in methanol for 1 h at −30°C. The embryos were rehydrated, treated with proteinase-K and re-fixed with 4% PFA. Pre-hybridization and hybridization were carried out at 67°C for snail1b and 65°C for rest of the probes. RNA probes were synthesized using a DIG labeling kit (Roche, 11175025910). Samples were hybridized overnight with RNA probes and color was developed using BCIP/NBT substrate (Roche, 11383221001, 11383213001). For immunohistochemistry, embryos were fixed in 4% PFA overnight and blocked in PBT (phosphate-buffered saline with 0.1% Tween 20) with 5% goat serum. Antibodies were incubated in the same buffer overnight with anti-GFP (Abcam, ab13970; 1:1000) and anti-ZO1 (Thermo Fisher Scientific, 33-9100; 1:500). Embryos were washed with PBT and incubated with respective secondary antibodies [goat anti-chicken-488 (Invitrogen, A-11039; 1:500) and goat anti-mouse-568 (Invitrogen, A-11031; 1:500)] and imaged on a Leica SP5 confocal microscope.

For time-lapse acquisition, Tg[cldnB:lynGFP] embryos were anesthetized at ∼20 hpf in 600 μm Tricaine (Sigma) and mounted in 0.8% low melting agarose (Nusieve GTG, Lonza) in a glass-bottomed chamber slides (Nunc). Movies were acquired until 48 hpf at intervals of 5 min using an inverted Leica SP5 confocal microscope with 10× lens. The data were analyzed using ImageJ (Rasband, 1997-2016).

We thank Angela Nieto for the snail1b mRNA clone. We thank Chitnis lab members for critical comments.