Zebrafish Cdx4 regulates neural crest cell specification and migratory behaviors in the posterior body

Abstract

The neural crest (NC) is a transient multipotent cell population that migrates extensively to produce a remarkable array of vertebrate cell types. NC cell specification progresses in an anterior to posterior fashion, resulting in distinct, axial-restricted subpopulations. The anterior-most, cranial, population of NC is specified as gastrulation concludes and neurulation begins, while more posterior populations become specified as the body elongates. The mechanisms that govern development of the more posterior NC cells remain incompletely understood. Here, we report a key role for zebrafish Cdx4, a homeodomain transcription factor, in the development of posterior NC cells. We demonstrate that cdx4 is expressed in trunk NC cell progenitors, directly binds NC cell-specific enhancers in the NC GRN, and regulates expression of the key NC development gene foxd3 in the posterior body. Moreover, cdx4 mutants show disruptions to the segmental pattern of trunk NC cell migration due to loss of normal leader/follower cell dynamics. Finally, using cell transplantation to generate chimeric specimens, we show that Cdx4 does not function in the paraxial mesoderm—the environment adjacent to which crest migrates—to influence migratory behaviors. We conclude that cdx4 plays a critical, and likely tissue autonomous, role in the establishment of trunk NC migratory behaviors. Together, our results indicate that cdx4 functions as an early NC specifier gene in the posterior body of zebrafish embryos.

Graphical Abstract

Introduction

The neural crest (NC) is a transient, multipotent cell population that gives rise to an array of cell types, including chondrocytes, sensory neurons, glia, and pigment cells (Le Douarin and Kalcheim, 1999). NC development proceeds in an anterior to posterior (AP) fashion and consists of induction and specification at the neural plate border (NPB), delamination from the neural epithelium, migration, and differentiation (Le Douarin and Dupin, 2018; Rocha et al., 2020b; Simões-Costa and Bronner, 2015). Careful analyses of early NC development in chick have revealed that NC induction and specification begin during gastrulation (Basch et al., 2006; Ezin et al., 2009) and continue concurrently with neurulation. Our current understanding of the molecular mechanisms that orchestrate NC induction and specification stems largely from studies carried out at gastrulation stages or shortly thereafter. However, much of the post-cranial body is generated over an extended period of time following the completion of gastrulation (Wilson et al., 2009). Precisely how NC fate is specified during posterior body outgrowth remains unknown.

The family of caudal-related homeodomain transcription factors, Cdx, are good candidates to orchestrate the specification and subsequent development of NC cells in the posterior body. Cdx factors are expressed in the posterior of the developing body and play a conserved role in axial elongation and patterning of embryos along the AP axis (Houle et al., 2003; Lohnes, 2003; Young and Deschamps, 2009). Cdx factors are necessary for spinal cord development and patterning in mouse (Metzis et al., 2018) and zebrafish embryos (Skromne et al., 2007). In addition, in mouse embryos Cdx factors directly activate key genes necessary for NC specification, including Pax3, Msx1, and Foxd3 (Sanchez-Ferras et al., 2016, 2014, 2012). Further, based on stem cell studies it has been hypothesized that Cdx factors are necessary for the establishment of neural plate border and NC identity in the posterior (Frith et al., 2018). Together, these previous studies led us to hypothesize a potential role for Cdx factors in the specification of zebrafish trunk NC cells.

In addition to a role in directing the early steps of NC cell specification, we also hypothesized that Cdx factors might regulate the characteristic properties of NC cells in the posterior body. Pioneering quail-chick chimera experiments revealed that the NC comprises distinct subpopulations along the AP axis (Le Douarin and Kalcheim, 1999). These populations exhibit important differences in cellular behaviors, differentiation potential, and underlying transcriptional networks (Rocha et al., 2020a). In particular, cranial and trunk NC cells exhibit distinct cellular behaviors during migration. Live imaging of NC cell migration in zebrafish embryos has revealed that cranial NC cells migrate in broad streams and dynamically rearrange without regard to their initial position. By contrast, trunk NC cells traveling along the medial pathway—between the neural tube and adjacent somites—migrate in single-cell chains and depend on a leader cell for directionality (Richardson et al., 2016). These differences may exist as a consequence of the unique challenges posed by the migratory environments encountered by cranial versus trunk NC cells. In the cranial region, zebrafish NC cells migrate ventrolaterally from the dorsal neurepithelium out into the adjacent pharyngeal arches (Schilling and Kimmel, 1994), largely passing through loosely packed mesenchymal cells. By contrast, in the trunk, zebrafish NC cells migrate along two distinct pathways: the medial pathway between the neural tube and the adjacent somites and—commencing at a slightly later stage—a lateral pathway between the somites and the overlying ectoderm (Raible et al., 1992; reviewed in Rocha et al., 2020). The NC cells on the medial pathway must squeeze through a narrow gap between two epithelial tissues: the neuroepithelium of the developing neural tube and the adjacent epithelialized somites. While we know navigation of trunk NC through this environment requires a specialized “leader” cell, the molecular mechanisms by which leader and follower identities are established are yet to be uncovered. Nevertheless, the work of Richardson and colleagues (2016) suggests that leader/follower dynamics represent a defining property of zebrafish trunk NC cells.

In this study, we investigated the role of zebrafish cdx4 in the development of NC cells in the posterior body. We show that cdx4 is expressed in premigratory trunk NC cells, as well as in foxd3-expressing cells in the tailbud. Our bioinformatics analysis of published Cdx4 ChIP-seq data (Paik et al., 2013) revealed that Cdx4 binds putative enhancers of key genes of the NC GRN. Moreover, Cdx4 binds at NC-specific enhancers of foxd3 and regulates its expression in the posterior body. In addition, we demonstrate that cdx4 is required for the establishment of trunk-specific migratory behaviors. In cdx4 mutants, the segmental patterns of trunk NC cell migration are disrupted, and cells fail to reach their ventral destinations. Using time-lapse microscopy to interrogate the details of trunk NC cell behavior, we reveal that disrupted trunk NC patterning in cdx4 mutants is likely a consequence of the loss of leader/follower dynamics. Finally, by generating chimeras via cell transplantation, we demonstrate that the defects in NC cell migration following loss of Cdx4 function are not due to its absence from the mesodermal cells that contribute to the somites past which the NC cells migrate. We conclude that cdx4 plays a critical, and likely tissue-intrinsic, role in the specification of trunk NC and its characteristic cellular behaviors during migration.

Results

Zebrafish cdx4 is expressed in NC cell progenitors

Previous studies have reported that zebrafish cdx4 is expressed exclusively in the presumptive posterior of the embryo during late gastrulation, becoming localized to the tailbud and spinal cord as segmentation and neurulation proceed (Davidson et al., 2003; Joly et al., 1992; Shimizu et al., 2005). The anterior limit of cdx4 expression in the neural tube corresponds to the boundary between the spinal cord and hindbrain, at the level of the third somite (Chang et al., 2016). Whether zebrafish cdx4 is expressed in the nascent trunk NC cells, which derive from the dorsal-most part of the developing spinal cord, has not previously been addressed. However, mouse Cdx1 is detected in premigratory and early migratory NC cells (Meyer and Gruss, 1993), and human Cdx2 is highly expressed throughout the differentiation of trunk NC cells from hPSCs (Frith et al., 2018; Gomez et al., 2019; Hackland et al., 2019). As NC cells derive from the neuroepithelium, where cdx4 is already known to be expressed, we hypothesized that zebrafish cdx4 transcripts are present in premigratory trunk NC cells.

To test this hypothesis, we simultaneously visualized the expression of cdx4 and a marker of premigratory NC cells, foxd3, using in situ hybridization chain reaction (HCR) (Choi et al., 2018). As expected, cdx4 RNA transcripts were detected in the neural tube and tailbud of 15 hpf embryos, with an anterior limit of expression adjacent to the third somite (Fig. 1A, arrow), as well in the tailbud and lateral-most regions of the developing somites, consistent with previous reports. In addition, foxd3 RNA transcripts were detected in premigratory NC cells located in the dorsal neural tube, as well as in the posterior somites and tailbud (Fig. 1B), as previously described (Odenthal and Nüsslein-Volhard, 1998). Notably, we found that cdx4 and foxd3 expression overlap in the dorsal spinal cord region (Fig. 1C). To evaluate whether cdx4 and foxd3 are co-expressed in individual cells, we visualized the transcripts in transverse sections. Our analysis of 5 μm optical sections confirmed that individual premigratory NC cells in the dorsal spinal cord do indeed express both cdx4 and foxd3 (Fig. 1A’–C’).

Figure 1. Zebrafish cdx4 is expressed in NC cells.

A-C. Lateral views of a 15 hpf embryo; A’-C’ are views of a single 5 μM optical section of a hand-cut transverse section from the trunk. A. cdx4 mRNA detected in the tailbud and spinal cord by HCR. The anterior limit of expression corresponds to the hindbrain/spinal cord boundary (arrow). A’. cdx4 expression in a 5 μM optical section from the trunk (line in A); n.t. indicates the neural tube, s. indicates the somite. B. foxd3 mRNA detected in NC cells along the AP axis, the posterior somites, and the tailbud. B’. foxd3 expression in the 5 μm optical section (line in B); arrowhead indicates foxd3 expression in the somitic mesoderm. C. co-detection of cdx4 and foxd3 at the dorsal neural tube, scale bar = 100 μm. C’. cdx4 and foxd3 co-expression in the premigratory trunk NC cells in the 5 μm optical section (line in C); scale bar = 25 μm. D-F. Dorsal view of the tailbud of a 12 hpf embryo. D. cdx4 mRNA transcripts in the tailbud detected by HCR. E. foxd3 mRNA transcripts in the tailbud detected by HCR. F. Merged image of A and B reveals co-expression of cdx4 and foxd3 transcripts at the tailbud. Dashed lines demarcate the outline of the neural plate border (inner line) and the yolk; scale bar = 100 μm. G, H. Lateral views of a 12 hpf embryo, scale bar = 100 μm; G cdx4 mRNA detected in the tailbud and spinal cord by HCR. The anterior limit of expression is indicated (arrow). H. foxd3 mRNA detected in NC cells along the entire AP axis and in the tailbud (arrowhead) by HCR. I. Dorsal view (as indicated by arrowhead in H) of a high magnification (40x objective) 0.75 μm single optical section of the tailbud of a 12 hpf embryo, colabeled with DAPI (grey), showing that cdx4 and foxd3 transcripts colocalize to single cells, scale bar = 25 μm.

In zebrafish embryos, foxd3 is also expressed in the developing tailbud (Lukoseviciute et al., 2018; Odenthal and Nüsslein-Volhard, 1998), in a region that—at least in amniotes—has been shown to contribute to the NC cells of the posterior body. As cdx4 transcripts have similarly been reported in the tailbud (Davidson et al., 2003; Joly et al., 1992; Shimizu et al., 2005) we assayed whether cdx4 and foxd3 are co-expressed in the tailbud of zebrafish embryos at 12 hours post fertilization (hpf). Dorsal views of the tailbud revealed that cdx4 transcripts are indeed detected in the foxd3-expressing region of the developing tailbud (Fig. 1D–F). As at 15 hpf, lateral views of 12 hpf embryos confirmed that cdx4 transcripts are localized posteriorly, to the trunk and tailbud (Fig. 1G), consistent with previous reports, whereas foxd3 transcripts extend along the full AP extent of the premigratory NC cell domain in the dorsal neural tube (Fig. 1H). To evaluate co-localization of cdx4 and foxd3 in the developing tailbud we co-labeled with the nuclear marker DAPI and imaged single 0.75 μm optical sections at high magnification. This analysis confirmed that individual cells of the tailbud do indeed express both cdx4 and foxd3 (Fig. 1I).

Our results establish that cdx4 RNA transcripts are present in foxd3-expressing cells of the tailbud and later in premigratory NC cells of the dorsal spinal cord. Notably, these findings are supported by RNA sequencing data on foxd3-expressing cells at various stages of development (Lukoseviciute et al., 2018), which concur that cdx4 is expressed in these cells at the stages during which NC specification is expected to take place. We did not, however, detect cdx4 RNA transcripts in migrating NC cells. This finding again corresponds with the available data from transcriptomic analysis of sox10-expressing migrating NC cells, which showed that cdx4 is not highly expressed in these cells (Lukoseviciute et al., 2018). We conclude that cdx4 is expressed at the right time and place to orchestrate the initial development of NC cells in the posterior body.

Cdx4 regulates the expression of NC specifier genes in the posterior body

The molecular underpinnings of NC development have been described in terms of a gene regulatory network (GRN) (Martik and Bronner, 2017; Meulemans and Bronner-Fraser, 2004; Sauka-Spengler and Bronner-Fraser, 2008; Simões-Costa and Bronner, 2015). This genetic circuit, which is organized as a series of interlinked modules, integrates cell-extrinsic signals and cell-intrinsic gene regulatory interactions. While the core components of the NC GRN are present in NC cells along the body, axial-specific elaborations have been shown to mediate the unique properties of spatially restricted subpopulations (Gandhi et al., 2020; Simões-Costa and Bronner, 2016; Soldatov et al., 2019). Thus, to investigate the role of cdx4 in the development of NC cells in the posterior body, we analyzed its function within the NC GRN.

As zebrafish cdx4 is expressed in NC cell progenitors at the dorsal aspect of the neural tube as well as in the developing tailbud, we hypothesized that it may function as a NC specifier in the posterior body. As such, we would expect Cdx4 to regulate the expression of other genes of the NC GRN. To evaluate this hypothesis we examined the binding activity of Cdx4 at genes of the NC GRN as described by Simões-Costa and Bronner (2015). In particular, we focused on the genes that comprise the following modules: induction, NPB (neural plate border), premigratory NC, and migratory NC. To achieve this, we analyzed a previously reported ChIP-seq dataset generated using Myc-tagged Cdx4 in 10 hpf zebrafish embryos (Paik et al., 2013), a developmental stage that marks the end of gastrulation proper (Kimmel et al., 1995). To uncover putative cis-regulatory regions, we assigned the 4,965 Cdx4-binding regions identified by Paik et al. (2013) to 2,407 zebrafish genes using GREAT (McLean et al., 2010).

To identify biologically-relevant binding sites, we incorporated previously reported ATAC-seq data (which identifies regions of open chromatin) from foxd3-expressing cells in 12 hpf embryos (Lukoseviciute et al., 2018). In addition, because the NC GRN exhibits extensive cooperation between its different components, we compared the targets of Cdx4 binding with those of Foxd3. To achieve this, we analyzed previously reported data from ChIP-seq of Foxd3 in 10 hpf and 12 hpf zebrafish embryos (Lukoseviciute et al., 2018). Using GREAT, we assigned 531 Foxd3-bound regions at 10 hpf and 2,955 regions at 12 hpf to 428 genes and 789 genes, respectively.

We analyzed 69 unique zebrafish genes comprising the orthologs of genes belonging to the four relevant modules of the NC GRN (see Supplemental Table 1 for a complete list). Of these 69 genes, 14 were bound by both Cdx4 and Foxd3 at 10 hpf and 28 were bound by both Cdx4 at 10 hpf and Foxd3 at 12 hpf (Fig. 2A; Supplemental Table 1). It should be noted that there are two important limitations of this analysis. First, it uses two distinct datasets from ChIP-seq performed at slightly different developmental stages. Second, these data are derived from populations of cells rather than single cells. Therefore, we cannot definitively state that both factors bind simultaneously at these putative enhancers in any individual cell. Nevertheless, the data do reveal that Cdx4 and Foxd3 share target genes within the NC GRN and suggest that they may co-bind within the loci of multiple genes.

Figure 2. Cdx4 and FoxD3 bind near early NC cell specification genes.

A. Schematic representation showing number of genes of each zebrafish NC GRN module bound by both Cdx4 at 10 hpf and Foxd3 (dark circle) at 10 and 12 hpf out of the total number of genes in the module (light circle). B. Schematic representation of Cdx4 and Foxd3 binding near genes of the NPB module (left) and premigratory module (right). C-F. Analysis of Cdx4 and Foxd3 binding near the NPB genes msx1a (C), pax3a (D), prdm1a (E), and tfap2a (F), as well as the premigratory and migratory NC genes sox9b (G), snai1b (H), id2a (I), and sox10 (J). In all panels, the plots are laid out as follows (from top to bottom): ATAC-seq profiles from foxd3-expressing cells, with called peaks below, Cdx4-Myc ChIP-seq peaks at 10 hpf, Foxd3-biotin-ChIP-seq peaks at 10 hpf, and 12 hpf, and Ref-seq annotated genes.

Notably, half of the genes (13 out of 26) that comprise the neural plate border module (Simões-Costa and Bronner, 2015) were bound by both Cdx4 and Foxd3 at one or both of the stages analyzed. We therefore examined the binding activity of Cdx4 and Foxd3 near candidate NPB genes—msx1a, pax3a, prdm1a, and tfap2a (Fig. 2C–F)—and observed that the two transcription factors often bind the same putative cis-regulatory regions (Fig. 2B). Shared binding sites were uncovered at both proximal and more distal regions of open chromatin, and sometimes within intronic regions. We also observed binding of both Cdx4 and Foxd3 near genes of the premigratory NC module, including sox9b, snai1b, and id2a (Fig. 2G–I). However, in these genes the two factors do not occupy the same putative cis-regulatory regions (Fig. 2B). Finally, the later-acting NC specifier gene sox10 was not bound by either Cdx4 or Foxd3 (Fig. 2J). We provide a comprehensive summary of the binding site data in Supplemental Table 1.

Foxd3 is a central component of the NC cell GRN (Simões-Costa and Bronner, 2015) and functions as a key specifier of NC cell fate (Hromas et al., 1999; Kos et al., 2001; Lister et al., 2006; Montero-Balaguer et al., 2006; Stewart et al., 2006; Teng et al., 2008). Given the importance of Foxd3 in priming the initial stages of NC cell specification, we elected to conduct a more detailed analysis of the foxd3 locus, as its regulatory logic in zebrafish has recently been described (Lukoseviciute et al., 2018). Our analysis revealed that Cdx4 binds four regions of open chromatin near the foxd3 locus (Fig. 3A). Of note, three of these enhancers drive reporter expression in NC cells (Lukoseviciute et al., 2018). Moreover, all four of these regions have previously been reported as targets of Foxd3 binding (Lukoseviciute et al., 2018), and are thus potential sites through which Foxd3 auto-regulates its expression (Hromas et al., 1999; Lister et al., 2006; Lukoseviciute et al., 2018; Pohl and Knöchel, 2001). Finally, our zebrafish results are consistent with a report that mouse Cdx factors activate the expression of a Foxd3 luciferase reporter in vitro (Sanchez-Ferras et al., 2016).

Figure 3. Cdx4 regulates posterior expression of foxd3.

A. Analysis of Cdx4 and Foxd3 binding near the foxd3 locus. B-E. Dorsal view of 15 hpf WT or cdx4^ch107+/− heterozygous embryo shows foxd3 RNA transcripts detected by HCR in the dorsal spinal cord and tailbud (B) counterstained with DAPI (C). D. Merged image of B and C; arrow indicates tailbud, arrowhead labels neural tube. Scale bar = 100 μm. E. Magnified view of tailbud region from D; scale bar = 50 μm. F-I. Dorsal view of 15 hpf cdx4^ch107−/− mutant embryo shows foxd3 RNA transcripts detected by HCR in the dorsal spinal cord and lack of expression in the tailbud (F) counterstained with DAPI (G). H. Merged image of F and G; arrow indicates tailbud, arrowhead labels neural tube. Scale bar = 100 μm. I. Magnified view of tailbud region from H; scale bar = 50 μm.

Having established that Cdx4 binds the putative NC enhancers of foxd3, we then investigated whether cdx4 is necessary for its expression in the posterior body. To facilitate this analysis, we used CRISPR/Cas9 to generate a new mutant in the cdx4 gene, cdx4^ch107. This allele is characterized by an insertion/deletion after codon 28 that results in a frameshift to create a premature stop at codon 102. Importantly, this mutation yields a truncated protein that lacks the homeodomain and is thus a predicted null (Supplemental Figure 1). As expected, cdx4^ch107−/− mutant embryos closely resemble those with previously described point or insertional mutations in the cdx4 gene: all are characterized by a posterior truncation accompanied by lack of the normal hind yolk extension leading to a rounded (‘kugelig’) shape (e.g. cdx4^{tl240a/kugelig} (Davidson et al., 2003), cdx4^hi2188a (Golling et al., 2002) and cdx4^ch107 (Supplemental Figure 1)). By 15 hpf the cdx4^ch107−/− phenotype can be unambiguously identified based on visual inspection of gross morphology. We therefore assayed foxd3 expression at this stage using HCR on the progeny of crosses between cdx4^ch107+/− heterozygous fish. Notably, the morphology and gene expression of heterozygous siblings were indistinguishable from wild type (WT) specimens. In WT or heterozygous sibling embryos foxd3 mRNA was detected in both the tailbud and in premigratory NC cells at the dorsal aspect of the neural tube (Fig. 3B–E) (n=16). By contrast, in cdx4^ch107−/− mutant embryos foxd3 expression was reduced in the dorsal neural tube and was undetectable in the developing tailbud (Fig. 1F–I) (n= 16). We conclude that zebrafish Cdx4 is necessary for the early tailbud expression of foxd3.

Together, these data demonstrate that Cdx4 directly regulates the posterior expression of foxd3 and are consistent with Cdx4 playing a broader role in the regulation of the early modules of the NC GRN. This finding is especially notable given that mouse Cdx2 has been hypothesized to function as a ‘pioneer factor’ (Amin et al., 2016; Neijts et al., 2016), a class of transcription factor capable of binding to targets in silenced regions of chromatin and establishing competence for transcription (Zaret and Mango, 2016). Our results suggest the possibility that zebrafish Cdx4 may similarly act as a pioneer factor to initiate the expression of foxd3 in the tailbud and subsequently in the dorsal spinal cord, a tissue that derives from the tailbud during axial outgrowth (Kanki and Ho, 1997). However, the presence of foxd3 transcripts in the premigratory NC cells at the dorsal aspect of the neural tube of cdx4^ch107−/− mutants, albeit at reduced levels, implies that early tailbud expression of foxd3 can be bypassed, and foxd3 expression can eventually be activated by other means.

Cdx4 function is necessary for trunk NC cell segmental migration

One of the more remarkable differences between cranial and trunk NC cells is their distinct cellular behaviors during migration. In the trunk—but not in the cranial region—leader cells are necessary for the directed migration of NC cells traveling in segmentally arrayed chains along the medial pathway (Richardson et al., 2016). Importantly, leader and follower identities are established prior to the onset of migration and remain fixed throughout (Richardson et al., 2016). Cdx factors have previously been shown to promote posterior identity in the developing neural ectoderm (Skromne et al., 2007) and we now know that cdx4 is expressed in premigratory trunk NC cells. Therefore, we hypothesized that cdx4 might be necessary for the establishment of trunk-specific cellular behaviors during NC cell migration.

To investigate the role of cdx4 in trunk NC migration, we visualized the location of trunk NC cells in the progeny of crosses between cdx4^ch107+/− heterozygous fish. The embryos were co-immunolabeled with antibodies against Sox10, a NC marker, and Myosin Heavy Chain (MHC), a somitic muscle cell marker. As expected, in WT and cdx4^ch107+/− heterozygous sibling control specimens, trunk NC cells migrated in segmental chains between the neural tube and somites (Fig. 4A–C, A’–C’). Notably, a single cell was present at the migratory front of each segmentally-organized chain of NC cells (Fig. 4A–A’, C–C’; arrows). By contrast, in cdx4^ch107−/− mutant embryos, the segmental distribution of trunk NC cells was disrupted; specifically, trunk NC cells were not confined to single cell-wide segmental chains and failed to reach ventral locations (Fig. 4D–F, D’–F’). Of note, the ventral-most position reached by the majority of trunk NC cells in mutant embryos is adjacent to the myoseptum that divides the dorsal and ventral compartments of the somite, which also corresponds with the boundary between the neural tube and the notochord.

Figure 4. cdx4 mutants show impaired segmental migration of trunk NC cells.

A-C. Lateral view of 24 hpf WT embryo labeled with anti-Sox10 (A, NC cells, magenta) and anti-Myosin Heavy Chain (B, somites, green) antibodies reveal segmental migration pathways of trunk NC cells, scale bar =100 μm. A’-C’. High magnification images of boxed area from A-C (arrows = individual segmental chains), scale bar = 50 μm. D-F. Lateral view of 24 hpf cdx4^ch107−/− embryo labeled with anti-Sox10 (D, NC cells, magenta) and anti-Myosin Heavy Chain (E, somites, green) antibodies reveal defects in trunk NC cell migration. Scale bar = 100 μm. D’-F’. High magnification images of boxed area from D-F, scale bar = 50 μm. G. Schematic representing approach for mapping the position of trunk NC cells relative to the adjacent somite; analysis was performed at the level of somites 5–10. H, I. Plots of migrating trunk NC cells relative to the adjacent somite in WT embryos (H) and cdx4^ch107−/− embryos (I). Data from 18 total segments were superimposed on each plot.

To analyze this phenotype in more depth, we plotted the location of migrating NC cells relative to the adjacent somite, focusing our analysis at the level of somites 5–10 (schematized in Fig. 4G). In WT and cdx4^ch107+/− heterozygous sibling controls (n= 3 embryos, 18 segments, 120 cells), the dorsal-most NC cells exhibited a broad distribution along the AP axis. As the NC cells migrate ventrally and approach the myoseptum (indicated by 0 on the DV axis), their AP locations become progressively more restricted, with NC cells positioned close to the center of each somite. Finally, cells migrating adjacent to the ventral portion of the somite tend to remain within a narrow spatial domain (Fig. 4H). In cdx4^ch107−/− mutants (n= 3 embryos, 18 segments, 129 cells), however, NC cell locations are not restricted to the center of the somites, failing to converge towards the center as they near the myoseptum. Further, we found only two NC cells located adjacent to the ventral portion of the somite (Fig. 4I). These results indicate that cdx4 is necessary for normal segmental migration of trunk NC cells along the medial pathway.

Loss of cdx4 disrupts leader/follower NC cell dynamics

As leader and follower dynamics are necessary for directed migration of trunk NC cells, we hypothesized that the defect in trunk NC cell migration that we observed in cdx4^ch107−/− mutants might be due to defects in establishing these two cell types. This model was further supported by our observation that by contrast to control specimens—which had a single leader cell at the front of each segmental chain (Fig. 4A, A’) (Richardson et al., 2016)—we detected multiple NC cells in the ventral-most positions of cdx4^ch107−/− mutants (Fig. 4D, D’). Therefore, we investigated whether cdx4 is necessary for the establishment of leader and follower cell identities in trunk NC cells.

To investigate the role of cdx4 in establishing leader/follow cell dynamics we visualized the migration of trunk NC cells in live embryos by single plane illumination microscopy (SPIM). For this analysis we generated a line of cdx4^ch107+/− heterozygous zebrafish that also carries the Tg(sox10:mRFP) transgene, which labels NC cells with membrane-localized RFP. In otherwise WT Tg(sox10:mRFP) embryos, trunk NC cells migrate in single-cell chains and the first cell to migrate—the leader—retains its position at the front of the chain throughout migration (Supplemental Movie 1). By tracking the position of leader cells during the migration from the premigratory region to a ventral position adjacent to the myoseptum (n = 17 cells, 3 embryos), we found that none of the leaders were overtaken by a follower cell during this period (Fig. 5A–C), consistent with previously reported results (Richardson, 2016). In cdx4^ch107−/− mutants, trunk NC cells also migrated ventrally from their site of origin but were unable to continue their migration past a position adjacent to the myoseptum (Supplemental Movie 2). In addition, we found that migrating cdx4^ch107−/− trunk NC cells failed to converge into a single-cell stream, instead migrating in broader AP domains, consistent with our analysis of fixed specimens (Fig. 5D–F). By tracking the first cell to migrate (n = 15, 3 embryos), we found that 12 (80%) of these were overtaken by a cell that began its migration later. Of these 12 cells, 10 were overtaken before reaching a position adjacent to the myoseptum and 2 were overtaken after stalling at this location. Further, while leader cells in WT embryos migrate alone at the front of each chain, we found that 8 (53%) of the cells that migrated first in cdx4^ch107−/− mutants migrated alongside another cell for at least 30 minutes. We conclude that cdx4 is necessary for the establishment of leader and follower identities.

Figure 5. cdx4 mutants exhibit aberrant cell behaviors during migration.

A-J. Timelapse microscopy of trunk NC cell migration in Tg(sox10:mRFP) embryos visualized by SPIM reveals cellular behaviors. A-C. In WT embryos, NC cells migrate as single-cell segmental chains. Tracking of leader and follower cells during migration shows that leader cells are not overtaken. Scale bar = 25 μm. D-F. In cdx4^ch107−/− embryos tracking of NC cells shows that the first cells to migrate are overtaken by later-migrating cells. Scale bar = 25 μm. G, H. High magnification view in a WT specimen of the leader cell (G) (blue arrow) and follower cell (H) (orange arrow). Scale bar = 5 μm. I, J. High magnification view in a cdx4^ch107−/− mutant specimen of the first cell to migrate (I) (blue arrow) and a later-migrating cell (J) (orange arrow). Scale bar = 5 μm. K, L. Box plots of the area (K, units = μm²) and aspect ratio (L) of migrating NC cells in WT and cdx4^ch107−/− embryos.

To understand the role of cdx4 in orchestrating trunk-specific behaviors during NC cell migration more fully, we analyzed whether loss of cdx4 disrupted the establishment of leader identity, follower identity, or both. To address this question, we measured the size and shape of migrating trunk NC cells from our in vivo time lapse analyses (Fig. 5G–J). Previously, Richardson and colleagues (Richardson et al., 2016) reported that leader cells are larger than follower cells and are polarized in the direction of migration. Consistent with this previous report, we found that in WT embryos the leader cells had a larger area and aspect ratio (n = 7, median area = 328 μm², median aspect ratio = 3.12) than follower cells (n = 7, median area = 173 μm², median aspect ratio = 1.82) (p = 0.011 and 0.0024, respectively). In cdx4^ch107−/− mutants, however, the first cell to migrate (n = 9, median area = 191 μm², median aspect ratio = 2.41) did not exhibit significant differences in area (p = 0.61) or aspect ratio (p = 0.93) compared to cells that migrated later (n = 8, median area = 167 μm², median aspect ratio = 2.32). In summary, by comparing cells from cdx4^ch107−/− mutants with WTs, we found that both the first mutant NC cells to migrate and those that followed were significantly different from WT leaders, but not from WT followers, in terms of both their area and aspect ratio (Fig. 5K–L).

Cdx4 function is not required in the somites for proper trunk NC cell migration

In zebrafish embryos, the migratory behaviors of trunk NC cells that migrate along the medial pathway are influenced by both NC cell-extrinsic and -intrinsic interactions. Specifically, the segmental pattern of trunk NC cell migration in regulated by the slow muscle precursors—adaxial cells—in the adjacent somites (Honjo and Eisen, 2005), possibly via signaling interactions mediated by MuSK and its ligand Wnt11 (Banerjee et al., 2011). On the other hand, cell-cell contacts between NC cells are equally critical for directing cell migration. When these contacts are disrupted, either by laser ablation (Richardson et al., 2016) or loss of the proto-cadherin encoding gene pcdh10a (Williams et al., 2018), migration along the medial pathways is halted.

Given that cdx4 expression is not restricted to trunk NC cells, we considered the possibility that the defects we observed in cdx4^ch107−/− trunk NC migration might result from Cdx4 action beyond the NC. In addition to expression in the spinal cord, cdx4 RNA transcripts are also present in the margin during gastrulation stages (Davidson et al., 2003; Joly et al., 1992) and in the tailbud (Fig. 1A) during segmentation stages, both of which contribute to the developing somites. Further, loss of Cdx function has previously been shown to impact mesoderm development, including the position and development of the pronephros (Wingert et al., 2007) and lateral plate mesoderm derivatives (Quintanilla and Ho, 2020). Consistent with previous reports, we find that the shape and size of the paraxial mesoderm-derived somites is altered in cdx4^ch107−/− mutants ((Davidson et al., 2003) and see Fig. 4E, E’). As NC cells migrate between the somites and the neural tube (Raible et al., 1992), and somite cells are in turn known to play an important role in regulating NC migration (Banerjee et al., 2013, 2011; Honjo and Eisen, 2005), it is possible that loss of Cdx4 function in the mesoderm could disrupt the environment through which trunk NC cells migrate, and thus disturb their behavior in a non-cell-autonomous manner.

To test the possibility that Cdx4 function in the paraxial mesoderm-derived somites is necessary to allow normal NC cell migration, we generated chimeric specimens via a cell transplantation approach (Ho and Kane, 1990). We elected to use a morpholino knockdown approach for these experiments because the transplants were performed prior to the stage at which mutants can be phenotypically recognized; the knockdown approach ensured Cdx4-deficiency in all experimental specimens. Importantly, Cdx4 MO-injected embryos fully recapitulate the phenotype of cdx4 mutants, and have been extensively used in previous studies (Chang et al., 2016; Davidson et al., 2003; Davidson and Zon, 2006; Hayward et al., 2015; Paik et al., 2013; Quintanilla and Ho, 2020; Shimizu et al., 2006, 2005; Skromne et al., 2007). Donor embryos were also labeled with 10 kDa lysinated Alexa Fluor 647 dextran, to enable tracking of donor-derived cells within the chimeras. The host embryos were double transgenics, carrying both the Tg(sox10:mRFP) transgene to label NC, and the Tg(His-GFP) transgene to label all nuclei.

We first generated chimeras in which Cdx4 function is absent from the somites but intact in the NC cells. Specifically, cells from blastula-stage (4 hpf) embryos injected with morpholinos targeted against cdx4 (Cdx4 MO) were transplanted into WT blastula-stage hosts (Fig. 6A). By transplanting Cdx4-deficient donor cells into the margin, the region of the fate map that contributes to paraxial mesoderm, we were able to reliably generate chimeric embryos in which a subset of somite cells was Cdx4-deficient, while the NC cells remained WT. We selected nine specimens in which donor-derived cells contributed to >20% (range = 21–78%, average = 47%) of the dorsoventral extent of three to six of the somites, providing a total of 37 segments for further analysis. Moreover, to ensure consistency with the analysis of cdx4^ch107−/− NC migration defects presented in Fig. 4, we limited our analysis to the somite 4–10 level. The chimeric embryos were analyzed at 25 hpf and appeared morphologically normal (Fig. 6C). Despite the presence of Cdx4-deficient cells in the somites, including 12 individual segments in which donor-derived cells contributed to >60% of the somitic area, trunk NC cell migration was always unaffected as indicated by the normal organization of segmental NC streams (Fig. 6B–E, B’–E’) (n= 9 embryos, 37 segments). Importantly, previous reports have implicated the adaxial mesoderm, which later contributes to the myoseptum, as a potential cue for NC migration (Honjo and Eisen, 2005). However, 10 (27%) of the segments we analyzed included a donor-derived Cdx4-deficient contribution to the myoseptum (e.g. Fig. 6D, arrowhead), confirming that even when Cdx4-deficient cells populate this important domain, NC cell migration remains normal. We conclude that loss of Cdx4 function in the mesoderm cells that generate the somites does not disrupt trunk NC cell migration.

Figure 6. Cdx4 function is not necessary in the somitic mesoderm.

A. Schematic of the transplantation approach to generate chimeric embryos with Cdx4-deficient cells in the somites. B-E. Lateral image of NC migratory streams (magenta), cell nuclei (green) showing the overall morphology of a 25 hpf chimeric embryo with Cdx4-deficient cells (cyan) in the somites. Scale bar = 150 μm. Arrowhead indicates donor cells at the myoseptum. B’-E’. Higher magnification images of boxed region from B-E. Scale bar = 50 μm. F. Schematic of the transplantation approach to generate chimeric Cdx4-deficient embryos with WT cells in the somites. G-J. Lateral image of trunk NC cells (magenta), cell nuclei (green) showing the overall morphology of a 25 hpf chimeric Cdx4-deficient embryo with WT cells (cyan) in the somites. Scale bar = 150 μm. G’-J’. Higher magnification images of boxed region from G-J. Scale bar = 50 μm.

In reciprocal experiments, we generated chimeras in which Cdx4 function was exclusively limited to the somites. To this end, labeled cells from WT blastula-stage donor embryos were transplanted into the margin of blastula-stage, Cdx4 MO-injected hosts (Fig. 6F). As expected, the chimeric embryos exhibited a shortened tail and aberrant morphology, consistent with the overall mutant phenotype of the host embryos (Fig. 6H). In this case, we selected five specimens in which donor-derived cells contributed to >20% (range = 21–80%, average = 45%) or more of the dorsoventral extent of one to seven of the somites, providing a total of 22 segments for further analysis. Once again, to ensure consistency with the analysis of cdx4^ch107−/− NC migration defects presented in Fig. 4, we limited our analysis to the somite 4–10 level. Although donor-derived cells, in which Cdx4 function was intact, were present in these somites, including 6 individual segments in which donor-derived cells contributed to >60% of the somitic area, trunk NC cell migration remained aberrant in all cases (n = 5 embryos, 22 segments) (Fig. 6G–J, G’–J’). Trunk NC cells were found adjacent to the ventral half of the somite in only a single segment, and even in this case the NC did not form a typical WT stream. We conclude that restoring Cdx4 function in the cells that generate the somites is not sufficient to rescue the defects in trunk NC migration observed in Cdx4-deficient embryos.

In summary, based on our transplantation results, we conclude that Cdx4 function is not required in somitic cells for proper segmental migration of trunk NC cells. While these data do not directly address whether cdx4 acts cell-autonomously in NC cells, the results are nevertheless fully consistent with that hypothesis.

Discussion

NC cell development is a multi-step process that we understand to be governed by a GRN comprised of nested modules. However, this framework of the gene-regulatory basis of NC development is largely derived from the analysis of cranial NC cells. Because cranial and post-cranial tissues have distinct developmental origins, a more detailed investigation of the molecular mechanisms that regulate NC cell development in the posterior body is needed. Here, we have presented evidence for the role of cdx4 in regulating the development of NC cells in the posterior body of zebrafish embryos. cdx4 is expressed in NC cell progenitors in the trunk and tail and directly regulates the expression of genes necessary for the early steps of NC cell development. In addition, cdx4 is necessary for the establishment of leader/follower behaviors that drive the segmental migration of trunk NC cells. We conclude that Cdx4 functions as an early NC specifier in the posterior of the developing embryo.

cdx4 functions as a NC specifier in the posterior body

We have demonstrated that cdx4 is expressed at the appropriate time and place to act as a NC specifier. cdx4 is expressed in foxd3-expressing premigratory NC cells in the trunk located at the dorsal aspect of the neural tube. In addition, cdx4 is also expressed in the developing tailbud, including a region that contains foxd3-expressing cells. While the expression of cdx4 is not limited exclusively to prospective NC cells, these expression patterns are consistent with those observed for other early NC specifier genes. Moreover, we did not detect cdx4 transcripts in migrating NC cells, which parallels the published descriptions of other early NC specifiers (Khudyakov and Bronner-Fraser, 2009).

The early NC specifier genes function to promote NC cell fate in a region between the neural plate and the non-neural ectoderm termed the neural plate border (NPB). However, the NPB does not only give rise to NC cells, but also contributes to placodal cells in the cranial region, epidermis, and the dorsal neural tube (Pla and Monsoro-Burq, 2018; Thawani and Groves, 2020). During gastrulation and neurulation, the NPB border emerges as a distinct territory defined by the overlapping expression of several transcription factors (Pla and Monsoro-Burq, 2018), which include the early NC specifier genes that make up the NPB module of the NC GRN (Simões-Costa and Bronner, 2015). Critically, the expression of most of these genes, in a variety of species, often extends beyond the NPB region (Pla and Monsoro-Burq, 2018), as discussed in more detail below.

Studies in Xenopus embryos were the first to determine that the transcription factors Msx1 and Pax3 integrate FGF, WNT, and BMP signals to promote NC specification at the NPB (Monsoro-Burq et al., 2005). However, while expression of Pax3 is restricted to the NPB region, expression of Msx1 extends all the way from the ventral side of the embryo to the edge of the neural plate (Monsoro-Burq et al., 2005). In chick embryos, similar expression of Msx1 in the ventral non-neural ectoderm as well as the NPB has been described (Khudyakov and Bronner-Fraser, 2009), whereas Pax3 and Pax7 are largely restricted to the NPB during gastrulation and early neurulation (Basch et al., 2006; Khudyakov and Bronner-Fraser, 2009), although they are later present at the dorsal neural tube (Otto et al., 2006). In zebrafish, msx genes are largely expressed in bilateral stripes at the edges of the neural plate at the beginning of neurulation (Phillips et al., 2006), whereas pax3 and pax7 exhibit a broader expression domain, which extends into the neural plate and later the dorsal neural keel (Seo et al., 1998). Another key specifier of NC cell fate is the transcription factor Tfap2a (De Crozé et al., 2011; Luo et al., 2003). Recent genomics analysis in chick embryos has revealed that Tfap2a heterodimerization with the paralogous Tfap2c and Tfap2b drives the genetic programs that orchestrate NPB formation and NC specification, respectively (Rothstein and Simoes-Costa, 2020). Yet, Tfap2a expression initially extends beyond the NPB and into the non-neural ectoderm and only later becomes restricted to the neural folds (Rothstein and Simoes-Costa, 2020).

These examples illustrate that the NPB border does not represent a single cell population defined by a restricted set of marker genes. Rather, the NPB is a heterogeneous region where a suite of transcription factors, with both restricted and more expansive expression domains, overlap. Within this region, feed forward interactions between these transcription factors stabilize and refine the NPB domain (Pla and Monsoro-Burq, 2018) and drive the expression of subsequent components of the NC GRN (Simões-Costa and Bronner, 2015). We conclude, therefore, that although the expression of cdx4 is not limited to the neural plate border, nor later to the dorsal neural tube where NC cells are specified, its expression domain is nevertheless consistent with a role as an early NC specifier.

NC specifier genes are defined functionally by their role in regulating the expression of other genes involved in the establishment of NC cell fate (Prasad et al., 2019). Consistent with this model, our results indicate that Cdx4 does indeed regulate the expression of early NC specifier genes, namely those of the NPB module. Specifically, we have shown that Cdx4 binds at four regions of open chromatin near the foxd3 locus that have previously been demonstrated to act as NC-specific enhancers (Lukoseviciute et al., 2018). Moreover, in cdx4^ch107−/− mutants, foxd3 expression is not detected in the tailbud and is also downregulated at the dorsal neural tube. Together, these results suggest that cdx4 is necessary for normal expression of foxd3 in the posterior body.

The finding that Cdx4 regulates the expression of foxd3 parallels results following loss of other early NC specifiers. In zebrafish, Prdm1a directly binds to an enhancer of foxd3 to activate its expression (Powell et al., 2013), and in prdm1 mutants, expression of foxd3 is downregulated leading to a reduction in the number of NC cells (Hernandez-Lagunas et al., 2005). Moreover, injection of foxd3 mRNA rescues defects in NC cell development caused by morpholino knockdown of Prdm1a. Similarly, injection of morpholinos targeted against tfap2a and tfap2c into zebrafish embryos eliminates the expression of foxd3 (Li and Cornell, 2007), suggesting that these genes also function upstream of foxd3. Finally, the regulation of foxd3 by early NC specifiers has been well-documented in other vertebrate model systems, as summarized by Prasad et al. (2019) and Simões-Costa and Bronner (2015). Thus, the regulation of foxd3 expression in the posterior body by Cdx4 represents a hallmark of early NC specifier gene function.

Our analysis of previously reported ChIP-seq datasets suggests that Cdx4 and Foxd3 both bind near 13 out of 26 genes that make up the NPB module. Our detailed examination of these loci revealed that these two transcription factors may bind at the same putative enhancers. Based on these results, we speculate that Foxd3 and Cdx4 may function co-operatively to activate the expression of other NC specifier genes. In addition, both Cdx4 and Foxd3 bind near 11 out of 29 genes of the premigratory NC module, although generally at different regions of open chromatin. While possible regulatory interactions between Cdx4 and Foxd3 have yet to be experimentally evaluated, this reflects a common trend observed in the NC GRN. For example, dissection of the regulatory logic of Foxd3 (Simões-Costa et al., 2012) and Sox10 (Betancur et al., 2010) in chick embryos, has revealed that these depend on interactions between other early NC specifier genes. More recently, genomic and epigenomic approaches used to investigate the role of Foxd3 in zebrafish embryos revealed that transcription factor binding motifs for other NC specifier genes, including those of the Sox and Pax families, are highly enriched at Foxd3-bound genomic regions (Lukoseviciute et al., 2018). Thus, co-regulation between the transcription factors that make up the NC GRN, as proposed here between Cdx4 and Foxd3, appears to be a common feature of this genetic circuit.

A model in which Cdx factors function to specify NC in the posterior has been previously postulated based on results from hPSC differentiation into trunk NC cells. Specifically, in this in vitro system, CDX2 is expressed at high levels during the generation of trunk NC cells (Frith et al., 2018; Gomez et al., 2019; Hackland et al., 2019). Consistent with this model, mouse Cdx factors have been shown to function within the trunk NC GRN. Sanchez-Ferras and colleagues (2016) found that conditional expression of a dominant-negative Cdx protein in the murine neural tube and premigratory NC cells caused defects in pigmentation and enteric nervous system development (Sanchez-Ferras et al., 2016), characteristic of impaired NC cell development. Further, the authors showed that mouse Cdx factors directly regulate the expression of other NC specifier genes, namely FoxD3, Msx1, and Pax3 (Sanchez-Ferras et al., 2016). These findings, in addition to a previous report that mouse Cdx promotes the expression of Pax3 at the neural plate border downstream of Wnt signaling (Sanchez-Ferras et al., 2012), together with the NPB transcription factor Zic2 (Sanchez-Ferras et al., 2014), strongly suggest that Cdx factors play an essential role in mediating the early regulatory events that result in NC cell specification across distant vertebrates.

cdx4 is necessary for trunk NC cell migration behaviors

We have also uncovered a second important role for cdx4, in the regulation of trunk NC migration. Specifically, we demonstrated that cdx4^ch107−/− mutants exhibit defects in trunk NC cell migration characterized by failure to form segmental chains and reach ventral positions. Our analysis of trunk NC migration in vivo revealed that in cdx4 ^ch107−/− mutants the first NC cell to migrate is regularly overtaken by later cells—an extremely rare event in wild type specimens (Richardson et al., 2016)—and uncovered multiple instances where two NC cells migrate alongside each other. While cdx4 mRNA is not detected in migrating NC, its expression in premigratory NC cells is consistent with a role in regulating the establishment of leader and follower NC cell identities. Importantly, Richardson and colleagues demonstrated that leader and follower identities are acquired prior to the initiation of migration (Richardson et al., 2016).

Moreover, prospective leader cells in the premigratory region could be distinguished from prospective followers based on differences in size (Richardson et al., 2016). Here, our in vivo imaging of trunk NC cell migration in cdx4^ch107−/− embryos has revealed not only that that the first NC cell to migrate is overtaken by cells that initiate their migration subsequently, but also that all of the mutant trunk NC cells analyzed resembled follower cells, not leaders, in terms of both area and cell polarity. We interpret our results as indicative that cdx4 is necessary for the establishment of leader identity and the related cell behaviors that drive the highly coordinated migration of trunk NC cells.

While our in vivo imaging reveals that cdx4^ch107−/− trunk NC cells do maintain some migratory capacity, including the potential to exchange positions, we postulate that aspects of the deficit in migration might be related to altered cell adhesion and epithelial-mesenchymal-transition (EMT) behaviors. In support of this possibility, we have previously shown that Prickle1-deficient cranial NC cells have altered distribution of Cadherin proteins associated with a partial block in the EMT necessary for NC cell migration to commence (Ahsan et al., 2019). Moreover, in trunk NC cells, Powell et al. (2015) have reported a role for zebrafish cdon in regulating N-cadherin to allow proper migration. A future investigation of EMT in cdx4^ch107−/− trunk NC cells would benefit from single cell labeling to allow details of individual cell behavior to be fully evaluated.

While the role that we propose for cdx4 in establishing specific NC cellular behaviors might initially appear to lie outside the scope of a NC specifier, these roles parallel those ascribed to the chick cranial NC specifier gene Ets1. Ets1 is a component of the chick cranial-specific GRN (Simões-Costa and Bronner, 2016) and there acts to activate the cranial enhancers of Foxd3 and Sox10 (Betancur et al., 2010; Simões-Costa et al., 2012), thus functioning as a bona fide NC specifier. In addition, Ets1 imparts cranial-specific cell behaviors during delamination (Théveneau et al., 2007). Specifically, chick cranial NC cells are capable of delaminating regardless of cell-cycle stage, whereas those in the trunk delaminate during S phase (Burstyn-Cohen et al., 2004; Burstyn-Cohen and Kalcheim, 2002; Sela-Donenfeld and Kalcheim, 2000, 1999). Ets1 is necessary for the delamination of cranial NC cells, and its ectopic expression is sufficient to promote cranial-like delamination in trunk NC cells under experimental conditions (Théveneau et al., 2007). In summary, the proposed dual roles of zebrafish cdx4 in regulating the expression of other early NC specifiers as well as imparting trunk-specific migratory behaviors, are consistent with the largely similar dual roles of chick Ets1 in the cranial NC. Further, the parallels with Ets1 function suggest that analysis of the role of Cdx4 in regulating cell cycle, and the potential links between cell cycle control and follower versus leader behaviors, may prove a fruitful area for future study.

Finally, the results of our transplantation experiments allow us to speculate that cdx4 likely functions cell-autonomously in premigratory trunk NC cells to regulate migratory behaviors. Specifically, we have established that cdx4 is expressed in premigratory trunk NC cells, and have shown that the defects in segmental migration of trunk NC cells in cdx4^ch107−/− mutants are not due to loss of Cdx4 function in adjacent somitic mesoderm. We found that the presence of Cdx4-deficient cells in the somitic mesoderm did not impair segmental migration of trunk NC cells, and conversely, when WT cells were placed in the somites of otherwise Cdx4-deficient embryos, trunk NC cells failed to migrate properly. Thus, we conclude that Cdx4 function in the adjacent somites is dispensable for the segmental migration of trunk NC cells. While these data do not definitively establish that cdx4 functions cell-autonomously in trunk NC cells to establish the leader/follower dynamics necessary for their directed migration, they are certainly consistent with such a model. A direct evaluation of the cell autonomy of Cdx4 function in the trunk NC cells via transplantation approaches must await a more thorough fate mapping of trunk NC cell progenitors in the gastrula.

In conclusion, we have shown that Cdx4 is expressed in NC cell progenitors, where it directly regulates the expression of genes of the NC GRN. Moreover, we have established that cdx4 is necessary for the establishment of the leader/follower cell dynamics that drive the segmental migration of trunk NC cells. Together, these results indicate that cdx4 plays a critical role in the development of NC cells in the posterior body of zebrafish embryos.

Materials and Methods

Animal Husbandry

Zebrafish (Danio rerio) were maintained in accord with IACUC-approved protocols. Embryos were maintained in E3 solution (in mM: 5.0 NaCl, 0.17 KCl, 0.33 CaCl, 0.33 MgSO₄) and staged according to standard guidelines (Kimmel et al., 1995). Embryos were obtained from crosses of adult fish stocks of mutants and/or transgenics. Transgenic zebrafish lines Tg(−7.2sox10: mRFP)vu234 (referred to as Tg(sox10: mRFP) (Kirby and Hutson, 2010) and Tg(h2az2a:h2az2a-GFP) (referred to as Tg(His-GFP) (Pauls et al., 2001) have been previously described. The cdx4 mutant cdx4(ch107) line is characterized by a missense mutation and precocious stop (nonsense) codon, as described below.

Generation of the cdx4 ch107 allele

A guide RNA for CRISPR-based mutagenesis was generated by annealing a cdx4-specific oligo, 5′-GTGTGGAAACAAAGTTCTGTGG-3, to the trRNA sequence followed by in vitro transcription as described in (Gagnon et al., 2014). cas9 mRNA was in vitro transcribed from plasmid pT3TS-nCas9n (Jao et al., 2013) and purified as described (Gagnon et al., 2014). 100 pg cdx4 sgRNA and 300 pg cas9 mRNA were co-injected in a 1.25 nl volume into early 1-cell stage wildtype (*AB) zebrafish embryos. Injected F0 specimens were raised to adulthood and genotyped to identify genetically mosaic cdx4 mutant fish. F0 mutation carriers were then outcrossed to generate individual cdx4 mutant F1 fish. An F1 cdx4 mutant carrier, containing a 1 base pair deletion directly followed by a 5 base pair insertion after nucleotide 84 in exon 1, was identified and designated as cdx4(ch107). cdx4(ch107) is predicted to produce missense mutations after amino acid 28 and a precocious stop (nonsense) codon at amino acid position 102. The ch107 F1 founder was outcrossed to produce the F2 generation, and adult heterozygous F2 siblings were inbred to produce cdx4^ch107−/− homozygous mutant embryos for study.

Bioinformatics analysis

All zebrafish datasets were converted to the genome assembly danRer7 using the liftOver tool (Hinrichs et al., 2006) to enable comparison. Cdx4 ChIP-seq data were obtained from NCBI GEO: GSE48254; these data, reported by Paik et al. (2013), were produced using Myc-tagged Cdx4 from 10 hpf zebrafish embryos. To uncover putative cis-regulatory regions bound by Cdx4 we assigned the 4,965 Cdx4-binding regions identified by Paik et al. (2013) to 2,407 individual genes using GREAT v3.0.0 (McLean et al., 2010) by annotating peaks to the single nearest gene within 100 kb.

Foxd3 ChIP-seq data were obtained from NCBI GEO GSE106676; these data, reported by Lukoseviciute et al. (2018), were produced using a Biotin ChIPseq approach for Foxd3 from both 10 hpf and 12 hpf zebrafish embryos. Using GREAT v3.0.0, and again by annotating peaks to the single nearest gene within 100 kb, we assigned 531 Foxd3-bound regions at 10 hpf and 2,955 regions at 12 hpf to 428 genes and 789 genes, respectively. ATAC-seq data were also obtained from GEO: GSE106676 and coverage was converted to danRer7 using the liftOver tool in the ‘rtracklayer’ package (Lawrence et al., 2009).

Using these datasets, we analyzed 69 unique zebrafish genes that comprise the orthologs of the induction, NPB, premigratory NC, and migratory NC modules of the previously described NC GRN. See Supplemental Table 1 for a list of the 69 genes investigated and a summary of the binding site data. Analysis of bound genes was performed using custom R scripts (available on request). ChIP-seq and ATAC-seq data were visualized using the ‘Gviz’ package in R (Hahne and Ivanek, 2016).

Hybridization Chain Reaction (HCR)

Antisense DNA probes were designed against the full-length zebrafish cdx4 and foxd3 mRNA sequences as described by Choi et al. (2018) and purchased from Molecular Instruments. Embryos were fixed with 4% PFA (PFA; Sigma) at 4°C overnight and then stained as previously described (Choi et al., 2018). Embryos were mounted in 1% low melting agarose for imaging.

Immunohistochemistry

Embryos were fixed in 4% paraformaldehyde (PFA; Sigma) and immunohistochemistry was performed as previously described (Prince et al., 1998) using the following primary antibodies: rabbit anti-Sox10 antibody (1:100, GeneTex, GTX128374), mouse anti-myosin heavy chain (1:100, Developmental Studies, Hybridoma bank, IA, USA, A4.1025). The following secondary antibodies were used: goat-anti mouse highly cross-adsorbed Alexa Fluor Plus 488 (Molecular Probes A32723), goat anti-rabbit cross-adsorbed Alexa Fluor 546 conjugate (Molecular Probes A11010). Embryos were also stained with DAPI (Thermo Fisher, R37606) and then mounted in 1% low melting agarose for imaging.

Confocal Image Acquisition

For assays in fixed specimens, embryos were fixed in 4% PFA at 4°C overnight. Following overnight fixation, embryos were washed in 1X PBS five times for 5 min each. For long-term storage of embryos, embryos were washed in 30%, 60% and 100% methanol (diluted in 1X PBS) and stored in 100% methanol at −20°C. If stored in 100% methanol, embryos were progressively washed in 60%, 30% methanol as well as 1X PBS + 0.1%Tween-20 before mounting or staining. For transverse section analysis, 50–100 μM sections were cut by hand from the trunk of embryos in PBS and embedded in 1% low-melt agarose (MidSci IB70051 St. Louis, Missouri) for subsequent imaging.

Images were collected using an upright Zeiss LSM710 confocal microscope with a Plan-Apochromat 10x/0.45 (working distance: 2.1mm) objective or a 40x/1.0 W Plan-Apochromat (working distance:2.5mm) objective. Green fluorescent dyes (Alexa Fluor 488) were excited by a 488 nm laser. Red fluorescent dyes (Alexa Fluor 546) were excited by a 543 nm laser. DAPI dye was excited using a 405 nm laser. For a single fluorophore or a combination of fluorophores, spectral unmixing was used to define emission fluorescence range. Images were acquired and saved as .czi files using Zen (Zeiss) software and processed with Fiji (Schindelin et al., 2012).

Single Plane Illumination Microscopy (SPIM)

Zebrafish embryos were staged to 14 hpf and mounted in 1% low melting agarose (MidSci IB70051 St. Louis, Missouri) dissolved in E3 medium and 0.2 mg/ml tricaine using glass capillaries (Carl Zeiss Microscopy, 701904). Embryos were incubated at 28.5°C during data collection. Images were captured with a Zeiss Lightsheet Z.1 single-plane illumination microscope (Carl Zeiss Microscopy) with tandem PCO.edge sCMOS cameras (PCO.Imaging, Kelheim, Germany) and Zeiss Zen imaging software. A 20×/1.0 long working distance detection objective was used alongside a pair of 10×/0.2 dry illumination objectives, and the excitation sheet was narrowed to 2.0 μm. Volumes were acquired every 5 min between 15 and 21 hpf, with 119 ms exposure per slice for both green (488 nm, 7.5%) and red (561 nm, 7.0%) channels. Cell tracks were manually reconstructed using Imaris.

NC cell quantifications

To plot the position of migrating NC cells relative to the adjacent somite, the coordinates for the anti-Sox10 Ab immunolabel and the vertex of each somite were obtained in Fiji. The position of each Sox10 point was then normalized relative to the position of the vertex of the adjacent somite in R. Plots of the distribution and 2D density of these points were then generated using ggplot2 (Wickham, 2016). Fiji was also used to measure the cell area and aspect ratio. Box plots of the cell areas and aspect ratios were generated using ggplot2 and t-tests were performed using ggpubr (https://rpkgs.datanovia.com/ggpubr/index.html) in R.

Transplants

Cell transplantation experiments were performed as previously described (Ho and Kane, 1990). To knockdown Cdx4 function, antisense cdx4 morpholino (1 nl at 20 ng/μl, as described by Davidson et al., 2003 and Skromne et al., 2007) oligonucleotides (Gene Tools LLC) were injected into one-cell stage donor or host embryos, as described (Skromne et al., 2007). In addition, donor embryos were injected with 10 kDa lysinated Alexa Fluor 647 (Life Technologies D-22914; 0.5% in 0.2M KCl, as described by Love & Prince, 2015). ~ 30 cells from blastula-stage (4 hpf) donor embryos were transplanted into the margins of blastula-stage hosts. Chimeric embryos were allowed to develop to 25 hpf and fixed in 4% PFA at 4°C overnight. To avoid cardiac and vagal NC cells, only NC cells adjacent to somite four and posterior were included in this analysis. Moreover, only those specimens in which >20% of the dorsoventral extent of one of more somites was populated by transplanted donor-derived cells were selected for further analysis.

Supplementary Material

1. Supplemental Figure 1: Genotype and phenotype of the cdx4^ch107 allele.

A. Comparison of the amino acid sequences of full length Cdx4 protein (homeodomain sequence indicated in blue) and Cdx4 (ch107). B. Comparison of the protein domains of Cdx4 and Cdx4(ch107), with the site of mutation and surrounding nucleotides indicated; the altered nucleotides, as a consequence of an indel, are indicated in red; the homeodomain is again shown in blue. C. A clutch of 24 hpf offspring from a cross between cdx4^ch107+/− adults; a WT tail is indicated with a blue arrow and a ‘kugelig’ phenotype ch107 mutant tail is indicated with a black arrow.

2. Supplemental Movie 1. Trunk NC cell migration in a WT embryo.

Timelapse of trunk NC cell migration from 16.5–18 hpf of a WT Tg(sox10:mRFP) embryo, captured using SPIM. The transgene labels NC cells, but is also expressed in a small population of cells in the ventral neural tube past which the NC cells are seen to migrate. Tracks indicate the migration of two leader cells and two follower cells. Scale bar = 25 μm.

3. Supplemental Movie 2: Trunk NC cell migration in a cdx4^ch107−/− embryo.

Timelapse of trunk NC cell migration from 16.5–18 hpf of a cdx4^ch107−/− Tg(sox10:mRFP) embryo captured using SPIM. The transgene labels NC cells, but is also expressed in a small population of cells in the ventral neural tube past which the NC cells are seen to migrate. Tracks indicate the migration of two leader cells and two follower cells. Scale bar = 25 μm.

Highlights.

• Zebrafish cdx4 is expressed in premigratory trunk neural crest (NC) cells

• Cdx4 regulates expression of foxd3 and other NC specifier genes

• Segmental NC migration patterns are disrupted in cdx4^−/− embryos

• cdx4^−/− trunk NC cells resemble ‘follower’ cells, with ‘leader’ cells absent

• NC migration is independent of somitic Cdx4 function and is likely cell-autonomous