Cellular and molecular mechanisms of convergence & extension in zebrafish

Abstract

Gastrulation is the period of development when the three germ layers, mesoderm, endoderm and ectoderm, are not only formed, but also shaped into a rudimentary body plan. An elongated anteroposterior (AP) axis is a key feature of all vertebrate body plans, and it forms during gastrulation through the highly conserved morphogenetic mechanism of convergence & extension (C&E). As the name suggests, this process requires that cells within each germ layer converge toward the dorsal midline to narrow the tissue in the mediolateral (ML) dimension and concomitantly extend it in the AP dimension. In a number of vertebrate species, C&E is driven primarily by mediolateral intercalation behavior (MIB), during which cells elongate, align, and extend protrusions in the ML direction and interdigitate between their neighbors. MIB is only one of many complex cellular mechanisms that contributes to C&E in zebrafish embryos, however, where a combination of individual cell migration, collective migration, random walk, radial intercalation, epiboly movements, and MIB all act together to shape the nascent germ layers. Each of these diverse cell movements is driven by a distinct suite of dynamic cellular properties/activities, such as actin-rich protrusions, myosin contractility, and blebbing. Here, we discuss the spatiotemporal patterns of cellular behaviors underlying C&E gastrulation movements within each germ layer of zebrafish embryos. These behaviors must be coordinated with the embryonic axes, and we highlight the roles of Planar Cell Polarity (PCP) in orienting and BMP signaling in patterning C&E cell behaviors with respect to the AP and dorsoventral axes. Finally, we address the role of GPCR signaling, extracellular matrix, and mechanical signals in coordination of C&E movements between adjacent germ layers.

1. OVERVIEW OF ZEBRAFISH C&E

Convergence and extension (C&E) is a highly conserved morphogenetic mechanism by which tissues become simultaneously longer and narrower. This type of tissue deformation is common during development and contributes to the shaping of a number of embryonic tissues and organ rudiments. The focus of this chapter will be on C&E of the zebrafish embryo during gastrulation, where it functions primarily to extend the primary body axis in the anteroposterior (AP) dimension while narrowing axial tissues in the mediolateral (ML) dimension. C&E is preceded by the specification of the embryonic axes and three primordial germ layers as described in Chapter 2 of this volume, and it occurs simultaneously with the other gastrulation movements of epiboly and internalization, discussed in Chapters 12 and 13, respectively. During early gastrulation, epiboly thins and spreads the blastoderm over the syncytial yolk cell, driving the embryonic margin toward the vegetal pole. Once the blastoderm covers approximately 50% of the yolk surface (50% epiboly), the internalization of mesoderm and endoderm germ layers begins at the blastoderm margin first on the dorsal side, and then proceeds ventrally around the circumference of the margin. This mixed mesendoderm internalizes primarily as individual cells, which “turn the corner” to migrate animally (anteriorly) between the yolk cell and the overlying blastoderm. As gastrulation proceeds, epiboly continues to spread the germ layers vegetally (posteriorly) while the internalized mesendoderm continues to migrate animally, thereby contributing to the AP extension of the nascent primary body axis (Fig. 1A). Cells initiate convergence movements once the embryo reaches approximately 70% epiboly, contributing to axial extension together with continued epiboly and anterior migration (Fig. 1B and C). Blastoderm cells that remain on the surface of the embryo upon completion of internalization give rise to non-neural (ventrally) and neural (dorsally) ectoderm. As epiboly proceeds, these ectoderm cells continue to move vegetally while simultaneously undergoing C&E concomitant with similar movements in the underlying mesendoderm. At this point, the most ventral mesoderm, which does not engage in C&E, switches from animal- to vegetal-directed migration. When epiboly is complete, this population will undergo the subduction movements to contact the posterior dorsal mesoderm at the site of yolk plug closure and give rise to the tailbud and its derivatives. Each of these germ layers therefore exhibits a distinct, region-specific suite of cell behaviors which are coordinated to collectively drive C&E of the AP axis.

Fig. 1.

Cell behaviors underlying convergence & extension (C&E) of zebrafish mesoderm. Gastrulation movements prior to (A) and after the onset of C&E in lateral (B) and dorsal (C) view. The purple stripe in C represents the axial mesoderm. (D) Cell behaviors underlying C&E movements of mesoderm in each of the five regions indicated in (B) and (C): vegetal migration in region I, dorsal migration in regions II and III, cell intercalations in regions IV and V, and anterior migration in region V. Dorsal is to the right in (A), (B), and (D); anterior is up in all panels.

C&E is conserved throughout the animal kingdom and is achieved through a number of cellular mechanisms, including cell shape change, oriented cell division, directed cell migration, and cell intercalation (Keller, 2006). The best-described among these mechanisms is planar cell intercalation, by which a group of cells rearranges into a longer and narrower array through junctional exchange and/or interdigitation with neighboring cells termed “cell shuffling.” Junctional exchange was identified as a mechanism for germ band extension during Drosophila gastrulation and involves remodeling of apical junctions between epithelial cells (Bertet, Sulak, & Lecuit, 2004; Zallen & Blankenship, 2008) (Chapter 7). In contrast, cell shuffling is driven by polarized cell protrusions thought to gain traction on neighboring cells and the underlying extracellular matrix (Pfister, Shook, Chang, Keller, & Skoglund, 2016). This is best exemplified by mediolateral intercalation behavior (MIB) of mesenchymal mesendodermal cells within Xenopus gastrulae (Keller et al., 2000). Recent evidence suggests that these two modes of cell intercalation are not mutually exclusive, and that the same cells can employ both junctional neighbor exchange and polarized protrusions to cooperatively drive C&E in both epithelial and mesenchymal tissues (Huebner & Wallingford, 2018; Sun et al., 2017; Williams, Yen, Lu, & Sutherland, 2014).

In these “classic” examples of C&E, convergence and extension movements are coupled by ML intercalation, which links narrowing in one dimension (ML) to a corresponding elongation in an orthogonal dimension (AP during C&E, dorsal-ventral during convergent thickening) (Keller et al., 2000; Shih & Keller, 1992a). For this reason, C&E in Xenopus is often referred to as “convergent extension.” By contrast, convergence and extension movements within zebrafish gastrulae are relatively independent and often uncoupled from one another, and are driven by a number of cell behaviors in addition to ML intercalation (Glickman, Kimmel, Jones, & Adams, 2003; Sepich, Calmelet, Kiskowski, & Solnica-Krezel, 2005; Warga & Kimmel, 1990). Unlike in Xenopus and Drosophila, in zebrafish gastrulae, individual mesendodermal cells migrate over the surface of a large central yolk cell. Lateral mesendoderm cells in the equatorial region undergo directed dorsal migration to narrow the nascent primary axis, while the more animally- and vegetally-located cells each exhibit a bias toward their respective poles, causing their paths to “fan out” and contribute to extension as they converge dorsally (Fig. 1B) (Sepich et al., 2005). As these lateral mesendodermal cells migrate to drive C&E, they leave behind an “empty space” in the animal-ventral region termed the evacuation zone (Kimmel, Ballard, Kimmel, Ullmann, & Schilling, 1995). Concurrently, anterior migration of axial mesoderm cells extends the nascent axis without contributing to convergence (Concha & Adams, 1998; Ulrich et al., 2003; Warga & Kimmel, 1990). Such individual cell migration is facilitated not only by interactions with the yolk, but also by the reduced adhesion between zebrafish cells relative to the tightly-packed cellular sheets that comprise Xenopus and Drosophila gastrulae, which allows more freedom of movement. Convergence and extension are coupled in the dorsal axial and paraxial mesoderm of zebrafish gastrulae as observed in “classic” mediolateral intercalation (Fig. 1D, region V purple box), but even these movements can be uncoupled in certain mutant conditions (Glickman et al., 2003). The distinct suite of cell behaviors underlying zebrafish C&E will be described in further detail below.

As in nearly all complex developmental processes, gastrulation morphogenesis is intricately linked with embryonic patterning. This is intuitive: for cells to move to the anterior or dorsal region of an embryo, they must first be provided with cues for where anterior and dorsal are. Indeed, the signals that specify cell fate often also dictate cell behavior within zebrafish gastrulae. For example, C&E movements are most prominent within the dorsal-most tissues, whereas ventral tissues on the other side of the gastrula exhibit essentially no C&E (Fig. 1). Mutant analyses demonstrated that signals that specify ventral fates (namely, Bone morphogenic protein (BMP)) actually inhibit C&E cell behaviors (Myers, Sepich, & Solnica-Krezel, 2002; von der Hardt et al., 2007), whereas widespread expansion of dorsalizing signals that inhibit BMP expands both dorsal cell fates (Mintzer et al., 2001) and C&E behaviors around the entire circumference of the embryo (Myers et al., 2002), squeezing the yolk cell into an abnormally elongated shape, sometimes until it ruptures (Mullins et al., 1996). BMP is thought to limit C&E movements through negative regulation of genes encoding PCP components and adhesion molecules (Myers et al., 2002; von der Hardt et al., 2007), acting in parallel rather than downstream of cell fate specification. This illustrates the close ties between patterning and morphogenesis, but with the exception of a few examples discussed further below, the molecular basis of this coordination is largely unknown.

In this chapter, we will describe C&E of each germ layer and tissue within the zebrafish gastrula, the underlying cellular mechanisms, and the known molecular regulators of each. Finally, we will discuss how interactions between germ layers influence C&E morphogenesis.

2. C&E OF THE MESODERM

Mesoderm within the zebrafish gastrula can be divided into five regions, each of which exhibit a distinct suite of cell behaviors (Fig. 1) (Myers et al., 2002; Roszko, Sawada, & Solnica-Krezel, 2009). Region I encompasses ventral mesoderm, regions II and III are lateral, region IV comprises medial presomitic mesoderm (PSM), and region V refers to the axial mesoderm.

2.1. Axial mesoderm: Chordamesoderm

Axial mesoderm within region V is comprised of two different tissues that exhibit distinct gastrulation cell behaviors: chordamesoderm and prechordal plate (Fig. 1C and D, purple and blue boxes). At the onset of gastrulation, both internalize at the dorsal margin and migrate toward the animal pole, thus extending the tissue prior to the onset of convergence. At approximately 70% epiboly, the chordamesoderm, which gives rise to the notochord, begins to converge and extend by mediolateral cell intercalation similar to that observed in Xenopus (Concha & Adams, 1998; Glickman et al., 2003). These cells elongate, align, and extend protrusions mediolaterally to facilitate intercalation between their anterior and posterior neighbors to produce a longer and narrower array (Fig. 1D, purple box). This continues until the chordamesoderm is only a single cell diameter across at early somite stages.

ML alignment of both cell shape and cellular protrusions is essential for efficient intercalation and is largely regulated by planar cell polarity (PCP) signaling. This highly conserved signaling system was first identified in Drosophila, where it functions at both local and global levels to promote polarization of cells within the plane of a tissue, ensuring that both sub- and supra-cellular structures (like wing trichomes and body hairs) all point in the same direction (Strutt & Strutt, 2005; Vinson & Adler, 1987; Wong & Adler, 1993). Vertebrate and Drosophila PCP share several so-called “core” components, including the four-pass transmembrane protein Van Gogh-like (Vangl), the Frizzled (Fz) seven-pass transmembrane receptor, the hybrid cadherin/G protein coupled receptor (GPCR) Celsr (also known as Flamingo), PDZ-domain-containing intracellular Disheveled (Dvl), and the LIM domain-containing Prickle (Pk) (Carreira-Barbosa et al., 2003, 2009; Gubb et al., 1999; Jessen et al., 2002; Lu, Usui, Uemura, Jan, & Jan, 1999; Park & Moon, 2002; Takeuchi et al., 2003; Taylor, Abramova, Charlton, & Adler, 1998; Usui et al., 1999; Veeman, Slusarski, Kaykas, Louie, & Moon, 2003). In addition to these “core” proteins, vertebrate PCP also shares a subset of components with the Wnt/β-catenin pathway, including Wnt ligands (Heisenberg et al., 2000; Kilian et al., 2003), Fz receptors (Čapek et al., 2019; Wallingford, Vogeli, & Harland, 2001), and intracellular Dvl effectors (Wallingford et al., 2000), and is therefore often referred to as non-canonical Wnt/PCP signaling. Zebrafish mutations that disrupt PCP signaling yield embryos with shorter AP and wider ML body axes, indicative of reduced C&E, but few patterning or cell fate specification defects. These include vangl2 (trilobite), wnt5a (pipetail), wnt11 (silberblick), and frizzled7a/b; as well as PCP-related/interacting genes such as glypican4 (knypek) and ptk7 (Čapek et al., 2019; Hayes, Naito, Daulat, Angers, & Ciruna, 2013; Heisenberg et al., 2000; Jessen et al., 2002; Kilian et al., 2003; Sepich et al., 2000; Topczewski et al., 2001). Core PCP components like Prickle and Vangl2 become asymmetrically distributed or enriched at the anterior and Dvl to the posterior edge of intercalating mesoderm cells (Ciruna, Jenny, Lee, Mlodzik, & Schier, 2006; Roszko, Sepich, Jessen, Chandrasekhar, & Solnica-Krezel, 2015; Yin, Kiskowski, Pouille, Farge, & Solnica-Krezel, 2008) (Fig. 2A), leading to the hypothesis that PCP acts as a molecular compass that allows cells to orient with respect to the embryonic axes (Gray, Roszko, & Solnica-Krezel, 2011; Yin et al., 2008). Without this spatial “awareness,” cells cannot acquire the ML polarity required for effective cell intercalation, leading to disrupted C&E morphogenesis.

Fig. 2.

Signaling pathways regulating C&E cell behaviors. (A) Within regions IV and V, non-canonical Wnt/PCP signaling regulates ML cell elongation and ML-biased cell protrusions via a number of downstream modules including JNK, Ca²⁺, and RhoA. (B) Within region V, anteriorly-directed cell migration is regulated by sphingosine-1-phosphate (S1P), Wnt/PCP, and PDGF signaling that ultimately promote directed protrusions via localized Rac1 and RhoA activity. (C) Within region II, cells converge slowly by alternating phases of “tumble”—blebbing induced by loss of membrane-cortical attachment, and “run”—directed protrusion-driven migration requiring Apelin receptor signaling. Black hashes represent the actin cytoskeleton, green hashes represent non-muscle myosin.

Chordamesoderm cells within PCP mutant gastrulae exhibit inefficient intercalation, leading to wider and shorter notochords (Čapek et al., 2019; Heisenberg et al., 2000; Jessen et al., 2002; Kilian et al., 2003; Topczewski et al., 2001). This is due to reduced ML polarity of both cell shapes and cellular protrusions, which gain traction on neighboring cells to drive their intercalation. Wnt/PCP signaling is thought to regulate the polarity of these cell behaviors largely via an intracellular cascade that promotes phosphorylation of Myosin regulatory light chain (MRLC) to enhance Myosin contractility. Work in Xenopus found that upon binding of non-canonical Wnt ligands to Fz, Dvl is recruited to the membrane where it forms a complex with the Formin homolog protein Disheveled-associated activator of morphogenesis (Daam1) to activate RhoA GT-Pases (Habas, Kato, & He, 2001). RhoA in turn activates downstream effectors Diaphanous (Lai et al., 2008; Zhu, Liu, Korzh, Gong, & Low, 2006) and the Rho Kinase Rok2, which promotes C&E cell behaviors in a cell-autonomous and non-autonomous fashion (Marlow, Topczewski, Sepich, & Solnica-Krezel, 2002) likely by phosphorylating MRLC (Winter et al., 2001) (Fig. 2A). Additional work in Xenopus demonstrated that Septins restrict Myosin activation in a planar-polarized fashion to AP cell interfaces to ensure ML directional intercalation (Shindo & Wallingford, 2014), but whether similar mechanisms play a role in zebrafish C&E is unknown.

Wnts also regulate C&E cell behaviors via a variety of other non-canonical downstream signaling effectors, including Ca^{2 +}, Protein Kinase C (PKC), and c-Jun amino terminal kinase (JNK) (Fig. 2A). In zebrafish embryos, signaling by Wnt5b—but not canonical Wnt ligands—was shown to promote release of calcium ions from intracellular stores in a G protein-dependent manner (Slusarski, Corces, & Moon, 1997; Slusarski, Yang-Snyder, Busa, & Moon, 1997). Work in a combination of zebrafish and Xenopus embryos found that upon Wnt binding to Fz7, Dvl is recruited to the membrane together with the Ca^{2 +}-responsive PKC, which then stimulates JNK and its target c-Jun (Kühl, Sheldahl, Malbon, & Moon, 2000; Kühl, Sheldahl, Park, Miller, & Moon, 2000; Sheldahl, Park, Malbon, & Moon, 1999; Sheldahl et al., 2003). Wnt-induced Ca^{2 +} release also activates Ca^{2 +}/Calmodulin-dependent Kinase II (CamKII) in Xenopus (Sheldahl et al., 2003), and activated CamKII suppresses tail defects in wnt5b/ppt mutant zebrafish embryos (Westfall et al., 2003). The relationship between this Ca^{2 +}-dependent non-canonical Wnt pathway and Wnt/PCP signaling is not fully understood.

The “choice” of signaling cascade downstream of Wnt signaling and its effect on C&E morphogenesis is dependent upon a number of alternative/co-receptors, including Ryk, Ror2, and CD146 (Fig. 2A). CD146 is a transmembrane cell adhesion molecule that interacts directly with both Wnt5 ligands and Dvl2 and is required for Dvl2 phosphorylation and downstream activation of JNK, which promotes cell motility and C&E movements upon Wnt ligand binding (Ye et al., 2013). Related to Tyrosine Kinase (Ryk), a kinase-dead transmembrane receptor tyrosine kinase (RTK), also acts downstream of Wnt5b, but functions independently of Dvl and Fz receptors to promote release of intracellular Ca^{2 +}. Ryk-expressing cells extend protrusions away from sources of Wnt5b (but not Wnt11), implying that it mediates directional cell responses to non-canonical Wnt ligands through repulsive interactions (Lin, Baye, Westfall, & Slusarski, 2010). Another RTK, Ror2, was shown to act as an alternative receptor for Wnt5a in other vertebrate species (Ho et al., 2012; Schambony & Wedlich, 2007), but mediates Wnt11 signals during zebrafish gastrulation via downstream activation of Dvl and RhoA (Bai et al., 2014). The RTK Ptk7 also promotes planar polarization of both cell behaviors and intracellular distribution of PCP components (Hayes et al., 2013), and its loss of function, like all of the Wnt alternative/co-receptors discussed above, disrupts C&E gastrulation movements in zebrafish. While it is clear that such alternative/co-receptors are vital to the interpretation of morphogenetic cues, the nature of these signals—let alone how they are integrated by cells via multiple receptors to achieve specific behavioral outcomes—is not well understood. In particular, whereas it is clear that Wnt11 and Wnt5b are essential for ML cell polarity underlying C&E, it remains unclear whether they play an instructive or permissive role (see below).

G protein coupled receptors (GPCRs) have also been implicated in non-canonical Wnt/PCP signaling and C&E morphogenesis. Overexpression of the adhesion (ad)GPCR Gpr125/Adgra3, for example, disrupts polarization of PCP components and reduces ML cell polarity and C&E. While Gpr125 knockdown using antisense morpholino oligonucleotides does not cause any phenotypes in wild-type gastrulae, it exacerbates C&E defects in PCP mutants (Li et al., 2013). The intracellular domain of Gpr125 recruits PCP signaling components to cell membranes by clustering the PCP effector Disheveled into membrane sub-domains (Li et al., 2013), but how this activity is induced/regulated during normal C&E is unknown. One possibility is that it functions as a Wnt co-receptor, as is the case of the closely-related adGPCR GPR124/Adgra2, which acts together with Reck to confer receptor specificity for Wnt7 ligands in mammalian cells (Eubelen et al., 2018; Vallon et al., 2018).

The G proteins Gα12_/13 are also required cell-autonomously for proper C&E, as their loss- or gain-of function results in shorter and wider chordamesoderm and reduced ML cell polarity in zebrafish gastrulae without affecting embryonic patterning (Lin et al., 2005). Loss of Gα12_/13 exacerbates defects in C&E and underlying ML cell polarity in PCP mutant embryos, but it is unclear whether they function downstream of or in parallel to PCP signaling. G proteins have been implicated in non-canonical Ca^2+/PKC-mediated Wnt signaling (Slusarski, Corces, et al., 1997), and Gα12_/13 can function upstream of Rho GTPases to regulate shape and migration of cultured cells (Buhl, Johnson, Dhanasekaran, & Johnson, 1995; Sugimoto, Takuwa, Okamoto, Sakurada, & Takuwa, 2003). GPCRs like Gpr125 (Li et al., 2013), the core PCP component Celsr (Carreira-Barbosa et al., 2009), and chemokine receptor Cxcr4 (described further below) are all required for C&E of mesoderm and/or endoderm, allowing for the possibility that signaling via G proteins like Gα12_/13 is a critical component of Wnt-dependent signaling networks that drive C&E.

Beyond these complex, overlapping, and interacting signaling networks that regulate polarized cell behaviors and tissue morphogenesis, mechanical forces are also essential regulators of gastrulation. Surface tension, for example, contributes significantly to epiboly movements that occur contemporaneously with C&E (Morita et al., 2017). For more on the importance of additional physical properties during morphogenesis, such as tissue stiffness and tension, please see: Davidson (2017), Keller, Davidson, and Shook (2003), Miller and Davidson (2013), and Zhou, Pal, Maiti, and Davidson (2015). One physical element that is critical for C&E of the chordamesoderm are its boundaries with the adjacent presomitic mesoderm (PSM). This tissue boundary between the axial and paraxial mesoderm was first shown to influence the behavior of adjacent chordamesoderm cells in the form of “boundary capture” in Xenopus gastrulae (Shih & Keller, 1992b), the phenomenon by which contact with the boundary quiets protrusions on one side of the cell, converting bipolar protrusive activity into monopolar activity that continues to narrow the axial mesoderm. The notochord boundary was found to similarly influence polarized cell behaviors within the chordamesoderm of zebrafish gastrulae. Boundary-associated cues, including cell adhesion and tissue tension, promote ML polarity of cells within approximately two cell diameters of the boundary and, importantly, function independently of Wnt/PCP signaling (Williams et al., 2018). Boundary cues and PCP were similarly shown to cooperate in C&E of Ascidian chordamesoderm (Veeman et al., 2008). Another physical attribute of the axial mesoderm that influences axis extension is cell number. Indeed, even a moderate reduction in proliferation rate significantly reduces extension of the chordamesoderm and AP axis length (Liu, Sepich, & Solnica-Krezel, 2017; Riley et al., 2010), likely due to fewer “building blocks” with which to extend this tissue. These studies demonstrate that an embryo’s physical properties cooperate with PCP and other signaling to promote ML polarity and/or cell intercalation within the chordamesoderm.

2.2. Axial mesoderm: Prechordal plate

The anterior-most portion of the axial mesoderm, the prechordal plate (ppl), contributes to C&E of the AP axis via a different suite of cell behaviors than its neighbor to the posterior. Axial mesoderm cells are among the first to internalize at the dorsal margin and undergo highly directed migration toward the animal (future anterior) pole of the embryo. As described above, the chordamesoderm switches from anterior migration to ML intercalation at mid-gastrulation, while ppl cells continue to migrate as a cohesive population that produces polarized protrusions at its leading (anterior) edge while maintaining adhesions with fellow ppl cells and the chordamesoderm to its posterior (Ulrich et al., 2003) (Fig. 1C and D blue box). Migrating cells at the leading edge of the ppl exhibit a variety of protrusive behaviors, including filopodia, lamellipodia, and blebs. The balance between these modes of migration determines the speed and persistence of migration, and is achieved through precise regulation of attachment between the cell cortex and membrane (Diz-Muñoz et al., 2010). Notably, insufficient anterior displacement of the ppl (as is observed in many PCP mutants) can result from reduced C&E in the lateral mesoderm alone (Heisenberg et al., 2000). This suggests that full extension of the axial mesoderm requires both “pulling” by directed ppl migration at the anterior end and “pushing” by C&E of the chordamesoderm and lateral mesoderm populations.

Like C&E cell behaviors within the chordamesoderm, anterior migration of axial mesoderm cells is under the regulation of Wnt/PCP signaling (Fig. 2B). Wnt11, for example, is required for the formation of anterior-directed protrusions within newly-internalized mesendoderm populations (Ulrich et al., 2003). Loss of Wnt5b does not affect anterior migration in the same way, but injection of exogenous wnt5b RNA suppresses ppl defects in wnt11/slb mutants (Kilian et al., 2003), demonstrating that these ligands have redundant functions within the anterior axial mesoderm. It has been debated, however, whether PCP signaling plays a permissive or instructive role in C&E during zebrafish gastrulation. It was shown that Wnt5b is not only necessary for efficient C&E, but can also orient cell protrusions away from the source of the ligand (Lin et al., 2010), suggesting that Wnt5b signaling may instruct C&E cell behaviors. This is supported by reports that ubiquitous expression of wnt5b RNA does not rescue C&E defects in wnt5b/ppt mutant gastrulae (Kilian et al., 2003), consistent with an instructive rather than permissive role in these cell behaviors. This contrasts with another well-characterized Wnt/PCP mutant, wnt11/slb, whose C&E defects can be rescued by wnt11 RNA injection (Ulrich et al., 2003). Similarly, the Fz7 receptor with which Wnt11 is thought to interact is essential for proper gastrulation, but uniform optogenetic-driven activation of Fz7 in fz7 mutant gastrulae rescues anterior migration of the ppl (Čapek et al., 2019), again implying a permissive rather than instructive role for Wnt11 signaling in anterior migration of ppl cells. Wnt5b was shown to signal through a different receptor, Fz2 (Kilian et al., 2003), raising the possibility that each of these ligand/receptor pairs regulates C&E via a distinct mechanism: one instructive and one permissive. In Xenopus embryos, non-canonical ligands Wnt5a and Wnt11 were shown to be both necessary and instructive for planar-polarized distribution of core PCP components Pk and Vangl2 (Chu & Sokol, 2016; Ossipova, Kim, & Sokol, 2015), demonstrating similarities and potential differences with Wnt/PCP signaling in zebrafish.

Another well-characterized signaling pathway regulating anterior axial mesoderm migration in zebrafish is platelet-derived growth factor (PDGF) (Fig. 2B). PDGF receptor signaling activates PI3 Kinase, which signals through Target of Rapamycin 2 (TORC2) and/or AKT (Dumortier & David, 2015; Montero, Kilian, Chan, Bayliss, & Heisenberg, 2003) to locally activate Rac1 at the leading edge of ppl cells, possibly via the Rac Guanine nucleotide exchange factor (GEF) Prex1 (Campbell et al., 2013; Woo, Housley, Weiner, & Stainier, 2012). The sphingosine-1-phosphate GPCR S1pr2, encoded by the zebrafish miles apart gene, negatively regulates PI3K-dependent cell motility while promoting ppl cell cohesion to ensure cohesive, directed migration, but also regulates cell motility independent of PI3K (Kai, Heisenberg, & Tada, 2008). This may be mediated by Gα_12/13-dependent activation of Rho and inhibition of Rac GTPases, which has been observed in cultured cells (Sugimoto et al., 2003). The Diaphanous-related Formin2 is also critical for efficient ppl protrusive activity, but it appears to function in a Rac-independent manner and instead interacts with activated RhoA and/or Cdc42 (Lai et al., 2008).

Interestingly, although PDGF is essential for activation of this signaling cascade, it does not appear to instruct the direction of migration. This is supported by experiments in which isolated ppl cells are motile but fail to migrate anteriorly until they make contact with the intact ppl, at which point their migration becomes appropriately polarized. This instructive interaction requires not only Rac1 activity, but also E-cadherin (Dumortier, Martin, Meyer, Rosa, & David, 2012), indicating that cadherin-based contacts between ppl cells either directly provide directionality to this collective cell migration or are required for ppl cells to perceive additional directionality cues. This likely explains why cohesion of the ppl is so critical to proper morphogenesis, and why dispersal of these cells results in abnormal migration and reduced C&E (Kai et al., 2008). Cohesion of the ppl is regulated in part by Wnt11, which regulates endocytosis of E-cadherin through interactions with Rab5 to modulate cells’ adhesive properties (Ulrich et al., 2005). This is consistent with evidence that Wnt11/Fz7 signaling plays a permissive rather than instructive role in ppl migration (Čapek et al., 2019; Ulrich et al., 2003). Cell-cell adhesion within the ppl can even influence cell fate specification, as prolonged cadherin-based contacts between cells enhance Nodal signaling and ppl cell fate as part of a positive feedback loop (Barone et al., 2017).

2.3. Paraxial mesoderm

Like the neighboring chordamesoderm, the paraxial mesoderm (region IV) also migrates anteriorly upon internalization at the embryonic margin, then switches to C&E movements at mid-gastrulation (Fig. 1C and D, maroon box). This population of cells directly abuts the chordamesoderm and ultimately gives rise to adaxial and somitic mesoderm that contributes to embryonic muscle, skeleton, and skin. Although the onset of C&E behaviors in both paraxial and the neighboring chordamesoderm is essentially simultaneous and utilizes many of the same cellular mechanisms, the rate of extension within the paraxial mesoderm is smaller compared to the axial compartment, resulting in shearing between these tissues (Glickman et al., 2003). Paraxial mesoderm cells are less densely-packed than chordamesoderm cells, and exhibit two distinct cell behaviors that cooperate to drive C&E: polarized medial (planar) intercalations and polarized radial intercalations. During planar medial intercalation, cells converge toward the dorsal midline and intercalate within the plane between their anteroposterior neighbors. Polarized radial intercalation, by contrast, entails the movement of cells vertically into a deeper/shallower cell layer to preferentially separate neighboring cells along the AP axis, and thus contributing to tissue extension (Yin et al., 2008).

As in the chordamesoderm, PCP signaling components become asymmetrically localized to the anterior or posterior edges of intercalating PSM cells at the onset of C&E (Roszko et al., 2015; Yin et al., 2008) and regulate both polarized planar and radial intercalation behaviors. Cell intercalations occur in PCP mutants, but their directional bias is lost. For example, in tri/vangl2;kny/gpc4 double mutants, cells undergoing either planar or radial intercalations are significantly less likely to separate AP neighbors and significantly more likely to separate ML neighbors, resulting in embryonic axes that are severely shortened and widened (Yin et al., 2008). This intercalation behavior requires ML polarization of cell bodies and protrusions, both of which are reduced in paraxial mesoderm cells upon loss of the Wnt/PCP components Wnt5b, Vangl2, and Gpc4 (Kilian et al., 2003; Yin et al., 2008).

2.4. Lateral mesoderm

Regions II and III of C&E are contained within the lateral mesoderm, which gives rise to the heart, blood and kidney among other tissues. These cells are more loosely packed than those in regions IV and V, but still undergo dorsal convergence (Sepich et al., 2005). The speed of this convergence is the main distinction between region II, which converges slowly, and region III, which is defined by faster dorsal convergence (Myers et al., 2002) (Fig. 1B and D, pink boxes). Unlike the regions discussed above, slow dorsal convergence within region II is independent of PCP signaling (Jessen et al., 2002). These cells are not highly ML elongated, and instead migrate along complex trajectories (Jessen et al., 2002), exhibiting “front and back” polarity and a distinct “tumble-and-run” mode of movement during which individual cells alternate between intervals of protrusion-driven dorsally-directed migration (run) and less directed blebbing behavior (tumble) (Diz-Muñoz et al., 2016). Bleb-driven migration is promoted by reduced membrane-cortex attachment (MCA) and increased Myosin contractility, which is thought to increase intracellular pressure and push regions of cytoplasm outward into the space between the membrane and cortex (Diz-Muñoz et al., 2016; Ruprecht et al., 2015). As in the ppl, the ratio of run to tumble is regulated by levels of MCA, which is mediated by ERM (Ezrin/Radixin/Moesin) proteins (Fig. 2C), and tipping the balance toward either run or tumble disrupts migration speed and persistence during gastrulation (Diz-Muñoz et al., 2016). Indeed, excessive Myosin contractility induced by loss of the Myosin Phosphatase Mypt1 increases blebbing within lateral mesoderm, disrupting C&E morphogenesis (Weiser, Row, & Kimelman, 2009). Migratory behavior within the lateral mesoderm also requires signaling via the GPCR Agtrl1b/Aplnrb and its ligands Apelin and Toddler (also known as Apela and Elabela), the latter of which was shown to act as a “motogen” to promote cell motility without providing directionality (Pauli et al., 2014). Loss of Apelin/Toddler signaling disrupts mesendoderm internalization movements (Pauli et al., 2014) and later dorsal convergence of lateral mesoderm cells by impairing normal protrusive activity, resulting in wider and shorter body axes (Zeng, Wilm, Sepich, & Solnica-Krezel, 2007). Gα_12/13 are also required for convergence in this region, as cells with reduced Gα_12/13 are rounder in shape and migrate dorsally with reduced speed and persistence (Lin et al., 2005) (Fig. 2C). Mathematical modeling suggests that these convergence movements are regulated by a chemoattractant(s) emanating from dorsal midline (Sepich et al., 2005), but their identity remains elusive.

Fibronectin and laminin ECM components are laid down between nascent germ layers during zebrafish gastrulation concurrent with the onset of C&E (Latimer & Jessen, 2010). ECM assembly is abnormal in several different PCP loss of function mutants (Dohn, Mundell, Sawyer, Dunlap, & Jessen, 2013), implicating cell-matrix adhesion as an important regulator of gastrulation morphogenesis in zebrafish. Indeed, abnormal ECM organization and increased expression of the ECM receptor Integrinβ1b are thought to contribute to C&E defects resulting from mutations in the transcription factor pitx2c (Collins et al., 2018). The matrix remodeling enzyme MMP14 is also required for proper cell polarity, migration speed, and axial extension, as its knockdown produces phenotypes similar to PCP mutants (Coyle, Latimer, & Jessen, 2008). PCP vangl2/tri mutant zebrafish embryos contain less Fibronectin than WT while gpc4/kny mutants exhibit a denser ECM network (Dohn et al., 2013; Williams et al., 2012), implying that both increased and decreased cell-matrix adhesion can adversely affect dorsal mesoderm migration.

Despite the relatively low density of lateral mesoderm cells, cell-cell adhesions are also critical for proper convergence movements. Although the most dramatic phenotype of N-cadherin (cdh2) loss-of-function mutants is within the neuroectoderm, close inspection revealed that lateral mesoderm cells are less ML elongated and exhibit slower and less persistent migration (Warga & Kane, 2007). Similar but even more severe mesoderm phenotypes are observed in cdh2 gain-of-function mutants, which manifest shorter and wider mesoderm domains (Warga & Kane, 2007). Cadherin complexes are essential for cell-cell adhesion, as demonstrated by the catastrophic loss of embryonic integrity in cdh1 (E-cadherin) and ctnna1 (α-catenin) mutants (Kane et al., 1996; Kane, McFarland, & Warga, 2005), but they also have important signaling functions. Epithelial integrity in ctnna1 mutants can be rescued by a ctnna1 construct lacking the mechanosensitive Vinculin binding site, but these embryos still exhibit severe C&E defects (Han et al., 2016). This suggests that in addition to their adhesive function, adherens junctions regulate C&E cell movements (at least in part) through mechanosensitive signaling.

Unlike the slow convergence in region II, fast convergence within region III is PCP-dependent. PCP signaling is required for proper ML polarization and migratory speed/persistence of these lateral mesoderm cells, but notably, not their direction of migration (Jessen et al., 2002; Topczewski et al., 2001). This reinforces the notion that PCP is essential for cells to sense and/or efficiently respond to directional cues, but does not itself comprise the hypothesized dorsal convergence cue.

2.5. Ventral mesoderm

Like other mesodermal precursors, the ventral mesoderm (region I) initially moves toward the animal pole upon internalization, but unlike other regions of zebrafish mesoderm, it exhibits neither convergence nor extension and instead begins at mid-gastrulation to move vegetally as the blastoderm undergoes epiboly (Myers et al., 2002) (Fig. 1B and D, orange box). The ventral mesoderm contributes to the tailbud upon completion of epiboly, giving rise to posterior mesoderm in post-gastrulation stages. Evidence supports the notion that C&E is actively inhibited within this population by the BMP signals that specify ventral fates via at least two distinct mechanisms. First, BMP suppresses expression of PCP signaling components on the embryo’s ventral side in parallel to fate specification (Myers et al., 2002). Second, BMP signaling modulates cell-cell adhesion within the ventro-lateral mesoderm that generates an “adhesion gradient” that increases from ventral to dorsal (von der Hardt et al., 2007). Reduced adhesion between cells prevents neighbors from gaining traction on one another, thereby preventing dorsal convergence movements. This example illustrates how embryonic patterning cues regulate morphogenesis, and it is one of few in which underlying molecular mechanisms have been identified.

3. C&E OF THE ENDODERM

In zebrafish, mesoderm and endoderm lineages are intermingled as they internalize at the blastoderm margin during gastrulation (Warga & Nüsslein-Volhard, 1999). Despite their physical proximity, endoderm cells immediately diverge from the cell behaviors of their mesodermal neighbors. Unlike the dorsal- or anterior-directed migration exhibited by the mesoderm, upon internalization the sparser endoderm layer disperses evenly across the surface of the yolk by random walk of individual cells (Fig. 3A). This migration is characterized by high speed but low persistence and is poorly correlated between neighboring endoderm cells. Random walk behavior is cell-autonomous, as it is exhibited by isolated endoderm cells and requires neither other endoderm cells nor mesoderm to occur (Pézeron et al., 2008). This movement is achieved through the elaboration of abundant and randomly-oriented actin-rich protrusions which prefigure cell body translocation and are therefore thought to directly drive endoderm cell displacement (Pézeron et al., 2008; Woo et al., 2012).

Fig. 3.

Cell behaviors underlying convergence & extension of zebrafish endoderm. Endoderm cells move by random walk prior to the onset of C&E (A), then switch to dorsal migration at mid-gastrulation (B). (C) Endoderm migration is regulated by Nodal and Apelin/Toddler signaling and is tethered to the underlying mesoderm by increased Integrin-based adhesion downstream of Cxcl12/Cxcr4 chemokine signaling.

Random walk behavior is intrinsic to endoderm cells and is induced by the same signaling cascade that specifies endoderm fate: high levels of Nodal signaling and its downstream transcriptional target, sox32. Cell-autonomous activation of Nodal signaling is sufficient to induce random walk behavior in isolated endoderm cells, but not when endoderm specification is blocked by simultaneous loss of sox32 function (Pézeron et al., 2008). Nodal signaling promotes actin dynamics underlying random protrusive behavior through activation of Rac1 via the Rac-specific guanine nucleotide exchange factor (GEF) Prex1 (Fig. 3C), which is also a transcriptional target of Nodal signaling (Woo et al., 2012). This cellular mechanism is notably similar to that regulating migration of lateral and anterior axial mesoderm (ppl, described above), another tissue induced by high levels of Nodal signaling. Given that directionality of migration is conferred to the ppl by interactions with other ppl cells (Dumortier et al., 2012), lack of cell-cell contacts may contribute, together with sox32 expression, to the distinction between Nodal-induced random walk exhibited by endoderm and directed migration by ppl cells.

Although other cell types are not required for random walk behavior, interactions with the overlying mesoderm limits (or “tethers”) endoderm migration toward the animal pole (Nair & Schilling, 2008). This tethering is mediated by signaling between Cxcl12 secreted by the mesoderm and its cognate chemokine Cxcr4 receptor in the endoderm. Rather than mediating chemotaxis, Cxcl12-Cxcr4 signaling modulates Integrin expression levels within endoderm cells to enhance their adhesion to the underlying ECM and consequently to limit their migration (Nair & Schilling, 2008) (Fig. 3C). Cxcl12-expressing mesoderm can recruit endogenous endoderm to ectopic sites, providing potential evidence for chemotaxis (Mizoguchi, Verkade, Heath, Kuroiwa, & Kikuchi, 2008), but this result could also be explained by enhanced endoderm adhesion to sites of increased Cxcl12-Cxcr4 signaling. Toddler/Apelin signaling via their GPCR is also critical for endoderm migration, as its loss prevents the dispersal of cells away from the embryonic margin during gastrulation (Pauli et al., 2014). However, the effect of Toddler on endoderm migration appears to be indirect via its requirement for mesoderm migration. Toddler only participates in migration of the endoderm when it is tethered to the underlying mesoderm, and this requirement is removed upon loss of Cxcr4 (Norris et al., 2017), indicating that the primary role for Toddler/Apelin signaling is within the mesoderm (Fig. 3C).

At mid-gastrulation, the mode of endoderm migration switches from random walk to dorsal convergence (Fig. 3B). The latter is driven by actin-rich protrusions, which reorient toward the embryo’s dorsal side concurrent with the switch in migration behavior (Pézeron et al., 2008; Woo et al., 2012). Unlike random walk, endoderm convergence is not cell-autonomous, but rather requires mesoderm and the appearance of an environmental “convergence signal” at mid-gastrulation (Pézeron et al., 2008). Because convergence of the endoderm mirrors the movements of the overlying mesoderm and cannot occur in its absence, the switch from random walk is likely mediated by interactions with the mesoderm. In fact, the same Cxcl12-Cxcr4 signaling that tethers early endoderm is also required for convergence of late endoderm (Nair & Schilling, 2008) (Fig. 3C). Notably, mesoderm convergence is unaffected by loss of this interaction, indicating that Cxcl12 is not a “convergence signal,” but rather promotes convergence of endoderm by tethering it to the converging mesoderm (Nair & Schilling, 2008). Indeed, live in toto imaging of endoderm gastrulation morphogenesis revealed cell movement patterns that are inconsistent with chemotaxis, instead supporting a tethering model (Schmid et al., 2013). Endoderm convergence is variably affected in PCP mutants with vangl2/tri mutants showing mild defects and gpc4/kny mutants showing strong defects (Hu et al., 2018; Miles, Mizoguchi, Kikuchi, & Verkade, 2017), and may be partially explained by defective C&E of mesoderm to which the endoderm is tethered. However, Glypican4 may play a more direct and cell-non-autonomous role in endoderm convergence by promoting expression of MMP14, a matrix metalloproteinase, which degrades ECM. Because enhanced cell-matrix adhesion limits endoderm migration (Nair & Schilling, 2008), a reduction in MMP14 expression and resulting increase in ECM may contribute to the reduced endoderm convergence observed in gpc4/kny mutant embryos (Hu et al., 2018).

4. C&E OF THE NEUROECTODERM

4.1. Neural plate

Once nascent endoderm and mesoderm cells have internalized during zebrafish gastrulation, those blastoderm cells that remain in the superficial layer of the embryo beneath the enveloping layer are fated to become non-neural (ventral) and neural (dorsal) ectoderm (Fig. 4A). Unlike in mouse, chick, and frog embryos, early zebrafish neuroectoderm does not develop as a columnar or pseudostratified epithelium, but instead consists of a closely packed, cohesive sheet of non-epithelial cells that comprise the neural plate (Fig. 4B). During gastrulation, cells within the neural plate converge toward the dorsal midline to extend the AP axis in much the same way as the underlying mesoderm, exhibiting ML alignment, elongation, and intercalation (Concha & Adams, 1998; Jessen et al., 2002; Kilian et al., 2003). Despite their ectodermal nature, these cells are highly motile and extend a variety of polarized protrusions that drive their dorsal migration (Kilian et al., 2003; Love, Prince, & Jessen, 2018). As in the underlying mesoderm, these cell behaviors are regulated by PCP signaling, ECM organization, and N-cadherin (Hong & Brewster, 2006; Jessen et al., 2002; Love et al., 2018).

Fig. 4.

Cell behaviors underlying convergence & extension of zebrafish neuroectoderm. (A) The zebrafish neuroectoderm transforms from a neural plate into the neural keel after completion of gastrulation. Convergence of neuroectoderm cells toward the dorsal midline first drives extension (B), then extension stops while convergence continues, instead driving internalization of cells at the midline to form the neural keel (C). Once the neural keel is established, mirror-image divisions across the midline maintain the width of the neural keel/neural rod (D).

PCP components Prickle1 and Vangl2 become asymmetrically localized to the anterior edges of neuroectoderm cells during gastrulation (Ciruna et al., 2006; Roszko et al., 2009; Yin et al., 2008), indicating planar polarization of the tissue. Disruption of PCP signaling by mutation or morpholino knockdown of PCP components vangl2 or wnt5b reduces ML cell alignment and elongation, and loss of wnt5b reduces directed protrusive activity of neuroectoderm cells (Jessen et al., 2002; Kilian et al., 2003; Love et al., 2018), resulting in a shorter and wider neural plate. Although their ML elongation is reduced, vangl2/tri mutant ectoderm cells exhibit excessive protrusive behavior and are associated with reduced Fibronectin deposition between the ectoderm and mesoderm layers. Ectopic Fibronectin restores the number of protrusions in vangl2/tri mutant ectoderm to WT levels, but does not improve ML cell elongation or alignment, demonstrating a requirement for both PCP-dependent elongation and finely-tuned cell-matrix interactions for efficient migration (Love et al., 2018). Though less severe than the C&E defects in PCP mutants, both gain- and loss-of-function mutations in cdh2 also produce wider neural plates by the end of gastrulation (Warga & Kane, 2007). These defects worsen during later development, as described in Section 4.2. C&E of the neural plate is also severely disrupted upon loss of Nodal signaling due to reduced cell migration speed, persistence, and directionality (Araya et al., 2014). However, complete loss of Nodal signaling also results in severe mesoderm and endoderm deficiencies, and evidence suggests that Nodal regulates neural C&E largely via the mesoderm it specifies (Aquilina-Beck, Ilagan, Liu, & Liang, 2007; Araya et al., 2014; Gonsar et al., 2016; Smutny et al., 2017). This and additional interactions between germ layers will be discussed in Section 5.

4.2. Neural keel

Unlike neurulation in other vertebrate embryos, in which the epithelial neural plate folds along the medial (and in some species, lateral) hinge point(s) to fuse along the embryo’s dorsal side and form the neural tube, the zebrafish neural plate undergoes a series of complex shape changes to become the neural keel (Fig. 4), then a cylindrical solid neural rod, which subsequently opens a lumen to become the neural tube (Schmitz, Papan, & Campos-Ortega, 1993). As discussed above, tracking of neuroectoderm cells reveals robust C&E movements during gastrulation, but shortly thereafter, extension movements stop while convergence movements continue (Araya et al., 2019). Instead of contributing to extension, these converging cells begin to move deeper within the embryo as they “dive in” along the midline of the neural plate to thicken it in the dorsoventral (D-V) dimension (Fig. 4C) (Araya et al., 2019), after which this structure is known as the neural keel. This “diving in,” or internalization, and the resulting convergent thickening are accompanied by dramatic changes in cell shape and orientation of medial cells (Araya et al., 2019; Hong & Brewster, 2006), including reorientation along the D-V axis and shrinking of their apical surfaces as they internalize at the midline. This shape change is driven by Myosin II contractility, which is enriched along the dorsal midline and whose localization is regulated by N-Cadherin (Araya et al., 2019; Hong & Brewster, 2006). Disruption of C&E (by loss of PCP function) or of internalization (by loss of Myosin contractility or N-cadherin) both result in an abnormally-shaped neural keel (Araya et al., 2019; Ciruna et al., 2006; Hong & Brewster, 2006; Lele et al., 2002). Although neural plate to neural keel morphogenesis entails cell behaviors typical of epithelia, a more complete suite of epithelial characteristics is expressed only during late segmentation (Buckley et al., 2013; Girdler, Araya, Ren, & Clarke, 2013; Tawk et al., 2007).

Cell divisions play an increasingly important role at subsequent stages of neuroectoderm morphogenesis. During gastrulation stages, neural precursors within the neural plate divide parallel to the plane of the tissue (Concha & Adams, 1998; Geldmacher-Voss, Reugels, Pauls, & Campos-Ortega, 2003). Within the neural keel (and also later in the neural rod stages), however, cells switch to “mirror-image” divisions perpendicular to the tissue plane and across the midline (Fig. 4D), producing one daughter that remains on the ipsilateral side and one that crosses contralaterally (Ciruna et al., 2006; Clarke, 2009; Geldmacher-Voss et al., 2003; Hong & Brewster, 2006; Kimmel, Warga, & Kane, 1994; Tawk et al., 2007). Proper midline crossing of daughter cells requires N-cadherin (Hong & Brewster, 2006) and polarization of ParD3 and Rab11 (Buckley et al., 2013; Tawk et al., 2007), and the location and orientation of these divisions is, like many other C&E-related events, also under the control of PCP. In vangl2/tri mutants, convergence of the neural plate is delayed, but cell divisions are not, causing them to occur in lateral regions of the abnormally wide neural keel rather than at the midline (Ciruna et al., 2006; Tawk et al., 2007). The result is a mass of cells in the medial region of the neural keel which is separated from each of the lateral regions by an ectopic “midline” (Tawk et al., 2007). Although reduced C&E is the primary defect within the neural plate of vangl2/tri mutants, the resulting ectopic divisions are responsible for their dramatic “double neural tube” phenotype, as inhibiting mitosis substantially improves neural keel morphology (Ciruna et al., 2006). The symmetry of cell divisions is similarly lost within the highly disorganized neural keel of Nodal signaling mutants, but these appear to be a symptom rather than the cause of this phenotype, as normal morphology is not restored to these mutants by blocking cell divisions (Araya et al., 2014).

5. INTERACTIONS BETWEEN GERM LAYERS INFLUENCE C&E

We have described above the cell behaviors that drive C&E morphogenesis of each of the germ layers within the zebrafish gastrula and many of the molecular mechanisms known to regulate them. However, none of these tissue movements occur in isolation, and disruption of C&E within one cell layer often influences the behavior of cells in neighboring compartments. Some of these interactions are chemical in nature, like the Cxcr4-Cxcl12 signaling that tethers migration of endoderm and mesoderm (described in Section 3), while others, like the influence of the notochord boundary on axial mesoderm (Section 2), are thought to be largely mechanical. Here we discuss additional examples of interactions between germ layers that influence C&E morphogenesis.

5.1. Mesoderm-neuroectoderm interactions

As discussed briefly in Section 4, C&E defects within the neuroectoderm of Nodal signaling mutants have been largely attributed to the lack of mesoderm within these embryos. In a phenotypic series of Nodal mutants or Nodal inhibitor-treated embryos, the degree of anterior mesoderm deficiency correlates well with the severity of neural tube closure defects (Aquilina-Beck et al., 2007; Gonsar et al., 2016). Furthermore, restoring mesoderm to Nodal signaling-deficient embryos is sufficient to restore neural tube morphology (Araya et al., 2014), indicating that neuroectoderm morphogenesis is tightly coupled to the mesoderm. Indeed, cell movements of the neuroectoderm and underlying mesoderm are highly correlated during normal gastrulation, but this correlation is severely reduced upon simultaneous loss of the ECM components Fibronectin and Laminin, resulting in neural tube defects similar to those seen in Nodal signaling mutants (Araya, Carmona-Fontaine, & Clarke, 2016). These results indicate that coupling of meso- and neuroectoderm by the ECM is essential for proper neural C&E. The anterior axial mesoderm was further found to promote C&E of the overlying neuroectoderm through E-cadherin-mediated friction forces between the two tissues (Smutny et al., 2017). Together, these results demonstrate the critical role of mesoderm interactions in C&E of the neuroectoderm.

5.2. Endoderm-mesoderm interactions

We described above how molecular tethering to the mesoderm is essential for proper endoderm migration during early gastrulation, but evidence suggests that endoderm can reciprocally influence morphogenesis of the overlying mesoderm at later stages. The miles apart mutation, which disrupts sphingosine-1-phosphate signaling, was so named for its cardia bifida phenotype resulting from the failure of mesodermal cardiac precursors to converge at the midline (Stainier et al., 1996). This effect on cardiomyocytes is cell-non-autonomous, and was instead found to be secondary to defective endoderm convergence (Xie, Ye, Sepich, & Lin, 2016; Ye & Lin, 2013; Ye, Xie, Hu, & Lin, 2015). Mutations in Toddler/Apelin signaling components also produce severe heart defects due to reduced migration of cardiac precursors (Zeng et al., 2007), and the Apelin receptor was similarly found to function cell-non-autonomously during heart development (Paskaradevan & Scott, 2012). Together these data indicate that convergence of at least the lateral plate mesoderm is coupled to that of the underlying endoderm. However, given that endoderm relies on mesoderm migration at earlier stages, it remains unclear precisely which cells are guiding which, and when.

6. Concluding remarks

In this chapter, we have described the cell behavior programs that drive convergence & extension of the zebrafish embryonic axis and, to the extent that they are understood, the signaling events that regulate them. While a subset of the cell behavior programs employed by zebrafish gastrulae, such as “tumble-and-run” migration, are not known to contribute significantly to axis extension in other model organisms, many other cellular and molecular mechanisms are highly conserved between multiple tissue types and vertebrate species, including PCP-dependent mediolateral intercalation, coupling of germ layers by ECM, and chemokine-directed migration (as described in other chapters of this volume). Due to their rapid external development, availability of transgenes and other genetic tools, and their optical clarity, zebrafish have proven an advantageous system in which to characterize both the detailed behavior of individual cells and their exquisite coordination that drives embryo-wide gastrulation movements. Because of the high degree of conservation of these behaviors, this wealth of zebrafish-derived knowledge is likely to inform the mechanisms at play in other vertebrate species.