Early patterning in a chondrichthyan model, the small spotted dogfish: towards the gnathostome ancestral state

Abstract

In the past few years, the small spotted dogfish has become the primary model for analyses of early development in chondrichthyans. Its phylogenetic position makes it an ideal outgroup to reconstruct the ancestral gnathostome state by comparisons with established vertebrate model organisms. It is also a suitable model to address the molecular bases of lineage-specific diversifications such as the rise of extraembryonic tissues, as it is endowed with a distinct extraembryonic yolk sac and yolk duct ensuring exchanges between the embryo and a large undivided vitelline mass. Experimental or functional approaches such as cell marking or in ovo pharmacological treatments are emerging in this species, but recent analyses of early development in this species have primarily concentrated on molecular descriptions. These data show the dogfish embryo exhibits early polarities reflecting the dorso-ventral axis of amphibians and teleosts at early blastula stages and an atypical anamniote molecular pattern during gastrulation, independently of the presence of extraembryonic tissues. They also highlight unexpected relationships with amniotes, with a strikingly similar Nodal-dependent regional pattern in the extraembryonic endoderm. In this species, extraembryonic cell fates seem to be determined by differential cell behaviors, which lead to cell allocation in extraembryonic and embryonic tissues, rather than by cell regional identity. We suggest that this may exemplify an early evolutionary step in the rise of extraembryonic tissues, possibly related to quantitative differences in the signaling activities, which shape the early embryo. These results highlight the conservation across gnathostomes of a highly constrained core genetic program controlling early patterning. This conservation may be obscured in some lineages by taxa-specific diversifications such as specializations of extraembryonic nutritive tissues.

Introduction

In line with the hourglass model, early development is generally viewed as highly divergent even across relatively close species. This conception, initially inferred from general anatomical comparisons and later supported by the discovery of highly conserved genetic networks acting at the so-called phylotypic stage (Duboule, 1994; Raff, 1996), has recently gained support from high throughput comparisons in very distant phyla (Domazet-Loso & Tautz, 2010; Kalinka et al. 2010; Irie & Kuratani, 2011; Levin et al. 2012). In vertebrates, an important source of variation appears related to the embryo nutrition mode and the presence or absence of extraembryonic tissues. It is generally accepted that in line with their morphological diversity and highly variable mode of formation, these tissues have independently arisen several times during evolution. For instance, comparisons across actinopterygians unambiguously support the conclusion that the yolk syncytial layer (YSL) of teleosts is a derived characteristic of the group (Cooper & Virta, 2007). Similarly, amniotes, but also chondrichthyans, have independently developed overt extraembryonic annexes, which extensively vary in their extent and complexity within each of these taxa (Blackburn, 2005; Ferner & Mess, 2011). These independent origins and diversifications, however, are at odd with two observations, which have emerged from genetics analyses of early development in the mouse, chick and zebrafish. Firstly, far from being solely involved in embryo nutrition and protection, these tissues are sources of signaling molecules, which play key roles in the formation and patterning of the embryo proper. This has been most clearly shown for the mouse anterior visceral endoderm (VE) and its homolog, the chick hypoblast, both involved in the control of axis formation and various aspects of early head formation (Srinivas, 2006; Stern & Downs, 2012). In the mouse, the extraembryonic ectoderm and proximal VE are also involved in antero-posterior patterning and the specification of primordial germ cells, while the zebrafish YSL is essential for mesendoderm induction (Chen & Kimelman, 2000; de Sousa Lopes et al. 2004, 2007; Rodriguez et al. 2005; Hong et al. 2011). Secondly, the zebrafish dorsal YSL and mouse anterior VE share expression of several regional markers and signaling molecules, which turn out to also be markers of anterior mesendoderm (Ho et al. 1999; Jones et al. 1999). From an evolutionary standpoint, these similarities are difficult to explain in view of the independent emergence of extraembryonic tissues in the major gnathostome taxa. The prevailing view is that they may result from independent recruitments of anterior mesendoderm genetic programs in extraembryonic components secondarily to their emergence in teleosts, mammals or chondrichthyans (Stern & Downs, 2012).

The small spotted dogfish Scyliorhinus canicula is an interesting model to revisit these issues. Reproductive strategies greatly vary across chondrichthyans. Most sharks are viviparous but, like most Scyliorhinidae, this species is oviparous (Blackburn, 2005; Ebert et al. 2006). It develops from a telolecithal egg and, as the chick, exhibits a distinct yolk sac connected to the embryo proper by a narrow duct ensuring nutritive exchanges with the yolk (Ballard et al. 1993). This mode of reproduction most likely represents the ancestral condition in chondrichthyans, phylogenetic reconstructions supporting the idea that viviparity is a derived characteristic, independently arisen several times during elasmobranch evolution (Blackburn, 2005; Ebert et al. 2006). Its phylogenetic position as representative of chondrichthyans, the closest outgroup to osteichthyans, also makes it a key reference to infer the gnathostome ancestral state by comparisons with the traditional vertebrate model organisms (Coolen et al. 2008). Although experimental analyses are only emerging in this species, it has entered the molecular and genomic era, with large-scale expressed sequence tags and genomic sequencing in progress. Here, we review the results of recent molecular descriptions of its development from blastula to early gastrula stages. These data highlight the conservation of a core genetic program controlling early patterning in jawed vertebrates, and suggest that the evolvability of early development may primarily concern extraembryonic tissues and their specializations when present.

General characteristics of early development in the dogfish S. canicula

Morphological and histological aspects

Morphological and histological descriptions of the early dogfish embryo have been first described more than 50 years ago (Vandebroeck, 1936; Kopsh, 1950; Nelsen, 1953), and more recently revised starting from blastula stages (Coolen et al. 2007). Fertilization is internal in the dogfish as in all Elasmobranchii, and detailed analyses of the cleavage pattern following fertilization are still lacking. Egg deposition takes place after cleavage and initial blastoderm formation, just prior to blastocoel appearance as a translucid posterior zone. At this stage, the blastoderm consists of a 1-mm-diameter blastodisc lying on top of a large undivided yolk mass (Fig. 1A). Although not organized in a distinct layer as in teleosts, syncytial nuclei are present in the proximity of the yolk membrane and they persist at least until gastrula stages (Lechenaud & Mellinger, 1993). The presence of these syncytial nuclei may in part result from polyspermy but, as described below, deep cell fusions with the yolk could also be involved in their formation. Externally, blastula stages appear marked by an expansion of the blastoderm, such that in late blastulae, its size reaches about 4 mm diameter. Molecular analyses clearly show that the animal/vegetal and organizer/ab-organizer polarity axes are already specified at the earliest blastula stages (see below), but the cellular organization of the blastoderm and its changes remain poorly characterized until stage 10 (late blastula). In particular, an epithelial organization of surface blastomeres reported more than 50 years ago remains to be confirmed. More detailed histological descriptions are available starting from stage 10. At this stage, the blastoderm clearly comprises two cell populations: a deep mesenchymal one consisting of dispersed round-shaped cells; and a superficial cuboidal epithelium. The most obvious hallmark characterizing this stage is the formation of a thickening at the prospective organizer side of the blastoderm. Three major morphological changes then take place at the transition to early gastrulation (stage 11; Fig. 1C,F). First, the upper layer converts into a pseudostratified columnar epithelium over a broad territory spanning the posterior half of the blastoderm, while it remains cuboidal in the ab-organizer sector. Histological sections also show that a novel mesenchymal cell population, showing characteristic elongated shapes, appears adjacent to the posterior margin, seemingly intermixing with the inner cell population already present. Lastly, the pseudostratified epithelium starts folding inward along a 60 ° sector of the posterior margin, to form an involuting bilayered structure. Starting from this stage, this structure extends posteriorly, to form the mesendoderm overlying the archenteron, itself delimited ventrally by the yolk membrane. This overhang is not observed at the lateral and anterior margins of the blastoderm, which go on spreading over the yolk. The embryonic axis proper becomes morphologically visible, at stage 12 (Fig. 1D). At this stage, the columnar epithelium territory has become restricted to a well-delimited sector adjacent to the posterior (organizer side) margin and containing the forming neural plate. An axial depression prefiguring the neural groove becomes visible, but there is no sharp demarcation between prospective neural plate and prospective epidermis, nor any evidence of a boundary between embryonic and extraembryonic surface ectoderm. The internalizing mesendoderm now forms two bilateral overhangs extending posteriorly and termed the posterior arms. These structures lie on each side of a midline indentation that was referred to as notochordal triangle by Kopsh (1950) or, more recently, node, based on its molecular properties and continuity with axial mesendoderm (Sauka-Spengler et al. 2004). The first indication of a segregation between mesoderm and endoderm is also observed at this stage, at the level of the posterior arms. Subsequent stages are marked by the elongation of the embryonic axis (Fig. 1E), the morphogenesis of the foregut, which rolls inward to form the foregut diverticulum (stage 13), and the formation of a characteristic head enlargement (stage 14). The elevation of the neural folds only becomes visible in the anterior-most part of the embryonic axis by late stage 13. At this level, the embryo proper becomes delimited by two grooves, which mark a transition zone with the adjacent extraembryonic yolk sac. This morphogenetic process is the only visible indication of the segregation between embryonic and extraembryonic germ layers, the lateral-most mesendoderm and adjacent extraembryonic endoderm on the one hand and the prospective embryonic surface ectoderm and extraembryonic yolk sac ectoderm on the other lying in the continuity of each other.

Figure 1. General characteristics of the early dogfish embryo. (A) Telolecithal dogfish egg. The blastoderm lies on top of the undivided yolk mass. (B–E) Dorsal view of early embryos. The organizer side of the blastoderm (posterior relative to the forming embryo axis) is to the bottom. (B) Late blastula embryo (stage 10). The most obvious characteristic of this stage is the appearance of a posterior thickening. (C) Early gastrula embryo (stage 11). This stage is marked by the beginning of mesendoderm involution at the posterior margin. (D) Axis formation (stage 12). The anterior neural plate becomes visible, as well as a characteristic posterior indentation corresponding to the organizer and referred to as notochordal triangle or node. Mesendoderm internalization takes place at the level of the posterior arms lying on each side of this structure. (E) Axis elongation and head enlargement formation (stage 14). (F) Cellular organization of a stage 11 embryo at a median sagittal section plane. Posterior is to the right. Histological sections provide indications of cell ingressions from the upper layer and of cells fusing with the yolk. Direct cell labeling highlights anterior displacements of cells detaching from the leading edge of the involuting layer and migrating as single cells. Colored arrows in (B–F) show prevailing cell movements at different levels of the embryo: dotted blue arrows: individual cell movements in the lower layer, taking place prior and anterior to the involution movement; red curved arrows: involution of the upper layer; orange arrows: epiboly; green arrows: posterior convergence. A red point indicates the organizer (notochordal triangle) in (D) and (E).

Cell movements: a temporal regulation of cell behaviors driving cell allocation to embryonic vs. extraembryonic territories

Analyses of cell movements and cell fates at blastula to gastrula stages have been hampered by difficulties to maintain early embryos alive during extended culture times after opening the eggshell (Vandebroeck, 1936; reviewed in Sauka-Spengler et al. 2004). Using vital dye-marking analyses based on short-duration marking experiments carried out at successive stages, Vandebroeck (1936) could nevertheless obtain more than 50 years ago a relatively rough description of cell movements between stages 10 and 12. These analyses led him to suggest that, in line with histological descriptions, the prevailing cell movement at blastula stages was epiboly along the whole circumference of the blastoderm, followed by internalization of the mesendoderm by involution along a 60 ° sector of the posterior margin. Even though analyses of nascent mesoderm markers such as Brachyury confirm the margin and later posterior arms as the major sites of mesendoderm internalization (see below), these data should be taken with caution. In particular, the detail of the cell movements involved should certainly be critically reassessed using modern cell-labeling techniques. As a matter of fact, several lines of evidence obtained recently indicate that much more complex movements may additionally take place during dogfish early development (Fig. 1F). Histological analyses thus strongly suggest that cells from the upper epithelial layer may ingress into the deep mesenchyme layer of mid- to early gastrula embryos (Coolen et al. 2007). Along the same line, section analyses following fluorescent immunohistochemistry using an antibody directed against β-catenin evoke the occurrence of deep cell fusions with the yolk, which could contribute to the formation of syncytial nuclei, as previously proposed (Lechenaud & Mellinger, 1993; Godard BG, Coolen M, Gombault A, Ferreiro-Galve S, Laguerre L, Wincker P, Poulain, J, Da Silva C, Kuraku S, Carre W and Mazan S, manuscript in preparation). Finally, we have recently used DiI labeling to examine cell behaviors at the level of the posterior margin at late blastula stages (Godard et al. manuscript in preparation). This analysis shows that cells start to emigrate from the posterior margin by late blastula stages, before the start of anterior mesendoderm involution (stage 10 to early 11). During this phase, which correlates with the presence of elongated cells seemingly detaching from the margin, migrating cells appear characterized by a relatively high dispersal, with clear evidence for individual cell movements. Based on their loosely packed organization and final location in the deep mesenchyme in contact with the yolk, their fate is likely to be primarily extraembryonic. In contrast, cells internalizing later, just prior to the morphological appearance of the head, remain tightly clustered within the involuting layer, likely to essentially contribute to anterior mesendoderm. In summary, these data highlight complex, temporally regulated cell behaviors in the deep cell populations of the dogfish blastula, which may control cell allocation either to the extraembryonic deep mesenchyme or to the involuting anterior mesendoderm. Some of these cellular behaviors are reminiscent of those described either in amniotes or in teleosts. For instance, the formation of the late blastula deep mesenchyme, which may involve both ingressions from the superficial layer and active migrations from the posterior margin, echoes the dual origin of the chick hypoblast, both from the epiblast and Koller's sickle. Similarly, fusions of deep cells with the yolk seem to occur in the dogfish as in the zebrafish. In view of other examples of syncytium formation occurrence in the development of early extraembryonic endoderm in as distant taxa as mammals (Flechon et al. 2007), this however may reflect a propensity of early endoderm cells to fuse and form syncytial structures rather than a territory homology.

A highly conserved molecular pattern of the forming germ layers

Despite attempts based on short labeling carried out at successive stages now more than 50 years ago (Vandebroeck, 1936), the establishment of accurate fate maps of the dogfish blastula has remained tentative due to the difficulties mentioned above to keep embryos alive for protracted culture times after opening the eggshell. In the absence of these data, molecular characterizations using candidate gene approaches have nevertheless made it possible to compare the regional pattern of the dogfish early embryo with other traditional models. These analyses have revealed a high conservation of this pattern, independently of the presence or absence of extraembryonic tissues.

Axial to lateral mesoderm patterning: an archetypal anamniote embryo

The site of mesendoderm internalization has been identified by expression analysis of the dogfish Brachyury homolog, ScT (Sauka-Spengler et al. 2003). This study has confirmed that, as suggested by cell-labeling experiments, the posterior arms are the major site of mesendoderm internalization from the first appearance of the embryonic axis (stage 12) to axis extension stages, expression later persisting at the tailbud level. They have additionally revealed a thin marginal expression extending into the lateral and anterior margins, suggesting that a minor mesoderm population may also internalize at this level. This is consistent with the presence of extraembryonic blood islands, which have been reported to form along the lateral and anterior blastoderm margins (Ballard et al. 1993). Analyses of axial to lateral mesoderm markers have also made it possible to clarify the regional identity of the margin at the onset of gastrulation and early axis extension stages (stages 12–14; Coolen et al. 2007). Axial to lateral markers exhibit nested territories along the posterior arms, which actually reflects their future location in the forming embryonic axis (Fig. 2). Expression of all lateral mesoderm markers tested (Bmp4, MafB) also extends as a thin territory to the lateral and anterior blastoderm margins, where epiboly prevails and extraembryonic blood islands later appear. Two main conclusions can be inferred from these data. Firstly, there is no evidence for a segregation between nascent extraembryonic mesoderm and embryonic lateral blood islands, which lie in the continuity of each other at the margin and express the same molecular markers at the onset of gastrulation. Extraembryonic mesoderm is thus of lateral identity in the dogfish. Secondly, the regional pattern of the site of mesendoderm internalization in the dogfish is qualitatively very similar to the one reported along the blastopore lips of amphibians, the blastoderm margin of teleost or the primitive streak or its equivalent in amniotes. One difference, however, is that in the dogfish, the nascent mesoderm of lateral identity, which comprises a major presumptive extraembryonic component in addition to the embryonic one, may extend over a proportionally broader territory of the blastoderm margin than, for instance, in teleosts or amphibians. As noted by Arendt & Nubler-Jung (1999), this is related to a restriction of the forming embryo axis to the posterior, organizer pole of the blastoderm, which is itself an adaptation to the increase of the yolk mass.

Figure 2. Regionalization of the site of mesoderm internalization: comparison between the dogfish (A), xenopus/zebrafish (B), turtle (C) and chick (D). The sites of axial, paraxial and lateral mesoderm internalization are shown in red, orange and yellow, respectively. In the turtle, mesoderm internalization only takes place at early gastrulation stages at the level of the blastoporal plate (dotted line).

A conserved Nodal-dependent, antero-posterior pattern of early endoderm independent of the presence of extraembryonic tissues

Endoderm development appears extensively divergent across vertebrates, with the presence or absence of distinct extraembryonic components and the diversity of the cellular organization of the former when present (Takeuchi et al. 2009). In amniotes, which have developed overt extraembryonic tissues, the fact that endoderm specification takes place into two phases, respectively taking place at early blastula and gastrula stages and primarily contributing to extraembryonic membranes and definitive gut endoderm, has been known for a long time. Developmental genetics have unambiguously shown that the latter phase, but not the former one, is dependent on Nodal signaling (Hoodless et al. 2001; Yamamoto et al. 2001; Robertson et al. 2003; Vincent et al. 2003). Two successive endoderm specification phases, respectively dependent and independent of Nodal signaling, have more unexpectedly also been found in teleosts (Dickmeis et al. 2001; Kikuchi et al. 2001; Sakaguchi et al. 2001; Hong et al. 2010). In the zebrafish, the extraembryonic YSL is of endodermal identity, as assessed by expression of Sox32, which codes for a zinc finger transcription factor closely related to the general endoderm differentiation marker Sox17, and its specification precedes, and actually controls, embryonic gut formation (Sakaguchi et al. 2001; Aoki et al. 2002). Such a biphasic mode of endoderm specification may even also exist in amphibians, despite the absence of distinct extraembryonic tissues in this case (Luxardi et al. 2010; Skirkanich et al. 2011). In the dogfish, analysis of endoderm markers coupled to pharmacological treatments directed vs. Nodal/activin signaling indicate that, similarly, the deep mesenchyme and syncytial nuclei are of endodermal identity, already specified at early blastula stages, and there is no evidence for an involvement of Nodal in its formation, or maintenance. In contrast, the specification of the anterior mesendoderm only takes place at late blastula stages, in a Nodal-dependent fashion (Godard et al. manuscript in preparation). These data clearly bring support to the ancestrality of a biphasic mode of endoderm development in jawed vertebrates. While no evidence for such a two-step mode of endoderm specification has been reported either in amphioxus or tunicates, the presence of two distinct endoderm components, respectively corresponding to the extraembryonic nutritive yolk mass and an embryonic component contributing to the gut proper, has also been shown in the lamprey (Takeuchi et al. 2009). The relationship between the two phases described in the dogfish and the lamprey remains to be confirmed, but this raises the possibility that a biphasic mode of endoderm specification may be a vertebrate innovation. However, most apparent in the dogfish is that these two specification events do not concern compartmentalized, respectively extraembryonic and embryonic, cell populations. Nodal/activin signaling actually appears required for the formation of the thickened posterior margin, which at the earliest stage following its appearance also contributes to the extraembryonic deep mesenchyme (Godard et al. manuscript in preparation). Along the same line, Nodal/activin signaling controls the establishment of a regional pattern spanning a broad posterior (organizer side) sector of the deep mesenchyme and adjacent involuting mesendoderm (Fig. 3; (Coolen et al. 2007; Godard et al. manuscript in preparation). This molecular pattern is established following an accurate temporal regulation (Lim1, then Hex, followed by Brachyury and Chordin), which later translates along the antero-posterior axis possibly through a mechanism related to those demonstrated at trunk levels in amphibians (Wacker et al. 2004; Durston et al. 2010). But these expression territories show no evidence for a restriction to either extraembryonic or embryonic territories, the anterior-most ones (Lim1 and Hex) actually spanning the transition zone between the deep mesenchyme and anterior involuting mesendoderm, while the posterior-most ones (Brachyury and Chordin) lie posterior to it (Fig. 3; Coolen et al. 2007; Godard et al. manuscript in preparation). This molecular continuity suggests that in the dogfish, cell fate choices between extraembryonic vs. embryonic endoderm result from cell allocation to either territory related to differential cell behaviors, rather than from distinct specification events. Intriguingly, the earliest phases of Lim1 and Hex expressions in the dogfish are strikingly reminiscent by their relative timing and nested spatial organization to those of their orthologs in the mouse embryonic VE, which also depend on Nodal signaling (Fig. 4; Mesnard et al. 2006). In both species, they are also followed by Brachyury and Chordin expressions in the forming anterior mesendoderm (Bachiller et al. 2000). Although the chronology and relative organization may be less accurately defined, Lim1 and Hex are also expressed dorsally in the zebrafish YSL and the xenopus deep endoderm, while Brachyury and Chordin are restricted to adjacent nascent mesendoderm (Ho et al. 1999; Smithers & Jones, 2002). Taken together, these data support a strong conservation of endoderm patterning mechanisms across jawed vertebrates, independently of the presence or absence of extraembryonic components. We suggest that the molecular similarities between the mouse embryonic visceral endoderm and zebrafish dorsal YSL result from their common ontogenetic origin from a single ancestral territory of anterior mesendoderm regional identity, containing intermingling cells derived from the first and second phase of endoderm specification and allocated to extraembryonic or embryonic tissues as the result of differences in cell behaviors, such as motility or adhesiveness, exactly as is the case in the dogfish. In this view, the epithelial and syncytial organizations of the mouse VE and zebrafish YSL may be related to taxa-specific specializations obscuring the ancestral continuity between extraembryonic and embryonic components, which is most apparent in the dogfish. Analyses of the origin of the YSL of teleosts by comparative analyses across actinopterygians actually support this view (Cooper & Virta, 2007). In support of this model is also the fact that in the mouse, VE, long thought to exclusively contribute to extraembryonic annexes, actually also contributes to the gut (Kwon et al. 2008). This shows that, as in the dogfish, no strict segregation between embryonic and extraembryonic endoderm is established at the onset of gastrulation, which may be a remnant of the ancestral continuity between these tissues proposed here. Accurate analyses of cell behaviors from blastula to early gastrula stages will be important to assess the generality of this conclusion at the vertebrate scale.

Figure 3. Dynamic of gene expression at the posterior margin of stage 10–13 dogfish embryos. Organizer side (posterior referring to the orientation of the axis when it appears) is to the right. Lim1 (light blue), Hex (dark blue) and Chordin (red) chronological order of expression is translated into an ordered spatial organization of their territories along the antero-posterior axis at stages 11–12. When the morphogenesis of the foregut diverticulum takes place, Hex territory is found ventrally at this level.

Figure 4. Similar specification events in the mouseVE between mouse VE and dogfish endoderm layer. Column 1, a dorsal projection of the mouse VE. Column 2, schematized sagittal section of the mouse conceptus from 4.5 to 5.5 dpc, with posterior to the right and distal to the bottom. Column 3, a schematized sagittal section of a dogfish embryo, with posterior to the right. Column 4, dorsal view with posterior to the bottom. In the mouse (columns 1–2), specification of Lim1 first (light blue) in the embryonic VE and then Hex (dark blue) in the distal VE successively takes place under the control of Nodal signaling (after Mesnard et al. 2006). In the dogfish (columns 3–4), Lim1 and Hex expression are expressed at the margin starting from late blastula stages, in the same chronological order and with the same nested spatial organization. These expressions are also dependent on Nodal/activin signaling.

Ectoderm patterning: conservation of an ancestral pre-neural territory?

Neural induction has long been thought to essentially rely on a bone morphogenetic protein (BMP) inhibition by antagonists secreted by the node, converting the Bmp4-expressing ectoderm, fated to epidermis in the absence of other signals, into neural tissue. It is now clear that neural induction involves a more complex sequence of inductive interactions than those postulated by this default model. Initial evidence for the involvement of additional earlier signals was obtained in the chick. In this species, combinations of signals secreted from the hypoblast do induce expression of several anterior neural plate markers, such as Otx2 and Sox3, but their expression remains transient in the absence of additional signals, including those emitted by the node (Streit et al. 2000; Linker & Stern, 2004; Albazerchi & Stern, 2007). This transient specification state is clearly distinct from the one later stabilized in the anterior neural plate, and is generally referred to as ‘pre-neural’ or ‘pre-forebrain’ state. More recently, experimental evidence for a requirement for Fgf signals to elicit neural induction and for the presence of a similar ‘pre-neural’ specification state, independent of the presence of the organizer, has also been obtained in xenopus (Delaune et al. 2005; Marchal et al. 2009). While experimental data are still lacking in the dogfish, analysis of the dynamic of expression of anterior neuroectoderm markers, such as Sox2 or Otx2, also suggests that neural induction in this species might also involve several distinct steps (Coolen et al. 2007). No expression of these genes is observed until late blastula stages. Until stage 10, molecular analyses suggest a relatively simple molecular partitioning of the blastoderm, exhibiting an animal–vegetal polarity (reflecting the future dorso-ventral axis and relying on a partitioning between deep, Gata6/Sox17-expressing blastomeres and superficial, Oct4-positive ones) and an ab-organizer–organizer one (aligned with the future antero-posterior axis and materialized by the restriction of Bmp4 to a broad ab-organizer sector excluding a posterior crescent adjacent to the Nodal-expressing posterior margin; Coolen et al. 2007; Fig. 5A). This pattern markedly complexifies at the transition from blastula to early gastrula, just prior to the start of involution (Fig. 5B). At this stage, two novel territories individualize in the superficial layer, with the onset of Brachyury, Otx2 and Sox2 expressions. The first, Brachyury-positive one, is ring-shaped, adjacent to the margin proper. The second, Otx2- and Sox2-positive one, spans a broad posterior sector of the blastoderm excluding the Brachyury-expressing margin (Fig. 5B). Only in a third phase, when the axis and neural plate proper becomes visible (stage 12), do Otx2/Sox2 and also Brachyury become confined to the forming anterior neural plate and involuting mesendoderm, respectively (Fig. 5C). The earliest Otx2/Sox2-positive territory is clearly distinct by its molecular properties from the anterior neural plate, which also expresses Otx1 (Coolen et al. 2007) as well as Hesx-1 (Godard, unpublished data). In addition, comparisons of this territory with the one observed in the underlying endoderm (extraembryonic deep blastomeres and anterior mesendoderm) show that it is adjacent to the broad Lim1 initial territory in the underlying deep mesenchyme, while Chordin expression starts later and remains restricted to the posterior part of the involuting layer (Coolen et al. 2007). Whether this territory may correspond to the pre-neural state described in other vertebrates remains to be assessed by direct analyses of the signals controlling its specification, but these data support the conclusion that in chondrichthyans, as in osteichthyans, neuroectoderm specification involves a complex sequence of signals, including signals preceding the BMP repression mediated by Chordin. Finally, concerning the specification of epidermis, of note is the fact that Bmp4 expression provides no evidence of a segregation between embryonic and extraembryonic components until at least stage 13, the only difference with a species devoid of extraembryonic tissues such as xenopus being the relative extent of neuroectoderm vs. non-neural ectoderm (Coolen et al. 2007).

Figure 5. Expression territories of Brachyury, Otx2, LeftyA, LeftyB and Bmp4 in the dogfish embryo from blastula to early gastrula stages. (A) Early to mid-gastrula; (B) stages 10+ to 11; (C) stage 12. Brachyury expression (purple) is in the upper layer at stages 10+ to 11, at the margin and internalizing mesendoderm at stage 12. Otx2 expression (blue) is in the upper layer at all stages shown. LeftyB (orange) expression is restricted to the posterior margin and adjacent lower layer at blastula stages. LeftyA (red) and LeftyB expressions are located at the lateral and posterior margin, excluding the midline (upper layer) at stages 11 and 12. Bmp4 (green) expression is restricted to the upper layer at blastula stages and at stages 10+ to 11, and becomes expressed in an additional territory at the lateral and anterior margins at stage 12. A faint signal is also observed at this stage in the presumptive surface ectoderm, whether embryonic or extraembryonic.

In summary, from blastula to early gastrula stages, the molecular characterization of the dogfish blastula to early gastrula highlights a typical anamniote regional pattern in deep and superficial layers. In particular, we find no qualitative evidence of differential regional identity between extraembryonic and embryonic territories at these stages, the transition between the two appearing related by differential cell behaviors, such as those observed between the deep mesenchyme and involuting mesendoderm, or those that lead to the morphogenesis of the yolk duct. This supports the idea that at least in chondrichthyans, but possibly also in amniotes, changes related to the rise of extraembryonic tissues may have initially relied on gradual, quantitative differences. Quantitative modulations of the shape or relative extent of signaling gradients known to control both regional patterns and cellular behaviors in the early embryo could be instrumental in this process. For instance, in the dogfish, an expansion of the territory submitted to the influence of BMP signaling could be one parameter affecting the relative extent of surface vs. neural ectoderm, and an initial step towards the rise of distinct extraembryonic tissues. A deeper understanding of the complex relationships between the signaling activities shaping the early embryo, and the control of cell fates and cell movements, will be crucial to gain insight into adaptations to variations in yolk content.

Patterning mechanisms: which relationship with interactions involving extraembryonic tissues in amniotes?

In the dogfish, the early distribution of Bmp4, Nodal and LeftyB (one of the two Lefty paralogs present in this species; Fig. 5A) defines a polarity, which most obviously reflects the dorso-ventral, or organizer–ab-organizer, polarity of amphibians and teleosts (Coolen et al. 2007). A conservation of patterning mechanisms between these species remains to be assessed, but would not be unexpected in view of the molecular similarities observed at gastrula stages. The relationship to amniotes is less obvious. As best described in the mouse, complex reciprocal interactions have been shown to take place between embryonic territories in the epiblast and extraembryonic territories, such as the embryonic or distal VE and the extraembryonic ectoderm, prior to gastrulation (Rodriguez et al. 2005; Richardson et al. 2006; Ben-Haim et al. 2006). No equivalent of these interactions has been reported in teleosts or amphibians. Along the same line, the anterior VE/hypoblast of amniotes is known to play a succession of distinct roles in axis and head formation, including induction of an unstable pre-neural state, control of the time and site of streak formation, early repression of the streak caudalizing activities (Stern & Downs, 2012). Some of these functions, such as the repression of caudalizing activities, seem to be shared by the anterior endoderm of amphibians and teleosts (Schneider & Mercola, 1999; Pera & De Robertis, 2000; Smithers & Jones, 2002). However, most of them are generally thought of as amniote innovations, possibly related to the relocation of the site of gastrulation to the center of the blastoderm (Stern & Downs, 2012). Whether analyses in the dogfish could question this conclusion and reveal a conservation of these functions and molecular interactions awaits functional analyses. In the absence of these data, we nevertheless note that expression analyses of signaling molecules at late blastula to early gastrula stages highlight complex and highly dynamic patterns, some of them consistent with a conservation of the mechanisms described in the mouse or the chick (Coolen et al. 2007). For instance, LeftyB is restricted to the posterior margin until late blastula stages, at which time it withdraws precisely from the site of axis (and head) formation, now extending to the lateral and anterior blastoderm margins (Fig. 5A–B). This pattern evokes the conservation of an early role in the control of the time and site of axis formation. Similarly, Bmp4 shows a highly dynamic pattern at blastula to early gastrula stages (Fig. 5A–C; Coolen et al. 2007). Initially expressed in a broad ab-organizer sector as described above, the transcripts transiently withdraw from the rim of the blastoderm at the onset of Nodal, Lefty and Brachyury expression at this level with a sharpening and intensification of the signal at the borders of the expression territory. Slightly later, as the embryonic axis becomes visible, the signal reappears at the lateral and anterior margins, together with lateral mesoderm markers, also persisting with a faint signal intensity in the prospective embryonic and extraembryonic prospective ectoderm. The second phase of expression is reminiscent of the reciprocal interactions demonstrated in the mouse between the Nodal-expressing proximal epiblast and Bmp4-positive adjacent extraembryonic ectoderm. Finally, the relative locations of Bmp4 territory and of Hex second phase of expression are consistent with a restriction of the latter by BMP signals secreted by the former. This raises the possibility that some interactions between extraembryonic tissues and epiblast in amniotes may have an equivalent in the dogfish, and suggests an unexpected unity in early patterning mechanisms across jawed vertebrates (Fig. 6; Coolen et al. 2007). Testing these working hypotheses using, for instance, pharmacological treatments directed vs. candidate signaling pathways during relatively short time windows will be crucial to more clearly define the gnathostome ancestral state and infer the diversifications that may have taken place in the broad osteichthyan lineages and their representative models. The slow development and flat blastodisc morphology of the dogfish may turn out to be an advantage in such an analysis, as they may help to reveal transient phases overlooked in fast-developing species, such as amphibians and teleosts. In particular, it may help to assess the conservation in vertebrates of the early antagonistic interaction between BMP and Nodal/Vg1 signaling recently demonstrated in amphioxus (Onai et al. 2009). Whether such an antagonism may take place at early blastula stages remains unclear in traditional vertebrate models, but is suggested by the molecular descriptions observed in the dogfish (Coolen et al. 2007).

Figure 6. Relationships between the early polarities observed in the zebrafish (A), xenopus (B), mouse (C) and dogfish (D). Molecular characterizations highlight an early organizer–ab-organizer polarity in the dogfish, unambiguously related to the dorso-ventral polarity of teleosts and amphibians. Comparisons of the dynamics of expression in early endodermal tissues between the mouse and the dogfish suggest that, in the former, this polarity is translated into a radialized, proximal to distal organization.

Concluding remarks: revisiting the hourglass model

In conclusion, analyses of the dogfish blastula to early gastrula reveal a highly conserved regional pattern with amphibians and teleosts, without qualitative alterations related to the rise of extraembryonic territories. They also suggest unexpected relationships with amniote embryos, which may reflect ancestral molecular and cellular interactions obscured by lineage-specific diversifications of extraembryonic tissues. These results suggest the existence of a highly constrained core genetic program, preserved from the general high divergence of early development (Duboule, 1994; Raff, 1996; Domazet-Loso & Tautz, 2010; Kalinka et al. 2010; Irie & Kuratani, 2011; Levin et al. 2012). This conservation may be related to key roles in embryo patterning, as most of the genes referred to here are involved in this process. Such roles are likely to constitute a strong developmental constraint, contrasting with the high evolvability of embryo nutritive modes across vertebrate evolution. In this respect, it is perhaps noteworthy that the dogfish early embryo exhibits an archetypal regional pattern, independently of the presence of distinct extraembryonic tissues. This suggests that unlike other Scyliorhinidae endowed with more complex specialized extraembryonic annexes, this species has largely retained an ancestral mode of early development. It remains to be assessed whether quantified estimates of the divergence of early development may be sensitive and possibly biased by the sampling of species analyzed. Analyzing a broad range of species, beyond traditional models that unlike the dogfish may have substantially diverged from the ancestral mode, should be important to revisit these issues and further assess the broad evolutionary tendencies of the mechanisms controlling development.

Acknowledgements

This work was funded by Région Bretagne (EVOVERT grant n°049755), National Research Agency (grant ANR-09-BLAN-026201), CNRS and UPMC. B.G. was funded by a Région Bretagne doctoral fellowship.