|Home | About | Journals | Submit | Contact Us | Français|
Conceived and designed the experiments: AS EH FL VD ER RR TL. Performed the experiments: AS EH FM NB MQ FL VD ER RR AO LB TL. Analyzed the data: AS EH FL VD ER RR TL. Contributed reagents/materials/analysis tools: PW. Wrote the paper: TL.
Echinoderms, which are phylogenetically related to vertebrates and produce large numbers of transparent embryos that can be experimentally manipulated, offer many advantages for the analysis of the gene regulatory networks (GRN) regulating germ layer formation. During development of the sea urchin embryo, the ectoderm is the source of signals that pattern all three germ layers along the dorsal-ventral axis. How this signaling center controls patterning and morphogenesis of the embryo is not understood. Here, we report a large-scale analysis of the GRN deployed in response to the activity of this signaling center in the embryos of the Mediterranean sea urchin Paracentrotus lividus, in which studies with high spatial resolution are possible. By using a combination of in situ hybridization screening, overexpression of mRNA, recombinant ligand treatments, and morpholino-based loss-of-function studies, we identified a cohort of transcription factors and signaling molecules expressed in the ventral ectoderm, dorsal ectoderm, and interposed neurogenic (“ciliary band”) region in response to the known key signaling molecules Nodal and BMP2/4 and defined the epistatic relationships between the most important genes. The resultant GRN showed a number of striking features. First, Nodal was found to be essential for the expression of all ventral and dorsal marker genes, and BMP2/4 for all dorsal genes. Second, goosecoid was identified as a central player in a regulatory sub-circuit controlling mouth formation, while tbx2/3 emerged as a critical factor for differentiation of the dorsal ectoderm. Finally, and unexpectedly, a neurogenic ectoderm regulatory circuit characterized by expression of “ciliary band” genes was triggered in the absence of TGF beta signaling. We propose a novel model for ectoderm regionalization, in which neural ectoderm is the default fate in the absence of TGF beta signaling, and suggest that the stomodeal and neural subcircuits that we uncovered may represent ancient regulatory pathways controlling embryonic patterning.
Echinoderms (sea urchins, starfish, etc.) are marine invertebrates that share a close ancestry with vertebrates. Their embryos offer many advantages for the analysis of transcriptional circuits that control developmental programs. During early development of the common sea urchin Paracentrotus lividus, a signaling center located within the ventral ectoderm sends two key signals, Nodal and BMP2/4, that control patterning of the embryo along the whole dorsal-ventral axis. How this signaling center works is not understood. We have conducted a large-scale functional analysis of the genes responsible for patterning of the ectoderm along the dorsal-ventral axis. We identified direct targets of Nodal and BMP2/4 and identified several key regulators that mediate the effects of these factors and drive essential and probably ancient regulatory circuits that together constitute a transcriptional program controlling morphogenesis of the embryo. In addition, we uncovered a striking parallel between the mouse embryo and the sea urchin embryo by showing that in both models a neurogenic ectoderm is the default state of ectoderm differentiation in the absence of Nodal and BMP signaling. Our results support the idea that inhibition of Nodal and BMP signaling was probably an ancient mechanism to specify neural cells in the ancestor of vertebrates.
It is becoming increasingly apparent that most developmental processes are controlled by dozens or hundreds of regulatory genes assembled into complex gene regulatory networks (GRNs), rather than by a small number of master genes. By describing the functional relationships between these genes, GRNs allow integration of various levels of information on the activity of transcription factors and signaling pathways that regulate developmental processes. Over the last few years, a number of GRNs have been elucidated, including regulatory networks that drive specification of germ layers or organs in various organisms –.
Sea urchin embryos offer many advantages for GRN analysis . Unlike vertebrates, sea urchin embryos have a relatively small number of cells (about 800 cells in a gastrula) are fully transparent, and their embryos, available in huge number, develop rapidly as free-swimming larvae. A panoply of techniques is available for the functional analysis of developmental genes including treatments with pharmacological inhibitors and exogenous ligands, microinjection of antisense morpholino oligonucleotides for gene loss of function, and overexpression of mRNA for gain of function. Analysis of the first full sea urchin genome sequence from Strongylocentrotus purpuratus has revealed that echinoderms have a vast genetic repertoire but a low level of genetic redundancy, with almost all developmental regulatory genes being present as single copy . Furthermore the sea urchin embryo has a rich history of experimental embryology and a wealth of biological knowledge is available on various aspects of its development. Finally, echinoderms occupy a basal position within the deuterostome lineage and are more related to chordates than most other invertebrate phyla. These various properties mean that echinoderms are a key phylum to study the evolution of developmental mechanisms and to understand the evolutionary origin of certain features of the chordate body plan. Axis specification has been extensively studied in the sea urchin . Pioneer studies on endomesoderm patterning have shown that it is possible to dissect a complex GRN without the use of classical genetics by combining cis-regulatory and functional analysis, embryological, cell biological and genomic/computational approaches . However, while considerable knowledge is available regarding the functional relationships between genes controlling specification of the territories along the animal vegetal axis, much less was known until recently on the genes that regulate ectoderm patterning and morphogenesis of the embryo along the dorsal-ventral axis. This gap started to be filled recently by the identification in Paracentrotus lividus of the TGFβ Nodal, Univin, and BMP2/4 as key regulators of ectoderm patterning –. Nodal is expressed zygotically, starting at the 32-cell stage. Its expression is initially very broad then it is rapidly restricted to a discrete sector of the ectoderm that corresponds to the presumptive ventral ectoderm. The restricted expression of nodal is so far the earliest known regional difference in zygotic gene expression detectable along the dorsal ventral axis. However, experiments performed at the beginning of the century have shown that as early as the 8-cell stage, respiratory gradients, visualized by mitochondrial cytochrome oxidase activity, prefigure the dorsal-ventral axis of the early embryo . In addition, orientation of the dorsal-ventral axis can be biased by using respiratory inhibitors or by culturing embryos in hypoxic conditions –. Recent studies reported that mitochondria are asymmetrically distributed in some batches of eggs of Strongylocentrotus purpuratus with the ventral side displaying the highest concentration, and that microinjection of purified mitochondria can bias orientation of the dorsal-ventral axis , . A possible link between the transcriptional activation of nodal and these redox gradients is suggested by the finding that the stress activated kinase p38 is required for nodal expression . An attractive model therefore emerges in which an asymmetry in the distribution of mitochondria may generate a redox gradient, which would activate p38 anisotropically leading to the spatially restricted expression of nodal. However, strong experimental evidence supporting this model are presently lacking and experimental manipulations that perturb the redox gradient have very modest effects on the spatial expression of nodal  (Thierry Lepage unpublished results). If the role of redox gradients in the establishment of nodal expression is still unclear, in contrast, the role of a reaction diffusion mechanism, which involves a short range Nodal positive autoregulation and a long range inhibition mechanism by the Nodal antagonist Lefty, is probably essential to convert a subtle initial anisotropy into a sharply defined pattern .
Overexpression of nodal strongly ventralizes the embryos and largely mimics the effects of treatments with nickel chloride , knockdown of Nodal function using morpholinos or by overexpressing lefty, completely eliminates dorsal-ventral polarity and results in embryos with disorganized skeletal elements, no mouth and a straight archenteron. The same, strongly-radialized, phenotypes are obtained by blocking translation of the univin transcript which encodes a Vg1/GDF1 ortholog expressed maternally , suggesting that Univin may either act upstream of nodal expression or that it may heterodimerize with Nodal as suggested in vertebrates , . Intriguingly, in the absence of Nodal, not only is the expression of ventral marker genes such as brachyury, goosecoid or lefty abolished, but the expression of dorsal marker genes such as tbx2/3 and of the novel transmembrane protein 29D is suppressed as well . As a consequence, most of the ectoderm (except the ectoderm surrounding the animal and vegetal poles) of Nodal morphants differentiates into a thick ectoderm consisting of cuboidal ciliated cells that morphologically resembles the neurogenic ectoderm of the ciliary band. Injection of synthetic mRNA encoding either Nodal or an activated Nodal receptor into one blastomere of Nodal morphant embryos at the 8-cell stage is sufficient to rescue both the ventral and the dorsal side of these embryos, indicating that a distinct relay molecule specifies dorsal fates. This relay molecule was recently identified as BMP2/4, which is transcribed in the ventral ectoderm downstream of Nodal signaling, has a strong dorsalizing activity when overexpressed, and mediates the “rescue” of dorsal structures when Nodal signaling pathway is ectopically activated in a cell-autonomous manner in a Nodal loss of function background . Furthermore, despite its ventral transcription, BMP2/4 has been shown to trigger receptor mediated signaling exclusively on the dorsal side of the embryo. Based on this series of findings, a basic model for sea urchin embryo dorso-ventral patterning emerges in which the dorsal ectoderm is induced by BMP2/4 signals emanating from the opposite side of the embryo. The ventral side produces inducing factors such as Nodal and BMP2/4 but it is also a source of inhibitors such as Lefty, which restricts Nodal signaling to the ventral side, and Chordin, which prevents BMP2/4 signaling in the ventral ectoderm. In the absence of lefty function, Nodal signaling is unrestricted and propagates throughout a large belt of cells surrounding the embryo while in the absence of chordin, ectopic BMP2/4 signaling occurs on the ventral side and causes abnormal patterning of the embryo , . Therefore, in the sea urchin as in vertebrates patterning of the embryo critically relies on sequential inductive events mediated by Nodal and BMP2/4 and on the interplay between ligands and their antagonists. However, in the sea urchin embryo, both the ligands (Nodal and BMP2/4) and their antagonists (Chordin and Lefty) are co-expressed in the ventral ectoderm, which may represent a D/V organizer, and D/V patterning requires translocation of BMP2/4 from the ventral side where it is produced to the dorsal side where it activates its receptor.
Another pathway that plays a crucial role in ectoderm patterning is the Wnt pathway. Wnt signaling from the vegetal pole region is required to restrict formation of the animal pole domain. The animal pole domain is a small ectodermal territory made of thick ciliated ectoderm that forms in the apical region of the embryo. This six3 expressing neurogenic territory appears to be specified at mesenchyme blastula stage and is thought to be resistant to Wnt and TGF beta signaling , , , . When the Wnt pathway is blocked by overexpression of cadherin or of a dominant negative form of TCF, the animal plate expands towards the vegetal pole and most of the ectoderm differentiates into neuroectodem, which contains scattered serotonergic neurons normally restricted to the animal plate region . In contrast, inhibition of Nodal/Vg1/Activin signaling with a pharmacological inhibitor of the Nodal receptor causes formation of a thickened ciliated ectoderm, but this ciliated ectoderm does not appear to be specified as animal plate ectoderm since serotonergic neurons remain localized to the animal pole in these embryos. Instead, this ectoderm may have a ciliary band like identity as first proposed by Duboc et al. . This idea is supported by the finding that the ectoderm of Nodal morphants abundantly expresses the ciliary band marker tubulinß3  and by the presence of ectopic neurons as revealed by staining for the pan-neural marker synaptotagmin . However, more in depth analysis of the specification state of this ectoderm in the absence of Nodal signaling is required to further test this idea.
Deciphering the gene regulatory network that controls patterning of the ectoderm is of special importance for several reasons. The first reason is that patterning of all three germ layers relies on the activity of a signaling center located in the ventral ectoderm and analyzing how this signaling center works is essential to understand how dorsal ventral polarity of the embryo is established. Another reason is that, despite a wealth of information available on establishment of D/V polarity during normal and regulative development, the GRN that controls specification of the main ectodermal territories (ventral ectoderm, dorsal ectoderm and ciliary band) remains incompletely described and the molecular mechanisms involved in regionalization of the embryo along the D/V axis in normal and perturbed embryos have just started to be investigated . A third reason to study the D/V GRN comes from the basal evolutionary position of echinoderms within the deuterostome superclade, and of the notion that studying D/V axis formation in echinoderms will contribute to better understand the evolution of the patterning mechanisms that shaped the deuterostome body plan. Indeed, recent studies have shown that this GRN relies extensively on cell interactions mediated by TGF beta family members such as Nodal, Univin/Vg1 and BMP2/4, molecules that play crucial roles during vertebrate development , , . Finally, since major morphogenetic processes such as mouth formation, skeleton formation and elongation of the arms and apex of the larva occur along the D/V axis, dissecting the D/V GRN offers the promise to study how morphogenetic processes are encoded in the genomic program of development. This will help to fill the gap that presently exists between our understanding of cell fate specification and our knowledge of how genes work together to regulate morphogenesis.
We previously described the core of the GRN that acts downstream of Nodal and is responsible for patterning of the ectoderm along the dorsal-ventral axis . We showed that on the ventral side, Nodal acts at the top of this GRN by regulating the expression of lefty, bmp2/4, goosecoid and brachyury while on the dorsal side BMP2/4 activates the expression of tbx2/3. Although the functional relationships between these key genes was elucidated in this initial study, recent molecular screens conducted by us (Thierry Lepage unpublished) and others  revealed that many more downstream genes are likely involved in patterning of the ectoderm along the dorsal ventral axis. A large scale effort to dissect the ectoderm GRN in S. purpuratus was recently published by Su and colleagues who used the nanostring technology to monitor the effects of gene perturbations . However, this technique, which measures RNA concentrations in whole embryos, lacks the spatial resolution that is required to analyze the changes in the complex spatial expression patterns of many developmental genes.
To understand better how the ectoderm of the sea urchin embryo is patterned by Nodal and BMP2/4 signals and to expand our provisional GRN, we conducted a large-scale study. Using a combination of gain of function and loss of function studies, and taking advantage of the amenability of Paracentrotus lividus embryos to detailed phenotypic analyses and in situ hybridization studies, we analyzed at high spatial resolution the expression and regulation by Nodal and BMP2/4 of 18 transcription factors and 8 signaling molecules that displayed a restricted expression along the D/V axis. Using an assay with recombinant proteins, we identified direct targets of Nodal and BMP2/4. Finally, by conducting a large-scale analysis of the epistatic relationships between these genes, we were able to start ordering them into a hierarchy and to identify key regulators acting downstream of Nodal and BMP2/4. Not only our results uncover novel and probably ancient regulatory circuits that drive morphogenetic processes such as mouth formation and neural induction, but they elicit a model for patterning of the ectoderm in which two successive inductive events regionalize the ectoderm into three territories: the ventral ectoderm that is specified by Nodal, the dorsal ectoderm that is specified by BMP2/4 and the neurogenic ectoderm of the ciliary band, which forms between the ventral and the dorsal ectoderm in a region protected from Nodal and BMP signaling. In addition, these findings highlight a striking parallel between the mouse embryo and the sea urchin embryo by showing that in both models a neurogenic ectoderm is the default state of ectoderm differentiation in the absence of Nodal and BMP signaling. Our analysis provides a picture of this GRN significantly different from that proposed by Su et al. in S.purpuratus and stresses the importance of the spatial resolution level in the analysis of gene regulatory networks in early embryos.
To elucidate the gene regulatory network that controls specification and patterning of the ectoderm in Paracentrotus, we first performed large scale in situ hybridization screens. In addition to a random screen initiated several years ago, which allowed us to characterize the expression of 4000 randomly selected cDNAs (Thierry Lepage unpublished), we screened a P. lividus EST database against S. purpuratus sequences encoding transcription factors and signaling molecules and analyzed the expression of all those that were expressed during development of the sea urchin embryo – (Table 1). This allowed us to assemble a list of 36 genes displaying a robust expression in either the ventral ectoderm, the dorsal ectoderm or in the ciliary band territory (Table 1) (Figure 1A, 1B). Genes expressed in the animal pole domain were largely excluded from this analysis since most of them do not display a restricted expression along the D/V axis. The expression patterns of a number of the genes presented in this study had previously been described at various degrees in S. purpuratus – but they had never been described in Paracentrotus. In addition, the expression of several genes analyzed here, including smad6, gfi1, id, admp2, BMP1, and oasis has not been described previously in either species.
The earliest asymmetrically distributed transcript that we identified in the in situ screens is the maternal transcript encoding mitochondrial cytochrome oxidase, with cleavage stage embryos frequently displaying a graded distribution of transcripts in the presumptive ectoderm (Figure 1A1). This asymmetrical distribution of a mitochondrial transcript likely reflects the asymmetrical distribution of mitochondria previously reported by Coffman and colleagues , . At the zygotic level, the first signs of tissue regionalization within the ectoderm are seen at 64/128 cell-stage with nodal and lefty transcripts starting to accumulate in the presumptive ventral territory (Figure 1A2,3) , . The second wave of zygotic genes displaying a restricted expression along the D/V axis starts at the prehatching blastula with bmp2/4 and goosecoid starting to be transcribed in the ventral ectoderm rapidly followed by fgfr1, chordin and nk2.2 at the swimming blastula stage (Figure 1A4–8) , . In Paracentrotus, there is no known example of genes displaying a restricted expression in the dorsal ectoderm before the swimming blastula stage. The first genes to be expressed in the dorsal ectoderm are nk2.2 and tbx2/3, whose expression increases abruptly in the presumptive dorsal territory after hatching (Figure 1A8,9) , , . These genes are therefore good candidates as immediate early targets of Nodal or BMP2/4 signaling and are likely to play an early role in specification of these territories. Soon after ingression of the primary mesenchyme cells, when the embryo acquires its bilateral symmetry, a third wave of zygotic genes starts to be expressed. This includes the largest number of genes such as foxA, brachyury, foxG, Delta, NK1, in the ventral ectoderm (Figure 1A10–14) , –, onecut/hnf6 and fgfA (Figure 1A15–16) – in the lateral ectoderm and glypican5, irxA, hox7, dlx, smad6, msx, id, oasis, admp2 and cyIII in the dorsal ectoderm (Figure 1A17–26) , , , , . Based on the timing of their expression, genes in this category are likely secondary targets of Nodal or BMP2/4 signaling.
Starting at the early gastrula stage, additional genes start to be expressed with a restricted pattern along the D/V axis, with ptb transcripts accumulating in the ventral ectoderm (Figure 1A27), bmp1, deadringer (dri), otx, and rkhd being expressed in a broad domain encompassing the ventral ectoderm and ciliary band territory (Figure 1A28–31) , –, and gfi1, pax2/5/8, wnt8, univin transcripts starting to be expressed in the presumptive ciliary band (Figure 1A32–36) , , , . Similarly, atbf1, unc4, wnt5 start to be expressed in the dorsal ectoderm at the early gastrula stage. (Figure 1A37–39). Finally, at prism stage, tubulinß3 transcripts accumulate in the presumptive ciliary band while transcripts encoding the sea urchin specific transmembrane protein 29D accumulate in the presumptive dorsal ectoderm (Figure 1A 40,41) , .
At the mesenchyme blastula stage, foxG (also known as Brain factor1 or Bf1) is expressed in two broad ventro-lateral stripes that largely overlap with the goosecoid expression territory (Figure 1A12), while Delta is first expressed in the ectoderm in a cluster of cells at the animal pole as well as in individual cells, possibly neurons, first on the ventral side then on the dorsal side, within the vegetal part of the foxG expression domain. At the prism/early pluteus stage, the pattern of foxG resolves into a thin belt of cells on the ventral side of the presumptive ciliary band (Figure 1A42)  while Delta expression now occurs in a salt and pepper pattern within the ciliary band and facial ectoderm (Figure 1A43) .
For simplification we divide the ectoderm into three main territories along the dorsal-ventral axis, however there are additional regional differences in gene expression that show that more than three regions can be defined (Figure 1C). For example, the homeobox gene nk1 is expressed in the ventral-vegetal ectoderm in a region fated to become the ventral supra-anal ectoderm (Figure 1C1). Similarly, several dorsally expressed genes such as msx, id, oasis, admp2 or unc4 are strongly expressed in the dorsal-vegetal region fated to become the dorsal supra-anal ectoderm (Figure 1A22–25; 38; Figure 1D2,3). Thus, the ectoderm near the vegetal pole is divided into at least two sub domains along the D/V axis. Gene expression patterns also revealed that the ventral and dorsal ectodermal regions are progressively regionalized into different domains. This is best illustrated by the dynamics of goosecoid expression. goosecoid and brachyury are initially co-expressed within the ventral ectoderm (Figure 1A5,11), but during gastrulation, the expression domain of goosecoid is progressively cleared from the center of the ventral ectoderm (Figure 1C3). While goosecoid expression is progressively shifted at the periphery of the ventral ectoderm, forming a belt of cells abutting the ciliary band, brachyury and foxA remain expressed at the center of the ventral ectoderm, where the stomodeum will form (Figure 1C2). Similarly, analysis of gene expression within the dorsal ectoderm revealed the existence of nested patterns, with genes like nk2.2, tbx2/3 and dlx (Figure 1D4, Figure 1A,9,20) being expressed in a broader domain than genes like msx, wnt5 or smad6 (Figure.1A22,29,39; 1D3,4 see also ) and genes like irxA being expressed in a sub domain of the dorsal ectoderm that excludes the dorsal apex (Figure 1D1). Finally, sub regions can also be recognized within the ciliary band territory starting at the early gastrula stage, with genes like fgfA, vegf, pax2/5/8 and sprouty being expressed in the ventral lateral region (Figure 1A33,34; Figure 1D6 and data not shown) , , genes like onecut/hnf6 or gfi1 being expressed in the entire presumptive ciliary band territory (Figure 1C5; Figure 1D5), and genes like foxG, which in vertebrates is expressed in and required for specification of the ventral telencephalon , , being expressed in a ventral subdomain of the ciliary band (Figure 1A42).
Interestingly, several genes whose expression is later confined to the ciliary band are initially expressed much more broadly in the ectoderm (Figure 1E). This is particularly apparent for glypican5, fgfA, univin, and wnt8, which are expressed in a large belt of ectodermal cells at blastula stage and also for the neural marker onecut/hnf6 which is first expressed ubiquitously, then in a broad ventro-lateral domain, and only later in the ciliary band (Figure 1E1–6) , –. This suggests that the expression of these ciliary band marker genes is initiated by broadly distributed transcription factors and later repressed on the ventral and/or dorsal sides by additional factors. As a first step to dissect the ectoderm gene regulatory network, we analyzed the regulation of these broadly expressed ciliary band genes. Since SoxB1 plays a critical role in ectoderm patterning in the sea urchin  and in the specification and maintenance of neural regions in vertebrates , we tested if SoxB1 is required for expression of ciliary band marker genes (Figure 1F). Injection of morpholinos against SoxB1 abrogated the expression of most markers of the neurogenic ectoderm of the ciliary band including onecut, gfi1, foxG, egip, fgfA, pax2/5/8, univin, wnt8 and strongly affected the spatial expression of dri and otx  (Figure 1F1–20). This result supports the idea that transcription of at least a subset of ciliary band marker genes is initiated by broadly distributed transcription factors such as SoxB1 and later restricted to the ciliary band by zygotic factors induced by Nodal and/or BMP signaling.
We next tested how Nodal and/or BMP2/4 regulate the expression of the 36 genes identified in the in situ screen. We focused on Nodal and BMP2/4 since previous studies showed that these two ligands are essential for specification and patterning of the ventral and dorsal territories. We first analyzed the effects of overexpressing nodal or bmp2/4 on the expression of ectodermal markers. Embryos were injected with nodal or bmp2/4 mRNA and the expression of the ventral, dorsal, or ciliary band markers was monitored at different stages. In most cases, results were confirmed by treatments with recombinant mouse Nodal or BMP4.
Overexpression of nodal mRNA or treatments with recombinant Nodal protein dramatically expanded the expression of nodal, bmp2/4, chordin, lefty, goosecoid and brachyury as reported previously (Figure 2) , . Overexpression of Nodal also expanded the ectodermal domain of expression of foxA and fgfr1 at mesenchyme blastula stages. Similarly, the expression domain of nk1, which is normally restricted to the ventral vegetal ectoderm, became radial in nodal overexpressing embryos. Genes expressed in the ciliary band behaved differently depending on the gene. In the case of deadringer, bmp1 and univin, which are expressed in the ciliary band and in the ventral ectoderm, overexpression of nodal expanded their expression to the whole ectoderm. In the case of wnt8, which is expressed in two broad lateral stripes at gastrula stages, as well as in the case of fgfA and its downstream target pax2/5/8, which are expressed in the ventral sub domain of the ciliary band, all expression was eliminated by exogenous nodal. However, in the case of foxG, egip, onecut/hnf6, gfi1, otx, exogenous nodal suppressed expression in most of the ectoderm except in the animal and/or vegetal most domains of the ectoderm. Overexpression of nodal increased the number of ventral-vegetal cells that normally express Delta at the early gastrula stage and, at 48h, produced ventralized embryos in which most Delta expressing cells were located at the animal pole and in the vegetal most ectoderm. Largely similar phenotypes were obtained following treatments with nickel chloride (Figure S3) although we noted intriguing differences in the behavior of a few genes including wnt8, univin, fgfA and pax2/5/8, in response to these perturbations. Overall, these data are consistent with the idea that in nodal-overexpressing or nickel treated embryos, radially expressed Nodal promotes specification of ventral ectodermal fates and suppresses specification of the ciliary band in a large equatorial region but not in the animal pole region or in the ectoderm surrounding the blastopore. One likely reason that may explain why the vegetal ectoderm is refractory to Nodal overexpression or to nickel treatment is that in these embryos, Nodal signaling is restricted to the equatorial region . The vegetal ectoderm may therefore be protected from Nodal activity by Lefty which is thought to diffuse farther than Nodal , . Consistent with this idea, in Nodal treated embryos and in nickel treated embryos, nodal expression expands to a large belt of cells in the equator and a ciliary band differentiates in the vegetal most ectoderm while in lefty morphants, which also display unrestricted Nodal signaling, ciliary band marker genes such as tubulinß3 and onecut/hnf6 are expressed in the animal pole region but not in the vegetal ectoderm (Figure 2) . Taken together, these results suggest that a Lefty dependent inhibition of Nodal signaling is required for ciliary band formation in the vegetal pole region. Finally, as expected, overexpression of nodal eliminated the expression of all the dorsal marker genes we tested including, nk2.2, tbx2/3, smad6, msx, atbf1, wnt5, admp2, unc4, hox7, dlx, and 29D (Figure 2).
Reciprocally, overexpression of bmp2/4 or treatments with recombinant BMP4 protein eliminated expression of all the ventral marker genes we tested including nodal, bmp2/4, chordin, goosecoid, foxA, lefty (not shown), brachyury, and nk1 (Figure 3). As in the case of nodal overexpression, misexpression of bmp2/4 or of the activated Alk3/6 BMP receptor (Alk3/6QD)  strongly suppressed the expression of the ciliary band markers such as bmp1, foxG, onecut/hnf6, otx, gfi1, tubulinß3, egip, dri, univin, wnt8, fgfA and pax2/5/8. However, unlike in the case of nodal overexpressing or nickel treated embryos, which conserved expression of ciliary band markers in the animal pole and in vegetal ectodermal regions, overexpression of bmp2/4 or of the activated type I BMP receptor (Alk3/6QD) efficiently eliminated the expression of all the ciliary band markers at the animal pole and in the vegetal most ectoderm as well as the expression of animal pole specific markers such as foxQ2 highlighting the very strong antagonism existing between high level of BMP2/4 signaling and specification of the animal pole and ciliary band cell fates. Finally, misexpression of BMP2/4 dramatically expanded the expression of all the dorsal marker genes including tbx2/3, smad6, nk2.2, wnt5, oasis, msx, irxA, dlx, atbf1, hox7, unc4, admp2, id and 29D.
We next sought to determine which genes are direct targets of Nodal and BMP2/4 signaling. Based on the timing of expression of the ventral or dorsal markers genes, it was expected that only a subset would be direct targets of Nodal or BMP2/4 signaling. For example, only lefty, bmp2/4, chordin, goosecoid, nk2.2, fgfr1 and tbx2/3 are expressed at swimming blastula stage, the expression of most of the other starting only at mesenchyme blastula stage. We therefore tested whether the ventral marker genes are transcribed in direct response to Nodal and whether the dorsal marker genes are transcribed in direct response to BMP2/4 signaling or if transcription of these genes requires protein synthesis. To achieve this, we treated embryos at the hatching blastula, mesenchyme blastula or gastrula stages with recombinant mouse Nodal or BMP2/4 proteins in the presence or absence of a protein synthesis inhibitor (Figure 4), and analyzed the expression of all the ventral and all the dorsal marker genes.
Short treatments with recombinant Nodal protein at blastula stage strongly induced expression of nodal, lefty, bmp2/4, chordin, goosecoid, nk2.2 and fgfr1 throughout most of the ectoderm (Figure 4A). These effects were observed even in the presence of a translational inhibitor suggesting that these genes are direct targets of Nodal signaling. In contrast, short treatments with Nodal at either mesenchyme blastula or gastrula stages failed to induce any ectopic expression of the other ventral genes such as foxA (Figure 4A) foxG, nk1, or deadringer (data not shown), which are expressed in the ectoderm starting at or after mesenchyme blastula. This suggests that these genes are indirect targets of Nodal signaling that cannot be induced during the short interval of the treatment. Interestingly, in the case of brachyury, a weak but consistent broadening of the ectodermal domain of expression was observed following treatment with Nodal. However, this effect was abolished by treatment with the protein synthesis inhibitor, consistent with this gene being an indirect target of Nodal signaling. Similarly, among all the dorsal marker genes we tested, 3 genes were strongly induced by treatments with BMP2/4, even in the presence of protein synthesis inhibitors. These were tbx2/3, nk2.2 and smad6 (Figure 4B). Short treatments with high doses of BMP2/4 failed to induce expression of irxA, dlx, msx, atbf1, hox7, id, unc4, oasis, wnt5, admp2 or glypican 5 (data not shown) suggesting that these genes may be indirect targets of BMP signaling. The very good correlation between the results of this induction assay and the timing of expression of the downstream targets of Nodal and BMP2/4 indicates that this assay predicts with good confidence the direct, and probably also the indirect, target genes of these ligands at swimming blastula stage. It should be kept in mind however, that at later stages, this assay does not allow to rule-out completely the existence of a direct input from Nodal or BMP2/4 to downstream target genes. An alternative explanation for the fact that several genes appear to be refractory to induction by recombinant Nodal or BMP4 proteins is that after swimming blastula stage, the ventral and dorsal ectoderm may no longer be competent to switch their gene regulatory networks to a state that supports expression of dorsal or ventral genes respectively.
We next attempted to determine if the activity of Nodal and BMP2/4 accounts for the restricted expression of all of the ventral and all the dorsal genes. Embryos were injected with a nodal morpholino and the expression of ventral, dorsal or ciliary band markers analyzed at successive stages (Figure 5). Expression of all the ventral marker genes that we tested including, bmp2/4, goosecoid, fgfr1, nk1, chordin, brachyury, foxA and lefty disappeared in the Nodal morphants, consistent with previous results (Figure 5B) , , . Injection of the nodal morpholino also largely prevented expression of foxG, confirming that this gene is induced downstream of Nodal signaling . We also found that in Nodal morphants, the expression of all dorsal markers genes was strongly downregulated in most of the ectoderm, with responses falling into two categories: for some genes, e.g. glypican5, oasis, msx, dlx, hox7, wnt5, smad6, or unc4, expression completely disappeared in the Nodal morphants (Figure 5C). Others, e.g. tbx2/3, id, irxA, nk2.2, atbf1, admp2 and 29D displayed residual expression in the vegetal-most ectoderm and/or in the PMCs indicating Nodal-independent expression of these genes in the presumptive dorsal vegetal ectoderm.
A striking result was obtained when we analyzed the expression of ciliary band markers in the nodal morphants (Figure 5D). The expression of most ciliary band markers dramatically expanded to most of the ectoderm following inhibition of Nodal signaling. This was the case for fgfA, bmp1, univin, wnt8, otx, pax2/5/8, onecut/hnf6, gfi1, dri, as well as of the late ciliary band marker tubulinß3 and the ciliary band antigen 295. Importantly, expression of Delta, which at pluteus stages identifies individual neurons of the facial ectoderm and ciliary band region , , was expanded to the whole ectoderm in Nodal morphants, strongly suggesting that most of the ectoderm is converted into neurogenic ectoderm in these embryos. Largely similar results were obtained using a pharmacological inhibitor of the Nodal receptor  (Figure S4). Taken together, these results show that Nodal signaling is essential for expression of all the ventral and of all the dorsal marker genes within the ectoderm. In the absence of Nodal, expression of all the ventral and dorsal marker genes is abolished and ciliary band genes are ectopically expressed throughout most of the ectoderm.
We also examined the effect of knocking down BMP signaling on the expression of the ventral, dorsal and ciliary band markers (Figure 6). As expected, we found that expression of all the ventral markers that we tested was independent of BMP2/4 signaling: nodal, bmp2/4, chordin, brachyury or foxA were expressed at similar levels and in similar domains in the controls and in the alk3/6 morphants (Figure 6B). Removing BMP2/4 or Alk3/6 function affected the expression of dorsal marker genes in a way very similar to that caused by removing Nodal: expression of most genes including wnt5, atbf1, hox7, msx, dlx, smad6, tbx2/3, unc4 was abolished while for irxA, nk2.2 and id, residual expression was still observed in the vegetal most ectoderm on the presumptive dorsal side (Figure 6C). These results confirm that expression of all the dorsal ectodermal genes stringently relies on BMP2/4 signaling and that in the absence of Nodal or BMP2/4 signals, no other signals compensate for the lack of these inducers. Again, a striking result was observed when we analyzed the expression of ciliary band markers in the bmp2/4 or Alk3/6 morphants. For all of them, including gfi1, onecut/hnf6, otx, deadringer, pax2/5/8, foxG, wnt8, fgfA, univin, bmp1 and tubulinß3, loss of BMP2/4 signaling caused a dramatic ectopic expression in the dorsal ectoderm (Figure 6D). This ectopic expression transformed the normally bilateral expression domains of fgfA, pax2/5/8, foxG, gfi1, univin, and wnt8 into a horseshoe shaped domain covering the lateral and dorsal regions and caused the expression domain of deadringer and otx to become radial. These results reveal that in addition to promoting specification of dorsal cell fates, an essential function of BMP2/4 signaling is to repress ciliary band gene expression within the dorsal ectoderm.
To establish the functional hierarchy between key ventral, dorsal and ciliary band genes, we designed morpholinos against 17 transcription factors and 8 signaling molecules expressed within the ectoderm with a restricted pattern along the dorsal-ventral axis. Among these 25 morpholinos, 19 (alk4/5/7, alk3/6, brachyury, bmp2/4, chordin, foxA, foxG, fgfA, goosecoid, irxA, lefty, tbx2/3, dlx, msx, nodal, onecut/hnf6, soxB1, univin, wnt8) gave a clearly recognizable morphological phenotype (Figure 5–9).9). The expression of 15 transcription factors (goosecoid, brachyury, foxA, nk1, nk2.2, tbx2/3, msx, smad6, hox7, irxA, onecut, gfi1, dri, pax2/5/8, foxG) and 8 signaling factors (nodal, bmp2/4, fgfA, chordin, wnt8, univin, wnt5, glypican5) was analyzed at different stages in the 17 morphant backgrounds while in the case of nodal and bmp2/4 morphants we analyzed the expression of an additional set of 17 marker genes (Tables S1, S2). In addition, we overexpressed a subset of genes encoding transcription factors (goosecoid, foxA, foxG, deadringer, nk2.2, tbx2/3, msx, smad6) and signaling molecules (nodal, bmp2/4, chordin) and analyzed the expression of ventral, dorsal and ciliary band genes in these embryos. Since many of the genes identified in our screens including brachyury, foxA, otx, smad6, tbx2/3, wnt5, oasis, univin, wnt8, rkhd, ptb, fgfA, Delta are expressed not only in the ectoderm but also in the mesendoderm and since many other markers such as atbf1, irxA, nk2.2 or egip, oasis, wnt5, glypican5, wnt8, Delta, otx or bmp1 are expressed in more than one region and sometimes in both the ventral and dorsal ectoderm, we used in situ hybridization rather than QPCR to monitor the consequences of the perturbations. In situ hybridization is usually not used as the primary technique in large-scale projects such as gene regulatory network analysis since it is time and effort consuming and requires large numbers of injected embryos. However, we believe it is the only technique that provides the necessary spatial resolution to accurately analyze the expression of genes with complex expression patterns in perturbed embryos. Furthermore, when used with appropriate controls, in situ hybridization can provide a good estimate of the level of expression in perturbed embryos compared to controls. To provide a temporal view of the consequences of these perturbations and avoid secondary effects, the expression of the genes analyzed in response to nodal or bmp2/4 overexpression was examined at two different stages, soon after the onset of their restricted expression, and at a later stage, most often early or late gastrula stage depending on the gene analyzed.
Information derived from these perturbations analyses was combined with earlier results to build a provisional gene regulatory network. The main features of this gene regulatory network are described below.
Low levels of goosecoid transcripts are present maternally then their abundance increases sharply at swimming blastula stage, shortly after the peak of Nodal expression  (Figure S2). Expression of lefty, chordin, bmp2/4, fgfr1 and goosecoid, was unchanged in the goosecoid morphants consistent with these genes being direct targets of Nodal signaling and with previous studies  (Figure 7A and data not shown). Interestingly, at gastrula stages, strong ectopic expression of wnt8, univin and foxG was detected in the ventral ectoderm of goosecoid morphants indicating that one function of Goosecoid is to repress expression of these three genes in the ventral ectoderm between blastula and gastrula stages. In contrast, ectodermal expression of foxA and brachyury, two likely indirect targets of Nodal required for mouth formation, was lost in the goosecoid morphants, consistent with the lack of stomodeum in these embryos (Figure 7A) . Reciprocally, overexpression of goosecoid caused a dramatic expansion of foxA and brachyury (Figure 7B). Therefore, in the sea urchin as in vertebrates, brachyury and foxA are targets of Nodal signaling but unlike in vertebrates, in the sea urchin, they are not primary targets of Nodal since their expression depends on the zygotic expression of goosecoid –. Overexpression of goosecoid also expanded the expression of deadringer as reported previously by Bradham et al. , . In contrast, the two dorsal marker genes hox7 and msx failed to be expressed in the goosecoid overexpressing embryos consistent with previous studies showing that goosecoid overexpression suppresses expression of dorsal genes such as tbx2/3 and spec1 , . Overexpression of goosecoid also abolished the expression of all the other ciliary band genes that we tested including wnt8, univin, foxG, egip, gfi1 and onecut/hnf6. Taken together these observations suggest that goosecoid plays a double function, first by allowing expression of stomodeal genes such as foxA and brachyury and second by suppressing the expression of ciliary band and dorsal genes.
Once goosecoid and foxA have been turned on, Brachyury and FoxA cross regulate each other so that brachyury maintains foxA expression while foxA promotes brachyury expression (Figure 7C, 7D). When the function of either of the two genes was blocked with a morpholino, expression of the other gene was lost and the resulting embryos developed without a stomodeum. The role of these cross regulatory interactions between brachyury and foxA may be to stabilize and lock the specification of the ventral ectoderm that has been initiated by Nodal as described in the endomesoderm GRN, for example between the transcription factors hex and tgif .
Inhibition of tbx2/3 function strongly perturbed establishment of dorsal-ventral polarity resulting in embryos with a rounded shape, which lacked ventral arms and had a strongly reduced dorsal region (Figure 8A). Molecular analysis revealed that ventral markers such as chordin, foxA or brachyury were expressed in tbx2/3 morphants, albeit with reduced levels compared to controls (Figure 8A). A similar slight reduction was observed for the ciliary band markers onecut/hnf6, fgfA and pax2/5/8. In contrast, inhibition of tbx2/3 function abolished the expression of several dorsal genes encoding transcription factors including msx, dlx, irxA and atbf1 while the expression of other genes such as smad6, glypican5, oasis and wnt5 appeared unaffected. These results identify tbx2/3 as a key regulator of dorsal gene expression downstream of BMP2/4.
Since loss of BMP2/4 or Alk3/6 signaling causes ectopic expression of ciliary band genes in the dorsal ectoderm, it follows that in unperturbed embryos, a transcriptional repressor must act in the dorsal ectoderm downstream of BMP2/4 to prevent expression of ciliary band genes. Of the four transcription factors expressed in the dorsal ectoderm that we tested, only in the case of one of them did we observe robust ectopic expression of a ciliary band gene. This gene is irxA. In embryos injected with morpholinos against the irxA transcript, onecut/hnf6 expression was strikingly expanded in the dorsal ectoderm (Figure 8B). This effect was very robust and the territory in which the ectopic expression of onecut/hnf6 was observed was congruent with the expression territory of irxA. Interestingly, a small number of embryos injected with irxA morpholinos later developed with a thickened ectodermal region on the dorsal side that resembled an ectopic ciliary band (Figure 8B). This suggests that IrxA is a repressor of ciliary band genes downstream of BMP2/4.
onecut/hnf6 is of one of the earliest marker genes expressed in the presumptive ciliary band. onecut/hnf6 morphants developed with a slightly reduced D/V axis but they clearly displayed a D/V polarity and a well-developed ciliary band (Figure 8C). Nevertheless, we found that the expression of several marker genes of the ciliary band was affected in the onecut/hnf6 morphants. A reduced level of expression in the onecut/hnf6 morphants was observed in the case of pax2/5/8, foxG and dri while in the case of gfi1, no expression was detected. onecut/hnf6 is thus an upstream regulator of gfi1. Gfi proteins are conserved in C. elegans (Pag3), Drosophila (Senseless) and mice (Gfi1). In all three species, these zinc finger proteins play conserved roles in neural development . Mice mutant for gfi1 are deaf and ataxic while flies mutant for senseless lack sensory organs indicating that Gfi proteins regulate sensory organ development , . One can therefore anticipate that Gfi1 likely plays a role in neural development in the sea urchin embryo as it does in vertebrates and in flies. Since gfi1 is downstream of onecut, the ciliary band network therefore appears to be composed of at least two layers of zygotic factors.
To identify the Wnt ligands responsible for maintenance of nodal expression, we examined the expression pattern of candidate genes encoding Wnt factors and tested them for their requirement to maintain nodal expression. Of the various Wnt ligands examined, only wnt8 had a dynamic expression in the ectoderm at blastula stages (Figure 9A). Consistent with previous studies, we found that wnt8 morphants are radialized . In these embryos, nodal expression is strongly reduced or absent (Figure 9A). These findings indicate that, in addition to the essential nodal auto regulatory loop that is required early to maintain the expression of nodal, Wnt8 function is required zygotically to maintain nodal expression at blastula stages through a second positive feedback loop.
We showed previously that most of the ectoderm of chordin morphants differentiated into an enlarged ciliary band, a finding that was difficult to reconcile with the demonstrated activity of Chordin as an inhibitor of BMP signaling. We found that the abnormal patterning of the ectoderm in the chordin morphants is associated with a strongly reduced expression of ventral genes such as nodal, chordin, bmp2/4 and goosecoid and of dorsal genes such as msx and wnt5 (Figure 9B). We confirmed that following injection of the chordin morpholino, tbx2/3 was transiently ectopically expressed within the ventral ectoderm, consistent with the presence of ectopic BMP signaling in chordin morphants at mesenchyme blastula stage . However, following this transient ectopic expression, tbx2/3 expression was rapidly downregulated like that of other dorsal marker genes such as msx and wnt5 (Figure 9B). Strikingly, expression of onecut/hnf6 was de-repressed in the ectoderm of the chordin morphants at gastrula stages consistent with the reduced expression of nodal and bmp2/4 in these embryos and with the development of an expanded ciliary band at later stages.
Overexpression of smad6 phenocopied the BMP2/4 loss of function phenotype, resulting in partially radialized embryos with ectopic spicules and expanded ciliary band on the dorsal side (Figure 9C). This phenotype was correlated with the loss of msx expression and ectopic expression of onecut/hnf6. These observations together with the finding that expression of smad6 is regulated by BMP signaling, strongly suggests that in the sea urchin as in vertebrates, Smad6 acts in a negative feedback loop required for the fine tuning of BMP signaling. High-level overexpression of smad6 abolished the expression of Nodal target genes such as chordin and foxA and caused a massive ectopic expression of onecut/hnf6 throughout the ectoderm indicating that one function of Smad6 may also be to restrict Nodal signaling to the ventral side.
In this study, taking advantage of the detailed phenotypic analyses and robust in situ hybridization procedures available in Paracentrotus lividus, we analyzed with a high level of spatial resolution the expression, the regulation and the function of most of the zygotic transcription factors and signaling molecules displaying restricted expression within the ectoderm of the sea urchin embryo. This analysis allowed us to assemble a gene regulatory network, the D/V GRN, which describes the regulatory interactions between these genes and provides a framework for understanding the developmental program responsible for patterning the embryo along the dorsal-ventral axis. Several interesting conclusions emerged from the resultant GRN. First, it provides a clear demonstration that the activities of Nodal and BMP2/4 account fully for the spatially restricted expression of all the known genes of this network: Nodal controls the expression of all the genes expressed specifically in the ventral ectoderm, and through BMP2/4, the expression of all the genes expressed specifically in the dorsal ectoderm. Both overexpression of these ligands and corresponding loss of function experiments produced very strong, all or none, effects consistent with the idea that Nodal and BMP2/4 are critical inputs that drive the D/V GRN. It should be noted that despite their essential roles, Nodal and BMP2/4 are certainly not the only ligands involved in D/V patterning of the ectoderm and other ligands more broadly expressed likely cooperate with Nodal and BMP2/4 to specify the ventral and dorsal regions. In particular, Nodal may bind to its receptor as a heterodimer with Univin, a GDF1/Vg1 ortholog, as shown in other models ,  while BMP2/4 may heterodimerize with BMP5/8 to specify the dorsal ectoderm as shown in vertebrates and in Drosophila , . Nevertheless, the key roles played by Nodal in this GRN together with the essential function of Nodal factors in D/V axis formation in vertebrates and basal chordates  reinforce the hypothesis that an ancestral function of Nodal may have been in the regulation of D/V axis formation in deuterostomes.
A second key conclusion emerging from our D/V GRN is that in the sea urchin, Goosecoid is a key upstream element of a small regulatory circuit that controls mouth formation. In vertebrates ectopic expression of goosecoid promotes cell migration and induces incomplete secondary axes while loss of function studies implicate goosecoid in the function of the Spemann organizer and head formation . The function of goosecoid during development of other deuterostome embryos has not been studied. In the sea urchin, previous studies reported that both overexpression and loss of function of goosecoid strongly perturbed establishment of the dorsal-ventral axis, however the target genes of goosecoid were not known and the role of this repressor within the ventral ectoderm remained largely unclear , , . Our finding that goosecoid is a direct target of Nodal signaling strongly suggested that this gene could play a key role in specification of the ventral ectoderm downstream of Nodal. We have shown that Goosecoid likely regulates the expression of deadringer and foxG in the ventral ectoderm. Furthermore, we demonstrated that Goosecoid plays a critical role in mouth formation by regulating downstream target genes such as the stomodeal genes brachyury and foxA. This raises the possibility that an ancestral function of goosecoid may have been in the regulation of stomodeum formation. Consistent with this idea, goosecoid is expressed in the stomodeal region in both protostomes and deuterostomes and is co-expressed with brachyury and foxA in the oral region of cnidarians . Since Goosecoid is a transcriptional repressor , this suggests that zygotic goosecoid activates foxA and brachyury by repressing the expression of a transcriptional repressor, the identity of which is presently unknown (Figure 10). Similar double repression mechanisms have been described in different GRNs. For example, in the sea urchin the skeletogenic mesoderm GRN, the repressor pMar has been proposed to repress hes-C as well as unidentified repressors to allow expression of genes specific of the PMC lineage , . Similarly Schnurri, represses the expression of brinker to allow the expression of Dpp target genes in Drosophila imaginal discs . One candidate for a repressor acting downstream of goosecoid is the transcriptional repressor ZEB1/Smad Interacting Protein 1 (Sip1) . In the sea urchin embryo, Sip1 is expressed early in the presumptive ectoderm and its expression is downregulated at blastula stage, coincident with the onset of goosecoid expression  (see Figure S2 and S5). Experiments are currently being carried out in different labs to test this hypothesis.
Another important function of Goosecoid appears to be in the repression of ciliary band and dorsal genes. Overexpression of goosecoid potently repressed expression of ciliary band markers. Furthermore, knockdown of Goosecoid function caused ectopic expression of univin, wnt8 and foxG in the ventral ectoderm. However, additional repressors likely cooperate with Goosecoid in this repression since inhibition of goosecoid function, unlike inhibition of irxA on the dorsal side, was not sufficient to derepress ciliary band markers genes such as onecut within the ventral ectoderm.
Tbx2/3 has a special status amongst dorsal genes since it is one of the earliest zygotic genes expressed on the presumptive dorsal side , . Previous studies had shown that tbx2/3 is expressed dynamically in a broad dorsal territory in all three germ layers and that its expression is regulated by BMP signaling , , , . Indeed we showed that tbx2/3 is a direct target of BMP2/4 signaling in the ectoderm and that its function is required for expression of several dorsally expressed transcription factors such as msx, dlx, irxA and atbf1. Intriguingly, previous studies in Paracentrotus failed to detect any D/V polarity defect in tbx2/3 morphants . In contrast, we found that tbx2/3 is essential for D/V axis formation in this species. The reasons for this discrepancy are unclear. Interestingly, in vertebrates, tbx2 is also a target of BMP4 signaling during D/V patterning of the optic cup . Similarly, in hemichordates, which are positioned phylogenetically as the sister phylum of echinoderms, tbx2/3 is a target of BMP2/4 suggesting that key genes that drive the D/V GRN are conserved in these two closely related phyla . In vertebrates, tbx2 and tbx3, unlike brachyury, which is a transcriptional activator, act as transcriptional repressors due to the presence of a strong repressor domain in their C-terminal region , . It is therefore possible that the sea urchin Tbx2/3 protein also functions as a transcriptional repressor and that, like Goosecoid, it stimulates gene expression by relieving the repressive action of a transcriptional repressor. The identity of this hypothetical transcriptional repressor is presently unknown.
One of the most important findings of this study is the identification of irxA as a gene which acts downstream of BMP signaling to repress the ciliary band gene onecut. We previously reported that inhibition of BMP2/4 or Alk3/6 function causes an expansion of the presumptive ciliary band territory towards the dorsal side, and that this expansion is accompanied by the ectopic expression of the neural gene onecut/hnf6 . On the basis of this result we anticipated that one function of the BMP pathway in the dorsal ectoderm was to repress ciliary band gene expression and we postulated the existence of a BMP2/4 dependent repressor of ciliary band genes. We have now identified IrxA as one such repressor based on the following evidence. First, we showed that irxA expression is regulated by BMP2/4 signaling. Second, we showed that blocking irxA translation with morpholinos caused a robust ectopic expression of onecut in a sector of the dorsal ectoderm that coincides with the expression domain of irxA. Finally, it is established that Irx proteins can function as repressors by recruiting the Groucho Co-repressor , . Since irxA is downstream of tbx2/3 in the GRN, we might predict that blocking tbx2/3 function should also result in ectopic expression of ciliary band genes. Surprisingly, we never observed ectopic expression of ciliary band marker genes in tbx2/3 morphants. This observation is consistent with previous GRN studies, which reported that direct target genes are more strongly affected than indirect target genes or in other words, that when a perturbation affects the driver gene, it causes stronger effects on target genes than when the perturbation affects genes further upstream in the pathway . However, the simplest explanation is that our tbx2/3 morpholino may not be completely effective and that residual irxA expression may prevent ectopic expression of onecut in these embryos.
In vertebrates and in Drosophila, irx genes are involved in neural development . In Xenopus for example, irx1 promotes neural development by repressing bmp4 expression in the neural plate. It was therefore surprising to find that in the sea urchin embryo, irxA acts downstream of BMP2/4 to negatively regulate neural marker genes. Nevertheless, the identification of irxA as a BMP2/4 dependent repressor of ciliary band gene expression strongly supports our proposal that the default state of the ectoderm in the absence of TGF beta signaling is the ciliary band and that the ectoderm is patterned by two successive inductive events that repress the ciliary band fate on the ventral and dorsal sides.
Positive and negative feedback loops are essential components of gene regulatory networks. Indeed, in the sea urchin like in vertebrates, maintenance of nodal expression critically relies on autoregulation , , , . Another auto regulatory loop that is highly conserved in vertebrates is found in the BMP signaling pathway with BMP signaling stimulating BMP transcription. However, in the sea urchin unlike in vertebrates, BMP2/4 signaling does not stimulate its own expression since transcription of bmp2/4 occurs on the ventral side while BMP signaling occurs on the dorsal side . While BMP2/4 signaling did not induce bmp2/4 expression in dorsal cells, we found that BMP2/4 signaling instead induces expression of admp2 and promotes expression of glypican5, a positive regulator of BMP signaling . The dorsal expression of admp2 and glypican5 was expanded following overexpression of bmp2/4 (Figure 3) and abolished in Nodal and/or BMP2/4 morphants (Figure 5, Figure 6) indicating that expression of these genes is regulated by BMP2/4 signaling. The positive regulation of admp2 by BMP2/4 in the sea urchin contrasts with the regulation of admp genes in vertebrates. In Xenopus or zebrafish, bmp4 and admp are expressed at opposite poles of the embryo and are under opposite transcriptional control: high levels of BMP signaling stimulate bmp4 expression but repress transcription of admp, a property that has been correlated with the ability of dorsal halves of early embryos to regulate and re-establish a D/V axis , , , . In the sea urchin, both dorsal and ventral halves of partial embryos can regulate and regenerate a complete D/V axis . However, in this embryo, bmp2/4 and admp2 are not under opposite transcriptional control since BMP signaling does not repress but stimulates admp2 expression.
Whether admp2 cooperates with BMP2/4 and contributes to specification of dorsal cells is presently not known but one can speculate on the possible role of this ligand as part of positive feed back loop downstream of BMP2/4 signaling. Therefore, while a BMP2/4 synexpression group does not exist in the sea urchin, BMP2/4 may activate a positive regulatory loop by inducing expression of admp2 to reinforce BMP signaling within the dorsal territory. Future studies are required to clarify the function of ADMP2 during D/V patterning of the ectoderm and to determine if this BMP ligand contributes to the activation of pSmad1/5/8 signaling during normal and regulative development.
Previous studies had shown that when the canonical Wnt pathway is inhibited by using a dominant negative form of TCF, expression of nodal is lost at blastula stage . By analyzing nodal expression at early stages in cadherin injected embryos Angerer and colleagues showed that nodal expression is initiated normally in animalized embryos but that it is not maintained . They further provided evidence that the loss of nodal expression in these animalized embryos is caused by the persistence of the expression of the FoxQ2 repressor in the ectoderm. Here we have identified Wnt8 as a ligand required to maintain nodal expression. The exact period when Wnt8 signals are required to maintain nodal expression is not known. During cleavage and blastula stage, wnt8 is expressed in the vegetal pole region and starting at mesenchyme blastula stage, wnt8 expression in a large belt of ectodermal cells (Figure 9A). One possibility is that Wnt8 signals emitted by the ventral and/or lateral ectoderm are responsible for the maintenance of nodal at the beginning of gastrulation. Alternatively, Wnt8 signals may be required earlier for the restriction of foxQ2 expression at early blastula stages. Future studies are required to resolve this issue. It is remarkable that, during mouse embryogenesis, a Wnt signal is also required to maintain high levels of nodal expression in posterior cells through a TCF dependent enhancer and that in the absence of Wnt3, nodal transcription is initiated but it is not maintained . Therefore, both in the sea urchin embryo and in the mouse embryo, in addition to the nodal auto regulatory loop, a second positive regulatory input from a Wnt ligand is required to maintain nodal expression.
Lastly, this study helped to resolve a paradox regarding the chordin loss of function phenotype. Chordin morphants display an expanded ciliary band covering the animal hemisphere, a reduced dorsal side and a pair of parallel spicules . In vertebrates, chordin is expressed within the organizer and promotes neural differentiation by interrupting a BMP positive auto regulatory loop in the prospective neural ectoderm. In the sea urchin, chordin likely functions in a similar way by preventing BMP signaling within the ventral and dorsal ectoderm . Since the function of Chordin as an inhibitor of BMP signaling is conserved, the reason why most of the ectoderm adopted a ciliary band fate instead of an epidermal fate in the chordin morphants remained unclear. We showed that in addition to ectopic BMP signaling, chordin morphants suffer from a strong downregulation of nodal expression leading ultimately to derepression of ciliary band and neural genes. One possibility is that the transient ectopic BMP signaling observed on the ventral side of chordin morphants interferes with Nodal signaling and interrupts the Nodal auto regulatory loop causing the subsequent loss of bmp2/4 expression. Taken together, these findings highlight the crucial role played by chordin in D/V patterning of the ectoderm in the sea urchin embryo by showing that chordin is required for normal patterning along the whole dorsal-ventral axis.
The results obtained in this study largely support this idea that the default state of the ectoderm in the absence of Nodal and BMP signaling is a ciliary band-like ectoderm that expresses a number of neural genes and that Nodal and BMP2/4 restrict this ciliary band fate by specifying the ventral and dorsal ectoderm. The first hint that the default state of the ectoderm in the absence of TGF beta signaling is the ciliary band is that several genes whose expression is later restricted to the ciliary band territory are expressed throughout the ectoderm at earlier stages. This is for example apparent for fgfA, univin and wnt8, which are expressed in a belt of cells that includes most of the presumptive ectoderm at blastula stages. The expression of fgfA, univin and wnt8 is subsequently repressed on the ventral and dorsal sides during gastrulation thereby restricting the expression of these genes to the ciliary band domain. Several additional lines of evidence support the idea that the default state of the ectoderm in the absence of TGF beta signaling is a ciliary band and neural fate and that alternative ectodermal fates must be induced by active signaling. First, overexpression of both nodal and bmp2/4 strongly antagonized the expression of ciliary band and neural markers such as onecut, foxG and gfi1, with bmp2/4 leading to a very potent inhibition of ciliary band formation. Second, in the lefty morphants the ciliary band failed to form while in the absence of Nodal and BMP2/4 signaling, the ventral and dorsal ectodermal regions were not specified and most of the ectoderm differentiated instead into a thickened ciliated ectoderm that resembled the ciliary band ectoderm and expressed all tested ciliary band markers. These ciliary band markers were de-repressed throughout the ventral and dorsal ectoderm in the nodal morphants while in the absence of BMP2/4, which acts as a dorsal inducer, or of alk3/6, which is required to transduce BMP2/4 signals, only specification of the dorsal ectoderm was perturbed and ectopic expression of these ciliary band genes was detected only on the dorsal side. A third argument is that the presumptive ciliary band territory is also a region in which fgfA and vegf are expressed and where MAP kinase activity is high , , . Studies in vertebrates have shown that the activity of the MAP kinase ERK inhibits both BMP signaling and neuralization by phosphorylating Smad1 in the linker region thereby preventing its nuclearization. We thus predict that during normal development of the sea urchin embryo, the high MAP kinase activity present in the lateral ectoderm promotes neural fates within the presumptive ciliary band by inhibiting the activity of pSMAD1/5/8 and pSMAD2/3. Thus, in the absence of Nodal and BMP signaling, signals such as FGFA that are normally present at the level of the lateral ectoderm are ectopically expressed in the ventral and dorsal regions where they may promote ectopic neuron formation , . One last but crucial argument that supports our model of the ciliary band as a default state of the ectoderm in the absence of TGF beta signaling is that we identified irxA and possibly Goosecoid as repressors of a subset of ciliary band genes downstream of Nodal or BMP signaling. One read-out of Nodal and BMP2/4 signaling therefore appears to be active repression of the ciliary band fate as we had predicted .
Yaguchi and colleagues previously demonstrated that in the absence of Wnt signaling, most of the ectoderm differentiates as a neurogenic ectoderm that expresses markers of the animal pole . Since many ciliary band genes are also expressed in the animal pole, it could be argued that the ectopic expression of ciliary band marker genes observed following inhibition of Nodal or BMP signaling also reflects an expansion of the animal pole domain. This can be ruled out for several reasons. First, we showed that the expression of animal pole markers such as foxQ2, is unaffected in Nodal morphants or in embryos treated with a pharmacological inhibitor of the Nodal receptor. Second, Yaguchi et al. showed that the number and location of serotonergic neurons of the apical organ are unaffected by inhibition of Nodal signaling. Importantly, we showed that pax2/5/8, which is expressed in the vegetal part of the ciliary band but not in the animal pole region behaved exactly like the other ciliary band marker genes and was strongly derepressed in the ventral and dorsal ectoderm of Nodal morphants. Taken together these observations indicate that the lateral ectoderm of the prospective ciliary band, not the animal pole domain, is expanded in the Nodal morphants.
Our study suggests that specification of the ciliary band is likely initiated by a combination of maternal factors such as SoxB1 and by zygotic factors such as FGFA, Otx and Onecut/Hnf6 whose expression is initiated independently of the Nodal and BMP2/4 signals (Figure 10). These zygotic genes initially show a broad expression in the ectoderm, which then becomes restricted to the presumptive ciliary band by the activity of transcriptional repressors such as Goosecoid and IrxA expressed in the ventral or dorsal ectoderm downstream of Nodal or BMP2/4. Collectively our results suggest that the neural ectoderm of the ciliary band forms in a territory that is devoid of Nodal and BMP2/4 signaling (Figure 11). On the dorsal side, inhibition of BMP signaling appears to be sufficient to trigger formation of the ciliary band as was observed in BMP2/4 or Alk3/6 morphants or in embryos injected with low doses of smad6 mRNA. Similarly, on the ventral side, inhibition of Nodal signaling is sufficient to initiate formation of a ciliary band since BMP signaling does not occur on the ventral side but on the dorsal side . In this case, ectopic neural differentiation likely results from inhibition of ventral differentiation. This highlights that, in the sea urchin ectoderm, preventing ventral cells to differentiate downstream of Nodal signaling promotes neural differentiation just as efficiently as inhibiting BMP signaling on the dorsal side. Similarly, in zebrafish embryos, inhibition of Nodal signaling causes the transfating of prospective mesendodermal cells into neural cells ,  and in the mouse, lack of Nodal signaling causes precocious neural differentiation . Therefore, in the sea urchin embryo like in vertebrate embryo models, neural differentiation can result both from inhibition of BMP signals as well as from inhibition of other signals that regulate the fate of early blastomeres and allocate cells to embryonic territories and germ layers.
In summary, our results show that in the sea urchin embryo, the neurogenic territory of the ciliary band is not induced by an interaction between the ventral and dorsal territories as previously suggested , but that it represents the default state of the ectoderm in the absence of Nodal and BMP signaling. Nodal and BMP2/4 may therefore be regarded as factors that are required to prevent premature differentiation of ectodermal cells into neural cells as much as factors that are required for specification of the ventral and dorsal ectoderm.
Another recent GRN analysis of ectoderm specification in S. purpuratus was performed using nanostring technology . A comparison of the architecture of the gene regulatory networks derived from this study and ours reveals the expected similarities but also some major differences. A common central element in the architecture of both networks is the critical dependence of dorsal genes on non-autonomous signaling by BMP2/4, a feature already proposed previously . Another point of convergence is that both studies pointed to goosecoid and tbx2/3 as important early zygotic genes downstream of Nodal and BMP2/4: both studies identified brachyury as a downstream target of Goosecoid, and dlx and irxA as downstream targets of Tbx2/3. Finally, both studies identified foxG and deadringer as downstream targets of Nodal.
The first important difference in the architecture of the two proposed networks is that whereas our study defines the default state of the ectoderm in the absence of Nodal and BMP signals as a ciliary band-like ectoderm, the network proposed by Su et al. largely ignores formation of the ciliary band. Another important difference between the two studies concerns the dependence of ventral genes on Nodal. Su et al. argued that only part of the oral ectoderm specification system is downstream of Nodal . According to the authors, a number of regionally expressed genes including onecut/hnf6, otx2, lim1, and foxA, are activated “specifically in the oral ectoderm…exactly the same with or without nodal”, leading them to speculate that hypothetical Nodal independent early oral ectoderm signals regulate these genes in the ventral ectoderm. We do not agree with this interpretation, since from our in situ analysis, it is clear that these genes cannot be considered as oral-specific markers. Furthermore, we showed that the expression of onecut/hnf6, otx2, lim1, and foxA in the presumptive ectoderm region of Nodal morphants was not regionalized, consistent with the absence of any oral territory in these embryos. The expression of onecut/hnf6 and otx2 is first initiated in a territory much larger than the ventral ectoderm, before subsequently becoming restricted to either to the ciliary band (onecut/hnf6) or to a broader territory that also includes the ventral ectoderm (otx2). We thus interpret the continued expression of onecut/hnf6 and otx2 in the ectoderm as reflecting adoption of a ciliary band character by the entire ectoderm. Concerning foxA, the nodal-independent detection of the mRNA reported by Su et al is undoubtedly due to the abundant expression of this gene in a distinct endodermal territory, which, unlike the oral ectoderm expression, is largely Nodal-independent. The foxA example highlights the importance of using methods that allow spatial resolution to analyze the expression of genes with complex expression patterns in epistasis experiments.
According to Otim and colleagues and Su and colleagues, two genes, onecut/hnf6 and deadringer, play essential roles in the DV GRN. Using an “unconventional morpholino” that targeted a sequence 660 bp downstream of the first ATG but that did not target a splice junction, Otim et al. reported that “inhibition” of hnf6/onecut function eliminated D/V polarity and caused a radialized phenotype that strikingly resembled the Nodal loss of function. Using the same reagent, Su et al. expanded this analysis and further argued that a positive regulatory input from onecut/hnf6 is required for the expression of several key regulators such as nodal, goosecoid, lefty, chordin, and bmp2/4 , . These results are highly surprising since morpholinos are predicted to be ineffective at blocking translation when they target sequences after the first 25 bases following the initiator ATG , . Using two different and more conventional morpholinos targeting the 5′ leader or the translation start site of the P. lividus hnf6/onecut transcript, we were unable to reproduce either the striking hnf6/onecut morphant phenotypes originally reported by Otim and colleagues or the effects on nodal, goosecoid, lefty, chordin, and bmp2/4 reported by Su and colleagues. It is therefore very unlikely that onecut/hnf6, which is expressed only transiently within the ventral ectoderm, plays the crucial role proposed by these authors in this gene regulatory network. Regarding deadringer, Su et al. found that deadringer morphants display a much reduced expression of ventral genes such as goosecoid, NK1 and hes as well as a strongly reduced expression of dorsal genes such as irx, nk2.2 and tbx2.3. Again, these results are surprising since the published cDNA sequence of deadringer used by Su et al. to design their morpholino as well as the associated predictions of the translation start site of the protein are probably incorrect and correspond to a truncated protein sequence as suggested by our sequence analysis of the genomic S. purpuratus deadringer locus and the analysis of the deadringer cDNAs in Paracentrotus (Figure S1). In addition, using two different morpholinos against the P. lividus deadringer transcript, we were unable to reproduce the published drastic effects of deadringer morpholinos on the expression of ventral and dorsal marker genes. It is therefore also unlikely that deadringer plays the role that it had been previously attributed in the S. purpuratus GRN.
Finally, it has been argued that specific aboral differentiation genes such as CyIIIa and spec1 are transcriptionally activated in the aboral ectoderm long before late blastula and that this implied the existence of an early asymmetry in the aboral ectoderm that affected transcriptional activity. Su et al. postulated that this asymmetry may be a redox gradient that would directly regulate the transcriptional activity of aboral genes such as CyIIIa and tbx2/3. Our results oppose this view. In Paracentrotus, the ectodermal expression of tbx2/3 is essentially lost following inhibition of Nodal or BMP2/4 signaling. While it is true that a residual tbx2/3 expression is observed in the Nodal morphants at gastrula stage, this expression is restricted to the vegetal most regions and therefore likely reflects the response of this gene to signals that act along the animal-vegetal axis rather than response to a redox gradient along the D/V axis. Furthermore, in Paracentrotus, expression of CyIII genes is first ubiquitous and only becomes restricted to the dorsal ectoderm at mesenchyme blastula stage (see Figure S5), coinciding with the nuclear translocation of pSmad1/5/8 in dorsal cells. In other words, we never observed any marker gene that was expressed specifically in the dorsal ectoderm before the onset of BMP signaling i.e. at late blastula stage. Our observations therefore do not support the view that the asymmetrical CyIIIa or tbx2/3 expression is driven by an early red-ox gradient, at least not in Paracentrotus, but suggest that their expression is more likely driven by differential Nodal and BMP signaling along the dorsal-ventral axis.
A comparison of the mechanisms of neural induction in different species reveals both similarities and divergences regarding the signaling pathways involved. In Xenopus, inhibition of both Nodal and BMP signaling appears to be essential for neural induction, although FGF signaling is likely implicated in the early steps of this process , . Similarly, in mammals, both Nodal and BMP signaling have been involved in neural differentiation, the strongest evidence being that most epibast cells of mouse embryos mutant for nodal or bmpr1 display widespread and precocious expression of anterior neural markers , . In the chick and in zebrafish, there is strong evidence that FGF signaling regulates neural induction partly through the regulation of expression of BMP ligands and of BMP antagonists , , . In contrast in ascidians, which are basally branching but divergent chordates, FGF signals are the key players in neural induction by directly regulating the expression of neural markers such as otx –. Inhibition of BMP signaling does not appear to play a role in this process  while Nodal plays a distinct, inductive role in patterning of the neural plate . Similarly, in hemichordates, which together with the echinoderms form a sister group of the chordates and have a diffuse neural system, BMP signaling does not appear to play a role in the choice between neural and epidermis .
Our experiments in the sea urchin embryo show that inhibition of Nodal and BMP signaling is central to neural induction in echinoderms and that in the absence of Nodal or BMP signaling, most cells of the ectoderm differentiate into a neurogenic ectoderm. Since BMP signaling also regulates neural differentiation in insects  and annelids , it appears likely that inhibition of Nodal and BMP signaling may have been an ancestral mechanism to specify neural cells not only in deuterostomes but also perhaps in bilateria, and thus that the neural specification mechanisms used in ascidians and hemichordates have diverged during evolution.
Although in the sea urchin inhibition of Nodal causes the ventral ectoderm to adopt ultimately a neurogenic ectodermal fate, it should be kept in mind that our experiments also suggest that Nodal may have an early and positive role in specification and/or patterning of the neurogenic territory of the ciliary band since we showed that Nodal promotes the expression of Delta in a subpopulation of ciliary band cells and drives the early expression of the neural gene foxG. Therefore, in the sea urchin as in chordates, in addition to its general inhibitory role on neural induction, Nodal may also play a positive role in specification and/or patterning of the neural territory , , .
In conclusion, this large scale, systematic GRN analysis has allowed us to identify a number of key gene regulatory interactions and to build a provisional gene regulatory network describing specification of the three main ectodermal territories of the sea urchin embryo. It has not only uncovered key and probably ancient regulatory sub circuits that drive morphogenesis of the ectoderm, but has also allowed us to propose a new model of how specific regions of the ectoderm are induced over a default state, and of how the ectoderm is patterned by successive rounds of induction by TGF beta ligands. This relatively simple model captures most of the results derived from the functional analyses of Nodal and BMP2/4 in the sea urchin embryo and provides testable predictions for futures studies. Finally, our study illustrates the power of the GRN based approaches which can provide a global perspective on a set of genes regulating a biological process, explaining how this process works and what happens when it fails.
Adults sea urchins (Paracentrotus lividus) were collected in the bay of Villefranche-sur-Mer. Embryos were cultured as described previously , . When required, fertilization envelopes were removed by adding 2mM 3-amino-1,2,4 triazole 1 min before insemination to prevent hardening of this envelope followed by filtration through a 75µm nylon net. SB431542 (10 µM in sea water) was diluted from stocks solutions in DMSO, and embryos incubated in 24 well plates protected from light. In controls experiments, DMSO was added at 0.1% final concentration. NiCl2 was used at 0.5 mM. SB431542 and nickel treatments were performed continuously starting 30 min after fertilization. Continuous treatments with recombinant mouse Nodal (1µg/ml) and BMP4 proteins (0.5 µg/ml) (R&D) started at the 16-cell-stage and used embryos lacking the fertilization envelope. We verified with a set of 10 genes that RNA overexpression and recombinant proteins produced equivalent effects for both Nodal and BMP.
To determine if marker genes are direct or indirect targets of Nodal or BMP4 signaling, embryos at the swimming blastula/late blastula, early mesenchyme blastula stage or at gastrula stage from which the fertilization envelope had been removed were treated for 2h with recombinant proteins in the presence or absence of protein synthesis inhibitors. To block protein synthesis, puromycin or emetine was added at a final concentration of 360µM (200µg/ml) or 5µM (10µg/ml) respectively using stock solutions prepared in DMSO. In control experiments embryos were treated with 0.1% DMSO or with Puromycin at 200µg/ml or emetine at 10µg/ml. Development of the treated embryos was usually arrested 30 min after addition of the inhibitor, an indication of the effectiveness of the reagent and after 3–4h, all the treated embryos underwent a massive and brutal apoptosis, an effect characteristic of treatments with protein synthesis inhibitors. In the case of nodal, bmp2/4, lefty, goosecoid, fgfr1, chordin, nk2.2, tbx2.3, treatments were performed at the swimming blastula stage. In the case of nk1, foxA, brachyury, foxG, dlx, hox7, id, irxA, glypican5, cyIIIa, admp2, smad6 and msx, treatments were performed at the early mesenchyme blastula sage. In the case of deadringer, atbf1, msx, wnt5, irxA and dlx treatments were also performed at gastrula stage. Short treatments with Nodal or BMP4 failed to induce ectopic expression of any marker gene at gastrula stage suggesting that most of the genes expressed at this stage are indirect targets of Nodal and BMP2/4 or alternatively that at this stage, ectodermal territories are resistant to respecification by exogenous Nodal or BMP4.
Most of the genes analyzed in this study were discovered in the course of a random in situ hybridization screen using cDNA libraries from various stages (T. Lepage unpublished). Additional marker genes were discovered in a second in situ screen aimed at analyzing the expression profiles of all the transcription factors and signaling molecules expressed during early sea urchin development  using a Paracentrotus lividus EST library (http://goblet.molgen.mpg.de/cgi-bin/webapps/paracentrotus.cgi). When the isolated clones were incomplete, full-length cDNA sequences were obtained either by screening cDNA libraries with conventional methods and sequencing the corresponding clones. In certain cases, 5′RACE was performed using the Smart RACE kit (Clontech) to obtain the 5′ sequences. A list of all the Paracentrotus transcripts analyzed in this study with a summary of their temporal and spatial expression patterns is provided in Table 1 together with the corresponding accession numbers and original references describing these genes. Note that in the case of deadringer, the sequence of the Paracentrotus lividus clones diverged significantly from the published Strongylocentrotus purpuratus sequence. The published S. purpuratus deadringer transcript is predicted to encode a 490 amino acid protein. However, all the 13 independent deadringer cDNA clones that we sequenced encoded a protein 100 amino acids longer on the N-terminal side. Furthermore, translation of the S. purpuratus genomic sequence upstream of the predicted first ATG revealed the presence of a much longer open reading frame compared to the published deadringer protein sequence that encoded a protein highly similar to the deduced protein sequence from Paracentrotus (see Figure S1). This indicates that the previously published deadringer mRNA sequence was probably incorrect on the 5′ end and that the predicted deadringer protein sequence deduced from this mRNA was truncated. Since morpholinos fail to block translation when their target sequence is located after the first 25 bp following the initiator ATG , the conclusions derived from previous functional studies of deadringer in S. purpuratus, which relied on a truncated sequence, are probably erroneous.
For each gene of the network, a detailed analysis of the expression pattern was performed using digoxygenin labeled probes and in some cases, the temporal expression was analyzed by Northern blotting to verify maternal expression and to determine the exact onset of zygotic gene expression (Figure S2). In situ hybridization was performed following a protocol adapted from Harland  with antisense RNA probes and staged embryos. For marker genes expressed in ventral or dorsal territories at early stages, and for genes with complex expression profiles, double in situ hybridization was performed to confirm the orientations of the expression pattern. In this case, the two probes were hybridized and developed simultaneously. Probes derived from pBluescript vectors were synthesized with T7 RNA polymerase after linearization of the plasmids by NotI, while probes derived from pSport were synthesized with SP6 polymerase after linearization with SfiI. Control and experimental embryos were developed for the same time in the same experiments. Two color in situ hybridization was used following the procedure of Thisse et al. .
For overexpression studies the coding sequence of the genes analyzed was amplified by PCR with the Pfx DNA polymerase (Invitrogen) using oligonucleotides containing restriction sites and cloned into pCS2 . Capped mRNAs were synthesized from NotI-linearized templates using mMessage mMachine kit (Ambion). After synthesis, capped RNAs were purified on Sephadex G50 columns and quantitated by spectrophotometry. RNAs were mixed with Tetramethyl Rhodamine Dextran (10000 MW) or Texas Red Dextran (70000 MW) or Fluoresceinated Dextran (70000 MW) at 5 mg/ml and injected in the concentration range 100–800µg/ml. The nodal, bmp2/4, fgfA, univin, alk3/6QD, and chordin pCS2 constructs have been described in Duboc et al. (2004), Röttinger et al. (2008), Range et al. (2007) and Lapraz et al. (2009). The pCS2 goosecoid construct is described in . RNA derived from the following additional constructs were made (the cloning sites are indicated in parenthesis): pCS2foxA (ClaI-XbaI); pCS2deadringer (EcoRI-XhoI); pCS2foxG (ClaI-XhoI); pCs2smad6 (EcoRI-XbaI); pCS2pax2/5/8 (BamHI-XhoI); pCS2tbx2/3 (BamHI-XhoI); pCS2msx (BamHI-XhoI); pCS2nk2.2 (BamHI-XhoI).
Morpholino antisense oligonucleotides were obtained from GeneTools LLC (Eugene, OR). The nodal, BMP2/4, Alk4/5/7,Alk3/6, univin, lefty and soxB1 morpholinos are described in –, . Since morpholinos can have side effects or display toxicity or produce variable reductions in gene activity , we designed and tested several morpholinos for each gene. A pair of morpholinos that did not display toxic effects was selected for further use (a morpholino was considered toxic if it caused developmental arrest during cleavage or a massive cell death at the onset of gastrulation when injected at low doses (0.1–0.3 mM)). In the cases of nodal, bmp2/4, alk3/6, Alk4/5/7, univin and soxB1, the efficiency of the morpholino to downregulate the expression of previously characterized targets genes was systematically assessed in control experiments , , . The phenotypes observed for nodal, bmp2/4, brachyury chordin, foxA, fgfA, goosecoid, irxA, lefty, tbx2/3, dlx, msx, onecut/hnf6, soxB1, univin, wnt8 morpholinos were considered specific since they were confirmed with a separate, non-overlapping morpholino. In the case of alk3/6, alk4/5/7 and nodal, a rescue experiment had previously been performed demonstrating the specificity of these reagents , , . The phenotypes observed were always consistent with the zygotic expression pattern of the targeted genes and with previous well-established functional data , , , , , . We did not observe inconsistent phenotypes among several knockdowns except in one case, in which knocking down Tbx2/3, an upstream regulatory gene of irxA, did not cause the same effect on the IrxA target gene onecut/hnf6 as knocking down irxA itself suggesting that the tbx2/3 morphant phenotype is a hypomorphic phenotype and not a null. In the case of the ventrally expressed genes nodal and bmp2/4, we observed strong non autonomous effects consistent with the demonstrated translocation of BMP2/4 from the ventral to the dorsal ectoderm and with the role of BMP2/4 as relay downstream of Nodal . In contrast, we never observed strong effects on the expression of ventral markers by morpholinos targeting genes expressed dorsally. In three cases, (dlx, msx, foxG) a morphological phenotype was consistently observed but molecular analysis failed to detect significant perturbations in the expression of the genes analyzed. Other morpholinos pairs (deadringer, hox7, nk2.2, oasis, wnt5) gave very weak or not always reproducible phenotypes. Molecular analysis on embryos injected with these morpholinos failed to detect significant and reproducible changes in gene expression in any of the ventral, dorsal or ciliary band markers genes that we tested. In a few cases, (atbf1, klf2/4) all the morpholinos synthesized were highly toxic and were not studied further. The loss of function phenotypes of 29D, tubulinß3, egip, CyIIIa, admp2, fgfr1, pax2/5/8, unc4, nk1, id, rkhd and ptb and otx were not analyzed in this study and these genes were only used as markers in the following experiments. The sequences of all the morpholino oligomers used in this study are listed below. The most efficient morpholino of each pair is labeled with a star.
alk4/5/7 Mo 1: TAAGTATAGCACGTTCCAATGCCAT
alk3/6: Mo1: TAGTGTTACATCTGTCGCCATATTC
brachyury Mo1: AGCATCGGCGCTCATAGCAGGCATA
brachyury Mo2*: CTGGCAGAAGATGTACTTCGACGAT
bmp2/4 Mo1*: GACCCCAGTTTGAGGTGGTAACCAT
bmp2/4 Mo2: CATGATGGGTGGGATAACACAATGT
chordin Mo1*: GGTATAAATCACGACACGGTACATG
chordin Mo2: CGAAGATAAAAACTTCCAAGGTGTC
deadringer Mo1: TGCTCGCGGTAACAAGTGATTCCAT
deadringer Mo2: TTATATGGCAAAGGACTTCTACAGC
dlx Mo1: CCCACGTCAAATGAATACATCAACA
dlx Mo2: AAACACGTTTAGAATCCTCACGACT
fgfA Mo1: ACTTTCATCCATTTTCGCTTTCATG
fgfA Mo2*: ACACATTTTGGATACTTACAGCTCC
foxA Mo1: CATGGGTTCCTCCTTGAAATCCACG
foxA Mo2*: TGAAAGATTAAAGTAGCACAGTCAG
foxG Mo1*: TCCGATGAATGTGCATGAAAAACTG
foxG Mo2: CTTCTTGCTAAATACCAAGTTGGAG
goosecoid Mo1*: TGTCTGGAAGGTAATAGTCCATCTC
goosecoid Mo2: AGATCAGAGCTAACCACTTAGGACG
hnf6/onecut: Mo1: AGCCGCTGGACCTCAAACGCGAAGA
hnf6/onecut Mo2*: AAAATGATAATGTGGTCTCCGTCGC
hox7 Mo1: TGACGAAATACGAACTCGAACTCAT
hox7 Mo2: ACCACTTCATTAATAGCCAAAACCT
irxA Mo1: ATTGTGGATAACTGCTCGTCGTCAT
irxA Mo2: TTGTTGAAATCAACTTTGAGACGAT
Lefty Mo1: GGAGCGCCATGAGATAATTCCATAT
Lefty Mo2: GGAGATGGGCAAAATATGAAGATAC
msx Mo1: CGACTTGATGGAAGAAAATTATTCC
msx Mo2 : TTATCGCTTTAAGAATGACCAAGGA
NK1 Mo1: AAGCATTGAGAATCCCTAAAACTGC
NK1 Mo2: CATGTGCTCTGTTCAGACGGTCAAC
nk2.2 Mo1: ATCAACATTCATACGATGTCTCTAT
nk2.2 Mo2: ATAGTTAATTCCACACCACCCACTT
nodal Mo1*: ACTTTGCGACTTTAGCTAATGATGC
nodal Mo2: ATGAGAAGAGTTGCTCCGATGGTTG
tbx2/3 Mo1: TCGACGAACCACCAAATCTTGAGCA
tbx2/3 Mo2* : TCGGCAAAAGCCTCCGAGTCCAAAT
Oasis Mo1: CTCTTCACCTAAAAGCCCATCCATG
Oasis Mo2: CCAATTTGGGCCGTAGTCGAGGGAC
soxB1 Mo 1*: GACAGTCTCTTTGAAATTAGACGAC
soxB1 Mo2: GAAATAAAGCCAAAGTCTTTTGATG
univin Mo1*: ACGTCCATATTTAGCTCGTGTTTGT
univin Mo2: GTTAAACTCACCTTTCTAAACTCAC
wnt8 Mo1: GAACAACTGCCGTAAAGATATCCAT
wnt8 Mo2*: AACAGTCCAAATATGAAGTTCAAAC
As a control for defects related to injection and egg quality, we used morpholinos directed against the hatching enzyme gene: 5′-GCAATATCAAGCCAGAATTCGCCAT-3′ or against the Nemo like kinase transcript -5′-TCGGAGGCAGACCAGCAGCGAGAAA-3′. Embryos injected with either of these morpholinos at 1mM normally develop into pluteus larvae. Morpholinos oligonucleotides were dissolved in sterile water and injected at the one-cell stage together with Tetramethyl Rhodamine Dextran (10000 MW) at 5 mg/ml. For each morpholino a dose-response curve was obtained and a concentration at which the oligomer did not elicit non-specific defect was chosen. Approximately 2–4 pl of oligonucleotide solution at 0.5 mM were used in most of the experiments described here. For morphological observations, about 150–200 eggs were injected in each experiment. To analyze gene expression in the morphants a minimum of 50–75 injected embryos were hybridized with a given probe. All the experiments were repeated at least twice and only representative phenotypes observed in more than 80% of embryos are presented.
Sequence analysis of the deadringer cDNA. (A) Partial genomic sequence of the deadringer locus and predicted protein sequence encoded by the first two exons of the Strongylocentrotus purpuratus deadringer gene. The sequence represents the first two exons and the first intron of the deadringer gene. The translation start site predicted from the analysis of several Paracentrotus lividus cDNAs and the predicted protein sequence are shown. The solid bar in position 1285 of the nucleotide sequence indicates the ATG previously thought to encode the initiator methionine. (B) Sequence comparison of the predicted Paracentrotus lividus Deadringer protein sequence with the predicted Strongylocentrotus purpuratus Deadringer protein sequence deduced from the genomic sequence. Note that the published Spu-deadringer protein sequence starts at position 1285 and therefore is truncated of the first 100 aminoacids.
(0.92 MB PDF)
Northern blot analysis of nodal, bmp2/4, goosecoid and sip1 expression during development of the sea urchin embryo. Embryonic stages are : egg (E), 16 cells (16), 32 cell-stage (32), 60 cells (60), very early blastula (B1), early blastula (B3), swimming blastula (SB), mesenchyme blastula (MB), early gastrula (EG), late gastrula (LG), prism (Pr), Pl pluteus. Loading control is 28S mRNA.
(3.37 MB TIF)
Expression of ectodermal marker genes following ventralization with nickel chloride - Embryos were treated with nickel chloride starting after fertilization and the expression of ventral, dorsal and ciliary band genes was analyzed at the relevant stages. Treatment with nickel caused an expansion of all the ventral marker genes, and strongly repressed the expression of dorsal and ciliary band marker genes. The effects of Nickel treatments on marker gene expression are largely similar to those resulting from Nodal overexpression. However, a few intriguing differences can be noticed. For example, nickel treatment more efficiently suppressed the expression of markers of the animal pole region and ciliary band markers, such as gfi1 or onecut/hnf6, than nodal overexpression. However, in the case of fgfA, the opposite result was observed with nodal overexpression more efficiently repressing expression of fgfA in the animal pole and vegetal ectoderm regions than nickel treatment. lv, lateral view, vv, vegetal pole view, fv, frontal view.
(8.54 MB TIF)
Expression of ectodermal marker genes following inhibition of the Nodal receptor with the pharmacological agent SB432542. Embryos were treated with SB431542 at 10 µM starting after fertilization and expression of ventral dorsal or ciliary band marker genes was analyzed at the relevant stages. Blocking Nodal signaling at the level of the receptor abolished the expression of ventral and dorsal marker genes and caused ectopic expression of ciliary band genes. lv, lateral view, vv, vegetal pole view, fv, frontal view.
(6.02 MB TIF)
Expression pattern of CyIII and sip1 during early development of Paracentrotus lividus. (A) The probe used corresponds to the CyIIIb transcript and crosshybridizes with CyIIIa. CyIII genes are expressed ubiquitously at early and hatching blastula stages. Starting at late blastula stage, CyIII transcripts accumulate preferentially in the ectoderm on the dorsal side. At mesenchyme blastula and gastrula stages, CyIII expression is restricted to the dorsal ectoderm and ventral SMCs. (B) At early blastula stage sip1 is expressed in two thirds of the embryo. Double in situ hybridization a probe for skeT, a gene expressed in the skeletogenic mesodermal precursors indicates that sip1 is expressed mostly in the presumptive ectoderm derived from the animal hemisphere. lv, lateral view, vv, vegetal pole view.
(2.57 MB TIF)
Gene knockdown and overexpression experiments used to construct the network.
(0.07 MB PDF)
Gene knockdown without detectable effect as examined by in situ hybridization.
(0.06 MB PDF)
We thank members of our laboratory and colleagues at the marine Station of Villefranche for help and support. We thank David McClay for performing the immunostaining with the 295 antibody and for fruitful discussions and Evelyn Houliston for careful reading of the manuscript. We thank David Luquet and Laurent Gilletta for collecting and taking care of the sea urchins and the three anonymous reviewers for their insightful comments on the manuscript.
The authors have declared that no competing interests exist.
This work was supported by grants from the Association pour la Recherche contre le Cancer (ARC 3801 and 4908), the Agence Nationale de la Recherche (ANR), the CNRS, the University Pierre et Marie Curie Paris 6, and the Fondation Bettencourt Schueller. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.