Ancestral Regulatory Circuits Governing Ectoderm Patterning Downstream of Nodal and BMP2/4 Revealed by Gene Regulatory Network Analysis in an Echinoderm

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.

Goosecoid is required to initiate a stomodeal regulatory sub circuit and to repress ciliary band genes

Low levels of goosecoid transcripts are present maternally then their abundance increases sharply at swimming blastula stage, shortly after the peak of Nodal expression [35] (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 [38] (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) [42]. 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 [63]–[65]. Overexpression of goosecoid also expanded the expression of deadringer as reported previously by Bradham et al. [22], [66]. 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 [35], [40]. 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 [7].

Tbx2/3: an early regulator of dorsal gene expression downstream of BMP2/4 signaling

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.

IrxA: a negative regulator of onecut/hnf6 downstream of tbx2/3

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: an upstream positive regulator of ciliary band gene expression

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 [67]. Mice mutant for gfi1 are deaf and ataxic while flies mutant for senseless lack sensory organs indicating that Gfi proteins regulate sensory organ development [67], [68]. 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.

Positive and negative feed back loops downstream of Nodal and BMP signaling

Wnt8 signaling is required for maintenance of nodal expression

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 [53]. 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.

Chordin and Smad6 act as negative regulators of BMP signaling on the ventral and dorsal side respectively

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 [26]. 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.

Discussion

Essential roles of Nodal and BMP2/4 in patterning of the ectoderm along the D/V axis

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 [24], [25] while BMP2/4 may heterodimerize with BMP5/8 to specify the dorsal ectoderm as shown in vertebrates and in Drosophila [69], [70]. 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 [71] reinforce the hypothesis that an ancestral function of Nodal may have been in the regulation of D/V axis formation in deuterostomes.

Goosecoid, a repressor that drives a stomodeal regulatory sub circuit and represses ciliary band and dorsal genes

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 [72]. 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 [35], [38], [40]. 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 [73]. Since Goosecoid is a transcriptional repressor [74], 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 [75], [76]. Similarly Schnurri, represses the expression of brinker to allow the expression of Dpp target genes in Drosophila imaginal discs [77]. One candidate for a repressor acting downstream of goosecoid is the transcriptional repressor ZEB1/Smad Interacting Protein 1 (Sip1) [78]. 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 [31] (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 an early mediator of BMP2/4 signaling

Tbx2/3 has a special status amongst dorsal genes since it is one of the earliest zygotic genes expressed on the presumptive dorsal side [40], [41]. 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 [13], [26], [40], [41]. 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 [40]. 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 [79]. 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 [80]. 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 [81], [82]. 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.

IrxA- and BMP2/4-dependent repression of ciliary band gene expression

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 [26]. 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 [83], [84]. 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 [29]. 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 [85]. 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.

Extracellular and intracellular modulators of Nodal and BMP2/4 signaling

admp2 and glypican5: two components of a positive feedback loop downstream of BMP2/4?

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 [12], [13], [14], [86]. 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 [26]. 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 [26]. 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 [87], [88], [89], [90]. In the sea urchin, both dorsal and ventral halves of partial embryos can regulate and regenerate a complete D/V axis [91]. 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.

Wnt signaling and maintenance of nodal expression

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 [13]. 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 [92]. 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 [93]. 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.

Restriction of BMP signaling by Chordin is essential for D/V axis formation

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 [26]. 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 [26]. 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.

A new model of specification and regionalization of the ectoderm: the ciliary band as the default state of the ectoderm in the absence of Nodal and BMP signaling

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 [48], [56], [94]. 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 [26], [27]. 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 [26].

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 [27]. 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 [26]. 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 [95], [96] and in the mouse, lack of Nodal signaling causes precocious neural differentiation [97]. 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 [98], 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.

Comparison with previous GRNs

Another recent GRN analysis of ectoderm specification in S. purpuratus was performed using nanostring technology [29]. 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 [13]. 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 [29]. 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 [29], [46]. 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 [99], [100]. 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.

Nodal and BMP inhibition as an ancestral mechanism of neural induction?

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 [101], [102]. 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 [97], [103]. 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 [104], [105], [106]. 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 [107]–[109]. Inhibition of BMP signaling does not appear to play a role in this process [110] while Nodal plays a distinct, inductive role in patterning of the neural plate [111]. 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 [80].

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 [112] and annelids [113], 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 [111], [114], [115].

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.

Materials and Methods

Animals, embryos, and treatments with recombinant proteins and inhibitors

Adults sea urchins (Paracentrotus lividus) were collected in the bay of Villefranche-sur-Mer. Embryos were cultured as described previously [116], [117]. 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.

cDNA sequences and cloning of full-length transcripts

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 [30] 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 [100], the conclusions derived from previous functional studies of deadringer in S. purpuratus, which relied on a truncated sequence, are probably erroneous.

Characterization of the temporal and spatial expression of regulatory genes of the network

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 [118] 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. [119].

Overexpression analysis

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 [120]. 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 [40]. 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).

Perturbation analysis with morpholinos

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 [12]–[14], [26]. Since morpholinos can have side effects or display toxicity or produce variable reductions in gene activity [121], 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 [13], [14], [26]. 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 [13], [14], [26]. The phenotypes observed were always consistent with the zygotic expression pattern of the targeted genes and with previous well-established functional data [13], [14], [26], [35], [42], [122]. 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 [26]. 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.

Supporting Information

Attachment 1

Attachment 2

Attachment 3

Attachment 4

Attachment 5

Attachment 6

Attachment 7