A combinatorial code of maternal GATA, Ets and β-catenin-TCF transcription factors specifies and patterns the early ascidian ectoderm

With the possible exceptions of mammals and polyembryonic wasps, most animal embryos studied are pre-patterned by the action of localized maternal determinants (Goldstein and Freeman,1997). Following fertilization, inheritance of these factors initiates distinct developmental programs in embryonic cells. The subsequent transcriptional activation of combinations of zygotic genes relay, amplify and refine the maternal cues. This suite of events has been best described in Drosophila in which precisely localized maternal determinants shape the embryonic axes and determine the germline. The identification of axial maternal determinants and of how they combinatorially initiate tissue-specific zygotic gene expression has had a profound conceptual impact on developmental biology (Nüsslein-Volhard,1995).

Several key vertebrate maternal axial determinants and a few of their direct zygotic targets have also been identified in Xenopus and fish(reviewed by Heasman, 2006a; Schier and Talbot, 2005),which has led to the initiation of the reconstruction of gene regulatory networks for the mesendoderm (Koide et al., 2005; Loose and Patient,2004). These networks are at present very incomplete. First, some maternal determinants are most likely missing. Second, deciphering the precise cis-regulatory logic that enables combinations of maternal factors to activate their targets in distinct territories has proved difficult owing largely to the size and complexity of the vertebrate genomes. As an additional difficulty, the early genetic program can significantly differ between amphibians and fishes. For example, the zebrafish orthologue of the major Xenopus maternal determinant of vegetal territories, VegT, is not expressed in early fish embryos (Ruvinsky et al., 1998).

In this context, studies on ascidians, members of the chordate phylum, have recently emerged as particularly informative. First, ascidian embryos have an early gastrula fate map and tadpole structure remarkably similar to lower vertebrates (Kowalevsky, 1866; Kowalevsky, 1871; Kourakis and Smith, 2005), yet use a particular mode of development based on a fixed cell lineage(Conklin, 1905; Nishida and Satoh, 1985;Nishida, 1987a), extensive use of maternal determinants(Chabry 1887; Nishida, 1987b; Nishida and Sawada, 2001) and very local inductions (reviewed by Nishida, 2005; Tassy et al., 2006). Comparison of the ascidian and vertebrate strategies should therefore allow discrimination between crucial events expected to be shared, and adaptations to specific modes of developments. Second, studies on the ascidian Ciona intestinalis have progressed at an impressive pace since the release of the draft sequence of its genome (Dehal et al., 2002), culminating recently in the first whole embryo draft zygotic gene regulatory network in a metazoan(Imai et al., 2006). This progress was made possible by a low level of genetic redundancy(Dehal et al., 2002), the availability of efficient gene gain- and loss-of-function strategies(Corbo et al., 1997; Satou et al., 2001), and the ease of identifying the cis-regulatory modules that drive zygotic gene expression. These advances open the way to an understanding of the developmental genetic program at a similar resolution level to that in flies(e.g. Bertrand et al.,2003).

In ascidians, the third cleavage separates the ectoderm, which occupies the animal half of the embryo, from the mesendoderm, which occupies its vegetal half (reviewed by Nishida,1997a). Imai and colleagues(Imai et al., 2000) showed that maternal β-catenin progressively accumulates in the vegetal nuclei of cleaving Ciona embryos where it acts as a key determinant of the vegetal territories. Their work also suggested candidate direct zygotic targets of this determinant (Imai,2003). As in vertebrates, little is known about animal specification. Fusion of animal egg fragments to isolated vegetal blastomeres of eight-cell embryos drove epidermal development in the vegetal host cell,suggesting the existence of localized epidermal determinants (reviewed by Nishida, 1997b). These experiments also showed that rearrangements following fertilization redistribute and concentrate the epidermis determinants to the animal region. Extensive microarray analyses have, however, failed to detect maternal mRNAs enriched in early animal blastomeres, suggesting that the animal determinants are not encoded by localized mRNAs (Yamada et al., 2005).

In this study, we sought to identify animal determinants by following an alternate strategy, i.e. the analysis of the cis-regulatory logic of one of the earliest animal-specific zygotic genes, Ci-fog. We show that maternal Ci-GATAa (a vertebrate GATA4, 5 and 6 orthologue) is necessary and sufficient for animal gene expression from the 8-cell stage and for the subsequent differentiation of the two major ectodermal tissues, epidermis and anterior neural plate. We further show that the restriction of the activity to animal territories of this ubiquitous maternal protein is due to an antagonism by the β-catenin-TCF complex in vegetal territories. Finally, we determined the transcriptional logic that allows Ci-GATAa to sequentially activate pan-animal genes at the 8- to 16-cell stage and neural-specific genes in response to neural induction at the 32-cell stage.

The pSP72-1.27 NLS-lacZ expression vector(Corbo et al., 1997) or the pSP1.72-RfB::NLS-lacZ Gateway destination vector(Lamy et al., 2006)(Invitrogen) were used for analysis of regulatory sequences. Computational analysis was performed using Vista, ClustalW and Transfac8.2 programs as described previously (Lamy et al.,2006).

fog (friend of GATA, ci0100149797) regulatory sequences covering 1981 bp upstream of the ATG (pfog-2000) were isolated by PCR from sperm genomic DNA with primers 5′P-pFOG aaaactgcagGCAACTATTGTAACACCACACG and 3′B-pFOG cgcggatccTATGTGTGTTATTTTTGTATAGACC (flanking PstI or BamHI sites, underlined) using AccuPrime Taq DNA Polymerase(Invitrogen). The pfog-566::NLS-lacZ construct contains sequences downstream of an endogenous SspI site at position -566 and includes the conserved block, as shown in Fig. 1. Construct pfog-300-bpbra::NLS-lacZ contains evolutionarily conserved FOG 5′ sequences between nucleotides -566(SspI site) and -213 (HindIII site) cloned with SalI-BamHI upstream of the basal promoter of Ci-Brachyury (bpbra, -68/+26 fragment)(Yagi et al., 2004). Deletion constructs within this conserved region were obtained by Gateway technology(Invitrogen) using attB1 and attB2 flanked PCR primers. FOG-specific sequences at the 5′ end of constructs were: pfog-530: AGAAAACAACCTTGTTATTACTC;pfog-354: CAAGCAGTAACGAGAAAACAAG; pfog-314: CTGGAGAAGACCAAGATAAAG; pfog-m314:CTGGAGAAGACCAACATAAAGTATTTC (mutations in GATA binding sites, underlined), pfog-290: CTCAAAATTCAGGAAACGGTC, whereas the 3′ end was generated with a Gateway version of 3′B-pFOG. The dINT construct(positions -314 to -85) is as pfog-314 at the 5′ end and deletes intron 1 and downstream sequences at the 3′ end by using the pfog-specific reverse primer sequence: 5′ TGTGACCACCTGCCTTGTC 3′. 2GdINT is a modification of dINT where the two endogenous Ci-fog GATA sites were replaced by GATA sites from the otx neural a-element and contains the sequence CAATATCTAAGATAGGAAAATTCAGGAAACGGTCCAAG at its 5′ end (GATA sites underlined). The G-12 construct, under the control of bpbra, contained the following GATA site multimer XhoI and HindIII (inspired from the otx a-element)(Bertrand et al., 2003):CGTTATCTCTCAATATCTAAGATAGGACGTTATCTCTCAATATCTAAGATAGGACGTTATCTCTCAATATCTAAGATAGGACGTTATCTCTCAATATCTAAGATAGGAG.

The 12X TCF sequence was isolated from the dTF12 vector (N. Perrimon Lab)(DasGupta et al., 2005) using ScaI and SmaI sites and inserted into a HindIII-blunted site upstream pfog-215::NLS-lacZ (basal promoter with an activity slightly weaker than pfog-290).

To generate constructs with heterologous enhancers upstream of the fog basal promoter, the following enhancer sequences were inserted into XhoI and HindIII sites, upstream of pfog-290::lacZ: the reduced otx neural a-element(Bertrand et al., 2003), the Ets site mutated a-element CGTTATCTCTAACaGAAGTTTTCGAAAAaGAAATTGTTCAATATCTAAGATAGGA(GATA and Ets sites underlined, mutations in lower case letters), or the Ci-bra enhancer (eBra) covering position -470 to -62, using forward and reverse primers 5′CCGCTCGAGAACACACCCAACGTACAATAAAAC3′ and 5′CCCAAGCTTCTTCTTTTTGAAATTTTATGTTTG3′.

RACE PCR was performed with GeneRacer (version D; Invitrogen) according to the manufacturer's instructions from total RNA of 64-cell Ciona intestinalis embryos using nested reverse primers dINT and -143rev 5′attB2AAACTGCCTTCTTCTCTTGTC3′.

Animal handling, electroporations and injections were as described previously (Bertrand et al.,2003). Embryos were treated with bFGF (100 ng/ml; Sigma),cytochalasin B (4 μg/ml; Sigma) or U0126 (5 μM; Calbiochem) as previously described (Hudson et al.,2003; Hudson and Lemaire,2001), or 50-300 mM LiCl(Yoshida et al., 1998). For electroporations, volumes were scaled to half and for microinjections, 20-50 ng/μl DNA plasmid, 0.5 mM morpholino oligonucleotides (MO; Gene tools, LLC)and 0.3-0.5 μg/μl synthetic mRNA were used.

GATAa mRNA, EnR-GATAa mRNA, GATAa-MO and Ci-β-catenin-MO were as described previously(Bertrand et al., 2003; Satou et al., 2001). The CiTCF-MO (against ci0100131330) was TCATCCGAGTTTAACTGAGGCATTC. The GATAb-MO(against ci0100136326) was GTTGCTCGCTACTTGTTGGCATCAT (ATGs underlined). The ATG for GATAb was determined according to the conservation to vertebrate GATA2 and GATA3 protein and Ciona savignyicDNA sequences and upon verification by RT-PCR (details available upon request).

In situ hybridizations and β-galactosidase (β-gal) reporter assays (X-Gal stain) were performed as described previously(Bertrand et al., 2003) with digoxigenin-labeled probes generated from the following clones: lacZ(pSP1.72-lacZ) (Corbo et al.,1997), Ci-fog (GR1 cieg033m14), Ci-GATAa(Bertrand et al., 2003), Ci-epiB (GenBank AL666073) and Ci-epi1(Hudson et al., 2003). pfoxD is p(-3.6)Cs-FoxD/lacZ (Imai et al.,2002).

To identify maternal factors that define the animal hemisphere, we searched for a gene with an early animal-specific expression pattern and studied its regulatory sequences. Analysis of the expression pattern of potential early transcriptional regulators revealed that the Ciona friend of GATAgene (Ci-fog; ci0100149797) perfectly matched this criterion. Ci-fog is specifically expressed in the animal region from the 16-cell stage to the beginning of gastrulation(Fig. 1A and data not shown). To identify the cis-regulatory regions responsible for this early expression pattern, we PCR amplified a 2 kb genomic fragment, pfog-2000, extending 5′ from the ATG of Ci-fog and placed it upstream of a nuclearβ-galactosidase reporter gene (NLS-lacZ). This expression construct, pfog-2000::NLS-lacZ, was electroporated into Ciona intestinalis fertilized eggs and the reporter activity was assayed by X-Gal staining during cleavage stages. pfog-2000 contained all necessary information to drive animal-specific expression at the 110-cell stage (Fig. 1B). The earliest reporter expression was detected by X-Gal staining at the 32-cell stage(Fig. 1C). Taking into account that NLS-β-gal protein synthesis takes one cell cycle(Bertrand et al., 2003) (see also lacZ mRNA detection in Fig. 4A), this suggests that animal transcription is already active at the 16-cell stage, thus recapitulating endogenous Ci-fog(Fig. 1A).

Phylogenetic footprinting of pfog-2000 between Ciona intestinalisand Ciona savignyi pointed to the presence of a 294 bp highly conserved fragment (Fig. 1D). A 566 bp fragment including this region and extending up to the ATG of Ci-fog, pfog-566, had similar activity to pfog-2000, whereas deletion of the conserved region suppressed this activity (pfog-215). A construct,pfog300-bpbra::NLS-lacZ, in which this conserved fragment was placed upstream of the unrelated Ci-bra basal promoter(Erives et al., 1998) drove animal β-gal expression. The conserved fragment, therefore, contains all the information required to drive early animal expression.

Through 5′ deletion analysis of the 566 bp fragment, we narrowed down the animal cis-regulatory region to a minimal fragment. This region, pfog-314(Fig. 2), contains two consensus binding sites for GATA transcription factors conserved between the genomes of Ciona intestinalis and Ciona savignyi. Deletion or point mutation of these sites abolished all activity(Fig. 2A). Thus, GATA binding sites are critically involved in animal gene expression. Injection of mRNA for a dominant negative form of GATA (fusion of the GATAa-DNA binding domain to the engrailed-repressor domain) (Bertrand et al., 2003) blocked endogenous Ci-fog expression (92%Ci-fog-negative embryos, n=12), confirming that GATA factors are required for its activation.

We next tested whether GATA-binding sites, placed in front of a basal promoter, are sufficient to drive early pan-animal expression. We previously reported that a synthetic construct carrying six GATA-binding sites in front of the Ci-bra basal promoter drove FGF-dependent expression in animal neural precursors only, starting from the 32-cell stage(Bertrand et al., 2003). To analyze whether an earlier, weaker, GATA activity existed, we constructed a more efficient sensor, G12-bpbra::NLS-lacZ (or G12 in short), in which 12 GATA-binding sites were placed in front of the Ci-bra basal promoter. This construct drove an early animal-wide expression, which strongly resembled the pfog-314::NLS-lacZ expression. By the late 32-cell stage, two levels of expression could be observed: a strong expression in the animal neural precursors (a6.5 and b6.5) as well as a weaker pan-animal activity (other a and b cells), which appeared stronger in posterior animal lineages (b cells) (Fig. 2D). This pan-animal activity was detected from the 16-cell stage onwards(Fig. 4A), prior to the activation of FGF signaling in the embryo, and suggesting that it was FGF signaling independent. Consistently, in embryos grown in the MEK inhibitor U0126, which blocks the FGF signaling cascade, G12 activity was unchanged in uninduced animal blastomeres (a6.6-6.8 and b6.6-6.8, yellow and light purple,respectively, in Fig. 2D), and reduced in FGF-induced neural precursors (a6.5 and b6.5, red and dark purple)to the level found in the uninduced blastomeres(Fig. 2D, compare left and right panels). We conclude that, in addition to the previously identified FGF-dependent GATA activity in neural lineages, an earlier FGF-independent GATA activity exists throughout the animal hemisphere.

The maternal Ciona GATA factor, Ci-GATAa, was previously shown to mediate gene expression in neural progenitors in response to FGF(Bertrand et al., 2003). To test whether GATAa is also involved in the earlier animal-wide expression of Ci-fog, we blocked translation of GATAa with a morpholino (MO). Co-injection of GATAa-MO and the minimal pfog-314::NLS-lacZconstruct, or G12-bpbra::NLS-lacZ, led to a dramatic reduction in lacZ expression (Fig. 2B,C). By contrast, injection of morpholino against the only other, more weakly maternally expressed, Ciona GATA factor, GATAb,had no effect on G12-bpbra::NLS-lacZ activity (75% of GATAb-MO injected embryos, n=28, were X-Gal-positive at the 110-cell stage compared to 61% in control MO injections, n=44; data not shown). To further test the relative contributions of GATAa or GATAb in the activation of G12, we blocked G12 with GATAa-MO and attempted to rescue G12 activity by injecting mRNA encoding either GATAa or GATAb. G12 activity could be rescued by GATAa (mutated for its MO binding site)(Bertrand et al., 2003) but not GATAb mRNA (Fig. 2C)supporting the idea that GATAa but not GATAb is the essential GATA factor involved in this early process.

Taken together, these results indicate that GATAa is a maternal factor that is necessary and sufficient to drive animal gene expression before, and independently of, neural induction. GATAa has thus two distinct roles in the embryo, an early FGF-independent role throughout the entire animal region,including Ci-fog activation, and a later, FGF-dependent function in the neural precursors, leading to Ci-otx expression.

We next tested whether GATAa is required for the specification of the whole ectoderm. As GATAa activity is known to be required for neural tissue formation in the animal hemisphere(Bertrand et al., 2003), we focused on the role of GATAa in the formation of the other major ectodermal derivative, the epidermis. We injected a repressor version of GATAa(EnR-GATAa) into oocytes and analyzed the resulting embryos for epidermal marker expression at neurula stages. To better identify neural and epidermal precursors in these embryos, we arrested cleavage at the 64-cell stage using cytochalasin and tested for epiB or epi1epidermal marker expression. Uninjected control embryos showed strong epiB and epi1 expression in epidermal precursors at the equivalent of the neurula stage (Fig. 3A). By contrast, EnR-GATAa-injected embryos had severely reduced expression of both epidermal markers(Fig. 3A). A similar result was obtained in normally cleaving, EnR-GATAa-injected neurula (not shown,0/23 embryos expressed epiB, and 15/25 embryos weakly expressed epi1). At the tailbud stage, the epi1 residual expression was not maintained in normally cleaving, EnR-GATAa-injected embryos(Fig. 3A), indicating that GATA function is strictly required for epidermis formation.

To assess whether GATAa and GATAb can both contribute to epidermis formation we inactivated GATA function with either GATAa-MO, GATAb-MO or a combination of both (Fig. 3B). Individual MO injection reduced epiB expression, with a strong reduction in 29% of the injected embryos for GATAa-MO and 46% for GATAb-MO. Co-injection of both GATA MOs synergistically blocked the epidermal marker in 100% of embryos. Thus, GATAa and GATAb are both required for epidermal differentiation in the animal region of ascidians.

β-catenin-TCF Maternal GATAa mRNA is ubiquitous and is ubiquitously translated at the 16-cell stage(Bertrand et al., 2003). The animal restriction of GATAa transcriptional activity at this stage is thus likely to be due to post-translational regulation and may involve either a vegetal repressor, or an animal co-activator. To understand when this animal restriction occurs, we first carefully mapped in time and space the onset of activity of pfog-2000, pfog-314 and G12 by more sensitive lacZ mRNA detection. All three constructs displayed the same activity. As predicted from Fig. 1C, strong levels of animally restricted lacZ mRNAs were detected at the 16-cell stage(Fig. 4A). By the 8-cell stage,when animal and vegetal blastomeres are first defined, GATAa activity was weaker and imperfectly localized to the animal cells as demonstrated by a quantification of the fraction of embryos showing preferentially animal,vegetal or ubiquitous expression (Fig. 4B). At even earlier stages (2- and 4-cell stages), a weaker transcriptional GATA activity was detected at low frequency throughout all blastomeres (18% n=66 at the 2-cell, 20% n=215 at the 4-cell and 39% n=44 at the 8-cell stage, analyzed for G12). This indicates that in Ciona zygotic transcription is initiated very early and that the initial GATA activity occurs throughout the embryo. Thus, as development proceeds, GATAa transcriptional response becomes progressively restricted to the animal hemisphere, a process completed at the 16-cell stage.

The gradual exclusion of GATAa activity from the vegetal hemisphere parallels the progressive accumulation of the vegetal determinantβ-catenin in the vegetal nuclei (Imai et al., 2000). Analysis of the onset of activity of the regulatory sequences of the direct β-catenin target foxD(Imai et al., 2002), or of a 12-mer of DNA binding sites for the β-catenin co-factor TCF(DasGupta et al., 2005) placed in front of the Ci-bra basal promoter (12xTCF; Fig. 4A, right panel) revealed that the progressive restriction of GATAa activity to the animal hemisphere had the same kinetics as the increase in β-catenin transcriptional activity in the vegetal hemisphere (Fig. 4A,B). This perfect temporal and spatial correlation suggests that GATAa and β-catenin may cross regulate each other.

We first tested whether accumulating β-catenin could block GATAa in the vegetal region, thereby restricting its activity to the animal hemisphere. For this, we injected a morpholino against β-catenin into eggs and monitored GATAa activity using the G12-bpbra::NLS-lacZ reporter. This construct, normally restricted to the animal hemisphere, became ectopically expressed in the vegetal region when β-catenin translation was blocked from before fertilization, showing that β-catenin normally represses GATA activity in the vegetal hemisphere (Fig. 5A,B). Bringing further support for this notion, a morpholino targeted against the β-catenin cofactor TCF, also derepressed G12 in the vegetal region (Fig. 5A,B). Finally, activation of the β-catenin pathway by lithium treatment led to the suppression of G12 activity (Fig. 5C) indicating that ectopic activation of the β-catenin pathway is sufficient to block GATAa activity.

Conversely, we also tested whether GATAa could antagonize β-catenin transcriptional activity (Fig. 5D). For this, we lowered GATAa levels in the embryo with GATAa-MO or raised them by GATAa mRNA injection and monitoredβ-catenin-TCF activity with pfoxD::lacZ. Modified GATAa levels had, however, no significant influence on β-catenin-mediated gene expression.

Thus, early animal-vegetal patterning results from the progressive definition of a vegetal domain of activity of β-catenin, which restricts the activity of the initially ubiquitous GATAa to the animal territory, where it activates the animal program.

Finally, we addressed the molecular mechanism by which GATAa is able to sequentially activate Ci-fog and Ci-otx in nested territories. We were particularly puzzled by the lack of early pan animal activation of the Ci-otx neural element (a-element), in spite of its containing three GATA sites.

The simplest explanation would be a qualitative difference in the GATA binding sites present in the Ci-otx and Ci-fog regulatory sequences. We tested this by replacing 17 bp containing the two GATA sites and flanking sequence of pfog-314 (nucleotides -303 to -286) with a 17 bp fragment harboring two GATA sites and flanking sequence from the neural element of Ci-otx (nucleotides -1434 to -1418 of potx)(Bertrand et al., 2003). This permutation maintained early-animal-wide expression(Fig. 6B), suggesting that differences in the GATA sites present in the two genes cannot explain their differential activation.

An alternative explanation would be the presence of regulatory sequences in the cis-elements of Ci-fog, but absent in those of Ci-otx,that could synergize with GATAa. To test this, we first analyzed the transcriptional activity of pfog-290, obtained from pfog-314 by deleting the two GATA sites. This element displayed no significant transcriptional activity on its own (Fig. 2A). Next, we placed the early Ci-bra notochord enhancer(Yagi et al., 2004) (ebra)upstream of pfog-290. Electroporation of this construct(ebra-pfog-290::NLS-lacZ) led to robust notochord expression(Fig. 6C), demonstrating that the pfog-290 sequences did not introduce an animal bias in the response to an enhancer. Lastly, if sequences in pfog-290 could synergize with GATAa, placing the Ci-otx a-element (which contains three GATA sites) upstream of pfog-290 should lead to early animal-wide expression. However, this construct,a-pfog-290::NLS-lacZ, drove strong expression in neural precursors(as in its endogenous Ci-otx genomic context or when placed upstream of the Ci-bra basal promoter) (see Bertrand et al., 2003), but only weak sporadic expression in uninduced animal cells(Fig. 6D). Taken together,these experiments suggest that pfog-290 behaves as a basal promoter. It may harbor sequences weakly synergizing with GATAa, but these are not sufficient to override the transcriptional logic of the a-element and efficiently activate it in uninduced animal cells.

A major difference between the two elements is the presence of two Ets sites in the Ci-otx a-element. To test whether these Ets sites could have a repressive activity in uninduced animal blastomeres, we generated point mutations in the two Ets sites of the Ci-otx enhancer in a-pfog-290(amEts-pfog-290::NLS-lacZ). This led to a strong animal-wide expression similar to that obtained with G12bpbra::NLS-lacZ or pfog-314::NLS-lacZ (Fig. 6E).

Taken together, these results indicate that a major determinant of the differential responsiveness of Ci-fog and Ci-otx to GATAa is the presence of Ets binding sites in the a-element. Before the onset of neural induction and in uninduced animal cells, these sites cause active repression of the a-element. In response to the neural inducing FGF signal, Ets binding sites become activating cis-elements and synergize with GATAa sites to turn on Ci-otx in a6.5 and b6.5 neural precursors(Bertrand et al., 2003).

Here, we showed that Ci-GATAa, a maternal Ciona orthologue of the vertebrate GATA4, 5 and 6 transcription factors, is required for the formation of ectodermal territories in the ascidian Ciona intestinalis.Restriction of the activity of this protein to the animal hemisphere results from the progressive repression of its activity by β-catenin in vegetal territories between the 8- and 16-cell stages. Once GATAa activity is restricted to the animal hemisphere, it acts at successive stages of the ectodermal genetic regulatory network. From the 8- to 16-cell stage, it activates Ci-fog throughout the animal hemisphere. At the 32-cell stage, it additionally turns on Ci-otx in FGF-induced neural precursors. Cis-regulatory analysis allowed us to decipher the logic underlying GATAa target gene choice in these two territories. This work thus provides a mechanistic framework for the early patterning of the whole animal vegetal axis in a chordate and initiates the reconstruction of the early ectodermal gene regulatory network.

We found that the early animal-specific gene Ci-fog is directly activated by maternal Ci-GATAa, that a multimer of GATA sites is sufficient to drive animal-specific expression, and is selectively activated by GATAa but not GATAb, whereas inhibition of GATAa activity represses both epidermal and neural fates. The mRNA for this factor is, however, ubiquitously distributed and we have no indication of differential translation of this message by the 16-cell stage when its target gene Ci-fog is specifically activated in animal territories (Bertrand et al.,2003). In addition, GATAa activity appears permissive rather than instructive as interference with GATAa function has no effect on the early vegetal program as measured by the activation of the β-catenin target gene Ci-foxD. Notably, ectopic GATAa does not impose the ectodermal program on vegetal cells. Ci-GATAa can thus be considered a permissive maternal activator of the ectodermal program.

Our results also suggest the existence of a GATA network within the ectoderm to form epidermis with a likely scenario where maternal GATAa signaling is complemented by mostly zygotic GATAb signaling. This is consistent with the strictly maternal expression of GATAa, while GATAb is weakly maternal, then zygotically expressed from the 64-cell stage in the ectoderm (Imai et al., 2004). Furthermore, the two proteins appear to have distinct activities as GATAb overexpression is unable to make up for the early ectodermal GATAa function to activate pan-animal expression (Fig. 2C). Understanding the role of FOG will need additional work. In contrast to GATAa and GATAb, FOG-MO knockdown did not interfere with EpiB or Otx neural element expression (not shown) suggesting that this gene may either act redundantly with other factors or contribute to more subtle aspects of epidermal differentiation.

The permissive nature of GATA signaling is in keeping with the finding that the down regulation of β-catenin leads to the formation of ectopic epidermis in the vegetal territories (Imai et al., 2000). However, our results are surprising in light of embryological experiments carried out in another ascidian, Halocynthia roretzi, which suggested that an instructive animal maternal determinant was distributed as a gradient along the animal vegetal hemisphere of fertilized eggs (Nishida,1997b). Although we cannot exclude that additional localized animal determinants may act in ascidians, or that the Ciona and Halocynthia logics may differ, our discovery that vegetally localized maternal β-catenin represses GATAa activity may be sufficient to solve this apparent paradox. By analogy to the Xenopus, fish and sea urchin situation, β-catenin is probably also activated in ascidians(Yoshida et al., 1998) as a result of the action of cortically localized factors such as dishevelled(Miller et al., 1999; Rothbächer et al., 2000; Weitzel et al., 2004). The Halocynthia experiments were carried out by fusing egg fragments to vegetal blastomeres, and therefore led to the local replacement of the vegetal cortex by cortex of a more animal nature. This may have led to the local lack of activation of β-catenin, and thus to the local derepression in the vegetal cells of GATAa.

Overall, our study suggests that ascidian ectoderm is established permissively and that an instructive maternal animal determinant may not exist in ascidians. It also highlights that ectoderm is a ground state that needs to be repressed by vegetal β-catenin for other germlayers to form.

In a previous study (Bertrand et al.,2003) we found that ectopic GATAa-GFP was transiently restricted to the animal hemisphere during the late 32-cell stage, at the time of neural induction. This protein was, however, ubiquitously distributed up to the 16-cell stage, at the time when Ci-fog expression becomes restricted to the animal hemisphere. Regulation of protein stability is thus unlikely to be crucial for the early definition of the animal hemisphere. Instead, our results suggest an interference with the transcriptional activity of GATAa. We do not know at present whether this is achieved at the level of its ability to bind DNA or to activate transcription.

The β-catenin-TCF complex usually functions in transcriptional activation and the simplest mechanism would be that it activates a zygotic repressor of GATAa activity in the vegetal territory. This would, however,result in a restriction of GATAa activity occurring later than the activation of the primary β-catenin-TCF transcriptional targets. In contrast, we found that the animal restriction of GATAa activity, measured by G12 and Ci-fog activation, is simultaneous to the activation of two direct targets of β-catenin, 12xTCF and Ci-foxD. This goes against a transcriptional relay mechanism and suggests that the β-catenin-TCF complex may directly repress GATAa activity. It will be interesting to test,in the future, whether these proteins are found in a single complex.

In addition to turning on target genes such as Ci-otx in the FGF-induced a6.5 and b6.5 neural progenitors at the 32-cell stage(Bertrand et al., 2003), we found that Ci-GATAa has earlier pan animal targets such as Ci-fog. Our analysis further allowed us to shed light on the transcriptional logic for the selection of different target genes in nested territories at successive developmental time points (Fig. 7).

Our results point to a major role for the two Ets sites in this process. In the absence of FGF signaling, Ets sites act as repressive elements for the Ci-otx neural enhancer. Our previous analysis(Bertrand et al., 2003)indicated that upon reception of an FGF signal, they act as positive transcriptional elements and synergize with the neighboring GATA sites. Thus,conversion of Ets sites from ongoing repressive elements in uninduced cells to activating elements in individually induced cells is a key determinant of the activation of Ci-otx. Conversion of Ets factors from repressor to activator upon MAPK signaling has been described in vertebrates(Maki et al., 2004) and insects (Rebay and Rubin,1995). It will therefore be interesting to examine in the future whether Ci-Ets1/2, which mediates the activation of Ci-otx in FGF-induced cells, also acts as a repressor in uninduced cells or whether the repressive function is achieved by another Ciona Ets factor.

Interestingly, there does not seem to be a sharp transition between the maternal and zygotic programs, rather, distinct combinations of maternal factors are sufficient to collaborate with overlapping early zygotic signals to directly turn on distinct zygotic genes at successive developmental stages in nested territories of the ascidian ectoderm. This situation is somewhat reminiscent of the feed forward loops at work in the early regulatory network that sets the Drosophila anterior posterior axis, in which combinations of maternal factors collaborate with their first broadly expressed zygotic targets, the gap genes, to activate pair rule and segment polarity genes in smaller territories.

It is interesting to compare the ascidian ectodermal logic with that of other bilaterians. Our data point to a GATA network for epidermis formation in ascidians, where a maternal role for GATAa (GATA4/5/6) is complemented by probable zygotic ectodermal GATAb (GATA1/2/3). A large body of evidence is arising to give support to the idea that GATAb orthologues are implicated at various levels in epidermis formation across bilaterians(Gillis et al., 2007). The function of GATAb can be therefore considered a conserved feature. The novel role of maternal GATAa in the initiation of the ectodermal program, and its restriction/repression by β-catenin, has not been demonstrated before.

Ascidians are phylogenetically close to vertebrates, a relatedness apparent in the shared early gastrula fate map and larval body plan (reviewed by Nishida, 2005). Yet, the vertebrate situation appears to differ significantly from the ascidian one. The transition between maternal information and the zygotic program has been best studied in lower vertebrates, and in particular in Xenopus and zebrafish. Although maternal vegetal determinants may differ between these two systems, they converge at the level of the activation of nodal signaling,which is the major pathway inducing both mesoderm and endoderm (reviewed by Schier and Talbot, 2005; Heasman, 2006a). Analysis of the animal region of Xenopus or fish embryos identified several maternal mRNAs, such as zic2, the soxB1 family gene sox3 and ectodermin functioning in maternal inhibition of nodal signaling (Houston and Wylie,2005; Zhang et al.,2004; Dupont et al.,2005). They act via different strategies and are crucial to set the position of the boundary between mesendoderm and ectoderm in the animal hemisphere. These results suggest that maternal repressors of nodal signaling are necessary for the vertebrate ectoderm to form. In ascidians, Nodal is not a major inducer of endoderm and mesoderm(Hudson and Yasuo, 2006),which may explain why such a strategy has not been conserved between ascidians and vertebrates. It should be noted, however, that in sea urchin embryos maternal SoxB1, repressed by vegetal β-catenin is probably required for ectodermal development (Angerer et al.,2005), although nodal has so far only been shown to be involved in oral ectoderm rather than mesendodermal induction(Duboc et al., 2004).

As no regulatory sequences driving early vertebrate ectodermal gene expression have been reported, it remains unclear whether inhibition of Nodal signaling is sufficient for the definition of the ectoderm in vertebrates, or if additional molecules are involved. At first sight, the GATAa orthologues GATA4, 5 and 6, seem unlikely candidates, as they have been shown to be important for the formation of the endoderm rather than the ectoderm (reviewed by Patient and McGhee, 2002; Loose and Patient, 2004). This is, however, a zygotic function of these proteins, which is probably an ancestral deuterostome feature (Gillis et al., 2007) as ascidian Ci-GATAa, sea urchin and starfish gatae are zygotically expressed throughout the endoderm(Imai et al., 2004; Hinman et al., 2003) and sea urchin and starfish GATAe function is crucial for endoderm formation(Hinman et al., 2003). In addition to this conserved zygotic function, however, Xenopus GATA5is expressed maternally (Partington et al., 1997). It will be important to carefully test the effect of the early loss of function of GATA4, 5 and 6 proteins on the vertebrate ectodermal program.

Our work constitutes a unique example of direct regulatory interactions linking maternal factors and the cis-regulatory elements of genes transcribed in the ectoderm of a chordate. Together with the recently published zygotic Ciona whole embryo gene regulatory network(Imai et al., 2006), it constitutes a base to reconstruct the ectodermal gene regulatory network. Comparison of ascidian and vertebrate networks should permit the identification of crucial conserved features, and also an assessment of the extent of deviation that is compatible with the adoption of a common body plan. Comparison with the urchin networks may point to deuterostome-specific traits.

We thank N. Perrimon, Y. Satou, N. Satoh, Y. Kohara and the Japanese In Situ Consortium for sending materials. We thank members of the Lemaire lab, L. Kodjabachian, H. Yasuo and C. Hudson for helpful discussions and comments on the manuscript. Finally, we thank François Graziani and the animal collection service of the Roscoff Marine Biology Station for help with the collection and keeping of Ciona intestinalis. This work was supported by CNRS, the French Ministry of Research (ACI program and GIS `Institut de la Génomique Marine'), the `Transcode' Framework 6 STREP and Marine genomics Europe Network of Excellence. V.B. was supported by a fellowship from the Fondation pour la Recherche Medicale, C.L. was recipient of a PhD fellowship from the French Ministry of Research and P.L. and U.R. were members of CNRS.