Ci-FoxA-a is the earliest zygotic determinant of the ascidian anterior ectoderm and directly activates Ci-sFRP1/5 Free

In vertebrate embryos, the anteroposterior and dorsoventral polarities of the three embryonic layers are set up during gastrulation by the action of a signalling centre, the organizer. This organizer (known as the Spemann organizer in Xenopus, the node in mouse and chicken, the shield in zebrafish) emits factors to determine the polarity of the layers. These secreted factors antagonize the extracellular BMP, Nodal and Wnt ligands,which establish a stable gradient of intracellular signalling in the developing embryo during gastrulation movements (reviewed in Niehrs, 2004).

Vertebrates share the chordate body plan with two invertebrate chordate groups, the cephalochordates and the tunicates(Blair and Hedges, 2005; Delsuc et al., 2006). Among the tunicates, the ascidians have been proposed to be lacking an organizer(Dehal et al., 2002), a property that may represent an ancestral chordate state(Kourakis and Smith, 2005). The availability of two assembled Ciona genomes, of powerful molecular tools and of extensive gene expression atlases make Ciona intestinalis a seducing model to address how chordate anteroposterior patterning can be achieved in absence of an organizer.

Ciona intestinalis forms a typical chordate tadpole larva made of only around 2600 cells. The larval trunk/tail region contains the rostral sticky palps, the anterior neural sensory vesicle (thought homologous to the forebrain and midbrain of vertebrates), endoderm and mesenchymal tissue. The larval tail is organized with a dorsal nerve cord (thought homologous to the vertebrate spinal cord), a central notochord, ventral endodermal cells and flanking muscles (for reviews, see Nishida, 2005; Lemaire et al., 2002).

Ascidian embryogenesis is well described and can be followed at the cellular level until the late gastrula stage. Cell cleavages are regular and the arising cell lineages are invariant. From the eight-cell stage, there is a segregation between vegetal blastomeres (A- and B-lines) that give rise to the endodermal and mesodermal tissues of the larva, and the animal blastomeres (a-and b-lines), which mainly develop into neural and epidermal tissues.

Ascidians were until recently considered as a classical model for mosaic development, based on the inheritance of localized maternal factors in the developing embryo. The identification of Macho-1 as a maternal determinant of posterior vegetal fates(Nishida and Sawada, 2001; Kobayashi et al., 2003) gave cues to understand axial patterning of this chordate embryo for the mesendodermal fates. An anteroposterior difference also exists between the two ectodermal lineages (a- and b-lines) from the eight-cell stage(Rose, 1939; Reverberi and Minganti, 1947; Okado and Takahashi, 1990; Hudson and Lemaire, 2001). The anterior ectodermal a-line gives rise to the trunk epidermis and the sensory vesicle. The posterior ectodermal b-line gives rise to the tail epidermis and the dorsal row of the posterior nerve cord. Upon neural induction by endogenous FGF signals emitted by anterior vegetal cells or by treatment with recombinant FGFs, anterior a-line cells form pigment cells and express anterior neural markers (such as Ci-Otx and Ci-gsx genes) at tailbud stage. By contrast, posterior b-line cells rarely form pigment cells or express anterior neural markers in response to FGF neural inducers(Hudson and Lemaire,2001).

In this study, we made use of the extensive available molecular tools in Ciona (Satoh and Levine,2005) to address the molecular mechanism of the establishment of the anteroposterior difference in the animal hemisphere. We show that there is a global difference between the anterior and posterior ectodermal lineages set up autonomously from the eight-cell stage in the ascidian embryo. By analyzing the mechanisms of activation of the earliest anterior ectodermal specific gene, Ci-sFRP1/5, we show that Ci-FoxA-a, a member of the FoxA family of transcription factors is a key zygotic activator of the anterior ectodermal program and a repressor of posterior ectodermal fates.

Ciona intestinalis adults were collected at the Station Biologique de Roscoff, France. Fertilization, embryo culture, animal cells isolation,bFGF (100 mg/ml, Sigma) and U0126 (3 μM, Calbiochem) treatments were as described (Hudson et al.,2003; Hudson and Lemaire,2001).

Microinjection was performed as described(Bertrand et al., 2003) in unfertilized eggs. The sequence of the Ci-FoxA-a morpholino(MO-CiFoxA-a) targeted against the ATG region was:5′-TTGACGGTGGAGACGACAACATCAT-3′ (Gene Tools, LLC). The sequence of the control morpholino (MO-Control) was:5′-CCTCTTACCTCAGTTACAATTTATA-3′. The morpholinos were injected at a stock concentration of 0.7 mM.

DNA constructs (50 μg) were introduced by electroporation as in Bertrand et al. (Bertrand et al., 2003),except when analysing the effect of MO injection on the pSP1.72-sFRP-118::NLSLacZ and pSP1.72-pFOG::Ci-mFoxA-aconstructs. In this case, morpholino (0.7 mM stock) and DNA constructs (20-30 ng/μl) were co-injected in unfertilized eggs. DNA electroporation or microinjection leads to the mosaic inheritance and expression of the electroporated plasmid in developing embryos, explaining that a proportion of embryos does not express the construct. Expression of the transgenes was always included within the endogenous expression of the corresponding gene.

The Ci-sFRP1/5 regulatory region was analysed by phylogenetic footprinting using Vista(http://genome.lbl.gov/vista/)(Dubchak et al., 2000) and ClustalW(http://www.ch.embnet.org/)(Thompson et al., 1994)alignments. The search for transcription factor binding sites was performed using Transfac Professional 8.2(http://www.generegulation.com/pub/databases.html)(Matys et al., 2003). Ci-sFRP1/5 cis-regulatory fragments were PCR amplified from Ciona intestinalis sperm genomic DNA using Accuprime Taq Polymerase(Invitrogen) and inserted in pSP1.72-RfB::NLSLacZ or pSP1.72-RfA-pBra::NLSLacZ Gateway (Invitrogen) destination vectors. These vectors were generated by insertion of an RfB or RfA cassette in pSP1.72(Corbo et al., 1997) or in pBra pSP1.72 (Bertrand et al., 2003)vectors in BamHI-NotI and HindIII sites,respectively. A precise map of these vectors is available upon request.

The specific sequences of the different primers used for cloning Ci-sFRP1/5 cis-regulatory elements were: sFRP-proximal-for,5′-TGTATGCCGTACATATGCCAG-3′; sFRP-proximal-rev,5′-TTCGGGGCTCTGAAAAAACATAG-3′; sFRP-1067 distal-for,5′-TTACGCTCGATGGCGCAACG-3′; sFRP-1067 distal-rev,5′-AACTAAACCGCAACTATAGTATG-3′; sFRP-640-for,5′-CGCTGTATATACAACTTTGC-3′; sFRP-640-rev,5′-TGTCACTATCAAAGCATGTG-3′; sFRP-262-for,5′-CGGGTTCTCATAGAATCTAC-3′; sFRP-262-rev,5′-TTGCTATCAAATTAGTGTAGC-3′; sFRP-172-for,5′-CGGGTTCTCATAGAATCTAC-3′; sFRP-172-rev,5′-CAACACCAACAGATTACAATG-3′; sFRP-118-for,5′-GCAAACAAACGACTTGTTTAC-3′; sFRP-118-rev,5′-GCACGTTTGTTTAAATTGGAG-3′; sFRP-150-for,5′-CGGGTTCTCATAGAATCTAC-3′; sFRP-150-rev,5′-TCAATCCTCTGGCACCAAC-3′; sFRP-126-for,5′-CGGGTTCTCATAGAATCTAC-3′; sFRP-126-rev,5′-TAATCTGTTAACGGGACAAG-3′; sFRP-95-for,5′-GCAAACAAACGACTTGTTTAC-3′; sFRP-95-rev,5′-CAACACCAACAGATTACAATG-3′; sFRP-92-for,5′-TGTCCCGTTAACAGATTAGG-3′; sFRP-92-rev,5′-GCACGTTTGTTTAAATTGGAG-3′; sFRP-67-for,5′-TGTCCCGTTAACAGATTAGG-3′; sFRP-67-rev,5′-CAACACCAACAGATTACAATG-3′.

The 6-Fox construct was obtained by cloning:5′-TCAGCAAACAAACGACTTGTTTACTTTTCAGCAAACAAACGACTTGTTTACTTTTCAGCAAACAAACGACTTGTTTACTTT-3′(Fox sites underlined) into the pSP1.72-RfA-pBra::NLSLacZ vector.

The pSP1.72-pFOG::RfA destination vector for overexpression in animal cells was generated first by PCR amplification on genomic DNA of Ci-FOGcis-regulatory sequence that contains -1980 bp upstream of the ATG of Ci-FOG (Friend of GATA, ci0100149797). The amplified fragment was cloned into PstI/BamHI sites of a pSP1.72-based expression vector in which the lacZ-coding region was replaced by the Gateway RfA cassette.

The Ci-FoxA-a ORF was amplified from the cicl44j20 clone using T3/T7 primers containing the attB1/attB2 sequences: attB1-T3,5′-GGGGACAAGTTTGTACAAAAAAGCAGGCTATTAACCCTCACTAAAGGGA-3′; T7-attB2,5′-GGGGACCACTTTGTACAAGAAAGCTGGGTTAATACGACTCACTATAGGG-3′). The PCR product was recombined into the p221 donor vector (Invitrogen) and the resulting entry clone was recombined into pSP1.72-FOG::RfA destination vector,giving rise to pSP1.72-FOG::Ci-FoxA-a.

To rescue the Morpholino phenotype, the pSP1.72-pFOG::Ci-mFoxA-a construct was generated as above except the ATG region recognized by the morpholino was altered by amplification of the ORF with attB1-mFoxA-a(5′-GGGGACAAGTTTGTACAAAAAAGCAGGCTAAAAATGATGCTTAGCAGCCCTCCATCAAAGTACCAACCCTTCC-3′)and T7-attB2.

In situ hybridization and X-gal staining were performed as described(Bertrand et al., 2003). DIG probes were synthesized from the following cDNAs: Ci-sFRP1/5 (clone 12ZE09, GenBank AL666897), Ci-Otx(Hudson et al., 2003), Ci-FoxA-a (GR1:cicl044j20), Ci-RORa (GR1:cilv008g20), Ci-FoxC (GR1:cilv050a24), Ci-Delta2 (GR1:cieg005o22) and Ci-Msxb (GR1:cign067l18).

Expression patterns were retrieved from the GHOST(http://hoya.zool.kyoto-u.ac.jp/otherfr.html)and Aniseed(http://aniseedibdm.univ-mrs.fr)databases. In situ hybridization results from this study were submitted to Aniseed.

Scanning the C. intestinalis gene expression database(Imai et al., 2004), we identified a gene specifically expressed in the whole anterior animal a-line. This gene, Ci-sFRP1/5, is orthologous to vertebrate Secreted Frizzled Related Protein (sFRP) genes 1 and 5. Its expression starts from the 64-cell stage in all blastomeres of the a-line(Fig. 1A), persists throughout its progeny up to neurulation and disappears from the ectoderm by the early tailbud stage (Imai et al.,2004) (and data not shown). This restricted a-line expression from the 64-cell stage makes it the earliest known animal specific anterior marker.

Wada and colleagues (Wada et al.,1999) showed that vegetal interactions are necessary at least until the 32-cell stage for anteroposterior patterning of the epidermis. To address whether Ci-sFRP1/5 expression is autonomous to the a-line or induced by vegetal signals, we first inhibited the major vegetal inducer Ci-FGF9/16/20 (reviewed in Nishida,2005) by treating the embryos with the U0126 MEK inhibitor from the eight-cell stage. This treatment abolishes neural tissue formation(Hudson et al., 2003) but the expression of Ci-sFRP1/5 remained as wild type in all a-line cells(data not shown).

To further address the issue of the autonomous activation of Ci-sFRP1/5 in animal cells, we isolated animal blastomeres at the eightcell stage (when animal cells segregates from vegetal cells), cultured the explants until the 64-cell stage, and assayed Ci-sFRP1/5expression in explants by in situ hybridization. When a whole animal cap was explanted, half of the cells of most explants expressed strongly Ci-sFRP1/5 at the 64-cell stage (80% of explants, n=75, Fig. 1B). The remaining explants (20%) expressed a weak, diffuse background-like staining. To assay whether this staining corresponded to anterior cells, we independently isolated anterior aline or posterior b-line cells at the eight-cell stage. At the equivalent of the 64-cell stage, all 16 cells of a-line explants strongly expressed Ci-sFRP1/5 gene (93% of explants, n=15, Fig. 1C) while no or very few cells of b-line explants expressed Ci-sFRP1/5 (6% of explants showing any staining, n=18) (Fig. 1D).

Therefore, the ability to express Ci-sFRP1/5 at the 64-cell stage appears autonomous to the a-line.

Ci-sFRP1/5 is expressed from the 64-cell stage, while the difference between the anterior and posterior ectodermal lineages is set up as early as the eight-cell stage. This suggests the existence of localized upstream regulators of Ci-sFRP1/5 expression. To identify them, we analysed the cis-regulatory sequences of Ci-sFRP1/5. Phylogenetic footprinting between Ciona intestinalis and Ciona savignyigenomes identified three conserved blocks in the 2 kb upstream of the putative sFRP1/5 transcription start site(Fig. 2A).

We cloned two fragments of this region upstream of a NLSLacZreporter gene and revealed the expression of this reporter by X-gal staining at the 110-cell and neural plate stages. A proximal fragment (from the putative transcription start to -985 bp, Fig. 2A) containing two conserved blocks produced no specific staining (data not shown). By contrast,sFRP-1067, a distal 1 kb fragment (from -927 bp to -1993 bp, Fig. 2A), drove specific anterior animal expression from the 64-cell stage.

Deletion of 195 bp in the 3′ extremity of this fragment abolished the expression of the transgene. a-line-specific expression could, however, be restored by adding a minimal promoter (from Ci-Bra gene)(Bertrand et al., 2003),between the cloned fragment and the reporter gene, suggesting that the minimal promoter of Ci-sFRP1/5 is located in this deleted region (data not shown).

Next, we generated 5′ and 3′ deletion constructs by PCR (see Materials and methods) and assayed whether they drove MEK-independent expression throughout the a-line from the 64-cell stage as expected from the analysis of endogenous Ci-sFRP1/5 expression(Fig. 2B). The sFRP-67 construct had no activity (Fig. 2Bc). The sFRP-150, -126, -95 and -92 constructs drove a low frequency expression in only part of the a-line(Fig. 2Bb). Unlike Ci-sFRP1/5, this expression was MEK dependent (not shown). Finally,sFRP-640, -262, -172 and -118 drove pan-a-line(Fig. 2Ba), MEK-independent(Fig. 3A and not shown)expression that recapitulated endogenous Ci-sFRP1/5 expression.

To identify candidate factors acting in trans on sFRP1/5, we searched in silico combinations of putative binding sites, conserved between C. intestinalis and C. savignyi, and present in all enhancers recapitulating Ci-sFRP1/5 expression but not in the other fragments. We found a perfect correlation between the presence of at least three Fox-binding sites and the ability of the fragment to drive pan-a-line expression. To study the importance of these putative Fox-binding sites, we focused on the smallest fragment driving expression throughout the a-line,sFRP-118 (Fig. 2C). We point mutated the core of the Fox binding sequences (TGTTT→TGCTT)(Overdier et al., 1994). Mutation of the two 5′ binding sites resulted in a great reduction of the staining in a-line precursors (Fig. 3A). Most of the residual activity was MEK dependent and therefore distinct from the endogenous Ci-sFRP1/5 regulation. Point mutations in the 3′ putative binding site abolished a-line staining. These experiments (n>60 embryos, and at least two independent experiments for each condition) suggest that putative Fox-binding sites are necessary for the expression of the transgene from the 64-cell stage.

We next investigated if these putative Fox-binding sites were sufficient for the expression in anterior animal cells from the 64-cell stage. We trimerized the 5′ region of the sFRP-118 element containing two putative FOX binding sites to build a synthetic fragment of 81 bp containing six identified Fox sites (6-Fox, Fig. 3B), which was placed in the pSP1.72-RfA-bpBra::NLSLacZvector. This construct drove strong MEK-independent expression throughout the anterior animal a-cells from the 64-cell stage(Fig. 3A and not shown). Additional staining in the anterior vegetal A-line cells was also detected in a large proportion of electroporated embryos. This expression was weaker than a-line staining (Fig. 3A). Consistent with the idea that this anterior expression is due to the presence of the Fox-binding sites, similar results were observed in Halocynthia roretzi by Oda-Ishii and colleagues(Oda-Ishii et al., 2005), and repeated by us in Ciona intestinalis (not shown), with the Fox-BS construct. This artificial construct is made of four Fox-binding sites from the Hr-Otx cis-regulatory sequence and solely shares the Fox-binding sites with 6-Fox (not shown). The anterior vegetal expression observed with 6-Fox is not due to a generally higher level of activity of this construct compared with sFRP-118, as both elements drive similar levels of expression in animal cells. Thus, the restriction of the expression of the sFRP-118 element to the anterior animal region may involve additional trans-acting factors that could either be animal-specific activators or vegetal repressors.

Next, we looked for a Fox factor expressed at the right place and time to be an activator of Ci-sFRP1/5. This Fox factor should be present in the anterior half of the embryo before the 64-cell stage. There are 27 putative Fox genes in Ciona intestinalis genome(Imai et al., 2004). Four of these are presumably not expressed in embryos, as they could not be detected by RT-PCR (Imai et al., 2004). The remaining 23 genes were analysed by in situ hybridization and only two of them detected in ectodermal cells before the 64-cell stage(Imai et al., 2004). Ci-Orphan-Fox1 is ubiquitously expressed from the 16-cell stage,while Ci-FoxA-a expression was reported at the 16-cell stage in vegetal cells and in the a-line blastomeres.

We performed in situ hybridization to establish a kinetics of activation of the Ci-FoxA-a gene in the animal territories. At the eight-cell stage, its expression is visible in both anterior lineages, vegetal A4.1 and animal a4.2 pairs (Fig. 4A). This early expression is consistent with the expression of the C. savignyi orthologue (Shimauchi et al., 2001). Expression of Ci-FoxA-a in animal territories persists at the 16-cell stage in the anterior blastomeres (a5.3 and a5.4), and disappears from the ectodermal cells by the 32-cell stage (data not shown).

To test the function of Ci-FoxA-a in Ciona embryos, we first performed loss-of-function experiments. When we injected an antisense morpholino oligonucleotide (MO) against Ci-FoxA-a in Cionaeggs prior to fertilization, the activity of the co-injected sFRP-118::NLSLacZ construct was significantly decreased. Owing to the mosaic inheritance of the plasmid, lacZ was detected in 58% of MO-control injected embryos (n=60, Fig. 4B). This proportion was lowered to 13% in embryos co-injected with MO-Ci-FoxA-a (n=71, Fig. 4C). The number of positive cells in these embryos was reduced (not shown). MO-Ci-FoxA-a-injected embryos also showed a strong decrease in endogenous Ci-sFRP1/5expression, both at the 64-cell stage (MO-control, 87% positive embryos, n=47; MO-Ci-FoxA-a, 3% positive embryos, n=37). A similar effect was observed at the neural plate stage(Fig. 4B, n=30; Fig. 4C, n=33). Thus,Ci-FoxA-a is necessary for the activation of Ci-sFRP1/5 via the sFRP-118 enhancer.

We next sought to overexpress Ci-FoxA-a throughout the animal hemisphere. For this, we placed a full ORF cDNA under the control of the Ci-FOG gene regulatory sequences. The Ci-FOG gene is expressed in all animal cells from the 16-cell stage and is downregulated after the 64-cell stage (U.R. and P.L., unpublished).

We first used this approach to rescue the decrease of Ci-sFRP1/5expression in MO-Ci-FoxA-a injected embryos. Co-injection of MO-Ci-FoxA-a and pSP1.72-pFOG::Ci-mFoxA-a construct, which drives a mutated version of Ci-FoxA-a lacking the morpholino target site, restored expression of Ci-sFRP1/5 in ectodermal cells at the 64-cell stage, demonstrating the specificity of the effect of the morpholino (MO-Ci-FoxA-a, 20%, n=5; MO-Ci-FoxA-a + pSP1.72-pFOG::Ci-mFoxA-a, 86%, n=14; data not shown).

To assay whether Ci-FoxA-a expression is sufficient to direct Ci-sFRP1/5 in the posterior b-line, we co-electroporated pSP1.72-pFOG::Ci-FoxA-a with pSP1.72-sFRP-118-bpBra::NLSLacZ. X-gal staining was detected at the 110-cell stage in both the anterior and posterior animal blastomeres (54% of embryos showed b-line expression, n=48)(Fig. 4E). The same qualitative result was obtained when assaying endogenous Ci-sFRP1/5 expression at the 64-cell stage following pSP1.72-pFOG::Ci-FoxA-a electroporation(Fig. 4E). Ectopic b-line expression was also seen at neural plate stage(Fig. 4E). We conclude that in a normal embryonic context, animal Ci-FoxA-a expression is sufficient for activating Ci-sFRP1/5 in all animal blastomeres. To rule out that this activation requires some relay via the vegetal cells, we electroporated in Ciona eggs pSP1.72-sFRP-118::NLSLacZ in the presence or absence of pSP1.72-pFOG::Ci-FoxA-a, explanted the animal blastomeres at the eight-cell stage, and analysed reporter activity in the explant at the 110-cell stage. In the absence of ectopic FoxA-a, a-line explants expressed lacZ at the neural plate stage (43% of positive explants, n=14, Fig. 5C), but b-line explants did not (8%, n=13, Fig. 5E). In pSP1.72-pFOG::Ci-FoxA-a-electroporated embryos, the expression of Ci-sFRP1/5 reporter staining was observed at neural plate stage in both a-line (73% of positive explants, n=11, Fig. 5D) and b-line (69%, n=13, Fig. 5F)explants. This experiment shows that Ci-FoxA-a is sufficient for the autonomous expression of anterior ectodermal genes.

Taken together, these and the previous results suggest that Ci-FoxA-a activates Ci-sFRP1/5 at the 64-cell stage via the direct regulation of the sFRP-118 enhancer.

We next tested the extent of the requirement of Ci-FoxA-a for ectodermal patterning. For this, we tested first the effect, at the neural plate stage, of interfering with the activity of this gene on several anterior ectodermal markers. Ci-Otx is expressed in the precursors of the sensory vesicle (a9.33, 34, 37, 38, 49, 50 cells)(Hudson and Lemaire, 2001), Ci-FoxC is expressed in the precursors of the anterodorsal epidermis and palps (a9.35, 36, 39, 40, 51, 52 cells)(Imai et al., 2004), and Ci-Ror-a is expressed throughout the a-line(Bertrand et al., 2003)(Fig. 5G,J,L).

When we downregulated Ci-FoxA-a activity by injection of MO-Ci-FoxA-a,expression was abolished for Ci-Otx (wt, 96%, n=47;MO-Ci-FoxA-a, 0%, n=29) (Fig. 5I) and Ci-Ror-a (wt, 90%, n=49; MO-Ci-FoxA-a,0%, n=32) (Fig. 5N),suggesting that Ci-FoxA-a is generally required for the anterior ectodermal program.

This result is in keeping with two distinct scenarios for the establishment of the anterior identity in the Ciona ectoderm. The first one is an independent specification of the anterior and the posterior ectodermal territories, with Ci-FoxA-a promoting the anterior fate while other mechanisms specify the posterior ectoderm. The second one is the establishment of a posterior identity in the whole ectoderm in absence of anteriorizing signals, an identity repressed in the anterior region by the action of Ci-FoxA-a. To address this issue, we tested if Ci-FoxA-a participates to the inhibition of the posterior fate in the anterior ectoderm of the embryo.

When we overexpressed Ci-FoxA-a in all animal blastomeres by electroporating pSP1.72-pFOG::Ci-FoxA-a, we observed ectopic staining for anterior markers in b-line cells. Ci-Otx was expressed in the sensory vesicle precursors as in wild-type embryos, but also in lateral cells of the neural plate corresponding to the posterior neural cells (b9.33, 37, 38 cells, Fig. 5H). Ci-FoxC was expressed in anterodorsal epidermis and palp precursors,as well as in lateral cells of the embryo, adjacent to the posterior neural cells (b9.35, 36, 40, Fig. 5K). Ci-Ror-a expression was likewise extended throughout bline cells(Fig. 5M). Thus, neural plate markers (Ci-Otx and Ci-FoxC) are ectopically expressed in the posterior neural lineage and the general anterior marker Ci-Ror-ais ectopically expressed more widely in the posterior ectoderm.

We then looked at posterior ectodermal markers. At the 110-cell stage,expression of Ci-Msxb and Ci-Delta2 is restricted to the b-line neural precursors. When we decreased Ci-FoxA-a activity by injection of MO-Ci-FoxA-a, we observed a strong ectopic activation of Ci-Delta2 at the 64-cell stage in the anterior neural blastomeres (a7.9 and a7.10 cells)(wt, 0% of embryos with ectopic expression, n=14; MO-Ci-FoxA-a, 89%, n=9, Fig. 6B). Conversely, when we overexpressed Ci-FoxA-a throughout the animal hemisphere, we observed a strong reduction of the expression at the 110-cell stage of Ci-Msxb (wt, 95%, n=41;pSP1.72-pFOG::Ci-FoxA-a, 14%, n=60)(Fig. 6F) and Ci-Delta2 (wt, 96%, n=152; pSP1.72-pFOG::Ci-FoxA-a,23%, n=116) (Fig. 6D),suggesting that Ci-FoxA-a represses the activation of these genes.

Taken together, these results indicate that Ci-FoxA-a is an early global activator of the anterior genetic program in the ectoderm of Ciona intestinalis embryos. It actively promotes anterior identity and represses the expression of posterior ectodermal genes in anterior ectoderm. Moreover, in the absence of Ci-FoxA-a expression the anterior region of the ectoderm adopts a posterior fate. Thus, the most likely scenario is the establishment of an initial posterior identity throughout the ectoderm,antagonized in the a-line by the action of Ci-FoxA-a.

Our study uncovers the earliest event taking place in the ectoderm revealing that anteroposterior patterning occurs before inductions by endo-mesodermal cells. We dissected the cis-regulatory region of Ci-sFRP1/5, an ectodermal gene specifically expressed in the anterior region of Ciona embryos. We found that the anterior identity is autonomous to the a-line, and dominant over posterior fate. The Ci-sFRP1/5 regulatory region is activated in vivo, via Fox-binding sites, by Ci-FoxA-a, the earliest zygotic anterior determinant. Although we have not carried out in vivo binding assays, Ci-FoxA-a is the only Fox gene in the Ciona genome expressed in the ectoderm in the a-line but not the b-line, strongly suggesting a direct regulatory interaction between Ci-FoxA-a and the sFRP1/5 regulatory regions.

In this section, we discuss the emerging anterior cascade in ascidians and its evolutionary relevance.

Ci-FoxA-a expression appears as the first step of differentiation in the C. intestinalis ectoderm(Fig. 7). Therefore, it provides a molecular mechanism for the acquisition of different anteroposterior competence to respond to neural induction revealed by previous studies (Rose, 1939; Reverberi and Minganti, 1947; Okado and Takahashi, 1990; Hudson and Lemaire, 2001). However, the action of Ci-FoxA-a is not restricted to the definition of the type of response to neural induction. Rather, it appears as a major regulator of anterior ectodermal identity. Its absence leads to the acquisition of posterior identity, while its ectopic expression in posterior territories is sufficient to give them an anterior character. This role as a major anterior ectodermal determinant is likely conserved in other ascidians as Ci-FoxA-a orthologues are also expressed in the anterior, but not the posterior, animal blastomeres in Ciona savignyi from the eight-cell stage (Schimauchi et al., 2001) and in Halocynthia roretzi(a distantly related species) at the 16-cell stage(Shimauchi et al., 1997).

Although Ci-FoxA-a is both necessary and sufficient for a proper anterior ectodermal patterning, we cannot rule out the possibility that it may collaborate with other genes for the specification of the anterior region. Other genes have been indeed reported to be expressed in anterior lineages at the eight-cell stage in another ascidian Halocynthia roretzi(Miya and Nishida, 2003; Makabe et al., 2001).

Our findings suggest that, at the eight-cell stage, the animal territories have a posterior default identity, which is counteracted by the activity of Ci-FoxA in the a-line. This apparently contrasts with the current model for AP regionalization in the vegetal territories in which posterior determinants, found in the vegetal-posterior cortical region of the egg after fertilization, are thought to define the anteroposterior axis of endo-mesodermal fates by repressing a default anterior identity (for a review,see Nishida, 2005). As a result, the patterning of endo-mesodermal tissues is currently explained without the help of an anterior determinant. At first sight, this argues that independent mechanisms pattern the AP axis in the ectoderm and endomesoderm. Consistent with this proposal, Ci-FoxA-a expression is not sensitive to the knock-down of the major vegetal posterior determinant identified to date, Macho-1 (Yagi et al.,2004).

This apparent paradox may, however, principally stem from our lack of understanding of the maternal factors, which activate Ci-FoxA-a at the eight-cell stage. Two, non exclusive, scenarios are possible for restricting Ci-FoxA-a anterior expression. A maternal activator may be specifically active in the anterior region of the zygote. Such an activator is unlikely to be encoded by a maternal messenger localized in anterior territories as a large scale microarray analysis did not succeed in identifying such localized maternal mRNA(Yamada et al., 2005). Alternatively, the activation of Ci-FoxA-a could result from the combination of a ubiquitous activator, and a repressor, distinct from Ci-Macho1, present in the posterior region. In the latter case, the activation of Ci-FoxA-a would be a `default' property of the embryo, repressed in the posterior blastomeres, a situation similar to that observed in the vegetal hemisphere. The identification of an enhancer recapitulating the activation of Ci-FoxA-a in anterior territories during the cleavage stages (Di Gregorio et al.,2001) (C.L., unpublished) opens the way to a better understanding of this important issue.

The results presented here, combined to previous observations, give an overall view of ectodermal patterning in ascidians, which appears to be a multistep process. First, anterior and posterior identities are established autonomously from the eight-cell stage as a result of Ci-FoxA-aactivation. Secondary signals emanating from the vegetal hemisphere then act to refine this initial ectodermal patterning(Wada et al., 1999; Darras and Nishida, 2001; Hudson et al., 2003; Hudson and Yasuo, 2005). At the 32-cell stage, Ci-FGF9/16/20 induces both anterior(Bertrand et al., 2003) and posterior (Bertrand et al.,2003; Hudson and Yasuo,2005) neural tissue within the Ciona ectoderm. Interestingly, while Ci-FoxA-a specifies the anterior or posterior identity of ectodermal cells in response to this induction, it does not affect which blastomeres are induced by FGF. Upon Ci-FoxA-a overexpression, a new anterior neural plate thus forms in the presumptive posterior neural territory, with sensory vesicle markers expressed closest to the vegetal-inducing cells, palp marker expressed in a more lateral region and anterior epidermis marker expressed in the posterior non-induced cells. This result is in keeping with the proposal that the selection of the induced animal cells relies mainly on the extent of the contact between inducing and competent cells (Tassy et al.,2006). Subsequent, yet uncharacterized, vegetal signals are then required to pattern the neural plate along the AP axis at the beginning of gastrulation, discriminating the anterior palp identity from the more posterior rostral CNS (Nishida and Satoh,1989).

A conceptually similar cascade is at work in vertebrates. In Xenopus, the ectoderm autonomously acquires an early pregastrula patterning prior to the action of vegetal signals. This early pattern specifies the type of tissues induced by these signals(Sokol and Melton, 1991; Otte and Moon, 1992; Baker et al., 1999; Kodjabachian and Lemaire,2001) and is marked by the autonomous ectodermal activation of specific dorsal markers (Kodjabachian and Lemaire, 2001; Kuroda et al.,2004). Finally, explants experiments carried out in zebrafish embryos identified a pre-existing anteroposterior patterning in the ectodermal region of the zebrafish embryo before gastrulation(Koshida et al., 1998). Following this initial phase, vegetal signals induce neural tissue, and pattern it along the anteroposterior axis during gastrulation. Two major models for neurectodermal patterning during gastrulation have been put forward. The `two-inducer' model postulates that different signals induce the ectoderm to form anterior and posterior neural tissues(Niehrs et al., 2001). The`two-step' model postulates an initial induction of anterior neural fate throughout the neural plate, followed by the posteriorization of part of this tissue by signals emitted by the underlying mesendodermal organizer, including FGF, Wnt and Nodal inhibitors, and retinoic acid (reviewed in Stern, 2001).

It is currently unclear how deep the similarity between ascidians and vertebrates extends. Only part of the ascidian neural tissue is induced(reviewed by Lemaire et al.,2002), but a single inducer, FGF9/16/20, is at work,differentially interpreted by anterior and posterior cells. This superficially appears different from the `two steps' model as the anterior and posterior identity of the induced tissue are specified at the time of induction, prior to gastrulation. It should, however, be stressed that the precise timing of action of the two vertebrate signals has not been determined. Furthermore, our finding that Ci-sFRP1/5 and Ci-Ror-a, two antagonists of the canonical Wnt pathway (Dennis at al.,1999; Forrester et al.,2004) are specifically expressed in anterior territories, under control of Ci-FoxA-a and after the induction process, is in keeping with a two-step model. However, Ci-sFRP1/5 morpholino knock-down experiments have no apparent effect (C.L., unpublished), indicating that in a wild-type situation this secreted antagonist of Wnt ligands may play a less crucial role in the Ciona ectoderm than it does in zebrafish(Houart et al., 2002). A rigorous assessment of the role of Wnt signalling and other extracellular signals in ascidian ectodermal patterning, coupled to the deciphering of the regulatory networks at work in both ascidians and vertebrates, will be required to estimate the level of conservation of this event in chordates. However, at this stage, we can already point to the overall conservation of a logic combining an early autonomous subdivision of the ectoderm, which affects its competence to respond to a subsequent cascade of vegetal signals.

We found that Ci-FoxA-a is a crucial determinant of the ascidian anterior identity. This appears to be a shared property of ascidians and vertebrates. In vertebrates, FoxA orthologues are also expressed before gastrulation (Ruiz i Altaba and Jessell, 1992; Sasaki and Hogan, 1993; Perea-Gomez et al., 1999; Strahle et al.,1993) and important for the formation of anterior ectodermal territories (Suri et al.,2004; Perea-Gomez et al.,1999). Furthermore, ascidian and vertebrate embryos share conserved molecular events downstream of FoxA genes. In ascidians, Ci-FoxA-a is necessary for the activation of the Ci-Otxgene, that has conserved functions among bilaterians to specify the anterior ectoderm (Hirth and Reichert,1999). Likewise, in mouse and Xenopus embryos,FoxA2/HNF3-β are involved in the regulation of Otx2 (Perea-Gomez et al.,2001; Suri et al.,2004). Thus, among chordates, Otx is a conserved downstream target of FoxA necessary for the formation of anterior ectoderm.

Likewise, Ci-FoxA-a could play a role in the protection of the ascidian anterior ectoderm from posteriorizing signals known to caudalize the neurectoderm (Hudson et al.,2003), as proposed in mouse(Perea-Gomez et al., 1999) and in Xenopus in the `two-step' model. This protection could occur through the activation of Ci-sFRP1/5 and Ci-Ror-a, two antagonists of the canonical Wnt pathway (Dennis et al.,1999; Forrester et al.,2004).

These observations suggest that Ci-FoxA-a and vertebrates FoxA2 may trigger similar molecular pathways to pattern the ectoderm along the anteroposterior axis, although the relative individual importance of these pathways may vary. It will be interesting to identify and compare additional targets of FoxA genes in ascidians and vertebrates.

The above similarities are, however, accompanied by major differences in the mode of action, but also in the site of expression and action of FoxA genes in ascidians and vertebrates. First, in contrast to what we see in ascidians, Xenopus FoxA2 overexpression is not sufficient to induce anterior fate in the ectoderm (Ruiz i Altaba and Jessell, 1992). Second, although our experiments point to an autonomous role of FoxA-a in the ascidian ectoderm, vertebrate FoxA2 transcripts have only been detected in non-ectodermal embryonic territories,such as dorsal mesendodermal cells in Xenopus and Danio, and extra-embryonic AVE cells in mouse (Ruiz i Altaba and Jessell, 1992; Strahle et al., 1993; Sasaki and Hogan, 1993). As this was initially the case for several organizer genes that have since been shown to be also expressed and active in the presumptive ectoderm(Kodjabachian and Lemaire,2001; Kuroda et al.,2004), it will be important to revisit the expression pattern of FoxA genes at early stages of vertebrate development. Yet, the fact that FoxA2 in mice is required only in the AVE and not in the epiblast suggests a different mode of action in vertebrates and ascidians(Perea-Gomez et al.,1999).

A possible scenario for the acquisition of different roles by FoxA while retaining a general function in anteroposterior patterning, may be an ancestral expression and function in both animal and vegetal anterior territories, followed in the vertebrate lineage by a gradual amplification of its role in the vegetal territories at the expense of its animal role. Dorsoanterior vegetal territories in amniotes correspond to the organizer, a structure that has been proposed to constitute a vertebrate innovation whose function is to maintain during gastrulation an initial coarse pattern(Kourakis and Smith, 2005). In support to this model, an organizer-independent activation of Hox genes(markers of posterior tissues) is seen in Xenopus in the mesoderm from the beginning of gastrulation (Wacker et al., 2004). Wacker and colleagues proposed that, during gastrulation movements, the organizer activity stabilizes and synchronizes the anteroposterior boundaries of Hox expression between adjacent ectoderm and mesoderm. We can speculate that, in vertebrates, the coordination of two different tissues streaming past one another requires constant mesodermal signals to adjust and maintain the boundaries of an already patterned ectoderm. By contrast, in ascidians the restricted gastrulation movements,which do not involve change in cell neighbourhood relationships(Munro et al., 2006), may not require this control mechanism. The acquisition of organizing activity by the vertebrate dorsoanterior mesendoderm may be linked to a shift of the major function of FoxA in ectodermal patterning from the animal to the vegetal territories. Alternatively, FoxA may have been specifically recruited in the tunicate ectoderm to form autonomously anterior structures.

The importance of Ci-Foxa in anterior specification has been independently reported (Imai et al.,2006).

We thank Elia Stupka for help with in silico analysis, and Izumi Oda-Ishii and Hidetoshi Saiga for sending the Fox-BS construct. We are grateful to Hitoyoshi Yasuo, Clare Hudson, Sébastien Darras, Stéphanie Marini and Laurent Kodjabachian for critical comments on the manuscript and very helpful discussions. Finally, François Graziani and the animal collection service of the Roscoff Marine Biology Station provided crucial help with the collection and keeping of Ciona intestinalis. This work was supported by CNRS, the French Ministry of Research (ACI programme and GIS`Génomique Marine') and the `Transcode' framework 6 STREP European Network. C.L. was recipient of a PhD fellowship from the French Ministry of Research, P.L., U.R. and D.C. were members of CNRS.