Sea urchin goosecoid function links fate specification along the animal-vegetal and oral-aboral embryonic axes

Patterning of cell fates along the animal-vegetal (AV) and oral-aboral (OA) axes of the sea urchin embryo requires a signaling process that is initiated in the vegetal organizing center. Recent studies have established that the inductive processes that pattern mesoderm, endoderm and ectoderm fates along the AV axis are initiated by cell-autonomous activation of downstream components of the canonical Wnt pathway, i.e. β-catenin and TCF-Lef (Emily-Fenouil et al., 1998; Huang et al., 2000; Logan et al., 1999; Vonica et al., 2000; Wikramanayake et al., 1998). β-catenin/TCF-Lef activates, in at least one case directly (Howard et al., 2001), production of downstream transcription factors (SpKrl) and signaling ligands, including SpWnt8 (A. Wikramanayake and W. H. Klein, personal communication) and probably Delta (H. Sweet and C. Ettensohn, personal communication), which function in specifying vegetal fates.

The OA axis is established after fertilization by a mechanism that may involve polarized differences in redox potential (Coffman and Davidson, 2001). The point at which this axis is conditionally specified varies from the two-cell stage in Strongylocentrotus purpuratus (Cameron et al., 1989), the organism used in the studies reported here, to as late as the eight-cell stage in some other species (Kominami, 1988). OA polarity is displayed primarily in the ectoderm, by the differentiation of distinct oral and aboral epithelial cell types that are separated by a thin band of ciliated cuboidal cells. Although OA polarity is established after fertilization by a mechanism different from that which fixes the AV axis during oogenesis, patterning of different cell types arrayed along the OA axis nevertheless requires signaling from the vegetal pole. Animal halves of eight-cell embryos or intact embryos in which β-catenin function is blocked by injection of cadherin mRNA fail to differentiate aboral ectoderm or ciliary bands (Wikramanayake and Klein, 1997; Wikramanayake et al., 1998). Differentiation of these tissues can be rescued in animal half embryos by treating them with LiCl, which inhibits the β-catenin-destabilizing kinase, GSK3β, or by injection at the one-cell stage of mRNA encoding a stable form of β-catenin (Wikramanayake et al., 1998). The fact that animal halves and cadherin-expressing embryos express the EctoV epitope, which also accumulates late in differentiated oral ectoderm (both facial epithelium and ciliary band), has led to the proposal that β-catenin-dependent signals are required specifically for aboral ectoderm differentiation (Wikramanayake et al., 1998).

In vertebrate embryos, activation of the canonical Wnt pathway leads to the establishment of a gastrulation organizing center that expresses the goosecoid transcription factor (Lemaire and Kodjabachian, 1996). The fact that ectopic expression of goosecoid can induce a second dorsal axis in Xenopus embryos suggests that it is a critical upstream regulator of dorsal fate in this embryo (Cho et al., 1991). Echinoderms are a sister clade to the chordates, whose embryos have been shown to use a number of molecular pathways employed in vertebrate embryos. In addition to the molecules noted above, BMP2/4, whose function is antagonized by goosecoid in Xenopus embryos, plays an important role in ectoderm specification in sea urchin embryos (Angerer et al., 2000). These considerations suggest that goosecoid might also play an important role in early sea urchin development. We present our characterization of an S. purpuratus goosecoid homolog, SpGsc, and our investigations of its role in sea urchin embryogenesis. We have determined its time and sites of expression, characterized phenotypes produced in loss-of-function and misexpression experiments and determined whether its expression depends on the canonical Wnt pathway. Our data show that SpGsc is required for gastrulation and subsequent differentiation of endoderm and pigment cells, one of the derivatives of secondary mesenchyme. SpGsc also plays a central role in patterning cell fates along the OA axis, in that its restricted expression in oral ectoderm is required to repress gene(s) that promote aboral ectoderm differentiation. The data presented here support a model in which SpGsc function links patterning of cell fates along the AV and OA axes of sea urchin embryos.

Adult sea urchins (S. purpuratus) were obtained from Charles M. Hollahan (Santa Barbara, CA). Embryos were cultured (Angerer and Angerer, 1981) and cell separation experiments were carried out (Reynolds et al., 1992) essentially as described previously. Blastomeres were resuspended in Ca²⁺-free seawater and cultured in spinner flasks as described (Hurley et al., 1989).

Degenerate primers representing sequences conserved in the homeodomains of Xenopus and Drosophila goosecoid proteins were used with a very early blastula random-primed cDNA library template to amplify the homologous cDNA. Full-length cDNA containing a 320 amino acid residue open reading frame (GenBank Accession Number, AF315231) was generated by 5′ and 3′ RACE (Life Technologies). SpGsc cDNA was inserted into Tclone, derived from the plasmid vector, pSp64T (Angerer et al., 2000), and synthetic mRNAs were transcribed with Sp6 RNA polymerase (Sp6 mESSAGE mACHINE, Ambion) from templates truncated with XbaI. The SpGsc-VP16 fusion construct was prepared as follows. The SpGsc DNA-binding domain (DBD) was obtained by PCR from the full-length cDNA clone described above using forward and reverse primers containing SacI and NotI sites (underlined below), respectively:

forward, 5′CCC CGA GCT CTT GGT GAT GGA ATT CAA GAG AAA GAG GCG ACA;

reverse, 5′ CCC CGC GGC CGC TTA CCG TTT CTG CTT CCT C 3′.

This fragment was inserted into Tclone between the corresponding restriction sites and the VP16 transcriptional activation domain from clone pCS2-VP16ΔβXtcf-3 (corresponding to amino acid residues 411 to 490) (Vonica et al., 2000), which was contained on an EcoRI fragment, was inserted upstream of the SpGsc DBD at the EcoRI site in the forward primer (italics).

Blots of genomic DNA digested with either EcoRI or RsaI were prepared as described previously (Yang et al., 1989a) and probed with random-primed cDNA sequence encoding the homeodomain and 22 additional 5′ amino acid residues in a solution containing 5×SSPE (0.75 M NaCl, 20 mM sodium phosphate, pH 7.4, 5 mM trisodium EDTA), 0.1% sodium pyrophosphate, 5×Denhardt’s (0.5% polyvinylpyrrolidone, 0.5% bovine serum albumin and 0.5% Ficoll), 0.5% sodium dodecyl sulfate at 60°C and washed with 0.15×SSPE, 0.1% SDS at 45°C, which corresponds to a low stringency (T_m–45°C). RNA blotting was carried out as described previously (Howard et al., 2001) with 2 μg of mesenchyme blastula polyA⁺RNA and the blot was washed at moderate stringency (T_m–35°C; 0.1×SSPE, 55°C). For RNase protection assays, total RNA (10 μg) from normal or dissociated embryos was purified with TRIzol reagent (Gibco BRL, Bethesda, MD) and hybridized to probes for SpGsc (1×10⁸ cpm/μg), Spec2a (1×10⁸ cpm/μg) (Hardin et al., 1988) or SpHE (1×10⁷ cpm/μg) (Wei et al., 1999). Hybridization and analysis of RNase-resistant fragments were as described previously (Yang et al., 1989a). We used a radioactive method for in situ hybridization assays because it provides greater sensitivity and reproducibility for early embryonic stages than we have been able to achieve with whole-mount methods. Sections (5 μm) of embryos at selected developmental stages were hybridized with ³³P-labeled RNA probes for SpGsc (2.5×10⁵ dpm/ng) as described previously (Angerer et al., 1987).

A synthetic SpGsc protein was used to determine DNA-binding properties. To generate the pGST-GSC construct, pGEX-KG vector (Amersham Pharmacia Biotech, NJ) was digested with BamHI and NotI. The SpGsc cDNA coding sequence was amplified by PCR using primers containing BamHI and NotI sites:

Gsc-Bam forward, TGG GAT CCT GGA CTA TTA TCT CCC CGA CGT C;

Gsc-Not reverse, CGA TGC GGC CGC GGC GAG GAG ACC CCG ATG GTG AG.

The digested PCR fragment was ligated into pGEX-KG to produce the pGST-GSC construct. The fusion protein was expressed in E. coli BL21 cells. The GST-SpGsc fusion protein was purified using the GST Purification Module from Amersham. Briefly, an overnight culture derived from a single colony was diluted 100-fold with 500 ml NZY medium and incubated for 3 hours at 37°C. IPTG was added to 0.2 mM and the culture was incubated for an additional 2 hours at 37°C. Cells were harvested, resuspended in phosphate-buffered saline (PBS) with 1% N-laurylsarcosine (sarkosyl) and sonicated on ice for 1 minute (power level 4, 50% duty cycle). The lysate was clarified and Triton X-100 was added to the supernatant to a final concentration of 2%. Washed glutathione-agarose bead suspension (0.5 ml; 50% v/v in PBS) was added, and the lysate was incubated at room temperature on a shaker for 30 minutes. The beads were transferred to a supplied column and washed six times with ice-cold PBS. The fusion protein was eluted from the beads with 10 mM reduced glutathione and subsequently dialyzed into storage buffer (50 mM Hepes, pH 7.4, 100 mM KCL, 1 mM DTT, 1 mM PMSF, 10%(v/v) glycerol). The GST moiety on the SpGsc-GST fusion protein was removed by proteolytic digestion with Factor Xa for 10 hours at room temperature.

EMSA was performed as described previously (Yuh et al., 2001). The reactions contained 20-50 ng of the GST-SpGsc fusion protein or SpGsc protein with the GST moiety removed, 1×10⁴ cpm of ³²P-end-labeled oligonucleotide probe, 100 ng of the indicated competitor, and 0.5 μg of poly(dI-dC) in a final volume of 15 μl 1×EMSA buffer (12% glycerol, 20 mM Hepes (pH 7.9), 5 mM MgCl₂, 100 mM KCl, 1 mM DTT) for 20 minutes at 4°C. DNA-protein complexes were resolved in a 5% polyacrylamide gel in 0.5×TBE and signals from the dried gels were recorded by a phosphorimager (Molecular Dynamics: Image Quant).

Constructs for in vitro transcription of mRNAs were verified by sequencing. Synthetic mRNAs were suspended in 30% glycerol, quantitated by spectrophotometry and by gel electrophoresis, and microinjected as described previously (Angerer et al., 2000). Either 2×10⁵ or 6×10⁵ RNA molecules were injected into each egg. Morpholine-substituted oligonucleotides complementary to nucleotides –30 to –6 with respect to the translation start site of SpGsc mRNA and a control morpholino were obtained from Gene Tools (Corvallis, Oregon) and dissolved in diethylpyrocarbonate-treated water at a concentration of 8 mM. This stock solution was diluted to either 200 or 400 μM and 2 pl were injected to give a final concentration in the egg of 2-4 μM.

A promoter containing multimerized SpOtx cis elements linked to the CAT (chloramphenicol acetyl transferase) reporter gene (5C) (Mao et al., 1996) and synthetic SpGsc mRNA were microinjected into one-cell sea urchin zygotes. CAT assays were carried out as described previously (Wei et al., 1995).

Embryos were fixed in artificial sea water containing 4% paraformaldehyde and stained with a polyclonal antibody against SpSoxB1, with a monoclonal antibody against a PMC-specific epitope (6e10, kindly supplied by Dr Chuck Ettensohn, Carnegie Mellon University) as described previously (Kenny et al., 1999), with Sp1 monoclonal antibody obtained from the Developmental Biology Hybridoma Bank (Dieter Soll, University of Iowa), and with polyclonal antibodies against Spec1 (Carpenter et al., 1984) and EctoV (Coffman and McClay, 1990) as described previously (Angerer et al., 2000). Fluorescent signals were captured by sequential scanning using a LeicaTS confocal microscope.

RNA from 24-hour embryos injected with glycerol (control), synthetic mRNAs or morpholinos was purified with TRIzol reagent (Gibco BRL, Bethesda, MD), DNase I digestion, organic extraction and ethanol precipitation. One-step RT-PCR (Advanced Biotechnologies, Surrey, UK) using the primers listed below was employed to amplify SpGsc, Endo16, BMP2/4, or 12S mitochondrial rRNA sequences (Mito) as a load control for embryo RNA.

SpGsc forward primer: CCT GAG TAT CAC TTA GCT GCC

SpGsc reverse primer: TCG TTC CTC TTT GAG GTC GA

Endo16 forward primer: AAC AAG GTG CGT GCG GCT GC

Endo 16 reverse primer: GCT GGG GCG AGC ACA TTA TT

Mito forward primer: ACT CTC TCC TCG GAG CTA TA

Mito reverse primer: GTA TAA TTT TTG CGT ATT CGG C

SpBMP2/4 forward primer: 5 CAG GCC TAC TAT TGT CGC

SpBMP2/4 reverse primer: GGT ACT AGT GCT GGG TTG

Signals were compared within the linear phase of amplification, as verified by analyzing samples collected at different cycles. After electrophoresis of the samples through 6% polyacrylamide gels in TBE (0.1 M Tris, 0.1 M sodium borate, 2 mM EDTA, pH 8.3), signals were quantitated by phosphorimagery.

Full-length SpGsc cDNA was isolated as described in Materials and Methods. It encodes a predicted peptide that contains two conserved regions corresponding to the homeodomain (Fig. 1A) and the goosecoid/engrailed homology (GEH) domain (Fig. 1B). Alignment of the homeodomain sequence with homologous sequences from other organisms indicates clearly that the closest relative to SpGsc is Drosophila Gsc (Accession Number, S70617), followed by homologs in the polychaete annelid (Platynereis dumerilii; Accession Number, CAC19336) and amphioxus (Branchiostoma floridae; Accession Number, AAF97935) (Neidert et al., 2000). Interestingly, most of the sequence divergence between these homeodomains and those of vertebrate Gsc proteins (Fig. 1A, boxed sequences) maps to the central region; both N- and C-terminal regions are extremely well conserved among all the members of this family. Because sea urchins are evolutionarily much more closely related to vertebrates than to Drosophila, and because multiple Gsc genes have been identified in both chickens and mice, we investigated the possibility that a second sea urchin Gsc gene exists that is more closely related to those of vertebrates. However, no evidence for a second gene could be detected in blots of DNA from individual sperm samples that were probed with the SpGsc homeodomain sequence at low stringency (Fig. 1C). In addition, reduced stringency hybridization screens of an arrayed 20 hour blastula cDNA library did not identify other Gsc-like sequences, although they did recover SpGsc again, as well as other cDNAs encoding homeodomain-containing proteins distinct from the bicoid-like class of which SpGsc is a member. The strong conservation of the Gsc homeodomains among an echinoderm, an arthropod, a polychaete annelid and a cephalochordate suggests that this is the ancestral gene. We conclude that after the divergence of cephalochordates and vertebrates, this gene was duplicated several times in the vertebrate line, after which the ancestral gene was lost. Nevertheless, despite some sequence differences between the central regions of the ancestral and vertebrate Gsc homeodomains, these proteins probably remain functionally similar. For example, Drosophila Gsc can elicit secondary axes when ectopically expressed in Xenopus embryos (Goriely et al., 1996).

The GEH peptide sequence in Drosophila Gsc is required to bind a co-repressor that mediates the repressor activity of Gsc (Mailhos et al., 1998). As shown in Fig. 1B, this sequence is well conserved in SpGsc, in which it probably functions similarly because, as shown below, SpGsc also has repressor activity in sea urchin embryos.

SpGsc transcripts are not detectable in the maternal RNA population or during cleavage; they accumulate between the very early blastula and mesenchyme blastula stages and persist at similar levels in the pluteus larva (Fig. 2). Blots of polyadenylated RNA detect a single 3.6 kb mRNA species (data not shown). In situ hybridization verifies that SpGsc transcripts are absent from eggs (Fig. 3A,B), cleaving embryos and very early blastulae (data not shown), but are detectable in hatched blastulae (Fig. 3C-H). At the latter stage the major region of expression encompasses about one half of the presumptive ectoderm. The hybridization patterns of gastrulae (Fig. 3I,J) and plutei (Fig. 3K,L) identify this region as presumptive oral ectoderm. Lower signals are observed over a few cells in the vegetal plate region of early mesenchyme blastulae (Fig. 3D,F,H), which can be identified by the presence of a few adjacent, unlabeled, ingressed primary mesenchyme cells (PMCs). Comparison of adjacent sections (compare Fig. 3E,F with 3G,H) demonstrates the reproducibility of these vegetal signals, while hybridization patterns on fortuitously oriented sections (e.g. Fig. 3C,D) suggest that the positive cells are arranged in a torus around a few unlabeled cells at the vegetal pole. The size and position of the unlabeled region are characteristic of small micromere daughters, suggesting that the adjacent labeled cells are PMC precursors that have not yet ingressed into the blastocoel. SpGsc message accumulation in mesenchyme cells is transient and is downregulated as they enter the blastocoel. Based on their number and position, we favor the view that these cells are PMC precursors, but we cannot exclude the possibility that some presumptive secondary mesenchyme cells also transiently express SpGsc. The early and late expression patterns are indicated by black shading in the diagrams in Fig. 3M,N.

We examined the effects of loss of SpGsc function by morpholino-mediated inhibition of translation. We verified that the SpGsc morpholino inhibits translation in sea urchin embryos using the method described previously (Howard et al., 2001): SpGsc sequence complementary to the morpholino or the morpholino sequence itself (non-complementary) was inserted in the 5′UTR of the GFP cDNA template. Embryos simultaneously injected at the one-cell stage with the SpGsc morpholino and synthetic mRNA containing the complementary sequence did not produce detectable GFP, whereas embryos co-injected with mRNA containing the non-complementary sequence were brightly fluorescent (data not shown).

Loss of SpGsc function produced a striking phenotype. Most embryos failed to gastrulate or to differentiate endoderm, while a few produced only small gut rudiments (Fig. 4A, Table 1). As shown by RTPCR, 24-hour embryos from this population had greatly reduced levels of Endo16 RNA (Fig. 4B), whose expression in the vegetal plate (presumptive endoderm and secondary mesenchyme) normally begins about the same time as does that of SpGsc. These embryos also lacked any vegetal derivatives labeled with monoclonal antibodies against EctoV, which labels foregut cells (Fig. 4A). The differentiation of pigment cells, a subset of secondary mesenchyme, was strongly suppressed: the embryo shown in Fig. 4C illustrates the maximum level of immunostaining with the Sp1 antibody, which recognizes an epitope expressed early in the differentiation of these cells, (Gibson and Burke, 1985); however, at this dose of SpGsc morpholino, most embryos lacked detectable Sp1 staining.

In addition to these defects in differentiation of vegetal cell types, embryos lacking SpGsc did not establish OA polarity, but instead assumed a radialized morphology that lacked definable oral and aboral territories separated by a ciliary band. Immunostaining reveals that all cells in the ectoderm express the aboral Spec1 marker. Little or no signal was detectable after immunostaining for the EctoV epitope, which is expressed at high levels and confined to the differentiated oral ectoderm of normal plutei (Fig. 4A). These observations demonstrate that SpGsc function is required for differentiation of endoderm, at least some, if not all, secondary mesenchyme, facial epithelium and ciliary band. By contrast, early differentiation of PMCs, as indicated by the presence of ingressed cells that stain specifically with the 6e10 antibody, is not affected (Fig. 4A, right; Fig. 4C).

The role of SpGsc within presumptive oral ectoderm could be to activate genes required for oral ectoderm fate, to repress those required for aboral fate, or both. The fact that the domain of expression of Spec1, a late aboral ectoderm marker, expands in the absence of SpGsc function and the presence of a GEH domain suggest a repressor function. One potential mechanism would be for SpGsc to antagonize the function of activators, such as SpOtx, another bicoid-class factor that is an essential early transcriptional activator in aboral ectoderm (Li et al., 1999; Mao et al., 1996). When SpOtx is converted to an active repressor by linking its DNA-binding domain to the engrailed repression domain and this protein is expressed in sea urchin embryos, accumulation of aboral ectoderm-specific markers is strongly suppressed (Li et al., 1999). Because Otx and Gsc have been shown to bind to the same cis-acting elements in other systems (Mailhos et al., 1998), it is likely that SpGsc can compete for binding at SpOtx target genes in oral ectoderm.

If this hypothesis is correct, then misexpression of SpGsc should divert aboral ectoderm toward oral fate, while an activator consisting of the SpGsc DNA-binding domain linked to the VP16 activation domain should have the opposite effect. That this is, indeed, the case is shown in Fig. 5. The ectoderm of embryos injected with SpGsc mRNA expresses EctoV (oral), but little or no Spec1 (aboral). Conversely, embryos injected with SpGsc-VP16 mRNA express predominantly Spec1. Thus, SpGsc and SpOtx-Eng behave similarly (Fig. 5, middle) (Li et al., 1999), as do SpGsc-VP16 and SpOtx (Fig. 5, bottom) (Mao et al., 1996).

The severity of the ectodermal defects caused by misexpression of SpGsc-VP16, as measured by the relative intensities of Spec1 and EctoV staining, appears to be greater than that produced by blocking SpGsc translation. This is not unexpected as SpGsc-VP16, when present at high levels throughout the embryo, is likely to have more transcriptional activation activity than that which is elicited by normal endogenous levels of SpOtx in the absence of SpGsc.

Both SpGsc loss-of-function experiments and the effects of SpGsc and SpGsc-VP16 misexpression support the idea that SpGsc can bind at SpOtx target sites and act as a competing repressor. To test this model directly, we performed EMSA to characterize SpGsc DNA binding properties in vitro, and transactivation assays to test the ability of SpGsc to directly compete with SpOtx function in vivo. EMSA showed that a bacterially produced GST-SpGsc fusion protein binds specifically to a DNA element contained on a 39 bp fragment (CII) previously shown to bind SpOtx and to mediate SpOtx activation of the aboral ectoderm-specific Spec2a gene (Mao et al., 1994). The CII probe also contains another cis element (OER) recently shown to mediate the binding of an oral ectoderm repressor unrelated to SpGsc (Yuh et al., 2001). As expected, the complex formed with GST-Gsc can be shifted with an anti-GST antibody (compare lanes 2 and 4 in Fig. 6A). This protein binds specifically to the Otx cis element, as competition with either the wild-type CII sequence or an oligonucleotide with mutations in the OER site effectively inhibited SpGsc-DNA complex formation (Fig. 6A, lanes 3, 7). By contrast, sequences with mutated Otx sites incapable of binding to SpOtx were ineffective as competitors (Fig. 6A, lanes 5, 6). In particular, a mutation changing TAATCC to TAATTG, which is known to specifically affect binding of class K₅₀ homeobox proteins, including Otx and Gsc (Hanes and Brent, 1991), did not compete for SpGsc binding at the Otx site (lane 5). To ensure that SpGsc was responsible for complex formation, we removed the GST moiety by proteolytic digestion with Factor Xa and repeated the EMSA. As expected, an SpGsc-DNA complex formed that was effectively competed by the CII wild-type oligonucleotide (Fig. 6A, lane 9) but not by the oligonucleotides with mutations in the Otx site (lanes 10, 11).

These experiments suggest that SpGsc binds to the Otx site with an affinity roughly comparable with that of SpOtx. This was confirmed by gel shift assays, which showed that the relative binding constant of SpGsc is within fourfold of that of SpOtx in vitro (3.1×10³ versus 1.3×10⁴). Direct competition between SpGsc and SpOtx for binding at the Otx site is demonstrated in Fig. 6B. In order to better distinguish between SpGsc- and SpOtx-DNA complexes by EMSA, we compared SpGsc with the GST moiety removed and the entire SpOtx-GST fusion protein. SpGsc alone formed a complex that was unaffected by the addition of an SpOtx antibody (Fig. 6B, lane 3). SpOtx-GST formed two lower mobility complexes that were both effectively supershifted with the SpOtx antibody (Fig. 6B, compare lanes 4 and 5). The faster migrating of the two SpOtx-GST complexes is likely to be a degradation product of the intact SpOtx-GST. When we fixed the amount of SpOtx-GST at 5.6 pmol and added increasing amounts of SpGsc from 2.8 to 19.9 pmol, we observed a decrease in the amount of SpOtx-GST complexes with a corresponding appearance of the SpGsc complex (Fig. 6B, lanes 6-8). About equal amounts of SpGsc and SpOtx-GST complexes were observed at an input ratio of 2 to 3 (lanes 7 and 8). In addition to the SpGsc complex, we observed a slightly slower migrating complex that behaved in a manner generally similar to the SpGsc complex and was dependent on added SpOtx-GST (Fig. 6B, lanes 6-8). Although the origin of this complex is unclear, it suggests an interaction between SpGsc and SpOtx-GST.

Conversely, when we kept SpGsc levels constant at 2.8 pmol and added increasing amounts of SpOtx-GST (1.4 to 11.2 pmol), we observed a decrease in the SpGsc complex and a corresponding increase in the SpOtx-GST complexes (Fig. 6B, lanes 9-13). Again, equimolar inputs of SpGsc and SpOtx-GST (lane 11) yield about three times more of the latter complex. Taken together, the EMSA results demonstrate that SpOtx and SpGsc are able to compete with one another at an Otx site.

To test the functional significance of the interaction between SpGsc and the Otx cis element in vivo, we used a transgene construct driven by a promoter whose activity is dependent on SpOtx cis elements (Mao et al., 1994). When increasing amounts of SpGsc mRNA were co-injected with this construct at the one-cell stage, its promoter activity was progressively and strongly reduced (Fig. 6C). The straightforward interpretation of these results is that SpGsc downregulates the activity of this promoter by competition for binding at the SpOtx cis elements. An alternative mechanism, i.e. that SpGsc sequesters SpOtx in inactive heterodimers, is unlikely because in other systems formation of such heterodimers depends on the binding of both proteins to a palindromic paired homeodomain binding site (Mailhos et al., 1998), which is lacking in the target promoter used here.

Although SpOtx was identified as an activator of aboral ectoderm-specific genes, it accumulates in the nuclei of all ectoderm cells (Li et al., 1997). By contrast, SpGsc is expressed only in the oral facial epithelial region of the ectoderm where it can act to repress SpOtx target genes, as the data presented above demonstrate. This suggests that the ratio of SpGsc and SpOtx levels is an important factor in regulating oral versus aboral fates, and spatial regulation of SpGsc transcription is therefore an essential feature of oral ectoderm specification.

Differentiation of ectoderm requires signaling from vegetal cells. Thus, animal halves of embryos, or embryoids derived from egg animal hemispheres, remain as dauerblastulae and a variety of experimental manipulations that interfere with signaling by vegetal blastomeres also secondarily lead to radialization of the ectoderm (Angerer and Angerer, 2000). Such embryoids or embryos have been interpreted to differentiate as oral ectoderm because they express the late oral ectoderm marker, EctoV throughout. SpGsc is expressed in oral ectoderm beginning at hatching blastula stage, much earlier the EctoV synthesis can be detected, and it is required for establishing OA polarity. Therefore, it was of interest to test whether accumulation of SpGsc mRNA is activated cell autonomously in oral ectoderm and, if not, then whether these signals are dependent on β-catenin, which is a major component of the vegetal signaling mechanism.

To test whether SpGsc expression is dependent on signals from other cells, transcript levels were measured by RNase protection in RNA from cells of embryos continuously dissociated beginning at the two-cell stage (Fig. 7A). No SpGsc mRNA was detectable in dissociated embryos by this sensitive assay, while control embryos showed the expected temporal pattern of accumulation. These results indicate that SpGsc expression in oral ectoderm, the major site of expression, is not activated cell autonomously, but they do not reveal whether the low level of transient SpGsc transcription in presumptive mesenchyme is sensitive to dissociation. In contrast to the sensitivity of SpGsc transcription to embryo dissociation, SpHE transcripts accumulate cell autonomously as previously reported (Ghiglione et al., 1993; Reynolds et al., 1992). Levels of 26S and 18S ribosomal RNAs in the stained gel demonstrate that the quality and recovery of RNA are equivalent in these samples.

To determine how late cell-cell interactions are required for SpGsc transcription, we dissociated embryos at progressively later cleavage and blastula stages and assayed SpGsc RNA levels when controls reached the gastrula stage (Fig. 7B). SpGsc message levels were strongly reduced even when embryos were dissociated as late as the ∼200-cell early blastula stage, but not when they were left intact until the mesenchyme blastula stage (∼500 cells), by which time SpGsc transcription has already begun in normal embryos. In this experiment, SpHE provides a contrasting control pattern of expression: SpHE transcripts are not present in 500-cell normal gastrulae (Reynolds et al., 1992), but they do persist to this stage in dissociated cells (Fig. 7B, dissociation at 2-, 32-, 64- and 200-cell stages), because interactions among cells are required for SpHE mRNA turnover (Ghiglione et al., 1993; Reynolds et al., 1992). These results indicate that the signaling required to initiate SpGsc transcription must continue until just a few hours before the gene is normally activated.

To determine whether SpGsc transcription in oral ectoderm requires nuclear β-catenin-dependent signals, SpGsc RNA levels were compared by RT-PCR in RNA of embryos injected with either glycerol or cadherin mRNA. As described previously (Howard et al., 2001; Logan et al., 1999; Wikramanayake et al., 1998), cadherin mRNA-injected embryos developed into epithelial spheres that lacked endoderm and mesenchyme (data not shown). The results of two experiments using different egg batches were in excellent agreement and are quantitated at the bottom of Fig. 7C, which also shows the primary data for experiment 2. In both cases, controls in which reverse transcriptase was omitted were negative and signal intensities were consistent with sampling during the linear phase of amplification. These measurements show that SpGsc RNA concentration is reduced between five- and tenfold in cadherin mRNA-injected 24-hour embryos. As expected, mRNA encoding the early vegetal plate marker, Endo16, is undetectable (Li et al., 1999; Wikramanayake et al., 1998).

To control for the formal possibility that lack of SpGsc expression results from a general arrest of ectoderm differentiation in cadherin-expressing embryos, we analyzed these embryos for accumulation of BMP2/4 mRNA. In normal embryos, this message begins to be transcribed about the same time as SpGsc mRNA, and it accumulates throughout ectoderm with higher levels in the oral region (Angerer et al., 2000). As shown in Fig. 7C, BMP2/4 mRNA accumulates to similar levels in cadherin-expressing and control embryos (compare BMP2/4 signals with mitochondrial rRNA load controls). We conclude that blocking β-catenin function does not generally inhibit activation of at least some relatively late ectoderm-specific genes.

The experiments presented here support the following major conclusions: (1) SpGsc acts downstream of nuclear β-catenin in sea urchin embryos, as demonstrated by its marked down regulation in embryos injected with cadherin mRNA; (2) as shown by morpholino antisense translational inhibition, SpGsc is required for gastrulation, for expression of the Endo16 vegetal marker, and for differentiation of pigment cells; and (3) SpGsc is required for differentiation of oral ectoderm and establishment of OA polarity, as also demonstrated by these loss-of-function experiments. Our results strongly suggest that SpGsc promotes oral fate by antagonizing aboral-specific gene functions. Thus, loss of SpGsc function in presumptive oral ectoderm causes it to express the Spec1 aboral ectoderm marker and misexpression of a transcription-activating counterpart, SpGsc-VP16, diverts presumptive oral ectoderm toward aboral fate. As an example of this proposed mechanism, we showed that SpGsc binds with high affinity to cis elements recognized by a closely related factor, SpOtx, that activates Spec2a and is thought to function upstream of other genes required for aboral ectoderm differentiation (Li et al., 1999). Furthermore, a transgene promoter whose activity depends entirely on SpOtx cis elements is silenced in vivo by co-expression of SpGsc. Thus, SpGsc is essential for cell fate specification along both AV and OA axes.

Loss-of-function experiments demonstrate that SpGsc expression depends on β-catenin/TCF-Lef function and that it is essential for mesendoderm development. The observed phenotype indicates that SpGsc is required for early steps in differentiation of multiple mesendoderm derivatives. SpGsc knockdown embryos do not accumulate Endo16 mRNA, an early marker that initially is expressed throughout the vegetal plate of normal embryos. Pigment cells, the major SMC type do not differentiate as assayed by expression of the early Sp1 marker. No archenteron develops and no vegetal cells express the EctoV epitope that identifies differentiated foregut cells. Typically, SpGsc knockdown embryos have some ingressed cells that migrate away from the vegetal pole and display the 6e10 epitope, indicating that initial PMC differentiation is not blocked. Other cells clustered at the vegetal pole may be PMCs that have not yet begun to express the epitope or may represent other, arrested, vegetal cell types.

The vegetal domain in which we have detected SpGsc mRNA expression is restricted to a central torus in the vegetal plate that includes PMC precursors and perhaps some SMC precursors. As differentiation of other vegetal cells is affected, the outputs of SpGsc activity probably include non cell-autonomous signaling from either vegetal or oral ectoderm cells. The first, more traditional, possibility is that vegetal SpGsc is a downstream effector of the patterning mechanism initiated by vegetal β-catenin/TCF-Lef. An interesting alternative is that SpGsc function in oral ectoderm is required to activate signaling to the vegetal plate. Such a function could provide a feedback mechanism in which β-catenin-dependent signals cause expression of SpGsc in oral ectoderm, which then sends signals to mesendodermal tissues that promote their differentiation. However, because exogastrulae can develop relatively normal tripartite guts with pigment cells and muscle cells at their tip, such signals presumably would have to function transiently at the beginning of gastrulation. Of course, the potential mesenchymal and oral ectodermal roles of SpGsc in vegetal differentiation are not mutually exclusive and it should be possible to examine their separate contributions by constructing chimeras between macromere+mesomere tiers and micromeres in which SpGsc translation is blocked in one component by a morpholino knockdown.

Activation of SpGsc in oral ectoderm clearly depends on nuclear β-catenin function. As differentiation of ectoderm is known to require vegetal signals, the most likely hypothesis is that expression in oral ectoderm is a consequence of the activity of the canonical, vegetal Wnt signaling pathway. While the possibility that β-catenin also functions in ectoderm nuclei cannot be excluded, it has never been detected there, even when these cells are induced to form endoderm. We found it surprising that SpGsc expression was blocked when nuclearization of β-catenin was inhibited by cadherin, because cadherin-treated embryos and embryos derived from animal halves express EctoV strongly throughout the ectoderm, which has consequently been interpreted to be oral ectoderm (Li et al., 1999). Thus, we anticipated that SpGsc would be up regulated in cadherin-treated embryos. The fact that it was downregulated leads us to suspect that EctoV accumulation is not always an indicator of oral ectoderm fate. Although EctoV accumulates specifically in differentiated oral ectoderm (both facial epithelium and ciliary band) at late stages, its initial pattern of zygotic synthesis, which begins at least several hours after SpGsc expression at late mesenchyme blastula stage (Coffman and McClay, 1990), is not known. While further studies are required to resolve this question, the relatively high and uniform concentration of EctoV in cadherin-expressing or animal-half embryos may alternatively reflect its abnormally prolonged, uniform synthesis in embryos in which aboral ectoderm specification is inhibited. We suggest it is not the case, however, that these embryos are arrested at a stage before SpGsc transcription begins, because they do transcribe BMP2/4, which is activated at the same time as SpGsc in normal embryos (Angerer et al., 2000).

If the nuclear β-catenin-dependent signals that activate SpGsc transcription in oral ectoderm come from vegetal blastomeres, then this mechanism would help explain why elaboration of oral-aboral patterning depends on vegetal signals. SpGsc message begins to accumulate in the presumptive oral territory of hatched blastulae only after major regions of the vegetal plate are thought to be conditionally specified but before the endoderm-ectoderm border is established (Angerer and Angerer, 2000). Our cell dissociation experiments show that the signaling required for SpGsc transcription must occur (or persist until) shortly before activation occurs. Therefore, activation of SpGsc in oral ectoderm most likely depends on signals originating from the adjacent conditionally specified endoderm. As oral ectoderm differentiation continues throughout gastrulation and the endoderm-ectoderm border remains subject to respecification during this time (McClay and Logan, 1996), continued signaling may be required relatively late in development. The persistence of high levels of SpGsc expression in oral ectoderm through the end of embryogenesis suggests that it acts within this territory to help maintain oral ectoderm fate.

Our observations on SpGsc function lead to a three-step model for ectoderm patterning. First, an autonomously acting set of animalizing transcription factors (ATFs) can drive initial specification of a pre-ectoderm state (Angerer and Angerer, 2000). Embryonic ectoderm passes only transiently through this state during cleavage in normal embryos. If embryos are artificially locked in this pattern of gene expression, as are animal-half embryos or embryos deprived of β-catenin/TCF-Lef function, they develop as classic dauerblastulae, which have a thickened wall at the animal pole and eventually express EctoV over most or all of their surface (depending on the species). Second, β-catenin-dependent signals emitted uniformly from the vegetal hemisphere then up regulate transcriptional regulators required for ectoderm to progress to conditional specification as aboral ectoderm. One of these regulators is SpOtx, which is present at similar concentration in all ectoderm nuclei (Li et al., 1997). This factor is present maternally and is likely to act upstream in the aboral ectoderm specification pathway. In addition to its accumulating in nuclei during cleavage stages, exogenously supplied SpOtx can rescue uniform expression of aboral ectoderm-specific genes in embryos injected with cadherin mRNA, including actin CyIIIa, which is thought not to be activated directly by SpOtx. In the third step, patterning of ectoderm to form separate oral and aboral territories requires activation of repressors of aboral ectoderm genes in the oral territory. This proposed sequence of events is consistent with previous observations that many genes ultimately expressed in only the aboral ectoderm are initially activated at the end of cleavage throughout most or all of the ectoderm (Kingsley et al., 1993; Yang et al., 1989a; Yang et al., 1989b). Only subsequently are these genes downregulated in oral ectoderm, beginning around mesenchyme blastula stage, which is the time when SpGsc begins to accumulate. Interestingly, the other spatial regulators of aboral ectoderm-specific genes identified to date (p3A2) (Kirchhamer and Davidson, 1996) (OER) (Yuh et al., 2001) also function as repressors in oral ectoderm rather than as spatially restricted activators in the aboral territory.

We have provided one mechanism for SpGsc function in oral ectoderm differentiation by demonstrating that it can compete with SpOtx to repress genes promoting aboral ectoderm fate. Thus, establishment of discrete ectoderm territories does not occur until competitive levels of repressor activities, such as that of SpGsc, are reached just after hatching blastula stage. When SpGsc is misexpressed precociously and ectopically by mRNA injection, it presumably displaces SpOtx from its target genes throughout the presumptive ectoderm, leading to suppression of aboral ectoderm differentiation. Conversely, overexpression of SpOtx effectively competes SpGsc function in oral ectoderm, thereby allowing continued transcription of aboral-specific genes (Mao et al., 1996). This remarkable ability of individual factors like SpGsc and SpOtx to drive differentiation towards a specific cell type may suggest that these factors function far upstream in the pathways that specify ectoderm tissues. We think a more likely alternative is that each of these factors is a member of a network of crossregulating activators and repressors, and that these networks can be manipulated by altering the levels of individual members. Thus, it will be of interest to identify other regulatory factors operating in these tissues.

A critical component of this model that remains to be identified is the activity that restricts SpGsc expression to the oral region. The specific activation of SpGsc in presumptive oral ectoderm demonstrates that a unique transcriptional territory has been established there by about the 250-cell stage. This activity cannot depend on vegetal nuclear β-catenin, which exhibits no OA polarity. This polarizing component most probably arises from the poorly understood mechanism that initially specifies the OA axis. It might originate in the early redox gradient that predicts the alignment of this axis and has been shown to affect the activity of at least one transcriptional regulator (Coffman and Davidson, 2001). Although vegetal signals clearly are required for establishing OA polarity, it is not yet known whether these signals are instructive (i.e. the polarity is imposed from the vegetal cells) or permissive (i.e. uniform vegetal signals are required to implement an inherent OA polarity in the ectoderm). Evidence that a relatively early, vegetal OA polarity exists is that the Notch receptor is enriched on the apical surfaces of presumptive endoderm cells on the aboral side around the blastula stage (Sherwood and McClay, 1997). If the model we have presented is correct, then identification of the spatial regulatory elements of the SpGsc promoter should define the additional activities that specify oral ectoderm, bringing us much closer to understanding how the oral-aboral axis is established.

This work was supported by grants from NIGMS (EM25553) (R. C. A.), NICHD (HD22619 to W. H. K.) and the Robert A. Welch Foundation (G-0010 to W. H. K.). We would like to thank Athula Wikramanayake and Ruthie Zearfoss for generating the original SpGsc PCR product, and Drs Alan Vonica and Judy Venuti for a plasmid containing the VP16 domain.