Repression of mesodermal fate by foxa, a key endoderm regulator of the sea urchin embryo

The endomesoderm gene regulatory network (GRN) models the transcriptional control system defining vegetal specification of the sea urchin(Strongylocentrotus purpuratus) embryo during the first 30 hours of development. The GRN is a nuclear view of transcriptional regulation and incorporates data from many sources. In addition, it includes known signal transduction events that subdivide the endomesoderm into subnetwork modules during this initial 30-hour period. By gastrulation, mesoderm and endoderm have been defined, and have acquired competence to undergo the morphogenetic movements of gastrulation. The network model further provides the genomic regulatory code for subsequent steps of differentiation and morphogenesis.

The GRN is activated by maternal inputs at the vegetal pole, first seen as nuclearization of β-catenin in the micromeres and in macromeres(Logan et al., 1999). Between the fourth and sixth cleavage, the early signal, the molecular nature of which is not yet known, from the micromeres provides added input to the macromeres to accelerate endomesoderm GRN activation(Ransick and Davidson, 1995; Oliveri et al., 2003). At sixth cleavage, the veg2 tier of endomesoderm cells separates from their sister veg1 cells, the veg2 tier occupying a more vegetal position. Between the seventh and ninth cleavages, the innermost veg2 cells receive a Delta signal from the adjacent micromeres, and the Notch-activated cells become secondary mesenchyme (SMC) precursors. This Notch (N) signal functionally distinguishes SMC from endoderm specification(Sherwood and McClay, 1999). A second Delta signaling event later originates in the SMC precursors,contributes to SMC specification, and is required to define the adjacent veg2 endoderm subcompartment (Sherwood and McClay, 1999; Sweet et al.,2002; Peterson and McClay,2005). In the sea urchin embryo, at least 16 transcription factors are activated specifically in cells that become endoderm, and among these is the foxa gene.

Foxa belongs to the forkhead family of transcription factors. Orthologous genes have been isolated in many different species, including Drosophila (Weigel et al.,1989), C. elegans(Mango et al., 1994), mouse(Ang et al., 1993), ascidians(Corbo et al., 1997; Olsen and Jeffery, 1997),hemichordates (Taguchi et al.,2000) and cnidarians(Fritzenwanker et al., 2004; Martindale et al., 2004). A common feature is that Foxa factors are restricted to endodermal cells just prior to, or during, gastrulation, and are necessary for the specification and differentiation of endodermal structures. Later, these factors are required in other domains and other developing structures. In Xenopus,foxa2/hnf3β is initially expressed at the vegetal pole of the embryo, but is excluded from the mesoderm during gastrulation(Suri et al., 2004). In this embryo during gastrulation, foxa2/hnf3β has the major role of determining the mesodermal boundary by repressing mesoderm fate in the endoderm.

Sea urchin foxa was originally identified in Hemicentrotus pulcherrimus and named hnf3(Harada et al., 1996). The nomenclature used here is current for Forkhead transcription factors of this class (Kaestner et al., 2000). In situ hybridization showed Hpfoxa to be expressed initially in the vegetal plate, then surrounding the blastopore and finally in the gut of the embryo. Here, we characterize the functional role of foxa in the endomesoderm regulatory network by adding a high-resolution analysis of the foxa expression pattern in both S. purpuratus and Lytechinus variegatus, by performing experimental perturbations and embryological manipulations, and by using biochemical approaches to show that foxa has at least three major roles in the embryo: it assures that veg2 endoderm cells do not express mesoderm genes; it is required in the stomodeal region of the oral ectoderm for the production of the mouth; and it provides a controlling function for postgastrular development of the gut.

A 224 bp fragment of Spfoxa cDNA (from position 644 to 867 of the GenBank sequence, Accession number DQ459376), corresponding to the most conserved region of the forkhead domain (a kind gift of Cathy Yuh, Division of Biology, Caltech) was used as probe for screening Strongylocentrotus purpuratus and Lytechinus variegatus bacterial artificial chromosome (BAC) genomic libraries. The 150 kb 41I19 S. purpuratusclone (GenBank Accession number AC131450) and the 55 kb L. variegatusclone 4G18 (GenBank Accession number AC131500) were sequenced by the Joint Genome Institute and the Institute for Systems Biology, respectively. Sequences are publicly available at http://sugp.caltech.edu/resources/annotation.psp.

Whole-mount in situ hybridization was performed as previously described(Minokawa et al., 2004). An RNA probe for foxa was synthesized in vitro from 200 ng of PCR fragment amplified from the 224 bp cDNA clone using the primers pSport F(5′-GTG CTG CAA GGC GAT TAA GT-3′) and pSport R (5′-TGT GGA ATT GTG AGC GGA TA-3′). For gcm probe description, see Ransick et al. (Ransick et al., 2002),and for gatae probe, see Lee and Davidson(Lee and Davidson, 2004).

A morpholino antisense substituted oligonucleotide (MASO) was synthesized(Gene Tools) complementary to the sequence just upstream of the first possible methionine, namely 5′-TGGGTTCCTCTTTGAAATCCACGAT-3′. A morpholino standard control 5′-CCTCTTACCTCAgTTACAATTTATA-3′ was provided by Gene Tools. MASOs were injected in a 120 mM KCl solution, at final concentrations of 50 mM to 300 mM. The foxa coding construct was made using a vector derived from BlueScript (Statagene, La Jolla, CA), which contained the 3′UTR and 5′UTR of the globin gene(Lemaire et al., 1995). A fragment containing the entire coding sequence of the foxa gene was obtained by PCR, using the primers Foxa Cod F(5′-CATACACATCAGTGGAGGCT-3′) and Foxa Cod R(5′-TCCATCTATAACTGGTCGTG-3′). The 1659 bp PCR fragment was initially subcloned in pGEM-T easy vector (Promega). A foxa pGEM-T positive clone was digested with EcoRI to release the 1659 bp fragment with ends compatible with the EcoRI cloning site of the pBlueScript derived vector. The orientation of the cloned fragments was tested by PCR. To build the 5′foxa-GFP construct, a 540 bp fragment containing 150 bp of 5′UTR and 390 bp coding sequence was amplified by PCR, using the primers Foxa HindIII R(5′-CCCCAAGCTTGATACGATCAATAGA-3′) and Foxa KpnI F(5′CCCCGGTACCAGGCTGACACTATATACT-3′). The GFP coding sequence was subcloned in frame as described by Oliveri et al.(Oliveri et al., 2002). The template used for these PCR reactions was the BAC clone 41I19. Each construct was checked by sequencing. RNA used for injection was synthesized as described by Oliveri et al. (Oliveri et al.,2002). The injection solutions were concentrated to 20 ng/μl,50 ng/μl, or as indicated, and the RNA was injected together with fluorescein dextran or rhodamine dextran (10 pg/pl; Sigma).

Embryo cultures of L. variegatus or S. purpuratus, and microsurgical procedures, were carried out as described earlier(McClay, 2000; Oliveri et al., 2003; Sweet et al., 2004). L. variegatus embryos were placed into Kiehart chambers in calcium-free sea water, whereas S. purpuratus embryos were placed in calcium-free sea water after two washes in Hyaline Extraction Medium. After surgery the embryos were returned to sea water.

Total RNA was isolated from batches (100-200) of embryos injected with different MASOs and/or mRNA. The RNA was extracted using the RNeasy Micro Kit(Qiagen) according to manufacturer's instructions. First-strand cDNA was synthesized using random hexamers and the Taq Man Kit (PE Biosystems), as described by the manufacturer. The cDNA was used directly for quantitative PCR(QPCR) analysis. QPCR was conducted as previously described(Rast et al., 2000). For all QPCR experiments, the data from each cDNA sample were normalized against ubiquitin mRNA levels, which are known to remain relatively constant during development (Nemer et al.,1991; Oliveri and Davidson,2004a; Ransick et al.,2002). The primers used can be found on the website http://sugp.caltech.edu/resources/methods/q-pcr.pspand have been previously published by Davidson et al.(Davidson et al., 2002).

The S. purpuratus and L. variegatus foxa genes were identified and sequenced as described in the Materials and methods. In both species, the foxa gene consists of a single exon. A search for the foxa gene in the recently available S. purpuratus genome sequence showed that foxa is a single copy gene, as was previously determined for Hemicentrotus pulcherrimus foxa(Harada et al., 1996).

Sea urchin Foxa is a class A forkhead transcription factor. Predicted amino acid sequences for the sea urchin Foxa proteins are 98.1% identical between S. purpuratus and H. pulcherrimus, 94.5% identical between S. purpuratus and L. variegatus, and 95% identical between L. variegatus and H. pulcherrimus, over the whole length of the protein. There is thus no doubt that these genes are true orthologs. The phylogenetic relationship of Spfoxa, and thus Lvfoxa, to other forkhead class transcription factors has been resolved by Tu et al.(Tu et al., 2006).

The analysis previously performed on H. pulcherrimus(Harada et al., 1996) was repeated to provide a higher resolution whole-mount in situ hybridization(WMISH) expression of foxa (Fig. 1). At 18 hours of development, foxa is expressed at a low level in the veg2 endomesoderm; the micromeres are devoid of expression(Fig. 1A,B). By 21 hours,expression of foxa becomes restricted to the endodermal ring(Ruffins and Ettensohn, 1996). The gene is transcribed unequally in the two sides of the endodermal ring(Fig. 1C). At 21 hours and beyond (asterisk in Fig. 1E-G),the oral side of the endoderm, marked by the stomodeal expression, expresses a higher level of foxa than does the aboral side. The stomodeal expression domain was not identified in the previous analysis(Harada et al., 1996). Following primary mesenchyme cell (PMC) ingression(Fig. 1D), the endoderm expresses the gene at a high level. Thus, initially there is a transient expression in the endomesoderm, some of which will be specified as muscle and coelomic pouch SMCs, then, later, foxa is exclusively an endodermal and stomodeal gene.

A temporal expression profile at a 3-hour resolution is shown in Fig. 2. The foxa gene is not expressed maternally. Its zygotic expression begins at about 15 hours postfertilization, and continues throughout embryogenesis. The temporal expression profile displays a striking oscillatory periodicity, as has also been described for the starfish foxa gene(Hinman et al., 2003).

A morpholino antisense substituted oligonucleotide (MASO) complementary to the translational start site of both the S. purpuratus and L. variegatus foxa mRNA was tested for efficacy in arresting translation using a GFP construct (5′foxa-GFP), which contained in-frame the target site sequence of the gene. As illustrated in Fig. 3J,K, the foxaMASO very effectively blocks the translation of RNA containing the initial ATG of the foxa message, but fails to block the translation of an altered form (5 base changes) that no longer recognizes the MASO (data not shown). Two different concentrations of foxa MASO were then injected into fertilized eggs. The embryonic phenotypes observed were of increasing severity at higher concentrations, and were equivalent in both species. Up to PMC ingression, the foxA MASO-injected embryos appeared normal, but at gastrulation MASO treatment caused a delay of gut invagination, and, at increased concentrations, a complete failure of invagination(Fig. 3A,B). If invagination occurred in MASO-treated embryos, the foregut of the embryonic archenteron was truncated or missing altogether (Fig. 3E,H). This region of the archenteron is most sensitive to the absence of Foxa, consistent with the hypothesis that the high level of foxa expression in the foregut region during gastrulation is crucial for specification (Fig. 1I,J). At higher concentrations of MASO the whole gut is absent, and in its place is a small, everted, scar-like structure (Fig. 3D,G). In both species, the overall effect of foxa MASO treatment is a reduction in the mass of endoderm. Even if invagination occurs,the gut fails to connect with the oral ectoderm and a mouth never forms(Fig. 3P). In addition,MASO-treated embryos produce excess numbers of pigment cells(Fig. 3M,N; an average increase of 40-70%, three experiments, n=10 experimental and 10 control embryos counted at the same stage in each experiment). These data suggest that an absence or reduction of Foxa leads to a diversion of presumptive endodermal cells to mesodermal fate, and a reduction in endoderm specification. The MASO phenotypes are consistent with the expression data shown in Fig. 1, and together with that data confirm foxa as a key regulatory gene of the endomesoderm GRN.

Table 1 shows the effects,as reflected by QPCR, on genes that execute functions essential to endomesoderm specification. Absence of Foxa function has very specific effects, and the expression of the large majority of genes tested was not quantitatively perturbed in the timeframe studied. The level of foxatranscript itself increases sharply when foxa mRNA translation is inhibited. Thus, the foxa gene is subject to repression by Foxa protein. This result at least partially explains the oscillatory character of the temporal expression profile observed for the foxa gene(Fig. 2). An early repressive effect was also seen on gatae at the blastula stage, when foxa is still transiently expressed in the endomesodermal territory and gatae is expressed in territory that will become late SMC mesoderm (Fig. 1)(Lee and Davidson, 2004). By the time the expression of gatae occurs in definitive endoderm at mesenchyme blastula stage, there is no evident foxa repression of gatae. Expression of both gatae and endo16, a well-known endodermal marker gene, are controlled at this stage by other known endodermal regulators (Yuh et al.,1994) (Table 1). At late gastrula stage in foxa MASO-treated embryos, the gataeand endo16 genes show a strong decrease in the level of expression,an obvious consequence of the prior failure of endomesoderm specification(Fig. 3). These last results are not interpreted as evidence for direct regulatory gene interactions in late gastrula, but as a consequence of the earlier failure of Foxa to contribute to endoderm specification. Foxa is also an important regulator of hedgehog (hh) expression, which is known to be expressed in endoderm beginning just before mesenchyme blastula stage(Walton et al., 2006).

A second important result (see Table 1) is that the early SMC-specific gcm gene(Ransick et al., 2002) is revealed to be a target of foxa repression in a time window that begins after 24 hours and extends at least until 34 hours. This result provides a direct explanation for the essential role of foxa in confining endodermal cells to their appropriate fate, as we demonstrate in the following experiments.

foxa MASO reduces gut size or eliminates it, and additional pigment cells are observed (Fig. 3N,P). Because in normal embryos foxa is expressed exclusively in veg2 endoderm after very early stages of development, the implication is that a specific function of this gene is to prevent a number of veg2 progeny from executing mesodermal fates. To test this hypothesis directly, a mosaic analysis was performed (see Fig. 4).

The foxa MASO was injected into a group of eggs along with Rhodamine dextran. At the 60-cell stage, two fluorescent, MASO-bearing cells from each embryonic tier were transplanted to equivalent positions in place of two normal cells in an unlabeled host embryo, as shown in the diagrams to the left in Fig. 4. These mosaic embryos were then compared with the respective mosaic control embryos, in which the fluorescent transplanted blastomeres contained no MASO. As shown in Fig. 4, control and foxa MASO animal tier, veg1 and micromeres cells behaved identically in the host embryos. However, the fate of foxa MASO-injected veg2 cells was very different from that of their controls. Transplanted veg2 blastomeres unable to generate the Foxa protein produced only dispersed mesodermal cell types, whereas the control dye-injected veg2 cells became normal gut endoderm, as well as mesodermal cells. This same outcome was observed both in embryos of S. purpuratus and L. variegatus;24 cases of each tier in each species were scored. Some experimental veg2 transplant embryos had a few fluorescent cells in the gut, but these were always far fewer in number than in control archenterons, and these often differentiated as ectopic pigment cells or ectopic coelomic pouches, both SMC derivatives. Thus, one function of the foxa gene is to repress mesodermal specification in the subset of veg2 cells normally fated to generate gut endoderm.

WMISH was carried out on foxa or control MASO-injected embryos at blastula (data not shown), mesenchyme blastula, early gastrula and late gastrula stages (18 hours, 23-24 hours, 30-34 hours and 48 hours postfertilization for S. purpuratus), along with non-injected controls. The genes targeted in this series of experiments were those implicated as being downstream targets of foxa in the QPCR experiments (see Table 1),namely foxa itself and the mesoderm regulator gcm(Ransick et al., 2002). In addition, we looked at gatae expression as an indicator of endoderm regulatory state. Representative WMISH results are reproduced in Fig. 5. As expected(Table 1), gataeexpression appeared normal until the onset of gastrulation (data not shown). As the MASO phenotype indicates, the foxa gene is required as a positive regulator of later endoderm development(Fig. 5C,F). Even though gastrulation of foxa MASO-treated embryos was delayed or did not occur at all, they produced a higher level of foxa transcripts than did controls (Fig. 5G-L), just as had been demonstrated by the QPCR data(Table 1). As indicated earlier, one MASO effect is the maintenance of a high level of foxatranscripts in the veg2 cells, by cancellation of the foxaautorepression circuit.

The reallocation in foxa MASO embryos of endodermal cells to a mesodermal state is also visualized by WMISH, by expression of gcm. No difference from controls is observed until 24 hours(Fig. 5A,D). By the beginning of gastrulation (32 hours), there is an expansion in the number of cells expressing gcm in MASO-treated embryos(Fig. 5B,E; 33% more gcm positive cells, n=6 experimental embryos and 6 controls). This is consistent with the derepressive effect of foxaMASO on gcm expression seen by QPCR(Table 1). Furthermore, ectopic foxa mRNA overexpression (Fig. 2), depresses the expression of gcm(Fig. 5M-O) in a concentration-dependent manner. Only high concentrations of injected mRNA completely abolish gcm expression to a level equivalent to the endogenous spike of foxa expression.

In normal embryos, the mesoderm territory arises from those veg2 cells that are exposed to a Delta signal expressed in the PMCs during cleavage(McClay, 2000; Sherwood and McClay, 1999; Sweet et al., 2002). The Delta signal is transduced by the Notch (N) receptor in these cells. For early SMC specification, a primary regulatory gene target of N signaling is gcm. The genomic interaction is direct, as Su(H) target sites that mediate these N effects have been demonstrated in the cis-regulatory apparatus of the gcm gene (Ransick and Davidson, 2006). At the beginning of the mesenchyme blastula stage, the delta gene is expressed in the SMCs, and this later signal is received by both late-specified SMCs(Sweet et al., 2002) and veg2 endoderm (Peterson and McClay,2005). It is reasonable to suppose, therefore, that foxarepression of mesoderm fates in the endoderm is required to preclude the same response to N signaling in the endoderm. Embryo recombinants were designed as tests of this proposition.

We investigated whether the foxa MASO could produce a transformation to mesodermal fates in veg1 endoderm, as these cells are not exposed to Delta-Notch signaling. At the 60-cell stage, fragments from embryos injected with Rhodamine dextran and foxa MASO, were recombined with control embryo fragments, labeled with fluorescein(Fig. 6A,D). Veg1 plus the animal hemisphere of foxa MASO embryos were combined with control veg2 plus micromere fragments, and the reciprocal recombinations were also generated. When the veg2 tier contained the foxa MASO, excess SMCs were produced by this tier, and there was very little veg2 contribution to the gut (Fig. 6A-C). Instead, the control veg1 cells made up almost all of the gut, even the foregut, which normally receives almost no contribution from the veg1 descendants. The foregut patches in Fig. 6B,C(red) were observed frequently, and much of that tissue became mesodermal coelomic pouches. Essentially, this is the same as the result shown in the mosaics (Fig. 4). In the reciprocal combination (Fig. 6D-F), in which all veg1 cells contained the foxa MASO,most of the gut is derived from the control veg2 cells (green). Thus, there is a subnormal contribution of veg1 cells (red) to the gut. The foxaMASO-injected veg1 cells contributed a small amount of gut in some cases (the MASO-treated cells produced a subnormal amount of hindgut, see Fig. 6E,F), but in each case the control veg2 cells regulated to compensate for the reduced endoderm formed by foxa-MASO veg1 cells. It follows that, in the veg1 cells, the foxa gene contributes to postgastrular specification of the endoderm,as, in its absence, far fewer veg1 cells become hindgut and midgut endoderm. Notably, no excess SMC types of veg1 origin were observed. Veg1 cells do not receive the Delta-Notch signal, and therefore one function employed by Foxa in veg2 cells (gcm repression) is not detectable in veg1 cells.

The foxa gene is expressed in the oral ectoderm region, beginning at about 24 hours after fertilization (Fig. 1), and, in the whole embryo, foxa MASO precludes formation of the mouth (Fig. 3). To confirm directly that oral expression of foxa is necessary for mouth formation, chimeric recombinations were generated (see Fig. 7). Recombinants were produced at the 32- to 60-cell stage. Uninjected halves (green) were reciprocally recombined with foxa MASO-injected halves (red; see Fig. 7A,D). The embryos were imaged at 40 hours. Embryos with control (green) animal halves had a normal mouth (Fig. 7C). Embryos without Foxa in the animal half (Fig. 7E,F) make a normal gut but no mouth. Thus, the oral territory of foxa expression is indeed required for production of the mouth.

We have shown that the foxa regulatory gene executes three different classes of function in the development of the sea urchin embryo. Only the first of these can, at present, be understood in mechanistic detail,as only this function takes place in the context of the established endomesoderm GRN: foxa represses gcm and mesoderm specification in the presumptive endoderm of the mesenchyme blastula-early gastrula embryo. The two other roles of foxa, later specification of endoderm and specification of a mouth in the oral ectoderm, occur later in the postgastrular embryo, which has yet to be analyzed at the gene regulatory network level. Foxa involvement in endoderm specification has been observed in all bilaterarians examined to date. In addition, expression of foxais necessary for development of the stomodeum, a new observation.

The participation of a given regulatory gene in multiple developmental processes, controlled by separate GRNs, is an emergent theme in regulation molecular biology. From an external point of view, this is an obvious and general requirement stemming from the fact that all bilaterians use essentially the same regulatory gene toolkit, so that in their immensely diverse developmental processes the same genes have to be deployed over and over again for different purposes (for reviews, see Davidson, 2006; Erwin and Davidson, 2002). Furthermore, in the sea urchin embryo, 80% of all regulatory genes are used just to get to the late gastrula stage(Howard-Ashby et al., 2006),and multiple re-use of these genes is therefore an inescapable inference. Many particular examples of regulatory genes that execute totally distinct functions during sea urchin embryogenesis are indeed already in hand, for example hnf6 (Otim et al.,2004), otx (Li et al., 1997; Yuh et al.,2002), deadringer(Amore et al., 2003), gsc (Angerer et al.,2001), and blimp1/krox(Livi and Davidson, 2006a; Livi and Davidson, 2006b). The C. elegans ortholog of the foxa gene, pha4,regulates different gene batteries at different times(Ao et al., 2004; Gaudet and Mango, 2002; Gaudet et al., 2004), and a particular aspect of its multiple capabilities is that the Foxa transcription factor recognizes high and low affinity target sites according to its concentration. That the same is likely to be true for sea urchin foxais implied by the oscillatory time course shown in Fig. 2. This is clearly due to foxa autorepression (Table 1). But when the autorepression is blocked from occurring by the introduction of foxa MASO, there is no change in the location of expression (Fig. 5), so the only significance of the normal oscillation is to alter the concentration of the foxa gene product over time within the cells of the endoderm. Where the specific targets of foxa are known, i.e. in the early phase of its function, we can associate the level of foxa expression with a given function, and from the time course of the oscillation phase we can determine the temporal duration of that function.

During the blastula stage, foxa has no input into any known portion of the regulatory apparatus controlling endoderm specification(Table 1; genes not affected by foxa MASO), nor does the blockade of Foxa translation cause any digression from normal developmental morphology up to gastrulation. But nonetheless, this gene has a function that is essential for endoderm specification, namely, to permit this specification to occur at all. Fig. 8A shows the inputs that activate foxa in the veg2 endoderm. Following mesoderm specification,the delta gene becomes active in mesoderm cells(Fig. 8C), although this second phase of Delta expression is independent of the first PMC Delta signal(Sweet et al., 2002). The second signal is received in the adjacent veg2 cells, where it is essential for activation of the essential endoderm regulatory gene gatae. This has been shown to be a direct cis-regulatory input mediated by the Su(H)transcription factor (P. Y. Lee and E.H.D., unpublished). As the gcmgene is directly activated by N signaling via Su(H) as well (op cit),and because foxa expression in the endoderm normally represses gcm (Table 1, Fig. 5), the effect of preventing foxa expression is to promote ectopic gcmexpression in cells that normally would become endoderm. These cells now become respecified as mesodermal cells (Figs 4, 6). In other words, a means of preventing gcm expression in endoderm is essential as a device to permit endoderm specification resulting from the N input.

Seen in this light, the fine-tuned elegance of the foxa regulatory architecture (Fig. 8) is thought provoking when viewed in evolutionary terms. Both PMCs and precociously specified pigment cells are echinoid specialties, and so at least parts of this architecture are likely to have been installed since the divergence of euechinoids. In starfish, there is also a N input into gatae in the endoderm at the equivalent stage, and foxa also represses itself, but gcm is expressed quite differently and in cells arising elsewhere (Hinman et al.,2003) (V. F. Hinman and E.H.D., unpublished). The foxarepression function has different targets in the starfish, including gatae in the mesoderm (Hinman et al., 2003). Since divergence, not only has foxarepression of gcm in the endoderm been inserted in the euechinoid lineage, but its operation is temporally regulated by its own autorepression so as to operate at just the right time to control N signal effects; high levels of foxa gene product are apparently required for both gcm repression and foxa autorepression.

Notch signaling is directly or indirectly required for specification of the late delaminating SMC derivatives, muscle and coelomic pouches(Sweet et al., 2002; Oliveri et al., 2002; Peterson and McClay, 2005; Sherwood and McClay, 1999). The precursors of these cells are sorted out within the SMC domain during the blastula stage (Ruffins and Ettensohn,1996), when foxa is transiently expressed in an overlapping domain with gcm. Because N signaling activates the gcm gene, and indeed is the necessary input required earlier to activate gcm in the mesoderm cells in response to the initial Delta signal received from the PMCs (Ransick and Davidson, 2006), and because Foxa represses gcm, the early phase of foxa expression could affect fate allocation among SMCs as well. In this case, some of the excess pigment cells produced by foxa MASO treatment would reflect an alteration of SMC fate balance.

Only phenotypical evidence suggests later roles of foxa, and no direct gene targets in either the archenteron or the stomodeal area are demonstrated. Furthermore, it must be considered whether the effects of foxa MASO treatment on development of the archenteron (see Fig. 3A-I, Fig. 5A-F) could, at least in part, just be secondary effects of the conversion of veg2 cells to SMC fates. Alternatively, convincing arguments suggest that foxa directly promotes the gene expression required for further development of the archenteron. The postgastrular roles of this gene are most simply interpreted as the provision of positive inputs into gut and stomodeal genes, even though the pregastrular role, the only one of which we know the mechanism, is a repressive one. It is possible that the amplitude of expression seen in Fig. 2 is permanently damped after gastrulation gets underway, and, as we speculate above, perhaps Foxa acts as a repressor only when expressed at high levels. This would explain the autorepression revealed by increased foxa production in the presence of foxa MASO, and the repression effect on gcm expression only at a time when endogenous foxa is at its highest levels.

The conversion to an SMC fate affects only veg2 and not veg1 endoderm. In the absence of veg2 endoderm, or if veg2 is unable to express foxa,veg1 endoderm can form the entire gut. The chimera experiments(Fig. 6) show clearly that when veg1 contains the foxa MASO it fails to generate hindgut endoderm or contribute at all to the midgut, as it does in normal embryos(Logan and McClay, 1997; Ransick and Davidson, 1998). But because veg1 is not subject to conversion to SMCs (because it has not received the Notch signal), the foxa MASO must interfere with other functions of foxa needed in the postgastrular endoderm of the hindgut. Note that foxa is normally expressed in the late gastrula in all of the gut (Fig. 1). It is likely that the MASO effects on foregut development are also due to the failure of regulatory interactions needed for that aspect of gut morphogenesis and differentiation, this time in veg2 derivatives. These deductions predict postgastrular GRN subcircuits in which the foxa gene, no doubt together with other regulators, operates batteries of downstream genes that are required in the anterior and posterior domains of the gut,respectively.

We thank Dr Andy Ransick of California Institute of Technology for critical and fruitful discussions; Jina Yun whose extraordinary technical assistance made the network analysis as complete as possible; Drs Qiang Tu and Özlem Yüce for help with the WMISH. We acknowledge Dr Cathy Yuh, who generously provided the S. purpuratus clone; Dr Albert Erives, who designed the foxa MASOs and screened the S. purpuratus BAC library; Dr Jenifer Croce, who provided probes and offered many suggestions on Notch signaling; and Dr Jim A. Coffman for the cDNA sequence of foxa. This research was supported by NIH grants HD37105, HD14483, GM61464, and by the Office of Science (BER), the US Department of Energy, grant DE-FG02-03ER63584. P.O. was supported by the Camilla Chandler Frost Fellowship; K.D.W. by the US Department of Defense.