Notch and Nodal control forkhead factor expression in the specification of multipotent progenitors in sea urchin

The ability of certain cells to resist differentiation is an important feature of animal development. Most notably, sexually reproducing animals establish germ line stem cells that create the sperm and eggs required for fertilization (Spradling et al., 2011). In animals undergoing indirect development, embryogenesis results in a larval form that may differ dramatically from the adult organism (Minelli, 2009). This requires that some cells are either segregated away during embryogenesis and remain multipotential, or that they reacquire multipotency, in order to contribute to the diverse tissues of the adult, including the germ line (Arenas-Mena, 2010).

In sea urchins, embryogenesis gives rise to a feeding larva, which after a prolonged period of time will produce the adult. A prerequisite for rudiment formation is the specification of coelomic mesoderm, a group of cells derived from the veg2 lineage that gives rise to two pouch-like protrusions at the tip of the archenteron (Cameron et al., 1991). Rudiment formation commences with the invagination of a piece of ectodermal wall above the left coelomic pouch. A complex morphogenetic process ensues, during which the adult body plan with its pentameral symmetry is established (Pearse and Cameron, 1991). In Strongylocentrotus purpuratus, each of the coelomic pouches initially contains only about ten cells, but their descendants are major contributors to various adult structures (Pearse and Cameron, 1991); at metamorphosis, ∼90% of the 150,000 larval cells are located in the juvenile (Cameron et al., 1989; Song and Wessel, 2012), illustrating that coelomic pouch cells are multipotent progenitors.

In euechinoid sea urchins, coelomic pouch cell (CPC) progenitors are specified during gastrulation as a consequence of Delta/Notch (D/N) signaling within the non-skeletogenic mesoderm (NSM) (Sherwood and McClay, 1999; Sweet et al., 2002). Interference with NSM Delta abolishes coelomic pouch formation but does not disrupt the specification of other mesodermal cell types specified earlier (Materna and Davidson, 2012; Materna et al., 2013; Sweet et al., 2002). During archenteron formation, the CPC progenitors are located at the tip but then split into two groups that relocate laterally to either side of the foregut. Although little is known about the regulation of this morphogenetic process, the expression of Hedgehog (Hh) in the adjoining endoderm and of its receptors in CPC progenitors (Walton et al., 2009) suggests the possible involvement of this signal. The small micromeres (SMMs) are also located at the tip of the archenteron and are distributed to either coelomic pouch at the pluteus stage (Pehrson and Cohen, 1986). They are set aside and remain mitotically quiescent during embryogenesis (Tanaka and Dan, 1990). The SMMs express orthologs of numerous germ line/stem cell markers (Juliano et al., 2010; Voronina et al., 2008) and are required to form the germ line in Lytechinus variegatus (Yajima and Wessel, 2011), and thus might be primordial germ cells. However, whether they also contribute to other tissues remains unresolved.

The forkhead factor foxY, originally named foxC (Ransick et al., 2002; Tu et al., 2006), is the first transcription factor to be specifically expressed in any of the cells that contribute to the coelomic pouches. At early blastula stage, foxY is expressed in the SMMs in response to D/N signaling originating in the skeletogenic lineage (Materna and Davidson, 2012), then at the tip of the archenteron and in both coelomic pouches as they form. At late gastrula stage, foxY becomes strongly biased to the left coelomic pouch, presumably as a consequence of Nodal signaling establishing left/right (L/R) asymmetry across the embryo (Duboc et al., 2005). Although many transcription factors are eventually expressed in the coelomic pouches (Duboc et al., 2005; Howard-Ashby et al., 2006a; Howard-Ashby et al., 2006b; Luo and Su, 2012; Poustka et al., 2007; Tu et al., 2006), when CPC progenitors are first specified foxY is the only gene to become expressed in direct response to NSM Delta among 182 known regulatory genes, most of them with spatially restricted expression (Materna and Davidson, 2012; Ransick et al., 2002; Tu et al., 2006). This suggests that foxY plays an important role in coelomic pouch formation.

Here we examine the regulation and significance of foxY expression and find that it is essential for the specification of CPCs. Its precise expression pattern is the result of the integration of the D/N and Nodal signals within the foxY regulatory region: D/N signaling directly activates foxY expression, first in SMMs and then in CPC progenitors, while Nodal represses foxY in the right coelomic pouch via Pitx2. Reception of the Hh signal in CPC progenitors sets in motion coelomic pouch morphogenesis, but does not directly affect foxY expression. However, Hh is required for proper Nodal expression on the right side of the embryo and is thus involved in the establishment of L/R asymmetry. Our results link foxY directly to the specification of a cell type with great developmental potency en route to formation of the adult.

FoxY morpholino antisense oligonucleotide (MASO) (5′-CATGGCTC - CAAGTGCAGAACACTAC-3′) was obtained from Gene Tools (Pilomath, OR, USA) and injected at 300 μM in 0.12 M KCl. A random MASO (N₂₅) was injected as a control; injection volumes were 5 pl. DAPT and SB431542 were applied as reported previously (Materna and Davidson, 2012; Materna et al., 2013). Cyclopamine was dissolved in ethanol and added to a final concentration of 0.5 μM. A corresponding volume of solvent was added to controls.

Sea urchin (Strongylocentrotus purpuratus) embryos were cultured at 15°C. RNA was extracted using the RNeasy Micro Kit (Qiagen, Hilden, Germany) according to the manufacturer’s instructions. Expression levels were quantified with the NanoString nCounter (NanoString, Seattle, WA, USA) using a probe set for 205 genes. Samples were processed according to the manufacturer’s instructions and data processed as described (Materna and Davidson, 2012). Data were supplemented by quantitative PCR (QPCR) for nanos and fold changes calculated using poly-ubiquitin and hmg1 as references (Materna and Oliveri, 2008). A threshold of 2-fold difference was chosen as a significant change (Materna and Oliveri, 2008). QPCR primer sequences are included in supplementary material Table S1. Code set information and all perturbation data are available at the Dryad Digital Repository (http://dx.doi.org/10.5061/dryad.np321).

WMISH was conducted following standard procedures (Materna and Davidson, 2012). In brief, templates were amplified by PCR (for primers, see supplementary material Table S1) and used for transcription of digoxigenin (DIG)-labeled probes. Embryos were fixed in 2.5% glutaraldehyde, 32.5% sea water, 32.5 mM MOPS (pH 7) and 162.5 mM NaCl on ice overnight. Embryos were proteinase K treated for 5 minutes at room temperature [25 ng/μl in TBST (0.1 M Tris pH 7.5, 0.15 M NaCl, 1% Tween 20)] followed by 30-minute fixation in 4% paraformaldehyde, 32.5% sea water, 32.5 mM MOPS (pH 7), 162.5 mM NaCl at room temperature. Probes were hybridized overnight and detected using anti-DIG Fab fragments conjugated to alkaline phosphatase (Roche). Color was developed using NBT/BCIP reagents (Roche); fluorescent in situ hybridizations were developed using TSA amplification (PerkinElmer).

In situ stained embryos were colabeled by immunofluorescence for Vasa protein. Following a 30-minute blocking step (TBST, 10% sheep serum), embryos were incubated overnight with Vasa antibody (Voronina et al., 2008) diluted 1:250 in TBST with serum. Embryos were washed three times for 15 minutes each with TBST without serum, and incubated for 2 hours with Cy3 goat anti-rabbit secondary antibody (Invitrogen) diluted 1:250 in TBST with serum. Embryos were then washed three times in TBST buffer and imaged.

F-actin was visualized by Rhodamine phalloidin staining (Strickland et al., 2004). In brief, embryos were fixed for 1 hour in Millonig’s phosphate-buffered fixative followed by two washes in 50 mM HEPES, 50 mM PIPES, 0.6 M mannitol, 3 mM MgCl₂, 0.1% Triton X-100, pH 7.0. Embryos were transferred to TBST and stained with Rhodamine phalloidin in TBST (2 U/ml) for 1 hour. Embryos were washed three times and then imaged.

About 1500 molecules of reporter construct in ∼4 pl were co-injected with a 6-fold molar excess of carrier DNA in 0.12 M KCl. BAC reporters were constructed as reported previously (Smith, 2008) and injected with 400 copies of BAC per 4 pl without carrier. Spatial activity was assessed by fluorescence microscopy. All experiments were repeated in different batches of eggs. Cis-regulatory regions were identified as described (Smith, 2008). For target site disruptions, C bases were mutated to T, and A bases were mutated to G, and vice versa, for invariant bases in consensus sites. 5′- (C/T)(A/G)TG(A/G)GA(A/G/T)-3′ was used to identify candidate Su(H) sites (Ransick and Davidson, 2006). The TG at positions 3/4 and the GA at positions 6/7 were mutated to CA and AG, respectively. 5′-GGATTA-3′ was used to identify candidate Pitx2 target sites, and 5′-GATT-3′ was mutated to 5′-AGCC-3′.

At early blastula stage, foxY is expressed in the SMMs at the vegetal pole of the embryo. During gastrulation it is expressed at the tip of the archenteron, but eventually becomes restricted to the left coelomic pouch (Ransick et al., 2002). As has been suggested by a recent study (Song and Wessel, 2012), the foxY expression domain might not be restricted to the SMMs during gastrulation, but instead may expand to include the NSM. We set out to obtain a better understanding of foxY expression.

A time series of foxY gene expression (Materna et al., 2010) revealed a biphasic profile through late gastrulation (Fig. 1A). The early expression phase lasts from ∼9 hours postfertilization (hpf) until 23 hpf, i.e. from when the SMMs are born to the mesenchyme blastula stage. After a quick initial rise, foxY levels remain constant at ∼140 transcripts per embryo (∼35 copies in each SMM). Staining of foxY message overlaps with that of Vasa protein, an RNA helicase involved in translational control in primordial germ cells (Fig. 1B,C) (Gustafson and Wessel, 2010). Vasa is specific to SMMs (Gustafson and Wessel, 2010; Voronina et al., 2008), confirming that foxY expression is restricted to this cell type. The four SMMs are located at the vegetal pole and are in direct contact with the cells of the skeletogenic lineage. When the skeletogenic cells ingress into the blastocoel, the SMMs stay behind and come into contact with the NSM cells, which replace the skeletogenic cells in the vegetal plate (Fig. 1B,C).

During its second expression phase, foxY transcript levels increase significantly and reach a peak of ∼500 transcripts/embryo at 30 hpf (Fig. 1A), followed by a steady decline. The rise of foxY transcript levels coincides with an expansion of the foxY expression domain (Fig. 1D-F); at 30 hpf foxY is expressed in the mesodermal cells, i.e. the veg2-derived CPC progenitors that directly surround the Vasa-positive SMMs (Fig. 1D). foxY is expressed throughout the cap of the elongated archenteron, whereas the SMMs remain located at the center (they have divided once; there are now eight Vasa-positive cells). Towards the end of gastrulation the cap bulges laterally to form the coelomic pouches, and foxY expression becomes strongly biased to the left coelomic pouch (Fig. 1E). SMMs are still located at the center of the archenteron tip (Fig. 1F), but staining for foxY message in the SMMs becomes progressively weaker and is barely visible at 48 hpf (Fig. 1F). This might indicate that the expression of foxY is not maintained in these cells, in agreement with the diminished number of foxY-positive cells (six to eight) at the early larval stage observed by Ransick et al. (Ransick et al., 2002). After ∼3 days postfertilization (dpf), foxY is restricted to the left pouch (Ransick et al., 2002).

D/N signaling is active on three occasions during sea urchin embryogenesis: Delta is expressed in skeletogenic cells starting at ∼9 hpf, then in NSM once skeletogenic cells have ingressed; it is also expressed in the apical plate, although its function here has not been studied in detail (Materna and Davidson, 2012; Smith and Davidson, 2008; Sweet et al., 2002; Walton et al., 2006). Delta expression in skeletogenic cells activates gene expression in the NSM and SMMs, which are both in direct contact with the Delta source. Importantly, foxY is the only known target that is specifically activated in the SMMs (Materna and Davidson, 2012). Disruption of the second, i.e. NSM, phase of D/N signaling with the Notch inhibitor DAPT prevents formation of the coelomic pouches. Among 182 known regulatory genes, the only one to be severely depleted at 30 hpf was foxY (Materna and Davidson, 2012) (supplementary material Fig. S1). As we have shown above, the strong increase in transcript levels between 23 and 30 hpf is due to activation of foxY in the mesodermal cells directly adjacent to the SMMs constituting the CPC progenitor population. The likely cause for this expansion is activation of foxY by D/N signaling originating in the NSM.

The timecourse data in Fig. 1A reveal that other genes restricted to CPCs (foxF, pitx2, soxE) are transcribed only at 30 hpf or thereafter. To better understand the role of D/N signaling in the specification of CPCs we extended the analysis of NSM Delta function into gastrulation. We added DAPT at 17 hpf, just prior to the activation of Delta in NSM, and quantified the perturbation effects on transcript levels with the NanoString nCounter (Geiss et al., 2008). We made use of a code set containing probes for all relevant transcription factors known to be expressed in a spatially restricted fashion, plus probes for several signaling ligands and marker genes. Probe sequences and accession numbers for all 205 genes included in our code set are available from the Dryad Digital Repository (see Materials and methods).

Fig. 2A displays the quantitative perturbation data collected for a selection of genes. A table with perturbation data for all genes is available from the Dryad Digital Repository (see Materials and methods). As expected from the failure to form coelomic pouches, all genes specific to, and uniquely expressed in, CPCs were strongly depleted in perturbed embryos compared with controls. At mid-gastrula (38 hpf) this included the forkhead factors foxC, foxF and foxY (Tu et al., 2006), the Sox/HMG factor soxE (Duboc et al., 2005; Howard-Ashby et al., 2006a) and the zinc-finger gene scratchX (Materna et al., 2006). gataE and pax6 are expressed in CPCs, but are also expressed in the gut and ectoderm (Fig. 3E,F; supplementary material Fig. S2B-C″). Independent expression in these compartments obscures the perturbation effects in our quantitative analysis. The spatial expression pattern of scratchX was unknown, but it is first expressed in CPC progenitors at the tip of the archenteron and later in both coelomic pouches (Fig. 3G; Fig. 5E). These perturbation effects remain strong until late gastrulation (48 hpf) with the exception of foxY, which is no longer significantly affected, nor highly expressed at this time (Fig. 1A). The homeobox gene pitx2 and the bHLH gene myor2, which have become activated by 48 hpf, are also depleted. pitx2 is expressed in CPCs, but strongly biased to the right (Duboc et al., 2005; Luo and Su, 2012). The expression level of myor2 is extremely low and we were unable to determine its spatial expression pattern. Endoderm genes are not activated by NSM Delta throughout gastrulation (Fig. 2A) (Materna and Davidson, 2012), despite strong expression of the Notch receptor in the archenteron (Walton et al., 2006).

To better understand the timing of the D/N signaling input into CPCs, we attempted to determine the spatial expression of delta during gastrulation. Similar to results reported by Walton et al. (Walton et al., 2006), we observed delta expression in dispersed cells in the ectoderm and in the apical plate (not shown), where it is an ongoing input into neurogenic genes (Fig. 2A). But owing to its low abundance (150 transcripts/embryo at 48 hpf) (Materna et al., 2010) and complex expression pattern we cannot rule out the possibility that low-level expression persists in single cells at the archenteron tip. To assess whether active Notch signaling is required for maintenance of CPC gene expression, we added DAPT at 30 hpf and quantified transcript levels. Our results show that some apical genes are strongly elevated at 38 and 48 hpf, indicating that D/N signaling is active in this domain. By contrast, no CPC gene is affected (Fig. 2B). These data suggest that the mesodermal Delta signal does not function as a continuous input into CPC genes during gastrulation. However, most CPC genes are activated after 30 hpf; in fact, foxY and foxC are the only known genes expressed in this compartment at 30 hpf (Fig. 1A). Thus, our findings imply that Delta activation of most CPC genes is indirect.

foxY is the first known transcription factor to be specifically expressed in CPC progenitors downstream of Notch signaling. To test whether it is a potential regulator of other CPC genes we knocked down foxY expression with a MASO. FoxY MASO injection has no adverse effect on the health of the embryo and does not interfere with the general progression of development. The overall anatomy of the larvae is normal, including the skeleton and partitioning of the gut (Fig. 3A′). Observation of the larvae revealed very robust defects. None of the treated embryos formed coelomic pouches or their derivative structures (Fig. 3A′), including the water canal that normally emerges from the left coelomic pouch and connects to the hydropore in the ectodermal wall. FoxY MASO-treated larvae were unable to swallow, although food particles accumulated in their esophagus. Contraction of circumesophageal muscles is responsible for transport of food into the stomach. Rhodamine phalloidin staining of F-actin in 3.5-day-old larvae revealed that these muscles do not form following FoxY MASO injection (Fig. 3B′). Muscle cells are thought to migrate out of the coelomic epithelium (Burke and Alvarez, 1988), although their exact origin remains unknown. Our results indicate they derive from the group of foxY-positive CPC progenitors at the tip of the archenteron. Thus, foxY is required for all known functions of NSM Delta signaling, i.e. the specification of CPCs and muscle cells (Materna and Davidson, 2012; Sweet et al., 2002).

At blastula stage (15 hpf), when foxY expression is restricted to SMMs, NanoString analysis revealed that no genes are affected by FoxY MASO injection. However, our code set does not contain any genes other than foxY that are restricted only to SMMs. A recent study examined the relationship between foxY and nanos (Song and Wessel, 2012), a stem cell/germ line marker specific to the SMMs until the pluteus stage (Juliano et al., 2010). We supplemented the NanoString data with QPCR data for nanos (not shown) and confirmed the previous finding that, at 15 hpf, loss of FoxY function does not affect nanos abundance (Song and Wessel, 2012).

If FoxY is the effector of NSM Delta signaling in the CPC progenitors, its knockdown should affect a similar set of genes as observed in our DAPT experiments (Fig. 2A). At 38 and 48 hpf, our quantitative analysis identified a distinct set of genes affected in FoxY morphants that is restricted to CPCs (Fig. 2C; supplementary material Fig. S3A). The changes were indeed similar to those caused by disruption of NSM Delta signaling (Fig. 2A). As in our DAPT treatments, the responsive gene set included all genes known to be specifically and uniquely expressed at the tip of the archenteron at 38 hpf, i.e. foxC, foxF, foxY, scratchX and soxE. At 48 hpf some differences became apparent: scratchX was more strongly affected by FoxY MASO than DAPT treatment. By contrast, pitx2 was significantly depleted in DAPT-treated embryos but unaffected in FoxY MASO morphants. Although foxY is initially expressed in all CPC progenitors, at 48 hpf it becomes strongly biased to CPCs on the left side as a result of pitx2 repression on the right (see below). The perturbation results thus reflect the progressive regionalization at the archenteron tip. This divergence notwithstanding, the striking agreement at 38 hpf argues that foxY is the immediate early effector of Notch signaling within the tip of the archenteron/CPC progenitor population.

We further examined FoxY-responsive genes by WMISH, including six1/2, its co-factor eya, and scratchX, which were undetectable in all FoxY MASO-treated embryos (Fig. 3C,G,H). The transcript levels of gataE and pax6 were only slightly affected in our NanoString analysis (Fig. 2C), but WMISH confirms that the expression of both genes is lost specifically in CPCs (Fig. 3E,F). As mentioned above, gataE is strongly expressed in the mid- and hindgut and this expression masks its loss in CPCs in our quantitative analysis (supplementary material Fig. S2B-B″). Similarly, pax6 is also expressed in the ectoderm (supplementary material Fig. S2C-C″); it becomes predominantly expressed in CPCs only relatively late on (Luo and Su, 2012; Yankura et al., 2010). Our NanoString data showed that foxY message is no longer significantly depleted at 48 hpf following FoxY MASO injection. WMISH confirmed that foxY is detectable at the tip of the archenteron. However, following FoxY MASO treatment the foxY-positive CPCs are dispersed and not restricted to the sides of the archenteron (Fig. 3D,D′). Despite this exception, these results confirm that essentially all of the genes known to be specifically expressed in the CPC compartment are severely depleted when FoxY function is lost.

Towards the end of gastrulation, foxY expression becomes biased to the left coelomic pouch where eventually the adult rudiment will form. Establishment of L/R asymmetry across the embryo is widely conserved and is dependent on Nodal/Tgfβ signaling (Duboc et al., 2005; Luo and Su, 2012). During embryogenesis, Nodal is first expressed in the oral ectoderm and induces oral cell fate in recipient cells (Duboc et al., 2004; Li et al., 2012; Materna et al., 2013). Later in gastrulation, Nodal expression ceases in the oral ectoderm and is instead activated in the right lateral ectoderm and right coelomic pouch. Interference with this second phase of Nodal signaling results in loss of L/R asymmetry, which is most pronounced in the coelomic pouches where genes restricted to the left side are ectopically expressed on the right, resulting in the formation of two rudiments (Duboc et al., 2005; Luo and Su, 2012).

To identify more comprehensively the gene set responsive to late Nodal signaling, we performed NanoString analysis of embryos treated with the Nodal inhibitor SB431542 at 24 hpf, after the oral/aboral axis is firmly established. Among all 205 genes in our analysis only six were significantly depleted (Fig. 4A; supplementary material Fig. S3B). This included the conserved regulators of L/R asymmetry (nodal, lefty, pitx2) that are expressed in the right ectoderm and coelomic pouch, complementary to foxY (Fig. 5B-D) (Duboc et al., 2005; Luo and Su, 2012). myor2 is depleted following Nodal perturbation, as are chordin and the homeobox gene not. However, the effect on chordin and not is due to Nodal signaling within the oral ectoderm, which is still ongoing by the time Nodal inhibitor is added, where these genes are expressed during gastrulation (Li et al., 2012; Materna et al., 2013). Although we cannot exclude the possibility of having missed perturbation effects on genes with more complex expression patterns, the primary candidate as the downstream effector of late Nodal signaling in the establishment of L/R asymmetry is pitx2.

Hh signaling is known to regulate developmental processes as diverse as patterning, proliferation and morphogenesis (Ingham and McMahon, 2001) and has been linked to the establishment of L/R asymmetry (Tsiairis and McMahon, 2009). In sea urchin embryos the Hh ligand is expressed throughout the endoderm and archenteron starting at the mesenchyme blastula stage (Fig. 1A), concurrent with expression of its downstream effector gli1 (Walton et al., 2006). The receptors ptc and smo are present throughout embryogenesis and are expressed prominently at the tip of the archenteron and in the coelomic pouches (Walton et al., 2006; Walton et al., 2009). Experimental activation of Hh signaling has been shown to randomize laterality and cause defects in muscle patterning (Walton et al., 2009), but the function of the endogenous Hh signal with regard to coelomic pouch development has not been studied systematically.

We disrupted Hh signaling by addition of cyclopamine, a highly specific inhibitor of Smo (Chen et al., 2002), to embryos at 22 hpf, prior to the onset of hh transcription (Fig. 1A). We determined that a final cyclopamine concentration of 0.5 μM is sufficient to cause reproducible effects (supplementary material Fig. S4) without causing severe developmental delay or death, which were common at 5 μM or higher. Phenotypic inspection revealed several defects, but the most relevant here is the disruption of coelomic pouch morphogenesis. In control embryos the CPC progenitors at the tip of the archenteron split into two groups that relocate laterally to form pouches on either side of the foregut. By contrast, in cyclopamine-treated embryos these cells remained atop the archenteron as a uniform group and formed one pouch-like structure (Fig. 5). Cyclopamine treatment also resulted in failure to form the mouth and in a short gut; whether this is due to improper elongation of the archenteron or defective gut patterning remains to be addressed.

NanoString analysis of the perturbation effects identified a number of genes reacting strongly to the loss of Hh signaling (Fig. 4B). Some of the strongest effects are found for endoderm genes, highlighting the importance of Hh signaling in gut development. By contrast, and despite the strikingly different morphology, the expression levels of CPC genes were virtually unaltered, except for some minor effects (e.g. foxF, soxE) (Fig. 4B). In situ staining of foxY transcript at 66 hpf, more than 30 hours after the pouches separate in controls, revealed strong expression at the tip of the archenteron in agreement with our quantitative analysis (Fig. 5A′,A″). At this time, foxY is restricted to the left coelomic pouch in controls (Fig. 5A). Staining of scratchX and soxE message was strong at the tip of the archenteron, confirming their continued expression following treatment (Fig. 5E′,E″,F′,F″).

One CPC gene strongly depleted in cyclopamine-treated embryos is pitx2 (Fig. 4B; Fig. 5D′,D″). nodal transcript levels were also strongly reduced following cyclopamine treatment, suggesting that the cause for the aberrant pitx2 expression is loss of Nodal signaling (Fig. 4B). Indeed, in situ staining of nodal and lefty confirmed that their expression, like that of pitx2, is lost in the coelomic pouches and also the right lateral ectoderm (Fig. 5B′,B″,C′,C″). In contrast to Nodal perturbation, the not gene is unaffected in cyclopamine treatments, supporting the conclusion that loss of Hh signaling specifically affects the late expression phase of nodal, which is discontinuous from its earlier expression in the oral ectoderm; it also confirms that the not gene is not expressed in coelomic pouches at late gastrulation [it is eventually expressed in the right coelomic pouch during larval development (Luo and Su, 2012)].

In summary, wild-type levels of Hh signaling are required for coelomic pouch morphogenesis, unilateral expression of nodal and, as a consequence, for asymmetric gene expression in the coelomic pouches once they have formed.

foxY expression is closely linked to that of Delta. To test whether foxY is a direct target of Suppressor of Hairless [Su(H)], the downstream effector of D/N signaling that interacts with the Notch intracellular domain (NICD) once it has relocated to the nucleus upon signal reception, we generated reporter constructs of the foxY regulatory region. We identified a foxY BAC of ∼145 kb and inserted the GFP coding region into the second exon by homologous recombination downstream of the start codon to create the foxY:GFP BAC reporter (Fig. 6A). This construct replicates endogenous foxY expression, albeit in a mosaic fashion as is common in sea urchins (Smith, 2008). GFP expression is first observed in one or two of the SMMs, then in the vegetal plate prior to gastrulation (Fig. 6B), then at the tip of the archenteron, and eventually primarily in the left coelomic pouch (Fig. 6D). Reporter expression in the right pouch occurs at low frequency (see below), as expected based on its weak expression observed in in situ stainings.

We amplified successively smaller fragments using the BAC as template and examined reporter expression (Fig. 6A). Construct #3 contains the sequence located between -778 bp upstream of the transcription start site (TSS) and intron 2; its expression was identical to that of the full BAC reporter. Deletion of intron 1 did not alter its expression (Construct #6; Fig. 6C,D). Further shortening at the 5′ end resulted first in equal expression between left and right pouches (Construct #4; Fig. 6E), and then in complete loss of activation (Construct #5). Thus, the region between -400 bp and the TSS contains the regulatory element necessary for expression, and the region between -778 bp and -400 bp contains an element required for the repression of expression in the right coelomic pouch.

We identified a potential Su(H) binding site in the region proximal to the foxY TSS (Fig. 7A; supplementary material Fig. S5). To test whether this site is required for activation of foxY we mutated it in Construct #3, which was then co-injected with an alx reporter that drives GFP expression in skeletogenic cells (Damle and Davidson, 2012) as a marker of integration. Injection of wild-type Construct #3 results in GFP expression within the vegetal plate (100%, 44/44) (Fig. 7C,C′). Expression of Construct #3 is clearly distinguishable from GFP-positive skeletogenic cells within the blastocoel. By contrast, when the Su(H) site was mutated [#3-ΔSu(H)], GFP expression was absent from the vegetal plate (95%, 57/60) (Fig. 7D-D″). Similarly, we observed no foxY reporter expression during archenteron formation (Fig. 7F,G). Thus, mutation of the Su(H) site within the foxY regulatory region confirms that foxY is a direct Notch target.

Our data suggest that Nodal might repress foxY in the right coelomic pouch via pitx2. Unfortunately, two Pitx2 MASOs had toxic side effects resulting in abortive development 24 hours before pitx2 is first transcribed. We therefore took a more direct approach and set out to identify potential Pitx2 binding sites. Pitx2 binding sites are similar to those of other homeobox factors, most notably Gsc and Otx, with a consensus motif of GGATTA (Berger et al., 2008). However, of these three genes, only pitx2 is expressed in the coelomic pouches at the appropriate time (Li et al., 2012; Morris et al., 2004); neither gsc nor otx is affected in our Nodal perturbations. We identified two potential Pitx2 binding sites in the distal foxY fragment (Fig. 7A; supplementary material Fig. S5). Mutation of these sites in Construct #3 (#3-ΔPitx2; Fig. 7A) resulted in equal reporter expression in the left and right coelomic pouches, similar to Construct #4 (Fig. 6A; Fig. 7B,B′). In controls, 76% of embryos (40/53) expressed GFP reporter exclusively in the left coelomic pouch, 9% (5/53) in both pouches, and 15% (8/53) exclusively in the right pouch at 3 dpf. We never observed endogenous foxY transcript exclusively in the right pouch, indicating that the fraction of embryos expressing GFP reporter on only one side is an overestimate due to mosaic incorporation. By contrast, mutation of both putative Pitx2 binding sites (#3-ΔPitx2) caused a significant shift to the right: 32% of embryos (22/68) showed expression of GFP exclusively in the left pouch, 25% (17/68) in the right pouch, and 43% (29/68) in both pouches. This resembles the bilateral expression of genes otherwise restricted to the left pouch (e.g. soxE, six1/2) as observed following Nodal perturbation (Duboc et al., 2005; Luo and Su, 2012), and thus provides evidence that pitx2 is the repressor of foxY in the right coelomic pouch.

The experiments presented in this work aimed to elucidate the earliest regulatory events required for adult formation in sea urchin larvae. Although much remains to be discovered, our results afford several new and important insights. First, we show that the primary function of NSM Delta expression is to directly activate foxY expression in a subset of mesodermal cells that constitute the pool of CPC and muscle cell progenitors. Second, our work demonstrates that foxY is the principal effector of the NSM Delta signal. It is essential for specification of the two late mesodermal derivatives and thus constitutes a central node in the regulatory network underlying this process. Third, we find that Hh signaling drives coelomic pouch morphogenesis and that Hh is essential for establishing L/R asymmetry by activating Nodal expression on the right side of the larva. As a consequence, foxY is repressed in the right coelomic pouch by Pitx2. Our results show that the initial specification of CPCs, the morphogenesis of the pouches, and the establishment of laterality are three intertwined processes; however, despite their concurrent unfolding they are regulated independently and are thus separable.

The exact origin of late mesodermal derivatives has remained elusive. Fate mapping experiments resolved the origin of pigment and blastocoelar cells within the vegetal plate, but the results were less clear for CPCs and muscle cells (Ruffins and Ettensohn, 1996). The demonstration that NSM Delta is required for specification of both CPCs and muscle cells (Sherwood and McClay, 1999; Sweet et al., 2002) led to the identification of foxY as a potential downstream effector (Materna and Davidson, 2012). As our expression analysis shows, foxY is likely to be the first transcription factor specifically expressed in the entire veg2-derived progenitor pool of CPCs and muscle cells. At early mesenchyme blastula stage, delta transcript is present throughout the NSM (Materna and Davidson, 2012; Sweet et al., 2002; Walton et al., 2006), including the cells directly adjacent to the SMMs that go on to activate foxY expression. This contradicts the idea that Delta presentation and active Notch signaling are mutually exclusive (del Álamo et al., 2011) and suggests that Delta expression might be more dynamic within the NSM than currently assumed. The exceedingly low transcript levels of delta during its expression in NSM (Materna et al., 2010) and the rearrangements following ingression of pigment and blastocoelar cells make disentangling the different mesodermal lineages challenging.

The dramatic phenotype caused by FoxY MASO injection, i.e. the complete lack of coelomic pouches as well as muscle cells, demonstrates the importance of FoxY. Our observation that expression of essentially all CPC genes is abolished identifies foxY as the linchpin in the regulatory network underlying the specification of CPCs. Exceptions are few: pitx2 is affected only in D/N perturbations but not following injection of FoxY MASO; but this is not surprising given the likely role of Pitx2 as the repressor of foxY in the right coelomic pouch. Nevertheless, the fact that pitx2 indirectly requires D/N signaling but not FoxY indicates that the functions of NSM Delta signaling and FoxY, although largely overlapping, are not perfectly congruent. The transient expression of Delta in the NSM requires that foxY receives additional inputs to assure its continued expression. foxY is likely to be part of a positive-feedback loop involving genes that it first activates, as is common for pioneering transcription factors establishing a new regulatory state (Davidson, 2010). Such recursive wiring might be the cause of the mostly normal, albeit low, transcript levels of foxY following its own perturbation at late gastrulation, i.e. only after a long delay. Additionally, foxY might also be activated by Bmp/Tgfβ signaling, as has been suggested for other CPC genes (Luo and Su, 2012). Nevertheless, restriction of foxY expression to the left side, where it remains expressed (Ransick et al., 2002), is one of the first events in the progressive regionalization at the tip of the archenteron and presages where the adult rudiment will form. Although muscle cells are derived from the same group of foxY-positive progenitor cells, they give rise to between 20 and 40 fully differentiated cells in grown larvae (Burke and Alvarez, 1988; Cox et al., 1986). By contrast, the CPCs are a major contributor to the adult; at metamorphosis, their descendants number in the tens of thousands.

The separation of CPC progenitors and their relocation to the sides of the foregut is dependent on Hh signaling. This morphogenetic process resembles the separation of midline structures in vertebrates, which is strongly dependent on Hh signaling (Schachter and Krauss, 2008). Improper regulation of Hh signaling leads to severe defects such as cyclopia and incomplete separation of brain structures (Roessler et al., 1996). In sea urchins, loss of Smo activation results in the formation of only one centrally located coelomic pouch. However, none of the transcription factors expressed specifically at the tip of the archenteron is affected by loss of Hh signaling. The only exception in our dataset is pitx2 due to interference with late Nodal expression. Thus, the CPCs now positioned atop the foregut display the regulatory state that is normally reserved for the left coelomic pouch only. Additional Hh-responsive transcription factors might await discovery but Hh signaling may also directly regulate morphogenetic effector genes, such as those encoding cell motility or adhesion molecules. When misregulated, these genes cause similar craniofacial defects as loss of Hh signaling (Chen et al., 2006).

Loss of Hh signaling causes failure to activate nodal transcription in the right coelomic pouch and the right lateral ectoderm. The importance of Nodal signaling in L/R asymmetry has been studied extensively (Bessodes et al., 2012; Duboc et al., 2005; Luo and Su, 2012; Nakamura and Hamada, 2012) and a close link between Hh and Nodal signaling is well established (Zhang et al., 2001). Nevertheless, how the two signaling systems relate to each other mechanistically remains unknown. Laterality defects in mice caused by loss of Hh signaling can be rescued by expression of a constitutively active Smo receptor in lateral plate mesoderm resulting in unilateral expression of nodal (Tsiairis and McMahon, 2009). In sea urchins, overexpression of a constitutively active Smo receptor randomizes the position of the adult rudiment (Walton et al., 2009). It has recently been shown that unilateral expression of Nodal at the archenteron tip, and thus in close proximity to the Hh source, determines the sidedness of Nodal expression in the ectoderm (Bessodes et al., 2012). However, in contrast to our Hh perturbations, disruption of Nodal signaling specifically within the archenteron leads to randomization of nodal transcription in the ectoderm and not to loss of expression; whether Hh itself is received in the ectoderm is currently unknown. Regardless, although the initial cause of symmetry breaking remains elusive, it is well recognized that Nodal signaling is part of the conserved downstream mechanism establishing laterality across the embryo (Duboc et al., 2005; Grande and Patel, 2009; Nakamura and Hamada, 2012). Our experiments demonstrate that in sea urchins this function includes repression of foxY in the right coelomic pouch by Pitx2, thus confining foxY expression, and adult rudiment formation, to the left side.

Coelomic pouches are a conserved feature of echinoderm embryos (Chia and Walker, 1991; Pearse and Cameron, 1991), but whether the molecular mechanisms controlling their formation are conserved is currently unknown. Nevertheless, in the distantly related starfish Patiria miniata, coelomic pouch formation is dependent on D/N signaling within the mesoderm just as it is in euechinoids (V. Hinman, personal communication). Whether starfish and other echinoderms possess a foxY ortholog that is employed downstream of D/N signaling is an open question.

A second conserved feature is the localized expression of Vasa protein. Although initially localized to SMMs, Vasa eventually becomes expressed throughout the coelomic pouches (Juliano et al., 2010). Despite its connection to germ cell specification, recent evidence indicates that Vasa is more broadly connected to pluripotency (Gustafson and Wessel, 2010) and is expressed in mammalian embryonic stem cells (Ingham and McMahon, 2001; Ramalho-Santos et al., 2002) and in adult stem cells in planarians and Hydra (Rebscher et al., 2008). In polychaete annelids, Vasa is expressed throughout the mesodermal posterior growth zone, a group of multipotent stem cells that, similar to the coelomic pouches in echinoderms, gives rise to adult structures and germ cells (Rebscher et al., 2007).

Coelomic pouch development thus provides a unique opportunity to understand how developmental potency is maintained throughout successive patterning events in cells that are embedded in an environment that otherwise promotes rapid differentiation.

We are grateful for support and advice from Eric Davidson in whose laboratory most of the work was performed. Many thanks to Celina Juliano and Andy Cameron for numerous insights and to Gary Wessel for Vasa antibody.