Fez function is required to maintain the size of the animal plate in the sea urchin embryo

Nervous system development in bilaterians is composed of a number of developmental processes, such as partitioning of the ectoderm into neuroectoderm and non-neuroectoderm, patterning of the neuroectoderm, and differentiation of neural cells including neural network formation. Among these, partitioning of ectoderm is one of the earliest and the most crucial events because all of the following processes of neurogenesis depend on it. In vertebrates, the prospective neuroectoderm is formed at the future dorsal side by the function of the Spemann organizer, which can induce an ectopic neuroectoderm when grafted to the ventral region of the embryo (Spemann and Mangold, 1924). From a number of previous studies analyzing the inductive function of the organizer, we now know that the molecular entities mediating this induction are diffusible extracellular molecules, such as Chordin, Noggin and Follistatin (De Robertis and Kuroda, 2004). These molecules antagonize the function of bone morphogenetic proteins (BMPs), members of TGF-β superfamily, on the dorsal side thereby protecting this region from BMP-mediated induction of non-neuroectoderm (Khokha et al., 2005). When one or two of them are knocked down, the neuroectoderm is normally partitioned and the morphants have normal neural cell differentiation. However, when the expression of all three is knocked down, the embryo contains no neuroectoderm because of enhanced and expanded BMP signaling. Conversely, when the translation of three BMP members (BMP2, BMP4 and BMP7) is blocked, the embryo has expanded neuroectoderm (Reversade et al., 2005). Therefore, it is widely accepted that protection against BMP activity is the mechanism that specifies the dorsal region as neuroectoderm during early development in vertebrates (De Robertis and Kuroda, 2004). However, BMPs are diffusible extracellular molecules that form an activity gradient that is reinforced by positive feedback (Shen et al., 2007), such that a small initial change can have a large influence on the effective signaling range (Gardner et al., 2000). Therefore, it has been unclear how BMP activity can be precisely controlled to partition ectoderm fates at a single-cell level. Because the size of the neuroectoderm is identical in individual embryos of the same species at the same developmental stage, there must be additional intrinsic mechanism(s) that downregulate BMP signaling and establish precise domains of neural and non-neural ectoderm.

The neurogenic ectoderm of the embryo of the sea urchin, a basal deuterostome, consists of two major parts. One is the ectoderm at the animal pole, which is highly related to the vertebrate anterior neuroectoderm, judged from the temporal and spatial expression patterns of genes encoding regulatory factors, and from the fact that canonical Wnt signaling is involved in positioning the region anteriorly during cell-fate specification during early embryogenesis (Wikramanayake et al., 1998; Kiecker and Niehrs, 2001; Wei et al., 2009). The second forms within the ciliary band located at the border between oral and aboral signaling territories (Duboc et al., 2004; Yaguchi et al., 2006; Wei et al., 2009; Lapraz et al., 2009; Yaguchi et al., 2010a; Saudemont et al., 2010). The anterior neurogenic region, called the animal plate, is the focus of the present study. FoxQ2 is among several genes that are the earliest to be expressed specifically in neurogenic ectoderm. FoxQ2 transcripts appear in the entire animal half at the 32-cell stage and are progressively restricted to the animal pole region in pre-hatching blastulae (Tu et al., 2006; Yaguchi et al., 2008). This process restricts at least several genes required for animal plate development and depends on the Wnt/β-catenin pathway, which functions in the vegetal hemisphere (Logan et al., 1999; Yaguchi et al., 2008). After the restriction of the FoxQ2-positive region is completed at the blastula stage, the prospective neuroectoderm starts to express additional genes that are specific for nervous system development (Wei et al., 2009). However, at around this stage, nodal and bmp2/4, which are common anti-neural factors during early development, are transcribed on the oral side and then BMP2/4 diffuses to the aboral ectoderm. Both Nodal and BMP2/4 induce non-neurogenic ectoderm fates (Angerer et al., 2000; Duboc et al., 2004; Bradham et al., 2009; Lapraz et al., 2009; Saudemont et al., 2010). The prospective animal plate, as well as ciliary band, neuroectoderm must have robust mechanisms that antagonize these signals and maintain their precise sizes during these stages. Here, we report that, in the animal plate, FoxQ2 induces production of an intracellular transcription factor, Fez (forebrain embryonic zinc finger), at the blastula stage that specifically attenuates BMP signaling, thereby maintaining and stabilizing the border of the anterior neuroectoderm in the sea urchin embryo.

Adult sea urchins, Strongylocentrotus purpuratus, were obtained from The Cultured Abalone (Goleta, CA, USA) and the embryos were used only for the microarray experiment. For all other experiments, embryos of Hemicentrotus pulcherrimus collected around the Shimoda Marine Research Center (University of Tsukuba, Shizuoka, Japan) and around the Marine and Coastal Research Center (Ochanomizu University, Chiba, Japan) were used. The gametes were collected by intrablastocoelar injection of 0.5 M KCl and the embryos were cultured with filtered natural seawater (FSW) at 15°C. For some experiments, γ-secretase inhibitor (DAPT; Sigma-Aldrich, St Louis, MO, USA) was added at the hatching blastula stage to a final concentration of 20 γM.

Sample preparation, microarray and data processing were described previously (Wei et al., 2006). The double-strand cDNAs synthesized from control embryos and SpFoxQ2 morphants were labeled, hybridized and scanned by Roche Nimblegen microarray services. The microarray was prepared based on S. purpuratus genome sequence information (Wei et al., 2006; Sea Urchin Genome Sequencing Consortium, 2006) and hybridized with amplified cDNAs isolated from S. purpuratus mesenchyme blastulae either containing or lacking the SpFoxQ2-MO1 previously described (Yaguchi et al., 2008) (see below). FoxQ2-dependent genes identified in the microarray were confirmed by in situ hybridization in H. pulcherrimus embryos, as described below and by quantitative polymerase chain reaction (QPCR) as previously described (Yaguchi et al., 2010b).

Microinjection was performed as described previously (Yaguchi et al., 2010b). We used the following morpholinos (Gene Tools, Philomath, OR, USA) at the indicated concentrations in 24% glycerol in injection needles: SpFoxQ2-MO1 (800 γM) (Yaguchi et al., 2008), FoxQ2-MO (200 γM) (Yaguchi et al., 2010b), Fez-MO1 (1.9 mM), Fez-MO2 (1.6 mM), BMP2/4-MO (400 γM) and Smad1/5/8-MO (1.0 mM). All morpholinos were designed against H. pulcherrimus genes except for the SpFoxQ2-MO1 that blocks S. purpuratus FoxQ2 translation. The morphant phenotypes were the same as those previously published (Duboc et al., 2004; Yaguchi et al., 2008). The morpholino sequences were the following: Fez-MO1: 5′-GAATGCTTTTTTCATGCACTAAAGA-3′; Fez-MO2: 5′-GCGTTCAAATCTACTTAAGGAGTGT-3′; BMP2/4-MO: 5′-GACCCCAATGTGAGGTGGTAACCAT-3′; and Smad1/5/8-MO: 5′-GCCATCGACATAGTGGAGCAAGCTT-3′.

mRNAs were synthesized from linearized plasmids using the mMessage mMachine kit (Life Technologies, Carlsbad, CA, USA) and injected at the indicated concentrations in 24% glycerol in injection needles: Δ-cadherin (0.3-0.6 γg/γl) (Logan et al., 1999), Fez-mRNA (2.0 γg/γl), BMP2/4-mRNA (2.5 γg/γl) and actSmad2/3-mRNA (2.5 γg/γl) (Yaguchi et al., 2007).

Whole-mount in situ hybridization was performed as described previously (Minokawa et al., 2004; Yaguchi et al., 2010b). The foxQ2-positive cell number was determined by counting DAPI signals in the foxQ2 region through serial optical sections. Immunohistochemistry for detecting serotonin, synaptotagminB (synB) and Nk2.1 was performed as described previously (Yaguchi et al., 2006). To detect FoxQ2 and pSmad1/5/8 immunochemically, embryos were fixed with 4% paraformaldehyde in FSW for 10 minutes. After washing with phosphate-buffered saline containing 0.5% Tween-20 (PBS-T) five times, the embryos were blocked with 1% skimmed milk (Difco; BD, Franklin Lakes, NJ, USA) in PBS-T and incubated with primary antibodies diluted as follows with Can Get Signal Immunostain Solution B (TOYOBO, Tokyo, Japan): mouse anti-FoxQ2 IgG at 1:100, rabbit anti-pSmad1/5/8 (9511; Cell Signaling Technology, Danvers, MA, USA) at 1:1000. The primary antibodies were detected with secondary antibodies conjugated with Alexa-568 for FoxQ2 and Alexa-488 for pSmad1/5/8 (Life Technologies). The specimens were observed with a Zeiss Axio Imager.Z1 equipped with Apotome system, and optical sections were stacked and analyzed using ImageJ and Adobe Photoshop. Panels and drawings for figures were made with CANVAS software.

cDNA encoding full-length FoxQ2 was cloned into the pET vector (Novagen, Darmstadt, Germany) and histidine-tagged FoxQ2 was induced in bacteria with IPTG. The bacteria were lysed with 6 M urea, and the fusion protein was purified by Ni-column chromatography and used to immunize three mice. Antisera were screened by whole-mount immunohistochemistry, and IgG was purified with Melon gel (Thermo Fisher Scientific, Waltham, MA, USA).

FoxQ2 is expressed very early in the prospective animal plate of sea urchin embryos and is required for the development of neurons in this region of the embryonic nervous system. We searched for regulatory genes downstream of FoxQ2 that are required for formation of the animal plate using a microarray approach to identify genes strongly downregulated in FoxQ2 morphants at the blastula stage of Strongylocentrotus purpuratus. Previous studies have shown this approach to be successful in identifying genes whose expression depends on upstream regulatory proteins (Wei et al., 2009; Yaguchi et al., 2010b). One candidate FoxQ2-dependent regulatory gene encoded an apparent ortholog of Fez, forebrain embryonic zinc finger, (SPU_027491; Sea Urchin Genome Sequencing Consortium, 2006). We recovered this gene from the genome of the Japanese sea urchin, Hemicentrotus pulcherrimus, and found that it contains sequences encoding an engrailed homology repressor motif 1 (EH1) at the N-terminal region and six zinc finger domains at the C-terminus (accession number: AB610478), as previously reported for vertebrate Fez or Fez-like proteins (e.g. mice; AK014242 and AB042399) (Hirata et al., 2006b). Because of the presence of EH1, Fez is thought to function as a transcriptional repressor (Hirata et al., 2006a). After confirming the microarray result that, during blastula stages, H. pulcherrimus FoxQ2 morphants have significantly decreased fez mRNA levels by in situ hybridization and QPCR (Fig. 1M-O), we focused on the function of Fez in neurogenesis.

To examine the gene expression pattern of fez, we performed in situ hybridization at developmental stages from egg to pluteus larva. No signals are detected before hatching (Fig. 1A), consistent with the temporal profiling data conducted by Wei et al. (Wei et al., 2006) using microarrays. After hatching, fez transcripts accumulate throughout the entire animal plate (Fig, 1B) until the early gastrula stage (Fig. 1C). After the embryo reaches the mid-gastrula stage, the uniform expression fades and is progressively replaced by stronger signals in a few individual cells within the animal plate (Fig. 1D-F, asterisks). To compare fez and foxQ2 expression patterns in detail, we employed fluorescent in situ hybridization. fez is expressed at low levels in the animal plate at the hatching blastula stage (Fig. 1G) but is expressed at much higher levels at the mesenchyme blastula stage and in a pattern that is almost identical to that of foxQ2 (Fig. 1H). As shown using chromogenic detection (Fig. 1D), fez transcripts accumulate in individual cells at the gastrula stage, whereas foxQ2 expression remains uniform throughout the animal plate (Fig. 1I,J). The expression domain of foxQ2 is restricted to the oral side of the animal plate at later stages (Yaguchi et al., 2008) whereas three to five individual fez-positive cells are located on the opposite (aboral) side at that time (Fig. 1K; insets in Fig. 1I′,J′,K′). To examine which cell type(s) express fez mRNA, we compared the distributions of tryptophan 5-hydroxylase (tph) and fez mRNAs by double fluorescent in situ hybridization. Because TPH is the rate-limiting enzyme in serotonin synthesis, it is a specific marker for serotonergic neurons in sea urchin embryos (Yaguchi and Katow, 2003). At the prism stage, fez is clearly expressed in the serotonergic neurons, although each mRNA localizes to a slightly different position in the same cell (Fig. 1L). Because the expressions of fez and foxQ2 do not overlap at later stages, fez expression does not depend on FoxQ2 at that time (Fig. 1O). In summary, fez is expressed throughout the animal plate between the hatching blastula and gastrula stages, and in serotonergic neural precursors and neurons after gastrulation.

To examine Fez function in sea urchin development and neurogenesis, we performed knockdown and mis-expression experiments by injecting morpholino anti-sense oligonucleotides (MOs) and mRNA encoding full-length Fez protein, respectively. In Fez morphants, the timing of cleavage and invagination is slightly delayed but otherwise no morphological defects are detectable (data not shown). However, Fez morphants have fewer serotonergic neurons than do normal embryos and the region in which they differentiate is smaller, being restricted to an area between the bases of anterolateral arms at the 72-hour pluteus stage (Fig. 2B). By contrast, in normal embryos, the left and right borders of the serotonergic neural complex extend up to the middle of those arms (Fig. 2A). This phenotype is specific because it was also obtained with a second morpholino targeted to a non-overlapping sequence in the mRNA (see Fig. S1 in the supplementary material). Because the cell bodies and axon network of synaptotagminB-positive, non-serotonergic neurons beneath the ciliary band in Fez morphants are almost identical to those of normal embryos (compare Fig. 2B′ with 2A′) (Nakajima et al., 2004; Burke et al., 2006), the loss of Fez function probably affects only the serotonergic neurons on the aboral side of the animal plate. There are two possibilities for the decreased number of serotonergic neurons in Fez morphants. One is that the animal plate itself becomes smaller than that of normal embryos, and another is that the size of animal plate is the same but the differentiation of serotonergic neurons is perturbed at the periphery of this region. To test this, we focused on the position of the first pair of serotonergic neurons that develop near the aboral-lateral margins of the animal plate, which are marked with Nk2.1 expression (Fig. 2D,F) (Yaguchi et al., 2000; Yaguchi et al., 2006). Although Fez morphants have fewer serotonergic neurons than do control embryos at later stages (Fig. 2A,B), the timing of the differentiation of the first pair of serotonergic neurons is the same. In Fez morphants, these serotonergic neurons are closer to each other than those in normal embryo are, and there are fewer Nk2.1-positive cells between them (Fig. 2D′,F′). Here, the serotonergic neurons are stained with anti-SynB antibody because at this stage all SynB neurons are serotonergic. In addition, the expression of homeobrain (hbn) suggests that the animal plate is smaller. hbn expression is cleared from the animal plate and surrounds it at the gastrula and pluteus stages (Wei et al., 2009). However, the central region of clearance is smaller in Fez morphants than in controls (compare Fig. 2G with 2E). Moreover, treatment of embryos with the γ-secretase inhibitor DAPT, which suppresses Notch signaling, supports these results. In sea urchin embryos, delta-Notch signaling suppresses differentiation of serotonergic neurons through lateral inhibition (Wei et al., 2011). Thus, if Notch signaling is blocked, additional neural precursors can differentiate and contribute to a cluster of serotonergic neurons. If the primary function of Fez, the early expression of which is in all cells of the animal plate, is to support the differentiation of each cell as a neuron, then the additional neurons that develop in DAPT-treated embryos (Fig. 3C) would not appear in Fez morphants. However, this is not the case, because the number of neurons increases and they form a cluster in DAPT-treated embryos, regardless of whether Fez is present or absent (Fig. 3A-E). This indicates that Fez is not required just for the differentiation of serotonergic neurons, but rather for determining the size of the neurogenic ectoderm. Furthermore, mis-expression of Fez can significantly increase the size of the animal plate and the number of serotonergic neurons, which extend all the way to the tip of anterolateral arms (Fig. 2C, Fig. 3F,G). In addition, because the size of the neural cluster along the oral-aboral axis in DAPT-treated fez mRNA-injected embryos is wider than that in control embryos, Fez mis-expression can expand the animal plate along both the ciliary band and the aboral ectoderm.

The loss-of-function experiments suggest that Fez is required to maintain the size of the neurogenic animal plate in the sea urchin embryo. To investigate when Fez executes this function, we examined the expression patterns of foxQ2 and its downstream gene, nk2.1. fez begins to be expressed between the hatching and mesenchyme blastula stages in control embryos (Fig. 1), and the size of the animal plate, marked by foxQ2, is almost identical at both stages in control embryos (Fig. 4A,B). The FoxQ2-dependent gene nk2.1 begins to be expressed at the mesenchyme blastula stage in a similarly sized region of the neurogenic ectoderm (Fig. 4C). By contrast, in Fez morphants, the size of the foxQ2-expressing domain at the mesenchyme blastula stage becomes smaller than that observed at the hatching blastula stage (Fig. 4D,E), and significantly smaller than that observed in control embryos at the mesenchyme blastula stage (compare Fig. 4E with 4B). As well, the size of the nk2.1-expressing region decreases (Fig. 4F). Taken together, these data indicate that Fez is required to maintain the precise size of the neurogenic animal plate as soon as it is expressed between the hatching and mesenchyme blastula stages.

Fez could regulate the size of the animal plate by supporting positive regulators of this domain or by inhibiting negative regulators such as BMP signaling, which suppresses development of other neuroectoderm territories in the sea urchin embryo (Lapraz et al., 2009; Saudemont et al., 2010; Yaguchi et al., 2010a). To investigate the first alternative, we employed Δcadherin (Δcad) mRNA-injected embryos that lack Wnt/β-catenin and TGF-β signals, and develop a greatly expanded animal plate and flanking region (Logan et al., 1999; Duboc et al., 2004; Yaguchi et al., 2006; Wei et al., 2009). In these embryos, fez is expressed throughout the expanded animal plate (see Fig. S2 in the supplementary material). When Fez translation is blocked in Δcad mRNA-injected embryos with a Fez morpholino, foxQ2 expression still covers half or more of the embryo, and its distribution is not distinguishable from that of Δcad only-injected embryos (see Fig. S2 in the supplementary material). This indicates that Fez does not function as an essential feedback inducer of foxQ2, an upstream regulator for animal plate development, and suggests the that it functions instead as an inhibitor of animal plate-repressing signal(s).

To test whether BMP signaling might be the postulated signal that suppresses animal plate development and that is suppressed by Fez, we examined the effects on the expression of foxQ2 and fez of knocking down or mis-expressing Fez throughout the embryo. As shown in Fig. 4, loss of BMP2/4 does not expand the animal plate or significantly alter the expression of these two genes (Fig. 4G-J), but mis-expression strongly suppresses the expression of both (Fig. 4K,L, arrowhead and double arrowheads). We then asked whether the reduction of the animal plate in Fez morphants, as marked by foxQ2 expression, requires BMP2/4. To test this, we injected Fez and BMP2/4 morpholinos simultaneously. In Fez morphants in which BMP signaling is normal, the foxQ2 domain is reduced compared with control embryos (Fig. 4M,N), but this is not the case in a double Fez-BMP2/4 morphant (Fig. 4O,P), indicating that in the absence of Fez, BMP signaling changes the size of the animal plate.

Although the data strongly suggest that it is BMP2/4 signaling that reduces the animal plate in the absence of Fez, they do not rigorously rule out the possibility that, in the context of Fez-BMP2/4 double morphants, ectopic Nodal, which diffuses to the aboral side of the animal plate (Yaguchi et al., 2010a) might affect its properties or interfere with some other signal that antagonizes the animal plate. To eliminate ectopic Nodal signaling in double morphants from regions where BMP2/4 normally signals, we used an experimental embryo described previously (Yaguchi et al., 2007), in which Nodal signaling is restricted to one side of the embryo.

Endogenous Nodal is eliminated with a morpholino injection at the one-cell stage, and Nodal signaling is supplied on only one side of the embryo by injecting into one blastomere at the two-cell stage mRNA encoding constitutively active Smad2/3 (actSmad2/3) along with mRNA encoding a myc marker. This treatment forces oral and aboral fates on the injected and non-injected sides of the embryo, respectively, because actSmad2/3 activates the oral program that produces BMP2/4 (see Fig. S3 in the supplementary material), which diffuses to and signals to cells on the aboral side and activates the aboral ectoderm program (Yaguchi et al., 2007) (Fig. 5A-C). In this embryo, the number of foxQ2-positive cells in the aboral half is about 20 (Fig. 5A-C,M; see Materials and methods), but this number is cut approximately in half in Fez morphants (Fig. 5D-F,M), showing a similar difference to that observed between normal and Fez morphant embryos. However, this reduction did not occur in the experimental embryo (doubly injected with Nodal morpholino and actSmad2/3) in which BMP2/4 is also knocked down. The size of neurogenic ectoderm is unchanged in those embryos, whether or not Fez protein is present or absent (Fig. 5G-M). These data indicate that the signal that Fez blocks in the animal plate is BMP2/4. Thus, FoxQ2 causes production of Fez, which attenuates BMP2/4 signaling in the aboral animal plate ectoderm.

To determine if and when BMP2/4 signals in the animal plate, we monitored the presence of phospho-Smad1/5/8 (pSmad1/5/8) (Lapraz et al., 2009) and FoxQ2 protein by immunostaining between the unhatched blastula and mesenchyme blastula stages. The specificity of each of these antibodies in H. pulcherrimus was confirmed because all signals were absent in embryos injected with the corresponding morpholinos (see Fig. S4 in the supplementary material). Furthermore, all pSmad1/5/8 signals at this stage are attributable to BMP2/4 activity, because there are no signals in BMP2/4 morphants (see Fig. S4 in the supplementary material). Before the hatching blastula stage, no BMP signaling is detected, but when the embryo hatches, a clear signal appears in the nuclei of cells in the animal half of the embryo, including the animal plate marked by FoxQ2 expression (Fig. 6A,B). However, a few hours later, at mesenchyme blastula stage, BMP2/4 signaling increases dramatically in the future aboral ectoderm and establishes a decreasing gradient from aboral to oral through animal plate (Fig. 6C) (Lapraz et al., 2009). Based on the immunohistochemical data, the relatively strong BMP signaling area is closer to the animal pole in Fez morphants, compared with that in control embryos (Fig. 6D,E). To analyze this further, we measured pixel values on images and used ImageJ software to quantitate relative immunofluorescent signal intensities that reflect the relative levels of FoxQ2 and pSmad1/5/8 proteins in nuclei in cells in a line from the animal pole to the center of the aboral side (Fig. 6F, white dots). The data were normalized in each embryo, then averaged among individual batches and shown as relative intensity. The measurements were performed in 10-14 embryos in three independent batches to check reproducibility. In normal embryos, the concentration of FoxQ2 drops from maximal levels to background in the area three to six cells from the center of the animal plate and within that same area an inverse relationship exists for pSmad1/5/8 (Fig. 6G, upper panel). A similar inverse relationship exists in Fez morphants, but the corresponding drop in FoxQ2 occurs several cells closer to the center of the animal plate (Fig. 6G, lower panel). In both normal and Fez morphants, maximal levels of FoxQ2 are achieved when pSmad1/5/8 levels are 30-40% of maximal. The clear difference in Fez morphants is that, within the critical region of the animal plate that is defined by half maximal levels of FoxQ2 in control and Fez morphant embryo (vertical dotted dashed lines), the pSmad1/5/8 levels are ∼20-40% of maximum in controls, but 50-80% of maximum in Fez morphants. Because maximum levels of FoxQ2 and pSmad1/5/8 are the same in control and Fez morphants, and because Fez is not required for FoxQ2 expression, these data indicate that, in this outer region of the animal plate, maximal FoxQ2 expression occurs if Fez suppresses BMP signaling below a threshold level ∼30-40% of maximal.

The data presented here establish that Fez functions as soon as it is produced downstream of FoxQ2 to attenuate BMP signaling on one side of the animal plate, thereby maintaining the size of this neuroectodermal domain in the sea urchin embryo. This mechanism is superimposed on the well-known pathway that produces a gradient of extracellular BMP activity through the activities of its antagonists or their activating proteases (Lamb et al., 1993; Sasai et al., 1994; Piccolo et al., 1997; Lee et al., 2006; Plouhinec and De Robertis, 2009). Fez produces a significant decline of BMP2/4 activity in the outer regions of the animal plate of the early embryo along which the serotonergic neurons will subsequently develop.

The model for the mechanism and timing of Fez function is described in Fig. 7. bmp2/4 begins to be expressed in prospective oral ectoderm of an unhatched blastula as a result of Nodal signaling, prior to the appearance of fez mRNA in the animal plate (Fig. 7A) (Duboc et al., 2004; Materna et al., 2010). Although no pSmad1/5/8 signals were observed in this embryo, it is likely that BMP2/4 protein starts to be secreted there and to signal at low levels because only a few hours later at hatching blastula stage low levels of pSmad1/5/8 can be detected in both aboral and anterior neurogenic ectoderm (Fig. 6B). At this time, FoxQ2 and perhaps other factors induce fez transcription in animal plate cells (Fig. 7A). During mesenchyme blastula stages, BMP2/4 moves towards the aboral side, away from the site of bmp2/4 transcription, possibly aided by Chordin, and establishes a gradient of activity between aboral and oral sides of the embryo (Fig. 6C, Fig. 7A) (Ben-Zvi et al., 2008; Lapraz et al., 2009). The BMP activity gradient in the neurogenic ectoderm is shallow and continuous in the absence of Fez function (Fig. 6G) and probably reflects the normal morphogen gradient pattern in the extracellular space. However, in the normal embryo, the continuous gradient is perturbed at the border between neurogenic and non-neurogenic ectoderm by Fez (Fig. 6G, Fig. 7B,C), sharpening the border between the animal plate and the aboral ectoderm.

The abnormal shift of the animal plate border in Fez morphants occurs on the side where BMP signals, which reduces the size of the animal plate size in these embryos. This conclusion is supported by the observation that foxQ2 expression is reduced where BMP2/4 signaling occurs, but Nodal signaling is prevented. The consequence is that Fez functions to stabilize the border on the aboral side along which the serotonergic neurons develop; Fez is later also expressed specifically in these neurons, although it is not required to confer neuronal identity to them. Instead it might continue to protect them from BMP signaling and/or it might be involved in promoting normal axon guidance because elongation of axons in these neurons is poorly developed in Fez morphants (Fig. 2B), as has been observed in the olfactory bulb of Fez mutant mice (Hirata et al., 2006b).

Although the data presented here indicate that Fez is involved in attenuating BMP signaling along the outer regions of the aboral animal plate ectoderm, it is still not clear how the central part of this domain is protected against BMP2/4 signaling in Fez morphants (Fig. 6). Although pSmad1/5/8 activity is not as low as in normal embryos in this region of the animal plate, it might be below the threshold required to induce aboral, non-neurogenic ectoderm there. Alternatively, this central region might be protected by a Fez-independent mechanism(s). The latter possibility is more likely because neurons differentiate in this region of animal pole ectoderm even when mis-expressed BMP2/4 generates dramatic downregulation of foxQ2 and fez expression in these cells at blastula stage (Fig. 4) (Yaguchi et al., 2006).

Downregulation of BMP signaling along the aboral side of the animal plate by Fez function is probably cell autonomous, although the possibility that Fez induces a short-range signal cannot be strictly ruled out. So far, because the sequence of Fez suggests that it acts as a transcriptional repressor (Hirata et al., 2006a) rather than a Smad-interacting protein, such as SIP1 (Verschueren et al., 1999; van Grunsven et al., 2003; van Grunsven et al., 2007), it is unlikely that Fez interferes directly with pSmad1/5/8 activity in intracellular regions. Thus, it is possible that Fez regulates a gene(s) that is involved in pathways to maintain pSmad1/5/8 levels, such as degradation, dephosphorylation, and/or nucleocytoplasmic shuttling of Smad (Inman et al., 2002; Chen et al., 2006). Target genes repressed by Fez have not yet been identified in sea urchin, but, for instance, could include gene(s) encoding protein(s) that promote BMP signaling or regulatory gene(s) specifying aboral ectoderm downstream of BMP2/4 signaling. It might also support expression of several intracellular BMP signaling attenuators that have recently been identified, including SIP1 (Nitta et al., 2004), Smurf1 (Alexandrova and Thomsen, 2006; Sapkota et al., 2007), Ski and SnoN (Deheuninck and Luo, 2009). However, it is thought that the function of these intracellular factors is to pattern the neuroectoderm, not to establish it as a domain, which is what we focus on in this study. Whether they are also involved as mediators of Fez function to establish the precise border between neurogenic and non-neurogenic regions will require further investigation.

Based on previous studies, the Fez family functions at several developmental steps in the central nervous systems in other animals (Shimizu and Hibi, 2009), such as defining a subregion of anterior neuroectoderm (Hirata et al., 2006a) and controlling individual neural fate (Shimizu et al., 2010). With respect to molecular mechanisms, more is understood about its role in neural fate specification. In vertebrates, Fezf1 and Fezf2 control neural differentiation by suppressing Hes5 (Shimizu et al., 2010), and in flies, Erm, an ortholog of Fez in Drosophila melanogaster, maintains the restricted developmental potential of intermediate neural progenitors (Weng et al., 2010). Although it has been reported that Fezf1 and Fezf2 suppress development of the caudal diencephalon in mice, and expression of Irx1 and Wnt3a, the molecular mechanisms controlling these steps are unclear. It is also not known whether attenuation of BMP signaling is involved, as shown here in the sea urchin embryo.

A combination of extrinsic and intrinsic mechanisms controls the size of the animal plate. After restriction of the animal plate to the animal pole by canonical Wnt-dependent processes, further signaling through BMP signaling can affect its size. Here, we show that extracellular control of BMP signaling through antagonists is supplemented by intracellular control within the plate itself. Similarly, Nodal signaling on the other side of the plate suppresses serotonergic neuron development and it is blocked from the animal plate by the extracellular antagonist Lefty and the intracellular factor FoxQ2. What emerges from this and previous work (Yaguchi et al., 2008) is that a fascinating regulatory cassette, FoxQ2 and its downstream gene Fez, are involved in attenuating expression of Nodal and its downstream gene BMP2/4 sequentially to protect the neurogenic animal plate on its oral and aboral sides, respectively. These parallel mechanisms ensure that the edges of the field where neurons develop in the animal pole ectoderm are stably defined.

We thank Robert Angerer and Angerer laboratory members for fruitful comments on this work; Robert Burke, Yoko Nakajima, Yasunori Sasakura and David McClay for essential reagents and devices; and Julius Barsi and Eric Davidson for sharing unpublished data. We thank Masato Kiyomoto and Mamoru Yamaguchi for providing the adult sea urchins and Mrs Y. Tsuchiya, T. Sato, H. Shinagawa and Y. Yamada from the Shimoda Marine Research Center for collecting and keeping the adult sea urchins.