Adult S. kowalevskii were collected intertidally on Cape Cod, MA within the Waquoit Bay Reserve. Animal husbandry and culture techniques were comprehensively described previously [15,49]. Embryos were staged by the normal tables of Bateson [50,51] and Colwin and Colwin [52,53].

Classical embryology experiments and microinjections were carried out as described previously [15]. Targeted blastomere injections were performed under a stereomicroscope using a back-filled needle connected to a glass syringe with plastic tubing. The entire system is filled with mineral oil and the injection is performed manually with the syringe under visual control (injected solution is colored by 1% fast green FCF (F-7252; Sigma-Aldrich, St. Louis, MO). Injection was performed into identified blastomeres at cleavage stages. Injection success was monitored by co-injection of 1% rhodamine dextran (D-1817; Molecular Probes, Eugene, OR). siRNAs targeting β-catenin and Fz5/8 were described previously [15,48]. The open reading frames of S. kowalevskii wnt3, sfrp1/5, and dkk1/2/4 were cloned into pCS2+, then linearized and used for in vitro synthesis of capped RNA using the SP6 Message Machine kit (Applied Biosystems/Ambion, Foster City, CA). The number of experimental embryos examined is indicated in figure panels.

The cWnt pathway was activated using the GSK-3β inhibitor 1-azakenpaullone [54] (191500; Calbiochem, Sigma-Aldrich) or the recombinant Wnt3a protein (R&D Systems, Minneapolis, MN). Treatments were carried out as described previously [15,55]. The treatments led to robust and homogeneous effects as assessed by in situ hybridization on 2 to 20 embryos per probe.

A total of 200,000 expressed sequence tags (ESTs) were screened from six libraries previously described [47,55,56]. Partial sequences were further cloned through library screening using PCR. Genbank accession numbers are as follows: wnt1, EU931645; wnt2, EU931646; wnt3, EU931647; wnt4, GU224244; wnt5, GU076159; wnt6, GU076160; wnt7, GU076161; wnt8, GU076162; wnt9, GU076163; wnt10, GU076158; wnt11, GU076158; wnt16, EU931648.1; wntA, GU224245; fz5/8, GU075997; fz1/2/7, MG711509; fz4, MG711510; fz9/10 MG711511; dkk1/2/4, GI:259013424; sclerostin, GU076111.1; sfrp1/5, GU076117.1; sfrp3/4 MG682447; r-spondin, GU076102.1; notum, GU076072.1; wif, GU076157.1; dkk3, MG682446. Tentative orthology of ESTs was assigned by BLAST. Sequences were aligned with sequences from cnidarians and bilaterians using clustalX [57]. Gene tree analyses were carried out using Bayesean [58] and Neighbor-joining [59] algorithms to assign orthology relationships (See S1 and S2 Figs).

Experimental samples of 40 to 50 embryos each were frozen in liquid nitrogen and stored at −80°C. RNA was extracted using the RNAqueous-Micro Total RNA isolation Kit (Life Technologies, Carlsbad, CA) using a motorized pestle for initial homogenization of samples.

cDNA synthesis was performed with Superscript III (Life Technologies) using 50 ng/μL total RNA in a 20 μL volume, using the oligo(dT)20-primer according to manufacturer’s instructions. qPCR reaction-mix was set up using SsoAdvanced Universal SYBR Green Supermix (Bio-Rad, Hercules, CA) with 0.4 μM final concentration per primer and 0.066 ng/μL final cDNA concentration (cDNA concentration is based on the initial 50 ng/μL total RNA concentration used during cDNA synthesis and dilutions of the cDNA synthesis mix thereafter). qPCR was performed in 96-well plates (Bio-Rad) with 9 μL total reaction volume per well using the CFX-Connect Real-Time System (Bio-Rad) running the following program: 95°C for 3 min for an initial melting, and 95°C for 10 s, 55°C for 40 s for 40 cycles, followed by a melting point analysis.

All primer sets were initially optimized for efficiency at 55°C annealing and low probability of primer-dimer product in No-template controls. All cDNAs were tested in a dilution series to determine the area of linear amplification with a larger set of control primers. cDNAs generally behaved linearly with low standard errors to a final dilution of 0.00833 ng/μL or lower.

Three technical replicates of each sample, a No-Template control, and RT⁻ controls with a subset of primers, were performed for each plate and cDNA. actin, beta-tubulin, odc, and G3PDH were used for sample normalization. Actin was used for inter-plate normalization. Quantitative values of gene up- or down-regulation relative to control were calculated using the 2^{−ΔΔ CT} method [60] using the Bio-Rad CFX Manager software version 3.1.

In situ hybridization was carried out as described [47,49]. Stained embryos were post-fixed in 10% formaldehyde in 1 X PBS overnight and then sequentially dehydrated into 100% EtOH, followed by several washes in 100% MeOH before clearing in MurrayClear reagent (2 parts Benzoyl benzoate and 1 part benzyl alcohol) and mounted in Permount. Pictures were taken on a Zeiss Axioimager Z1 using a Zeiss Mrc5 camera. Image panels and figures were constructed with Adobe Photoshop and Adobe Illustrator.

To begin our investigation of the function of cWnt signaling during the development of hemichordates, we cloned and described the expression of key components of the signaling pathway (ligands, receptors, and modifiers) detected by whole mount in situ hybridization and reverse transcription PCR (RT-PCR), over a range of developmental stages, from oocytes to juveniles.

The Wnt ligands are a large family of secreted glycoproteins characterized by an invariant pattern of 22 to 24 highly conserved cysteine residues [61]. We cloned 13 Wnt genes from ESTs sequenced from three developmental stages, from both normalized and non-normalized cDNA libraries [56], which include wnt1, wnt2, wnt3, wnt4, wnt5, wnt6, wnt7, wnt8, wnt9, wnt10, wnt11, wnt16, and wntA. All members of the proposed ancestral complement of 13 Wnt subfamilies are present in the hemichordate genome, including Wnt11 and Wnt2 that are absent from the sea urchin genome, and WntA, which has been lost from chordates [62]. We cloned one representative from each Wnt subgroup, and have not detected any paralogy duplicates by systematic screening of ESTs and genome assembly (S1A Fig).

We first present expression data for Wnt ligands by RT-PCR to determine the temporal expression profiles of each gene from oocyte to day 3 of development, when the main features of the hemichordate body plan are established (Fig 1). Maternal expression of two ligands, wnt4 and wnt9, was detected in oocytes and early cleavage stages, and very low levels of wnt1 and wnt8 in oocytes. By late blastula/early gastrula and subsequent developmental stages, all Wnts except wntA were detectable by RT-PCR.

We describe the patterns of expression for all 13 Wnt genes (with the exception of wnt10, which we failed to detect) by whole mount in situ hybridization from mid-blastula (12 hpf [h postfertilization] at 20°C) to day three of development (Figs 2 and 3). The first evidence of zygotic Wnt expression is at mid-blastula: five of the Wnts are detected by in situ hybridization (wnt4, wnt6, wnt8, wnt2, and wnt11). Their expression is detected in circumferential bands, strongest at the intersection of the AV hemispheres in the region fated to become the blastopore (Figs 2Di, 2Ei, 3Ai, 3Bi and 3Ci). All five genes are expressed in broad overlapping domains in the animal hemisphere, but none are expressed at the animal pole, the prospective far anterior region. Expression is detected exclusively in the ectodermal, animal hemisphere precursors, not in the vegetal endomesodermal precursors at pregastrula stages. During gastrulation, the expression domains of most of the ligands become more spatially restricted, but generally they retain their relative spacing and order along the newly forming A/P axis of the ectoderm (Figs 2 and 3). As has been described in other species, many Wnt ligands are expressed around the forming blastopore; wnt1 and wnt4 (Fig 2Cii and 2Dii) are both expressed around the inside of the blastopore lip at the boundary of ectoderm and endomesoderm, whereas wnt3, wnt6, and wnt16 (Fig 2Bii, 2Eii and 2Fii) are expressed further anteriorly in the ectoderm at the exterior edge of the blastopore lip. The remaining expression domains are located further anteriorly in the ectoderm; wnt8, wnt2, wnt7, and wnt5 are all expressed anterior to wnt3 in broadly overlapping domains (Fig 3Aii, 3Bii, 3Dii and 3Eii). Wnt11 has the most anterior limit of ectodermal expression at this stage, but is still restricted from the most apical region (Fig 3Cii). As the embryos elongate after gastrulation, between 36 and 48 h of development, expression of the ligands becomes far more spatially restricted along the A/P axis. Most ligands remain expressed exclusively in the ectoderm with a few exceptions. Two of the ligands are expressed in the two posterior coeloms: wntA (Fig 3Fiv and 3Fv) and wnt9 (Fig 3Giii–3Gv), and others are expressed in the endoderm: wnt3 into the posterior endoderm (Fig 2Biii), wnt9 in dorsal, anterior endoderm, around the forming gill slits (Fig 3Giii–3Gv), wnt16 in the ventral posterior endoderm (Fig 2Fiii), and wnt5 at low levels throughout the endoderm at 48 h of development, which later refines to the forming gill slits (Fig 3Eiv and 3Ev). wnt1, 3, 4, and 6 (Fig 2Biii, 2Ciii, 2Diii and 2Eiii) all retain their early expression around the blastopore. However, wnt1, 4, 6, and 16 develop additional expression domains around the anterior metasome (Fig 2Civ, 2Diii, 2Eiii and 2Fiv). wnt8, 2, 11, 7, and 5 all begin to refine their single expression domains to the boundary between the prosome and metasome following gastrulation (Fig 3Aiii, 3Aiv, 3Biii, 3Biv, 3Ciii, 3Civ, 3Diii, 3Div, 3Eiii and 3Eiv), and by day three of development, some of these expression domains are sharply localized in narrow regions, such as in the case of wnt8, 2, and 7 (Fig 3Av, 3Bv and 3Dv). wnt7 is not expressed as a single circumferential domain, and is down-regulated along the ventral midline by 48 h of development (Fig 3Div), and up-regulated in a dorsal spot at the base of the proboscis (Fig 3Dv).

By day three of development, all ligands with the exception of wntA and wnt10 are detected in one or more of three ectodermal domains of expression; the base of the prosome, the boundary of the metasome and trunk at the developing first gill slit, and in the posterior ectoderm.

In summary, the zygotic expression of the Wnt ligands begins at the midblastula stage in the animal hemisphere, close to the boundary with the vegetal hemisphere. Throughout gastrulation, ligands are expressed broadly in the ectoderm from the blastopore to the more anterior regions of the ectoderm, but are always excluded from the most apical/anterior regions. By later developmental stages, this expression largely refines to spatially restricted domains marking a region in the posterior ectoderm, over the developing first gill slit, at the boundary between the trunk and collar, and finally at the boundary between the developing proboscis and the collar. Expression is also detected in the posterior endoderm and mesoderm.

An increasing number of modifiers of the cWnt pathway have been described in vertebrates [2,63]. Many of these genes encode secreted proteins, which act on the pathway by a variety of means and can act to both potentiate and inhibit cWnt signaling. Comparative studies on the expression and developmental roles of these proteins are far less extensive than studies on the Wnt ligands themselves. We have isolated and determined the expression pattern of several Wnt modifiers in S. kowalevskii and compared their expression and function to those reported in other animals in order to identify possible evolutionarily conserved roles of these pathway modifiers. Three antagonists are described below, and the other modifiers are described in S1 Text and S3 Fig.

We cloned four representatives of the Fz receptor family characterized by an extracellular domain, including a signal peptide, and a cysteine-rich Wnt-binding domain, seven transmembrane domains, and a cytoplasmic tail [78]. A genomic survey of the current genome assembly identified only four Fz genes. Previous studies have described four ancestral Fz subgroups: Fz1/2/7, Fz4, Fz5/8, and Fz9/10 [62]. The four Fzs from hemichordates exhibit robust orthology to these subgroups already present in cnidarians (S1B Fig) and build further support to the hypothesis that four Fzs are ancestral to the eumetazoan lineage [79]. Similar to the findings from echinoderms, which are a sister group to hemichordates, we were not able to isolate a member of the Fz3/6 group, which is present in vertebrates but absent from cnidarians and sea urchins. This supports the hypothesis that the group arose by duplication during chordate evolution [62].

Partial expression profiles of fz5/8 in S. kowalevskii and early embryonic and larval expression in P. flava have been previously described in the anterior of both embryos [48,73]. By RT-PCR, two Fzs, fz5/8 and fz1/2/7 are detected at high levels maternally in newly fertilized oocytes and early cleavage stages (Fig 1). By late blastula/early gastrula and subsequent developmental stages, all Fzs are expressed. Their expression, like the Wnt ligands, is also highly regionalized along the A/P axis during all developmental stages, suggesting that they play region-specific roles during the patterning of the embryonic A/P axis. The most anteriorly expressed of the receptors is fz5/8, detected midway through blastula from the animal pole throughout most of the animal hemisphere, but excluded from the most vegetal region of that hemisphere, the region fated as posterior ectoderm (Fig 5A). As gastrulation begins, the expression is increasingly restricted to the most anterior ectoderm. As the vegetal hemisphere invaginates and contacts the anterior ectoderm midway through gastrulation, fz5/8 expression is initiated in the anterior endomesoderm that is fated to become the proboscis/prosome mesoderm, at the same A/P level as the ectodermal expression (Fig 5C). As gastrulation continues, the expression is now clearly restricted in the anterior ectoderm of the prospective prosome, and continues in this domain [48] throughout all stages examined. This expression resembles the early expression of fz5/8 in both P. flava, sea urchin, and the anterior localization of fz8 during mouse development [72,73,80]. Short interfering RNA (siRNA) knockdown of fz5/8 resulted in abnormal patterning of the anterior ectoderm with the expansion of apical markers at the expense of more posterior proboscis markers, demonstrating the importance of Wnt signaling for the posteriorization of the anterior ectodermal territory [48].

fz4 is expressed in a highly restricted domain from the onset of zygotic expression during gastrulation (Fig 5F). Expression is uniquely ectodermal and is localized to a narrow ring in the more rostral region of the anterior ectoderm. This domain refines to a narrow ring at the boundary between the prosome and mesosome, at the posterior boundary of the expression limit of fz5/8 (Fig 5G and 5H). This to some extent resembles the expression of fz4 in chick and mouse, where it is localized in the diencephalon at the boundary with the telencephalon [80]. fz1/2/7 has a much broader domain of expression; it is first detected at blastula stage throughout the animal hemisphere, more broadly than that of fz5/8 (Fig 5I). By the beginning of gastrulation, its expression has been largely down-regulated in the anterior ectoderm and around the blastopore, but remains in the midectoderm (Fig 5J). Expression is also detected in the forming mesendoderm. By 36 h after fertilization, transcripts are detected in overlapping domains with both fz4 and fz5/8 in the anterior domain, with expression mainly in the prospective mesosome, expanding down into the prospective trunk, but not past the ciliated band into the blastoporal area (Fig 5K). By day three of development, most transcripts are detected in the mesosome and at the base of the proboscis, with only weak expression in the anterior trunk. At this late stage, expression is now visible in the anterior mesoderm of the proboscis (Fig 5L). fz9/10 expression begins at blastula in the most vegetal portion of the animal hemisphere (Fig 5M). At gastrulation, expression remains in a similar domain in the posterior ectoderm, just anterior to the forming ciliated band (Fig 5N), and by 36 h of development, the expression domain has expanded anteriorly to the entire trunk and part of the collar, overlapping extensively with the expression domain of fz1/2/7 (Fig 5O). By day three of development, expression is no longer detectable.

The interactions of Fzs and Wnts is poorly understood, and it is proposed that the overlapping expression of many of the vertebrate Fzs leads to functional redundancy [81,82]. In S. kowalevskii, Fz expression is highly regionalized along the A/P axis and suggests that the different receptors are responsible for patterning different regions of the axis. Notably, none of the receptors are expressed around the blastopore, where many of the ligands are localized. The functional significance of this observation is explored in later experiments.

The expression of Wnt ligands and their antagonists is strongly localized along the developing A/P axis of S. kowalevskii, and their relative domains are similar in location to their orthologues during vertebrate development; ligands are expressed posteriorly and antagonists are expressed anteriorly, with Fz receptors expressed in staggered domains (Fig 6). We carried out a series of experiments to test whether the cWnt pathway is involved in the early specification of the hemichordate A/P axis.

To test for the timing of Wnt activity during early A/P patterning, we further developed the 1-azakenpaullone experiments. Concurrent treatments were carried out by adding the inhibitor every 2 h, with the first beginning at early blastula (11 hpf), and the treatment ending at 64 hpf (S6 Fig). Embryos were cultured at 20°C and fixed at 96 hpf when all the major body divisions had formed, and the A/P and dorsoventral (D/V) axes were clearly morphologically differentiated. Graded phenotypes were observed and correlated with the treatment initiation time; the more severe phenotypes resulted from earlier and longer treatments (S6I Fig). The first treatment initiated at 11 h led to the most dramatic phenotype that we have observed. The embryos exhibited major morphological defects; all anterior markers down to anterior trunk failed to express, even the anterior trunk marker en (S6ii Fig), but msx—normally a broad trunk marker—was expressed up to the most anterior region of the embryo (S6Gii Fig). hox9/10, a posterior trunk marker, was expressed almost up to the anterior of the treated embryos (S6Hii Fig). Later treatments produced weaker phenotypes with a gradual restoration of molecular marker expression from posterior to anterior. hox9/10 expression was restricted posteriorly for treatments starting at 13 hpf or later, and msx expression did not expand anteriorly for treatments starting at 17 hpf or later (S6G and S6H Fig). These experiments suggest that the ectoderm is most sensitive to cWnt activation during the blastula and gastrula stages. It is the period when Wnt genes are expressed broadly in the ectoderm (Fig 6).

We have further analyzed the above experiment by examining gene expression at 21 hpf (onset of gastrulation), when little morphogenesis has occurred (S6B–S6D Fig). A similar gradation in phenotypes was observed for the anterior markers sfrp1/5 and six3. Interestingly, expression of the posterior marker hox9/10 was unchanged even for the earliest treatment starting at 11 hpf, while we observed an ectopic expression anteriorly when its expression was analyzed at 96 hpf (compare S6Dii and S6Hii Fig). To better understand this discrepancy, we treated embryos with a range of 1-azakenpaullone concentrations starting at early blastula (12 hpf) and analyzed A/P ectodermal markers expression using quantitative PCR (qPCR) at early gastrula stages (24 hpf; Fig 8A). This restricted the treatment to early establishment of the A/P axis. The second treatment again began at 12 h but was extended to 48 h into early embryo elongation from 24 to 48 h, after the early specification of the A/P axis (48 hpf; Fig 8B). For both treatments, proboscis/prosome markers were strongly repressed (Fig 8A and 8B). Collar markers were initially strongly down-regulated, whereas their later expression was either moderately activated or repressed depending on the inhibitor concentration (Fig 8A and 8B). This observation suggests that anterior suppression regulates fates down to the collar and its exact position varies according to the concentration of 1-azakenpaullone. All anterior/central trunk markers are activated in a dose-dependent manner at both stages. Early expression of posterior Hox genes, during the initial establishment of the posterior territory, is insensitive to cWnt activation, with the exception of hox11/13b. By contrast, the same genes are strongly activated when their expression is analyzed at 48 hpf. We confirmed this global gene expression analysis by in situ hybridization of selected markers (Fig 8C). The anterior gene six3 is not detected following either treatment. The expression of the trunk marker msx was expanded anteriorly for both treatments while the posterior marker hox9/10 was only ectopically activated when analyzed at 48 hpf during embryo elongation following gastrulation.

Overall, the results from this section indicate that A/P markers can be organized into three groups according to their sensitivity to cWnt activation: (1) anterior genes are repressed (proboscis) or activated or repressed depending on concentration (collar), (2) intermediate genes (anterior/central trunk) are activated, and (3) posterior genes (posterior trunk) are initially insensitive before being activated following gastrulation during embryo elongation.

In hemichordates, we have shown that endomesoderm is the source of early signals posteriorizing the ectoderm that otherwise adopt an anterior character (Fig 9) [15]. We tested whether activation of the Wnt pathway was sufficient to posteriorize naive ectoderm explants that lack inputs from the endomesoderm. Embryos were cut at the 32-cell stage and the animal hemispheres cultured in isolation. They developed anterior fates demonstrated by the expression of the apical marker foxQ2-1 throughout the explant (Fig 9Biii), and failed to express posterior transcriptional markers such as en, msx, and hox9/10 (Fig 9Bvii, 9Bxi and 9Bxv). However, treating these explants with 1-azakenpaullone from midblastula stages (12 h of development) resulted in repression of the apical marker foxQ2-1 (Fig 9Biv) and the ectopic activation of the more posterior markers, msx, and en (Fig 9Bvii and 9Bxii) throughout the ectoderm. However, we were unable to activate the most posterior marker hox9/10 (Fig 9Bxvi). The expression of this marker was actually unchanged upon cWnt activation in whole embryos (Fig 9Bxiv), whereas expression of both en and msx was expanded (Fig 9Bvi and 9Bx). These results are entirely consistent from the 1-azakenpaulone treatments in intact embryos, and further suggest that there is a differential response to Wnt signaling along the A/P axis, with three distinct domains.

We carried out reciprocal experiments to test for the effects of reducing Wnt activity. We first injected capped mRNA of the Wnt antagonists dkk1/2/4 and sfrp1/5. Both overexpressions led to very similar phenotypes; expansion of the anterior proboscis territory and reduction in the size of the trunk (Fig 10i, S1 and S2 Movies), which is the reciprocal phenotype from Wnt3 overexpression (Fig 7Hi–7Iiii). The phenotypes differed in that the proboscis took on a more bulbous appearance with a narrower trunk following dkk1/2/4 injection (Fig 10iF and 10iH) when compared to sfrp1/5-injected embryos (Fig 10iB and 10iD). This may be due to the difference in specificity of the secreted antagonists as Dkk1/2/4 specifically antagonizes the canonical pathway by binding to Kremen/lrp, whereas Sfrp1/5 binds the Fz receptor and inhibits all Fz-mediated Wnt signaling. The molecular markers support the phenotypic transformations with both foxQ2-1 and six3 expanding in injected embryos (Fig 10iB, 10iF and 10iH), and the trunk marker en expressed further posteriorly with weaker expression when compared to controls following injection of sfrp1/5 (Fig 10iD), consistent with down-regulation of this trunk marker. Previous data have revealed that knockdown by siRNA of fz5/8 results in an anteriorized phenotype, as assayed by the expansion of the apical marker SkFGF-1; the effect is restricted to the proboscis matching the expression domain of fz5/8 [48]. We further tested this phenotype by assaying the expression of another apical marker foxQ2-1 and show a similar expansion (Fig 10iI and 10iJ). Conversely, the expression domain in experimental embryos of rx, a proboscis ectodermal marker normally excluded from the most apical domain, is restricted to a more posterior domain of expression (Fig 10iK and 10iL) [48], suggesting that a gradient of Wnt activity is involved in patterning the proboscis ectoderm.

The overexpression of Wnt antagonists results in a phenotype complementary to cWnt activation, but trunk fates cannot be repressed entirely. Because this could result from an incomplete blockade of cWnt activity, we performed additional experiments utilizing targeted injections of siRNAs designed to β-catenin. However, β-catenin is necessary and sufficient to specify endomesoderm, and injection of siRNA against β-catenin before the first cell cycle disrupts the establishment of the AV axis, resulting in fully animalized embryos that fail to gastrulate [15]. To avoid disrupting germ layer formation, β-catenin siRNA was injected into single blastomeres at the 4-cell stage. The first two cleavage planes occur along the AV axis and define the left/right and dorso-ventral axes, respectively [53]. Gastrulation and endomesoderm formation (revealed by foxA expression) were relatively normal if injection was delayed until the 4-cell stage (Figs 10ii and S7). Expression of foxQ2-1, which is normally restricted to the anterior-most ectoderm (Fig 10iiA), expanded posteriorly in descendants of blastomeres injected at the 2,4, and 8-cell stage (Fig 10iiB and 10iiC and S7 Fig). Notably, expression did not expand down to the presumptive posterior ectoderm, despite the presence of the siRNA in all ectodermal descendants in the injected quadrant (Fig 10iiC). Expression of the trunk marker msx was lost in all descendants of injected cells (Fig 10iiE and 10iiF), while the posterior marker hox9/10 was activated even in the absence of cWnt signalling (Fig 10iiH and 10iiI).

These targeted β-catenin siRNA experiments further support that there is a threshold of Wnt sensitivity in the trunk ectoderm. Wnt inhibition is sufficient to ectopically expand the most anterior genes, but only the presumptive anterior ectoderm can expand following cWnt suppression. Anterior/midtrunk gene expression requires active cWnt signaling, whereas posterior genes such as hox9/10 are unaffected by suppression of cWnt during the early establishment of A/P pattern (Fig 11). The inability to anteriorize posterior ectoderm by elimination of Wnt signaling, and the early activation of posterior genes even in the absence of cWnt signaling is likely to be due to non-cWnt posteriorizing signals emitted by endomesoderm.

A variety of analyses from genome and EST projects have established that the ancestral bilaterian complement of Wnts was likely 13 distinct families. Comparisons between bilaterian clades have revealed both Wnt gene diversification and loss during bilaterian body plan evolution. Within the deuterostomes, in chordates, the ancestral complement of Wnt gene families is 12; amphioxus has 13 ligands, and humans have 19, which can be classified into 12 distinct subfamilies [83]. The only subgroup missing from the vertebrates and amphioxus is WntA. In the tunicates, data from ascidians suggest that several Wnts were lost during the extensive genomic rearrangements in the tunicate lineage. Outside of chordates, 11 Wnts have been isolated from sea urchins with wnt2 and 11 absent from the genome of S. purpuratus [62], and RNAseq data suggests this is conserved in another species of urchin [72]. From comprehensive arthropod sampling and recent work on onychophorans, along with limited lophotrochozoan sampling, the likely ancestral repertoire of protostome Wnt genes was 12, excluding wnt3 [84,85], with several groups showing secondary losses of individual ligands. The most parsimonious reconstruction of the Wnt gene family complement in bilaterians is 13. Sequence data from S. kowalevskii conclusively supports this hypothesis, and provides the first data for a bilaterian possessing all of the 13 Wnt ligand families.

Comparative studies from placozoans, sponges, and ctenophores suggest that two Fzs were ancestral to metazoans [6,13,79,86]. Data from sea urchins [62], sea stars [87], chordates, hemichordates (our study), and annelids [85] support the conclusions from cnidarians that the Fz complement fz5/8, fz1/2/7, fz4, and fz9/10 is the ancestral state for bilaterians and eumetazoans, and that the Fz3/6 group present in chordates represents a lineage-specific duplication from the Fz1/2/7 group [62,79].

We cloned and characterized a range of extracellular Wnt modifiers. Of the characterized Wnt antagonists, we cloned dkk1/2/4, dkk3, sfrp1/5, sfrp3/4 (frzb), WISE/sclerostin, wnt inhibitory factor (wif), and notum/wingful, and of the Wnt agonists, we cloned r-spondin. Comparative genome surveys that explicitly discuss extracellular Wnt modifiers are scarce, and outside of chordates there are few reports of the presence or absence of these genes in other groups. Notably, a comprehensive survey in sea urchins [62] has identified all the Wnt modifiers that we identified except sclerostin, and thus S. kowalevskii has many of the antagonists that have been described in a variety of vertebrate species.

The current data from hemichordates, when combined with data from other metazoans, provides strong support for the hypothesis that cWnt signaling was intimately involved in the early evolution of the A/P axis. In many bilaterian groups, Wnts have been proposed to play a conserved role in anterior suppression and posteriorization, and strong supporting experimental evidence has been generated from chordates, sea urchin, annelids, planarians, and acoels [20,23,34,37–40,43] (reviewed in [10,11,14]). Functional data from cnidarians also demonstrates that the evolution of the role of Wnts in axis polarization was likely a very deep innovation of metazoan embryos [77,88–90], even prior to bilaterians. We present comprehensive expression data for the entire complement of Wnt ligands throughout all early developmental stages (Figs 2 and 3), as well as of Wnt modifiers (antagonists and agonists) (Fig 4, S3 Fig) and Fzs (Fig 5). The relative expression domains of Wnt ligands and their antagonists resemble those of other bilaterians at opposite ends of the forming axis (Fig 6). In S. kowalevskii, the predominant expression of Wnt ligands posteriorly, excluded from the most anterior domains, the broad expression of the four fz genes in localized ectodermal domains down the A/P axis, and the predominant localization of Wnt antagonists in the anterior of the embryo, resembles expression patterns of many different metazoan embryos [11]. However, our data contrast with the comprehensive expression data from sea urchins, which shows most of the ligand expression domains in early development in the endomesoderm, with a subset of the ligands with localized domains in the ectoderm [72]. In this study, we focus on the role of Wnts in ectodermal patterning, but it is important to note that Wnt ligands, Fzs, and Wnt modifiers are also expressed in the endoderm and mesoderm (Figs 2, 3 and 5), and likely also play important developmental roles in those germ layers (S4 Fig). Although the relative localization of the ligands and their antagonists during A/P patterning is perhaps not surprising based on other comparative studies, the functional data we present, both embryological and molecular, show some unexpected results. While we have not addressed the function of individual Wnt ligands or identified which ligands activate the cWnt pathway, our data reveal close similarities to the regulatory role of cWnt during vertebrate neuraxis A/P patterning that are not shared by invertebrate chordates, and some surprising results for the role of cWnt in posterior specification.

In both vertebrates and S. kowalevskii, the ectoderm is initially fated to become anterior unless exposed to posteriorizing signals from the endomesoderm. Early embryonic manipulation of cWnt signaling, either positively or negatively, results in dramatic and complementary changes to the relative proportions of the major body regions, the proboscis, collar, and trunk (Figs 7 and 10). Over-activation of the pathway or overexpression of Wnt3 ligand results in reduction or loss of the proboscis and anterior collar (Fig 7), and overexpression of secreted antagonists of the cWnt pathway results in an over expansion of the anterior proboscis fates and reduction in the size of the trunk (Fig 10). The severity of anterior truncation from Wnt over-activation was concentration-dependent; a high level of 1-azakenpaullone resulted in loss of both anterior collar and proboscis fates, whereas lower concentrations resulted in loss of only the proboscis markers (Fig 7). This supports a model of a Wnt gradient, specifying the A/P pattern of the anterior embryo as has been proposed in vertebrates [20]. Despite the morphological disparity between vertebrates and hemichordates, cWnt is regulating the same region of the transcriptional network that defines the ectodermal A/P axis in both species. Proboscis/anterior collar markers, like forebrain/midbrain markers, are down-regulated, whereas posterior collar/pharynx/anterior trunk markers, like hindbrain markers, are up-regulated by cWnt signaling. The timing of Wnt activity during this early phase of A/P specification is also very similar between vertebrates and hemichordates. In S. kowalevskii, the onset of zygotic Wnt expression at midblastula (11 h) corresponds to the onset of posteriorizing activity. This was determined by timed treatments with the GSK3β inhibitor 1-azakenpaullone (S6 Fig). Treatments beginning at early blastula, just before the onset of zygotic Wnt expression, have the most severe anterior truncation, and this time period likely represents the onset of A/P patterning. Earlier treatments impact AV patterning and the amount of endomesoderm specified by β-catenin. Incremental delays of onset of exposure result in less severe anterior truncations through gastrulation, suggesting that the anterior ectoderm becomes increasingly refractive to cWnt with time.

A key finding of our work is the subdivision of the ectoderm into three domains regarding their sensitivity to cWnt (Fig 11). Below, we compare these observations with results from other metazoans.

The S. kowalevskii anterior-most ectoderm is defined by a set of transcription factors (six3, rx and, foxQ2-1), whose anterior expression is conserved over a broad phylogenetic range [40]. Early cWnt activation or ligand overexpression leads to anterior marker inhibition and a “proboscis-less” embryo (Figs 7 and 8, S5 and S6 Figs), and overexpression of anteriorly localized Wnt antagonists results in expansion of this anterior territory (Fig 10, S7 Fig). This demonstrates that cWnt needs to be actively inhibited to allow correct patterning of anterior territories. However, the repressive effect of cWnt is likely tightly regulated in normal development in the anterior ectoderm to produce graded Wnt levels and correctly pattern proboscis and collar fates. qPCR data indicates that the most anterior fates are the most strongly inhibited by cWnt activity, with more caudal proboscis and collar markers less strongly down-regulated supportive of a local repressive gradient in the anterior ectoderm (Fig 8B). In addition, loss-of-function data suggest some anterior markers require some cWnt signaling for transcription; knock down of fz5/8 demonstrated that rx expression contracts posteriorly when Wnt signaling is locally repressed in the anterior ectoderm, and anterior clones of β-catenin-deficit cells fail to express rx, but activate ectopically the apical marker FGFsk-1 [48]. Further, more sophisticated experiments would be required to further dissect this issue. Moreover, the initial broad expression of anterior markers in the ectoderm together with the observation that ectodermal cells separated from the endomesoderm at blastula stages develop into anterior-most ectoderm [15] suggests that anterior identity could correspond to a default or intrinsic fate. In comparison with vertebrates, it appears that the hemichordate embryo does not depend on an endomesodermal source of Wnt antagonists for anterior patterning; the anterior ectoderm suffices. The implication of cWnt in anterior suppression has been documented in a wide range of phylogenetically diverse groups of animals from chordates, echinoderms, arthropods, and annelids [20,36,40,41,97]. A similar situation may operate during regeneration in acoels and planarians and for the specification of the aboral pole in cnidarians [37–39,43,88].

We have shown that cWnt is necessary and sufficient for the activation of markers of the midaxial identities in the posterior collar/anterior trunk along with anterior and central class Hox genes (Figs 8 and 9). This territory corresponds transcriptionally to the chordate hindbrain and anterior spinal cord [47]. Although anterior suppression has been broadly demonstrated in bilaterians, a role for cWnt in promotion of more posterior territories by up-regulation of midaxial markers, rather than by repression of anterior ones, comes mainly from vertebrates in which Wnts promote hindbrain and anterior spinal cord fates. We propose that this region of both embryos represents a homologous embryonic midaxial territory and is regulated by a cWnt-dependent regulatory program. Importantly, this conclusion could not have been strongly supported with current knowledge from nonvertebrate deuterosomes. In indirect-developing larval species of echinoids, asteroids, and hemichordates, most of the larval ectoderm is fated to be anterior [34,91,92], without an equivalent midaxial trunk territory. Consequently, the experimental focus in echinoderms has been mainly on the establishment and patterning of the anterior neural territory [34,35]. Nevertheless, in urchin larvae, Wnt5 activates a posterior ectodermal marker in the larval ectodermal [93], suggesting some role of Wnt, even in a body plan without a trunk. Strong similarities have not been found for tunicates, and it is possible that this patterning device has been greatly modified or changed its function. In cephalochordates, however, cWnt activation by Gsk3β inhibitor treatments leads to an expected anterior truncation, but the reported effect is rather moderate, affecting only the most anterior region of the neural tube, and it is thought that most of the axis is patterned independent of cWnt signaling [29]. Our data is thus very similar to vertebrate neural data except that patterning influences the entire ectoderm rather than a localized central nervous system. We thus propose that cWnt has been instrumental in the specification of the anterior axial and midaxial identities at least at the base of the deuterostomes, and that secondary modifications have occurred in echinoderms and invertebrate chordate lineages. Outside of deuterostomes, there is support for a role of cWnt in trunk identity during planarian regeneration [94,95], but most other experimental studies in arthropods deal with posterior growth rather than initial establishment of the A/P axis [18].

We have previously reported that in S. kowalevskii, endomesoderm sends posteriorizing signals to the ectoderm in mid/late blastula stages that would otherwise adopt an anterior identity [15]. Gain- and loss-of-function experiments presented here demonstrate that initiation and early expression of posterior trunk markers such as hox9/10, hox11/13a, c, are insensitive to cWnt signaling. Moreover, targeted knock down of β-catenin by siRNA starting at the 2-, 4-, and 8-cell stages only expands the expression domains of anterior markers down to a region corresponding to the midtrunk, suggesting that the posterior trunk is resistant to anteriorization by cWnt inhibition (Fig 10ii, S7 Fig). These results suggest that posterior ectoderm identity is specified independent of cWnt and that other factors are involved in specification of posterior trunk. In vertebrates, Nodal, bone morphogenetic protein, fibroblast growth factor, and retinoic acid signaling are all involved in this process [96,97] and we are currently investigating the roles of these pathways during the early development of S. kowalevskii.

By combining the results of both gain- and loss-of-function, we tentatively mapped the boundary between cWnt-dependent and cWnt-independent ectodermal domains at the anterior limit of hox9/10 expression (Fig 11). Based on similarities in gene expression, this limit roughly corresponds to the posterior spinal cord in vertebrates. Our results are very similar to what is observed in comparable experiments performed in vertebrates (reviewed in [27,98]) When X. laevis animal caps are cut following RNA injection of xwnt3a and noggin, the forebrain markers XAG-1, xanf-2, and otxA were inhibited, and the hindbrain markers markers en-2 and krox-20 were induced. However, this approach does not activate the spinal cord marker hoxB9 [22,99]. Moreover, when the Wnt pathway is activated in whole X. laevis or chick embryos, it results in posteriorization of the anterior neural plate and expansion of hindbrain markers, whereas repression of cWnt has the reciprocal phenotype; repression of hindbrain fates and expansion of forebrain fates. However, posterior spinal cord markers were not examined in these studies, making it difficult to make firm conclusions about the role of Wnt signaling in posterior neural territories [20,22,23,25,96,99,100]. Our results suggest that the early specification of deuterostome posterior embryonic fates may be more dependent on the activity of other secreted ligands and that cWnt plays a less important role than previously proposed.

We argue that the consideration of the role of cWnt signaling in posteriorization largely conflates several developmentally distinct roles of Wnts during AV and A/P patterning. Three distinct roles of the cWnt pathway in early embryonic axial patterning of metazoans have been proposed to have deep evolutionary origins in animal evolution (reviewed in [10,11,14]), and so far no single species displays all three of these. A clear developmental distinction may help for making coherent hypotheses about the role of this complex pathway in the establishment of metazoan axes.

The role with the earliest onset in development involves the action of β-catenin in the establishment of the AV axis and the specification of the endomesoderm. This role has been described in sea urchins, hemichordates, cnidarians, and nemerteans [15–17,101]. We described elsewhere this conserved function in S. kowalevskii [15]. Importantly, we have shown that endomesoderm specification by β-catenin is essential for subsequent ectoderm A/P patterning. We propose that endomesoderm sends posteriorizing signals to the overlying ectoderm that on its own harbors a “default” anterior fate that is independent of Wnt signaling. These unidentified signals are likely controlling posterior identity and posterior trunk formation, and are responsible for the initial cWnt insensitivity of this part of the embryo (Fig 11).

The second role is dealt with explicitly in this study and involves the role of cWnt in the early establishment of the A/P axis. The role of cWnt in anterior suppression has broad phylogenetic support. In deuterostomes, posteriorization of ectodermal fates by cWnt activity has been firmly established in vertebrate CNS in multiple species [20,23,97], and demonstrated to a limited extent in echinoids [93], and now in hemichordates. Whether this function of cWnt is broadly conserved in bilaterians is still unclear because it has not been formally investigated during embryogenesis in animals in which anterior suppression is evidenced; however, work in planarian and acoel regeneration are supportive of a broader role in posterior specification [37,38,43,94].

Finally, the third role of the cWnt pathway is later in development, following the early crude establishment of the A/P axis. Both arthropods and vertebrates deploy a conserved regulatory network of genes localized in a terminal growth zone mediated by the action of Wnts, which likely represents an ancestral developmental strategy for posterior growth (reviewed in [11,18]). We are currently investigating a later role of Wnts in posterior extension of the trunk in hemichordates. Following the early establishment of A/P polarity, hemichordates undergo an extended period of posterior growth to elongate the trunk. Wnt genes continue to be expressed in the posterior and potentially mediate this morphological extension, which mechanistically may be homologous to posterior growth in arthropods and chordates. If this is supported by further experiments, then early development in S. kowalevskii would be regulated by all three of the proposed conserved roles of cWnt signaling in axial patterning, endomesoderm specification, early establishment of A/P axis, and posterior growth.