Spalt-like 4 promotes posterior neural fates via repression of pou5f3 family members in Xenopus

Nieuwkoop and Eyal-Giladi suggested that development of the amphibian central nervous system arises by ‘activation and transformation’ (Nieuwkoop, 1952; Nieuwkoop et al., 1952a, 1952b; Eyal-Giladi, 1954) whereby neural tissue is induced as an anterior state by the organizer and then posteriorized by additional signals from the mesoderm to specify the anterior-posterior (A-P) pattern of the neural plate. Activation, or neural induction, requires bone morphogenetic protein (Bmp) antagonists, such as Noggin (Lamb et al., 1993), Chordin (Sasai et al., 1994) and Follistatin (Hemmati-Brivanlou et al., 1994), from the organizer to induce the neural fate (Khokha et al., 2005). Indeed, any manipulation that blocks Bmp signaling in ectoderm results in anterior neural fates (Hemmati-Brivanlou and Melton, 1994). Caudalization, or transformation, occurs via signaling by retinoic acid (RA) (Durston et al., 1989; Sive et al., 1990; Ruiz i Altaba and Jessell, 1991; Blumberg et al., 1997; Kolm et al., 1997), Fgf (Cox and Hemmati-Brivanlou, 1995; Kengaku and Okamoto, 1995; Lamb and Harland, 1995; Ribisi et al., 2000; Fletcher et al., 2006) and Wnt/β-catenin (McGrew et al., 1995; Itoh and Sokol, 1997; Domingos et al., 2001; Erter et al., 2001; Kiecker and Niehrs, 2001). Despite the identification of these secreted factors as mediators of A-P neural patterning, the mechanism by which transduction of these signals results in the adoption of posterior fates remains poorly understood.

Given the interest in axial patterning, a few transcription factors that mediate A-P differentiation have been identified. The homeobox gene gbx2 is a direct target of canonical Wnt signaling and primarily serves to localize the isthmus and induce neural crest (Simeone, 2000; Li et al., 2009). The meis3 gene, which is required for hindbrain and neuronal differentiation, is directly activated by Wnt3a from the dorsal lateral marginal zone (Elkouby et al., 2010,, 2012). The caudal homologs Cdx1 and Cdx4 are direct Wnt targets in the mouse (Prinos et al., 2001; Pilon et al., 2006,, 2007) and have overlapping roles in posterior development of the three germ layers (Isaacs et al., 1998; Faas and Isaacs, 2009; van de Ven et al., 2011). In the neural plate of Xenopus, simultaneous knockdown of Cdx1, Cdx2 and Cdx4 is required to block adoption of the most posterior neural fates (Faas and Isaacs, 2009).

The Spalt-like (Sall) proteins are vertebrate homologs of the Drosophila protein Spalt. The four members of the Sall family of zinc-finger transcription factors in vertebrates contain an N-terminal C2HC zinc-finger domain followed by variable numbers of doublet and triplet C2H2 zinc-finger domains (Sweetman and Münsterberg, 2006; de Celis and Barrio, 2009). Sall1 and Sall4 function as either transcriptional repressors (Lauberth and Rauchman, 2006; Lauberth et al., 2007; Lu et al., 2009; Yang et al., 2012) or activators (Kiefer et al., 2002; Zhang et al., 2006; Yang et al., 2007; Lim et al., 2008). Mutations in human SALL1 and SALL4 cause the autosomal dominant Townes-Brocks and Okihiro syndromes, respectively, both characterized by limb and cognitive defects (Kohlhase et al., 1998,, 2002). Sall4 knockout mice fail to maintain a pluripotent inner cell mass (Sakaki-Yumoto et al., 2006); null embryos lack Pou5f1 (Oct4) expression in the ICM, increase Cdx2 expression and replace epiblast with trophectoderm (Wu et al., 2006; Zhang et al., 2006). Furthermore, knockdown of Sall4 inhibits induction of induced pluripotent stem cells (iPSCs) (Tsubooka et al., 2009).

Previously, sall2, sall3 and sall4 were shown to be expressed during early Xenopus embryogenesis (Hollemann et al., 1996; Onuma et al., 1999; Onai et al., 2004) and, with the exception of sall2 (Onai et al., 2004), were expressed in posterior neural regions. Conditional knockouts of Sall1, Sall2 and Sall4 result in mouse embryos with neural tube closure defects (Böhm et al., 2008), revealing a role for these genes in neural differentiation or morphogenesis. Despite their expression in posterior neural regions of vertebrate embryos, a role for the Sall genes in caudalization has not been elucidated.

Here we describe the results of an expression screen designed to discover targets of canonical Wnt signaling that determine neural posteriorization in Xenopus. This screen identified sall1 and sall4 as targets in neuralized tissue. We show that sall4 is required for caudalization and, importantly, spinal cord differentiation of neural tissue. Finally, we show that sall4 represses the stem cell factor pou5f3 to release cells from an undifferentiated state.

We used the inducible β-catenin analog TVGR (TCF/LEF DNA-binding domain fused to both the VP-16 transactivation domain and growth hormone receptor) to mimic a posteriorizing Wnt signal (Darken and Wilson, 2001). Having confirmed that this treatment effectively activates Wnt signaling in response to DEX using ventral vegetal injections (supplementary material Fig. S1A), we tested the activity of TVGR in posteriorizing neural tissue using ectodermal explants treated at later stages. Animal caps overexpressing Noggin expressed the anterior neural marker otx2 but not epidermal keratin, demonstrating that the explants had adopted a neural fate. By contrast, neuralized caps, which were injected with TVGR and induced with DEX, expressed the posterior markers krox20 and hoxb9. Ethanol vehicle did activate the hindbrain marker krox20, but DEX was required to induce spinal cord fates as assayed by hoxb9 expression (supplementary material Fig. S1B). Consistent with these results, activation of TVGR in neuralized animal caps induced convergent extension-like morphogenesis consistent with differentiation into spinal cord (Elul et al., 1997) (supplementary material Fig. S1C).

Next, we validated the use of activated TVGR in neuralized animal caps to enrich for transcriptional targets of Wnt signaling (supplementary material Fig. S2A). Treatment with the translational inhibitor cycloheximide (CHX) did not prevent activation of the direct target meis3 (Elkouby et al., 2010) but did block the indirect target hoxb9 (Domingos et al., 2001) (supplementary material Fig. S2B). Thus, these conditions induce neural tissue and posteriorize it via Wnt activation.

To screen for posterior neural genes, we harvested total RNA from animal caps treated with noggin alone (anterior neural), neuralized caps with activated TVGR (posteriorized neural), and neuralized animal caps treated with or without CHX prior to TVGR activation (enriched target sample). The RNAs from these samples were used for Illumina sequencing. The resulting reads were mapped to a collection of non-redundant full-length Xenopus laevis cDNA sequences (Xenopus Gene Collection, http://xgc.nci.nih.gov). By comparing read quantities between anterior neural tissue and tissue treated to enrich for direct targets, we found over 200 genes with expression that was increased greater than 2-fold (supplementary material Table S2). Importantly, the set included the direct targets meis3 (Elkouby et al., 2010) and cdx2 (Wang and Shashikant, 2007). To determine whether the expression of these genes was consistent with a posteriorizing Wnt signal, we stained Xenopus tropicalis embryos by in situ hybridization to identify candidates expressed in the posterior neurectoderm (Fig. 1A). Several candidates were expressed in dorsal tissues of gastrula embryos and posteriorly in early and mid-neurula stage embryos, consistent with the expression domains of known Wnt targets. Of particular note, the transcription factors spalt-like 1 (sall1) and sall4 showed strong expression in posterior neurectoderm (Fig. 1A).

We confirmed the results of our screen by qPCR (Fig. 1B). Incubation with CHX prior to activation of TVGR in neuralized caps resulted in increased cdx2 (supplementary material Fig. S3A), sall1 (supplementary material Fig. S4A) and sall4 (Fig. 1B) expression. Incubation with CHX alone did result in an increase in sall4 expression, but this was not statistically different from caps treated with noggin alone. However, injection of fgf8a RNA [a posteriorizing spliceform of fgf8 (Fletcher et al., 2006)] was sufficient to significantly induce sall4 expression in neuralized animal caps (Fig. 1C).

The activation of sall4 by TVGR in the presence of CHX prompted us to examine whether β-catenin binds to the genomic locus of sall4. We overexpressed a C-terminal FLAG-tagged version of X. laevis β-catenin and confirmed expression by immunoblotting (supplementary material Fig. S5A). Co-injection of FLAG-tagged β-catenin RNA restored dorsal structures in embryos injected with β-catenin morpholinos (MOs) (Heasman et al., 2000), demonstrating both the specificity and activity of this construct (supplementary material Fig. S2B). Consistent with a previous report (Yost et al., 1996), injection of 500 pg RNA encoding tagged β-catenin did not significantly alter dorsal structures as measured by the dorsoanterior index (Kao and Elinson, 1985) (supplementary material Fig. S2C). The sall4 locus in X. laevis contains four exons and three introns (supplementary material Fig. S2D), with six putative TCF/LEF binding sites (Elkouby et al., 2010; McKendry et al., 1997) within the first intron (supplementary material Fig. S6). Three of these sites are tightly clustered within a 150 bp span at positions +2347, +2387 and +2456 (relative to the predicted transcription start site) and are conserved in X. tropicalis. Using FLAG antibodies for ChIP, this region was found to be significantly enriched compared with a negative control (Xmlc2) region (supplementary material Fig. S2E). A −2.7 kb region upstream of meis3 was used as a positive control for β-catenin binding (Elkouby et al., 2010). Anti-FLAG pulldowns in uninjected control embryos resulted in negligible enrichment of any loci assayed. A 500 bp fragment containing these three TCF/LCF sites was cloned and used in luciferase reporter assays. This fragment was not sufficient to enhance expression upon Wnt activation alone (supplementary material Fig. S2F) but was found to be significantly responsive to Fgf (Fig. 1D). Additional experiments demonstrated that Fgf and Wnt did not result in synergistic activation of this fragment.

Taken together, these results suggest that sall4 is likely to be primarily regulated by the posterior Fgf signal and that Wnt signaling may play a minor or negligible role in its regulation.

During gastrulation, sall4 is expressed throughout the marginal zone and the animal pole (Fig. 2A). At stage 10, sall4 is restricted to the sensorial neurectodermal cells in animal dorsal regions (Fig. 2E). At the onset of neurulation, sall4 continues to be expressed in the sensorial neurectoderm (Fig. 2B,F,G). Neural expression of sall4 in stage 15 (mid-neurula) embryos is in the hindbrain and spinal cord anlage (Fig. 2C,H,I). In later stage neurulae (stage 18), sall4 spreads through the posterior neural tube, hindbrain, developing placodes and epidermis (Fig. 2D,J,K).

Similarly, sall1 is expressed in the dorsal ectoderm and involuting mesoderm during gastrulation (supplementary material Fig. S4B,B′). Expression becomes restricted to the notochord and circumblastoporal collar at the early neurula stage (supplementary material Fig. S4C-C″). Like sall4, sall1 is expressed in the spinal cord anlage at mid- and late neurula stages (supplementary material Fig. S4D-E″).

Given its neural expression, we hypothesized that loss of Sall4 would affect neural patterning. To test this, we knocked down Sall4 function with MOs. Morphant embryos had neural tube closure defects and began to disintegrate at mid-tailbud stages. The closure defect is consistent with defects in neural patterning, so we assayed several markers of neural differentiation. The pan-neural marker sox2 was expressed in the neural plate in uninjected and Sall4 morphants, demonstrating that the dorsal ectoderm of morphants still retained a neural identity (Fig. 3A,B). Conversely, the expression of n-tub, a marker for differentiating neurons, was markedly reduced although still present in the morphants, suggesting that Sall4 is required for the second wave of neurogenesis in the tailbud tadpole (Fig. 3C,D). Another marker for early motor neuron differentiation, nkx6.1, was expressed in the central nervous system of morphants, and neural crest cells were still induced as determined by the expression of snai2 (Fig. 3E,F). Although present, these markers were expressed in a pattern more similar to that of early neurulae, suggesting either a delay or failure of terminal differentiation. Sall4 morphants expressed the dorsal mesoderm marker myoD in a similar pattern to uninjected control embryos, and therefore the neural defects were not secondarily due to a loss of paraxial mesoderm (Fig. 3G,H).

As sall4 was identified in a screen for posterior neural genes, we predicted that Sall4 morphants would lose posterior neural identity. To test for this, we injected Sall4 MOs into one animal dorsal (A/D) cell of 4-cell stage embryos to allow for comparison between injected and uninjected sides. The injected side of embryos showed a posterior shift in expression of the hindbrain markers gbx2 (Fig. 4A,B), mafb (Fig. 4D,E) and pax2 (Fig. 4J,K). Sall4 loss resulted in loss of meis3 rhombomere expression and a reduction of its spinal cord expression domain (Fig. 4J,K). Surprisingly, overexpression of sall4 did not result in a change or shift in any of these markers (Fig. 4C,F,I,L), nor was it sufficient to rescue defects associated with Dkk1 overexpression (supplementary material Fig. S7).

The posterior shift of brain markers observed in Sall4 morphants suggested that knockdown of Sall4 results in an expansion of anterior neural identity at the expense of posterior neural differentiation. Accordingly, otx2 is expanded and krox20 is significantly shifted relative to the control side (Fig. 5A-C). Strikingly, the injected side had a significant reduction in the expression domain of the spinal cord markers hoxb9 (Fig. 5D-F), hoxc10 (Fig. 5G-I) and hoxd10 (Fig. 5J-L). However, Sall4 knockdown does not reduce expression of the Wnt target cdx2 to the same extent (supplementary material Fig. S3B,C).

The failure of Sall4 morphants to induce posterior neural identity suggested that the caudal tissue remained in an undifferentiated state. In mouse embryos, Sall4 positively regulates the stem cell factor Pou5f1 (Oct4) to maintain pluripotency (Zhang et al., 2006). One explanation for our results is that Sall4 negatively regulates the Pou5f1 homologs in neural tissue. In Xenopus, there are three class 5 Pou-domain genes that show similar sequence and ancient synteny to mammalian Pou5f1 (Morrison and Brickman, 2006). However, eutherian mammals and frogs retain different copies of the locus from the last tetrapod whole-genome duplication, and their Pou5 genes are not the simple orthologs of Pou5f1. Here, we use the term pou5f3 (as used by Xenbase.org, zfin.org) (Morrison and Brickman, 2006; Frankenberg et al., 2010).

If Sall4 negatively regulates pou5f3, then morphants should increase their expression. Indeed, knockdown of Sall4 in unilateral and bilateral injections resulted in ectopic expression of pou5f3.2 (oct25) (Fig. 6A-C), pou5f3.3 (oct60) (Fig. 6D-F) and pou5f3.1 (oct91) (Fig. 6G-I). Accordingly, the increase in expression of pou5f3.2 and pou5f3.1 was greatest in the neural tube, where sall4 is normally expressed. pou5f3 expression in Sall4 morphants relative to control embryos was quantified by qPCR and displayed a significant increase in all three pou5f3 genes (Fig. 6J-L). Co-injection of X. tropicalis sall4 RNA that is not targeted by the Sall4 MO resulted in a partial rescue of the pou5f3.2 expression level and a full rescue of the pou5f3.3 and pou5f3.1 expression levels.

Next, we asked whether ectopic pou5f3 expression is sufficient to block posterior neural differentiation by injecting RNA for the three pou5f3 genes unilaterally into embryos and assaying A-P neural gene expression. Neural plate cells expressing ectopic pou5f3 (as traced by β-galactosidase) had altered otx2 expression and failed to express krox20 (Fig. 7A,B), hoxb9 (Fig. 7C,D) and hoxc10 (Fig. 7E,F). This loss in A-P neural marker expression cannot be attributed to a loss of neural identity as the pou5f3-injected side of embryos broadly expresses sox2 (Fig. 7G,H).

The observed pou5f3 increase following knockdown of Sall4 suggested a mechanism for the loss of posterior neural identity whereby the ectopic pou5f3 expression prevents differentiation of neural tissue into spinal cord. We reasoned that knocking down pou5f3 in Sall4 morphants would restore posterior neural identity. To this end, we co-injected MOs targeting the three pou5f3 homologs (Morrison and Brickman, 2006; Livigni et al., 2013) along with Sall4 MOs. Consistent with the results described above, knockdown of Sall4 resulted in loss of posterior hoxb9 (Fig. 8A,E), hoxc10 (Fig. 8B,F) and hoxd10 (Fig. 8C,G) but not in a loss of pan-neural sox2 (Fig. 8D,H). Co-injection of the Pou5f3 MOs with Sall4 MOs restored the spinal cord marker expression lost by Sall4 knockdown alone (Fig. 8I-K). Knockdown of Pou5f3 in Sall4 morphants did not restore krox20 stripe expression, consistent with previous work showing that Pou5f3 MOs inhibit krox20 expression (Morrison and Brickman, 2006). Although reduced, sox2 was expressed in the neural plate of Pou5f3 morphants (Fig. 8N) and Sall4-Pou5f3 double morphants (Fig. 8L). Finally, measuring the Hox gene expression domains of Sall4 and Sall4-Pou5f3 morphants revealed a significant rescue of all three spinal cord markers (Fig. 8M).

Our results demonstrate that posteriorizing factors induce Sall4 expression, which represses pou5f3, thereby allowing posterior neural differentiation. Fgf and RA signaling also posteriorize the neural plate. Therefore, we tested whether repression of pou5f3 via Sall4 is required for both Fgf- and RA-induced caudalization. We treated embryos with either fgf8a RNA or incubation in RA. Again, Sall4 knockdown resulted in loss of hoxb9 (Fig. 9A,B) without major alterations to sox2 (Fig. 9E,F). Overexpression of fgf8a in the dorsal ectoderm resulted in expansion of sox2 and hoxb9, a lateral expansion of krox20, and repression of otx2 (Fig. 9C) (Fletcher et al., 2006). These expansions are due to the long-range effects of overexpressing the secreted Fgf ligand. However, overexpressing fgf8a in Sall4 morphants still resulted in otx2 (brain) repression, but hoxb9 (spinal cord) was lost (Fig. 9D). krox20 expression in rhombomere 5 was severely reduced in the Sall4 morphants despite fgf8a overexpression, whereas rhombomere 3 expression remained expanded, probably owing to the specific posteriorizing effects of Sall4. Morphants typically had a posterior shift and reduction in rhombomere 5 krox20 expression, whereas expression in rhombomere 3 was shifted but not reduced (Fig. 5B).

Increasing RA signaling results in severe loss of anterior neural tissue and expansion of posterior identities (Durston et al., 1989; Sive et al., 1990; Ruiz i Altaba and Jessell, 1991; Blumberg et al., 1997; Shiotsugu et al., 2004). To test whether Sall4 is required for posteriorization via RA, we treated control embryos and Sall4 morphants with all-trans retinoic acid (ATRA). Uninjected control embryos treated with 1 μM ATRA lacked otx2 and krox20 but sustained hoxb9 expression (Fig. 9I). However, 1 μM ATRA treatment of Sall4 morphant embryos repressed otx2 and krox20 but also failed to induce the caudally expressed marker hoxb9 (Fig. 9J). The reduction of these markers was not due to a loss of neural tissue as sox2 expression was similar between control embryos, embryos treated with ATRA, and Sall4 morphant embryos treated with ATRA (Fig. 9E,K,L).

Wnt, Fgf and RA signaling are caudalizing factors required for posteriorization of the neural plate. However, the transcription factors identified as mediating the patterning signals from these pathways have largely been restricted to those specifying midbrain and hindbrain fates. In this study, we identify sall4 as a posteriorizing factor target required for spinal cord differentiation. The primary role of Sall4 in neural patterning is to repress pou5f3 (oct4). This repression is necessary for spinal cord differentiation; Sall4 knockdown (Fig. 5E,H,K), as well as pou5f3 overexpression (Fig. 7D,F), results in loss of spinal cord fate. Furthermore, the posterior defects in Sall4 morphants can be rescued via pou5f3 knockdown (Fig. 8I-K). We suggest that repression of pou5f3 via Sall4 provides a permissive environment allowing cells in the neural plate to respond to instructive signals from Fgf, RA and Wnt. This model fits with the observation that overexpression of sall4 did not result in a perturbation of A-P hindbrain marker expression (Fig. 4). If the main role of Sall4 in neural patterning is to repress pou5f3, then overexpression is unlikely to have a significant effect on otherwise normal embryos. Further, this model predicts that Sall4 would not rescue a Wnt loss-of-function phenotype since it is functioning as a permissive and not as an instructive signal. Another prediction is that Sall4 is required for adoption of posterior fates by multiple posteriorizing signals. Therefore, an increase in pou5f3 expression after Sall4 knockdown would inhibit differentiation induced by other caudalizing factors. Indeed, we found Sall4 knockdown prevented induction of hoxb9 by Fgf or RA (Fig. 9D,J).

Our findings build upon previously described mechanisms of posterior neural patterning. Wnt activates cdx1 (Prinos et al., 2001; Pilon et al., 2007) and, in frogs, Cdx1 represses pou5f3 gene expression at the onset of gastrulation (Rousso et al., 2011). However, knockdown of Cdx1 does not result in a loss of spinal cord differentiation, and combinatorial knockdown of Cdx1/2/4 is required before hoxb9 and hoxc10 are reduced (Faas and Isaacs, 2009). There is, however, a dramatic loss of hoxb9, hoxc10 and hoxd10 in Sall4 morphants. In the absence of Sall4, pou5f3 expression remains high, resulting in neural cells being unable to commit to a posterior neural fate and differentiate into spinal cord. Several studies have shown that Cdx factors regulate posterior Hox gene expression in vertebrates (Isaacs et al., 1998; van den Akker et al., 2002; Gaunt et al., 2004,, 2008). Therefore, Wnt acts as an instructive signal through the activation of Cdx genes to induce posterior Hox genes and thereby transform the neural precursors into a posterior fate. Here, we find that sall4 represses pou5f3, providing a parallel, permissive signal for posterior Hox gene expression. Wnt still signals in the posterior neural regions of Sall4 morphants, activating Cdx genes (supplementary material Fig. S3B,C), but the prolonged expression of pou5f3 prevents Hox gene expression. Conversely, it is likely that sall4 is still expressed in Cdx morphants, priming the neural plate to respond to other instructive signals. This could explain why knockdown of individual Cdx homologs results in unexpectedly mild phenotypes.

Our work found sall4 to be activated by Fgf signaling in the neurectoderm. However, our finding that a 500 bp fragment in the first intron of sall4 is enriched in β-catenin ChIP is consistent with it being a Wnt target (supplementary material Fig. S2E). However, this region does not mediate a Wnt-induced signal. Interestingly, we found that this region does show responsiveness to Fgf signaling. Taken together, these experiments show that Fgf is the primary posteriorizing signal that regulates sall4 expression and that Wnt either plays a minor role or does not regulate sall4 during early neural patterning.

The broad expression of sall4 at early neurula stages (Fig. 2B) and later in limbs during Xenopus development and regeneration (Neff et al., 2011) suggests regulation through different enhancers, each responsible for discrete expression domains. This is the case with the neural expression of Sox2 in the chick, which is regulated by five different enhancers, each responsible for a portion of the full expression domain (Uchikawa et al., 2003). Fgf signaling is sufficient to posteriorize neurectoderm (Kengaku and Okamoto, 1995; Lamb and Harland, 1995; Christen and Slack, 1997; Fletcher et al., 2006), and we found that this activity requires Sall4. Therefore, it is possible that Fgf and Wnt signaling converge on other, as yet unidentified, enhancers to regulate sall4 expression. Indeed, Fgf and Wnt signaling converge on one enhancer in the chick sox2 gene to mediate the most posterior expression of sox2 in the neural plate (Takemoto et al., 2006). Likewise, Wnt and Fgf response elements in the enhancers of pax3 and zic genes cooperatively regulate their expression (Garnett et al., 2012), and both pathways mediate expression of these genes at the neural plate border (Monsoro-Burq et al., 2005).

In amphibians, caudalization of the neural plate via Fgf and canonical Wnt signaling induces undifferentiated neural precursors to commit to posterior fates. This induction requires repression of stem cell factors and the activation of differentiation factors. pou5f3 (oct4) genes are first expressed animally in cleavage stages and throughout the mesoderm and ectoderm of amphibian gastrulae (Frank and Harland, 1992; Morrison and Brickman, 2006). Knockdown of Pou91 (Pou5f3.1), Pou60 (Pou5f3.3) and Pou25 (Pou5f3.2) results in precocious cell fate commitment in the three germ layers (Morrison and Brickman, 2006; Snir et al., 2006). Accordingly, pou5f3 overexpression prolongs the undifferentiated state (Morrison and Brickman, 2006; Archer et al., 2011). Our results suggest that pou5f3 expression must be downregulated in the neurectoderm to allow for cells to respond to instructive Wnt/Fgf/RA signals and commit to posterior fates.

Several studies have demonstrated the role for the pou5f3 genes in maintaining pluripotency in Xenopus. In the early embryo, Oct25 (Pou5f3.2) and Oct60 (Pou5f3.3) were found to antagonize VegT and Wnt/β-catenin signaling to prevent precocious germ layer fates (Cao et al., 2007) and overexpression of Oct25 activates Xvent-2B, resulting in a failure of neurectoderm to differentiate (Cao et al., 2004). Further, the histone methyltransferase Suv4-20h has been demonstrated to directly repress oct25 to allow for neural differentiation in Xenopus eye development (Nicetto et al., 2013). These studies all support a conserved role of pou5f3 genes in pluripotency (Morrison and Brickman, 2006; Cao et al., 2007). Our results are consistent with the model; we find that the ectopic expression of pou5f3 following knockdown of Sall4 results in the neurectoderm failing to differentiate in response to transforming signals.

Injection of pou5f3 RNA results in more severe anterior defects than does Sall4 knockdown. This is likely to be due to higher levels of pou5f3 expression following RNA injection (Fig. 7). Indeed, since the Pou5f3 family inhibits Fgf signaling (Cao et al., 2006; Snir et al., 2006), the disruption of krox20 expression in the pou5f3-injected embryos is likely to be due to ectopic Pou5f3 inhibiting hindbrain patterning mediated by Fgf from the isthmus.

The class 5 Pou-domain factors play a conserved role in maintaining pluripotency in Xenopus. Here, we show that Sall4 mediates the transition between pluripotency maintenance and differentiation in the neural plate via repression of pou5f3. How Sall4 regulates pou5f3and whether this is a general role for Sall4 or specific to the neurectoderm remains to be elucidated.

X. laevis embryos were obtained (Sive et al., 2010) and staged (Nieuwkoop and Faber, 1967) as described previously. Ectodermal explants (animal caps) were cut using fine watchmaker's forceps from stage 9 embryos and cultured in 0.75× NAM (Sive et al., 2010).

A cDNA clone of X. tropicalis sall4 (CT025472) was identified in a full-length cDNA collection generated from gastrula embryos (Gilchrist et al., 2004). The coding sequence was subcloned into CS-108 (DQ649433.1) with SalI and XhoI using primers (5′-3′): forward, CGATGTCGACGGACCATG-TCGAGGCGAAAGCAGCC; and reverse, ATCGATCCTCGAGTTA-cttatcgtcgtcatccttgtaatcGTTCACCGCAATATTTT. The coding sequence of X. laevis β-catenin was amplified with a FLAG epitope using: forward, GCATGAATTCCCACCATGGCAACTCAAGCAGATCT; and reverse, GCTAGCGGCCGCTTActtatcgtcgtcatccttgtaatcCAAGTCAGTGTCAAA-CCAGG; it was then subcloned into CS-108 with EcoRI and NotI. Lowercase sequence delineates the FLAG epitope and underlined sequences are restriction sites. X. laevis sall4 was PCR amplified with primers: forward, CTTGGTGCGCACTTATCTCA; and reverse, GCCTCAGATTGTGTGG-GACT; it was then cloned into pCR TOPO II (Invitrogen) for the generation of antisense RNA probes.

Capped RNAs were synthesized using mMessage mMachine (Ambion). sall4 CS-108, fgf8a CS-108, noggin CS-108 and β-catenin CS-108 were linearized with AscI and transcribed with SP6 RNA polymerase. The pou5f3 plasmids (a gift from Joshua Brickman, University of Copenhagen), TVGR (Darken and Wilson, 2001) and nuclear β-galactosidase CS2+ were linearized with NotI and transcribed with SP6. All RNAs were injected in 5 or 10 nl bursts along with GFP and lacZ RNAs, which served as tracers.

The Sall4 MO oligonucleotide (5′-GCCAATTATTCCCTTTCTCCACC-AC-3′; Gene Tools) and the Pou5f3 MOs (a gift from Joshua Brickman) (Livigni et al., 2013; Morrison and Brickman, 2006) were injected in 5 or 10 nl along with fluoresceinated control MO (Gene Tools) to serve as a tracer.

noggin (10 pg) and TVGR (4 pg), an inducible Wnt agonist (Darken and Wilson, 2001), RNAs were injected animally into both blastomeres of 2-cell embryos (Fig. 1A). At stage 9, animal caps were cultured with or without 10 μM dexamethasone (DEX) (Sigma) to activate Wnt signaling (Darken and Wilson, 2001). To block translation, caps were pre-treated with 5 μM cycloheximide (CHX) (Sigma) for 1.5 h prior to DEX addition (Obrig et al., 1971). Animal caps were cultured until stage 15 equivalent and total RNA was harvested using Trizol (Invitrogen).

Embryos were stained by in situ hybridization as described (Harland, 1991). β-galactosidase staining was as described (Fletcher et al., 2006). Embryos for sectioning were mounted in a PBS solution containing 20% sucrose, 30% BSA and 4.9% gelatin, and fixed with 1.5% glutaraldehyde. Embedded embryos were sectioned on a Pelco 101 vibratome.

RNA was isolated from whole embryos or animal caps using Trizol and 1 μg total RNA was reverse transcribed with either MMLV reverse transcriptase (Promega) for semi-quantitative PCR or iScript (Bio-Rad) for quantitative PCR (qPCR). Semi-quantitative PCRs included [³²P]dCTP (PerkinElmer) in the reaction and were analyzed during the log phase of amplification. qPCR reactions were amplified on a CFX96 light cycler (Bio-Rad). ornithine decarboxylase (odc) and elongation factor 1a1 (eef1a1) were used for internal controls. All primers annealed at 60°C and are listed in supplementary material Table S1.

RNA-seq was performed as described (Dichmann and Harland, 2012). Single-end 76-bp reads were sequenced on an Illumina Genome Analyzer II. All reads were mapped to an index created from a collection of full-length X. laevis mRNA sequences (NCBI, http://xgc.nci.nih.gov) using TOPHAT and BOWTIE (Langmead et al., 2009; Trapnell et al., 2009). Analysis of transcript abundance employed CUFFDIFF (Trapnell et al., 2010).

FLAG-β-catenin RNA-injected embryos were prepared for ChIP as described (Blythe et al., 2009). Chromatin shearing used a Branson Model 450 digital sonifier with a Model 102C probe for 24 ten-second bursts set at 30% amplitude. ChIP DNA was quantified with SYBR Green PCR mix (Bio-Rad) on a CFX96 light cycler (Bio-Rad). Enrichment was calculated by comparing the percentage input among ChIP samples. Uninjected embryos served as a control for non-specific binding. Xmlc2 (Blythe et al., 2009) and meis3 (Elkouby et al., 2010) served as negative and positive controls, respectively, for β-catenin binding.

A 500 bp fragment containing three putative TCF/LEF sites in sall4 intron 1 (Scaffold 1115: 234, 269-234, 644) was cloned into the pGL4.23 luciferase reporter (Promega) with SacI and XhoI (NEB). Each of the three sites was mutagenized using Pfx polymerase (Invitrogen) according to the manufacturer's instructions. HEK293 cells were transfected with 0.1 μg each of pGL4.23 and pLR-CMV (Promega) and treated with 0.1 μg mouse Fgf or 50 μM BIO (Cayman). Relative luciferase units were measured on a Turner Design TD-20/20 luminometer using the Dual Luciferase Assay Kit (Promega).

We thank Joshua Brickman for the Pou5f3 MOs and plasmids; Elena Casey for the pou91 clone; Rakhi Gupta and Julie Baker for technical assistance with ChIP; and Darwin Dichman and Sofia Medina-Ruiz for computational assistance.