Segregation of brain and organizer precursors is differentially regulated by Nodal signaling at blastula stage

When the dorsal lip of an amphibian early gastrula is grafted into the ventral side of a host, a secondary embryo develops with complete anterior-posterior and dorsal-ventral axis (Spemann and Mangold, 1924; De Robertis and Kuroda, 2004). Because of these properties, the dorsal lip region is known as the gastrula organizer (GO). The GO is preceded by an earlier dorsal signaling center, the blastula Chordin- and Noggin-expressing (BCNE) center, which expresses Chordin.1 (Chrd.1) and Noggin, two BMP antagonists proposed to initiate anterior neural induction directly in the BCNE (De Robertis and Kuroda, 2004). This center encompasses the animal/marginal dorsal cells of the blastula and gives rise to the forebrain and most of the midbrain and hindbrain, as well as to all endomesodermal tissues derived from the GO. Although chrd.1 transcripts are distributed throughout the entire BCNE, they are restricted during gastrulation to the endomesodermal descendant of this center (GO and its derivatives) but are absent from the neuroectodermal descendant (the presumptive brain) (Kuroda et al., 2004). It was proposed that dorsal accumulation of maternal nuclear β-catenin (nβ-cat) triggers chrd.1 expression throughout the BCNE, while Nodal signaling is only required later to maintain chrd.1 in the GO (Wessely et al., 2001). The siamois-related homeobox genes (sia) are directly activated by the dorsal Wnt/nβ-cat cascade. They encode the first transcription factors expressed in the BCNE and activate chrd.1 expression by directly binding to its promoter (Ishibashi et al., 2008; Reid et al., 2012). Thus, it appeared that only dorsal nβ-cat signaling initiates brain and GO development through the establishment of the BCNE, while Nodal would be required later for the maintenance of the GO and its descendants. These findings suggested that the BCNE behaves as a homogeneous cell population induced by dorsal nβ-cat and segregates later, during gastrulation, into brain and GO.

Accumulation of transcripts encoding Xenopus Nodal-related endomesodermal inducers (Xnrs) in the Nieuwkoop center (NC), located in the vegetal dorsal cells, requires the cooperative action of VegT and nβ-cat (Takahashi et al., 2000). Nodal activity can be experimentally blocked by the C-terminal fragment of Cerberus protein, known as Cerberus-short (Cer-S), which specifically binds to and antagonizes Xnrs (Bouwmeester et al., 1996; Piccolo et al., 1999; Takahashi et al., 2000). After blocking mesoderm induction with cer-S mRNA (hereafter, cer-S), embryos still develop head structures, including brain tissue and a cyclopic eye, express the pan-neural marker sox2 and forebrain, midbrain, and hindbrain markers (Wessely et al., 2001). These findings indicated that anterior neural tissue can be specified in the absence of mesoderm, lending support to the idea that anterior neural specification is initiated in the BCNE due to the transient expression of neural inducers in the presumptive brain territory (De Robertis and Kuroda, 2004).

In this work, we aimed to determine if the BCNE behaves as a homogeneous or heterogeneous cell population in response to Nodal. To this end, we inhibited the Nodal pathway with cer-S or with a dominant-negative form of FoxH1, a transcription factor with Forkhead domain which binds Smad2 and Smad4, the transducers of Nodal signaling (Chen et al., 1997; Watanabe and Whitman, 1999; Hill, 2001). We found that the BCNE is functionally and topologically compartmentalized, as revealed by a differential response of distinct cell subpopulations to the blockade of Nodal. We demonstrate that Nodal is already necessary as early as at blastula stage for restricting a subpopulation of brain precursors while favoring a subpopulation of chordal organizer precursors. We also found that during gastrulation, Nodal is required for maintaining the axial mesoderm (AM), with the chordal mesoderm (CM) being the subpopulation with highest sensitivity to Nodal depletion. Finally, we compare the requirement of Nodal signaling in the segregation of dorsal territories in Xenopus with those previously observed in other vertebrate models.

We blocked Nodal by injecting cer-S and analyzed at s9 by in situ hybridization (ISH) if this could result in spatial changes of the BCNE marker chrd.1 and its upstream regulator, the direct Wnt/nβ-cat target gene sia1 (Ishibashi et al., 2008; Reid et al., 2012). Notably, chrd.1 expression decreased in the marginal region of the BCNE (red asterisk, Fig. 1B) in comparison to control siblings (Fig. 1A), but persisted in the animal region (green asterisk, Fig. 1B) (Table 1). This suggests that the BCNE is composed of two chrd.1 subdomains, regarding their response to Nodal blockade. However, the domain of the up-stream regulator sia1 was not reduced after cer-S injection at any place at s9 (Fig. 1D,E, Table 1). Control experiments showed that cer-S significantly inhibited the expression of mix1, a direct target of Nodal signaling (Charney et al., 2017), both at s9 and s10 (Fig. S1A,B) and also, that of the paraxial mesoderm marker myod1 at neurula stage (Fig. S1C,D). Our results indicate that Nodal is required for chrd.1 expression in a subpopulation of BCNE cells, regardless of sia1.

Intriguingly, RT-qPCR analysis of whole embryos at s9 showed that despite that mix1 was significantly downregulated after cer-S injection, as expected (Fig. S1A,B), chrd.1 transcripts levels increased (Fig. 1C). Since this approach reflects a global measure of chrd.1 mRNA levels and does not distinguish spatial perturbations of chrd.1 expression within the BCNE, it is plausible that an increase in chrd.1 transcripts levels in a subpopulation of cells masked the downregulation in another subpopulation. To corroborate if there are different chrd.1 subdomains in the BCNE regarding Nodal regulation, we unilaterally injected foxh1-SID mRNA (hereafter, foxh1-SID), which encodes a dominant inhibitor of Smad2 activity and blocks Nodal signaling in a cell-autonomous manner (Chen et al., 1997; Müller et al., 2000). By comparing the injected- and non-injected sides at s9, we found that the animal expression of chrd.1 persisted (green asterisk, Fig. 1F,G) but was suppressed in the marginal subdomain of the BCNE (red asterisk, Fig. 1F,G; Table 2). Control experiments showed that foxh1-SID suppressed the pan-mesodermal marker tbxt on the injected side, as expected (Fig. S1E,G).

These results confirm that the BCNE is already compartmentalized into two chrd.1 subdomains: 1) an animal subdomain, which persists after blocking Nodal, and 2) a marginal subdomain, which depends on Nodal. This finding is in contrast to a previous proposal which stated that Nodal is not necessary for chrd.1 expression in the BCNE and it is only required later for its maintenance in the GO (Wessely et al., 2001).

To corroborate that during s9 chrd.1 transcripts are indeed found in the animal region fated to become neural ectoderm, we compared single ISH of chrd.1 in whole and hemisected embryos at s9 with double ISH of chrd.1 combined with hes4 at s9 and s10 (Fig. 2). At these stages, the strongest expression of hes4 is found throughout the animal hemisphere (Aguirre et al., 2013). At s9, chrd.1 is expressed in a large dorsal domain encompassing cells in the dorsal blastocoel roof, sometimes almost reaching the animal pole, well invading the hes4 domain, as well as in dorsal marginal zone cells (Fig. 2A–D). In contrast, at the onset of gastrulation (s10), chrd.1 expression does not overlap the hes4 domain in the animal hemisphere, being restricted to the dorsal marginal zone (Fig. 2D). Therefore, at s9, chrd.1 is expressed in derivatives of both the A1 and B1 blastomeres of the 32-cell stage (Fig. 2E), which are known to contribute mainly to the neural plate and the GO, respectively (Bauer et al., 1994). This large expression of chrd.1 at s9 encompassing both animal and marginal cells that defines a BCNE region is consistent with previous findings (De Robertis and Kuroda, 2004; Kurth et al., 2005; Harata et al., 2019).

We next analyzed the consequences of blocking Nodal on chrd.1 expression during gastrulation, when it is not expressed anymore in brain precursors but persists in the GO and in its AM descendants. At early gastrula, the presumptive prechordal mesoderm (PM), which is the first to involute, occupies a relatively more internal position than the pre-involuted, presumptive CM, and both express chrd.1 (Zorn et al., 1999; Kaneda and Motoki, 2012). At this stage, we found a less dense ISH staining for chrd.1 in cer-S-injected embryos in comparison to uninjected control siblings (Fig. 3A–C). Chrd.1 was downregulated or suppressed from more external GO cells, which correspond to pre-involuted prospective CM at the beginning of gastrulation. As these cells became more transparent due to the suppression of chrd.1 staining, we could observe that chrd.1 expression persisted in a cloud below them (Fig. 3B,C, Table 1). This deeper expression, more refractory to cer-S, corresponds to the presumptive, involuted PM at this stage. Unilateral injection of foxh1-SID (0.25 ng to 1 ng) allowed a better comparison of the behavior of these subpopulations between the injected and the non-injected sides. Chrd.1 persisted (green asterisk, Fig. 3G,I) or the domain was even expanded (light blue asterisk, Fig. 2H) in the deeper, involuted cells (Table 2), while in the pre-involuted, more external cells, chrd.1 was suppressed (red asterisk, Fig. 3G–I, Table 2). To corroborate the contribution of chrd+ cells to PM and CM, we analyzed the expression of this marker at s13. Normally, at this stage, all chrd+ cells have been internalized, and PM and CM have completely segregated. This allows the distinction of two chrd+ subdomains in the AM, with clear distinct shapes: a) an anterior fan-like subdomain, characterized by an active migration behavior, corresponding to the PM (green arrow, Fig. 3D); b) a posterior subdomain, corresponding to the notochordal cells (red arrow, Fig. 3D), which ultimately form a rod by convergent-extension movements (Murgan et al., 2014; Kwan and Kirschner, 2003). Cer-S-injected embryos showed that most chrd.1 expression consisted of the fan-shaped subdomain (green arrow, Fig. 3E,F). In contrast, the posterior subdomain was lost or reduced (red arrows, Fig. 3E,F), sometimes with chrd.1+ cells arrested on the blastopore, unable to involute (yellow asterisk, Fig. 3E) (Table 1).

It was previously shown that not all the PM express chrd.1 in Xenopus (Yamaguti et al., 2005). These authors showed that the PM consists of two cell populations, a more anterior one (APM) expressing the homeodomain transcription factor Goosecoid (Gsc), and a more posterior one (PPM) expressing chrd.1 (Yamaguti et al., 2005). The same occurs in the mouse (Anderson et al., 2002). Therefore, we analyzed the consequences of cer-S injection on the spatial expression of gsc and of the CM-specific transcription factor Tbxt/Brachyury, which control the characteristic movements of PM and CM cells, respectively, in an antagonistic way (Artinger et al., 1997; Latinkić and Smith, 1999; Kwan and Kirschner, 2003; Luu et al., 2008). Our ISH results show that, at gastrula stages, gsc expression decreased in most cer-S-injected embryos (Fig. 4D,E) (Table 1). We also found through RT-qPCR analysis that gsc was downregulated at the onset of gastrulation after overexpression of cer-S (Fig. 4F). ISH analysis showed that tbxt was affected at different degrees. The expression normally found in the extending notochord (red arrows, Fig. 5A,C,E) was suppressed in almost all injected embryos (Fig. 5B,D,F,G, Table 1). The expression in the GO presumptive mesoderm (dorsal part of the tbxt ring, yellow asterisk, Fig. 5A,C,E) often disappeared or was downregulated (Fig. 5B,D,F, Table 1) and that of the non-GO presumptive mesoderm (white arrows, Fig. 5A,C,E) was sometimes suppressed (Fig. 5B,G, Table 1). As an independent notochordal marker, we also analyzed the spatial expression of not, which encodes a homeodomain transcription factor normally expressed during gastrulation in the extending CM, GO, and in the limit of involution (von Dassow et al., 1993). cer-S injections affected the not pattern in a similar way to tbxt (Fig. 5H–M, Table 1). These observations indicate that, during gastrulation, among the mesodermal descendants, CM cells are the most sensitive to Nodal blockade, which disrupted notochord development almost completely. Although cells with PM characteristics persisted, as shown by ISH of chrd.1, a lower gsc expression was detected at gastrula stages, both by ISH and by RT-qPCR. This might be due to a differential sensitivity among APM and PPM subpopulations to Nodal blockade. In addition, the ventrolateral mesoderm was less sensitive to cer-S than the CM, as shown by a lower penetrance of the downregulation of tbxt in the circumblastoporal ring.

In conclusion, Nodal is required for the onset of chrd.1 expression in the marginal subdomain of the BCNE at the blastula stage. Later, the mesodermal descendants of the BCNE require Nodal for their maintenance during gastrulation, with the CM having the highest requirement.

We have observed that blocking Nodal distinguishes two chrd.1+ subpopulations in the BCNE. Since this center is composed of the brain and GO precursors, and the GO, in turn, segregates into PM and CM, we wondered if blocking Nodal could already discriminate between PM and CM precursors and the neural ectoderm at s9. For this purpose, we compared the expression of the pan-neural marker sox2 and the specific PM (gsc) and CM (not) markers in the BCNE between cer-S-injected embryos and uninjected control siblings at s9. Although it is known that gsc and tbxt are involved in an antagonistic relationship that controls PM and CM segregation during gastrulation (Artinger et al., 1997; Latinkić and Smith, 1999; Kwan and Kirschner, 2003; Luu et al., 2008), tbxt is weakly expressed in the presumptive CM at late blastula. Therefore, we chose not as an alternative spatial marker of CM precursors. While gsc and sox2 expression persisted and even increased in the case of the latter (Fig. 4A–C,G–I, Table 1), not expression in the BCNE subdomain corresponding to the presumptive CM territory decreased in cer-S-injected embryos (Fig. 4H,I, Table 1). Interestingly, RT-qPCR analysis showed that both, chrd.1 and sox2 expression at s9 increased after blocking Nodal (Figs 1C and 4I). These results indicate that the population of brain precursors expressing both, the neural inducer and the neural specification marker in the BCNE was expanded after blocking Nodal. Also, because the animal chrd.1 subdomain was essentially refractory to the blockade of Nodal signaling, while the marginal subdomain was suppressed, as shown by ISH (Fig. 1A,B,F,G), an increase in the level of chrd.1 transcripts in neural precursors might account for the global increase measured by RT-qPCR at s9 (Fig. 1C).

Thus, we propose that BCNE cells are functionally compartmentalized, with presumptive CM cells requiring Nodal for their specification, while neural specification is restricted by Nodal.

The results presented here indicate that Nodal is indeed required to trigger the full expression of the neural inducer chrd.1 in the BCNE (Fig. 6C), in contrast to a previous proposal (Wessely et al., 2001) (Fig. 6A,B). In fact, we found that chrd.1 was abolished by either cer-S or foxh1-SID in the marginal region of this center, whereas sox2 expression increased at s9 in cer-S-injected embryos. These results indicate that the population of brain precursors in the BCNE was expanded after blocking Nodal. Indeed, we notice that the chrd.1 domain in cer-S-injected embryos at s9 (shown in Fig. 3D,E in Wessely et al., 2001) is less extended in the animal-vegetal axis in comparison to control siblings, resembling the results shown in the present work, but the authors interpreted that chrd.1 expression was refractory to the blockade of Nodal. Besides, we found that the PM marker gsc persisted whereas the notochordal marker not was reduced in the BCNE after blocking Nodal. Therefore, mesodermal precursors in the BCNE are differentially regulated by Nodal.

Altogether, these results indicate that the BCNE does not behave as a homogeneous cell population. In fact, our data show that the BCNE is compartmentalized into the precursors of the prospective brain (pre-brain), the PM precursors, and the CM precursors, which can be distinguished earlier than previously thought since they are being differentially regulated by Nodal (Fig. 6C). Therefore, here we present a modified model of the originally proposed by Wessely et al. (2001), concluding that chrd.1 expression is triggered in the whole BCNE by Wnt/nβ-cat through sia (Ishibashi et al., 2008) and has different requirements for Nodal, depending on the BCNE subpopulation. Nodal is necessary to trigger chrd.1 expression in the chordal pre-organizer, whereas it is not required at the prechordal pre-organizer. At the pre-brain subpopulation, chrd.1 is restricted by Nodal (Fig. 6C). Since sia1 expression persisted in the BCNE after cer-S injection and sox2 expression increased, we suggest that Nodal restricts the pre-brain territory downstream of sia (Fig. 6C). Besides, within the presumptive mesodermal subdomain in the BCNE (pre-organizer), Nodal promotes the development of posterior AM derivatives but does not affect the presumptive prechordal subpopulation at this stage (Fig. 6C). Later, Nodal is necessary to maintain both the CM and the PM during gastrulation. The chrd.1, tbxt, and gsc patterns at gastrula stages suggest that, among all mesodermal derivatives, the CM is the most sensitive to the blockade of Nodal in Xenopus. Thus, it appears that CM maintenance during gastrulation requires more input from Nodal than other mesodermal cell types, like the PM.

The subdivision of the BCNE into two chrd.1 subdomains according to Nodal responsiveness shown here is consistent with the observation that, at stage 9, nuclear phosphorylated-SMAD2 (p-SMAD2), a bonafide indicator of active transduction of Nodal signaling, is highly accumulated in the dorsal-marginal and dorsal-vegetal region but not in the animal part of the BCNE area (Schohl and Fagotto, 2002) (Fig. 2G). Moreover, cycloheximide treatment prevented the reduction of the chrd1 domain in early gastrulae, which showed a larger chrd.1 domain very similar to the BCNE domain at stage 9 (Kurth et al., 2005). This indicates that during the blastula-to-gastrula transition, a mechanism dependent on protein synthesis normally represses the animal subdomain of chrd1 expression present in the BCNE, without affecting chrd.1 expression in the GO. This mechanism probably underlies the downregulation of chrd.1 in the neural descendants of the BCNE during gastrulation described by Kuroda et al. (2004). Interestingly, sia1 expression shows a dynamic patterning, first appearing in the animal and marginal region at s8/8.5, then adding a vegetal domain at early s9. By mid-stage 9, expression of sia1 has disappeared from the animal region but persists in the marginal and vegetal regions (Sudou et al., 2012). The combined knockdown of sia1+sia2 completely abolishes chrd.1 expression in the BCNE. Therefore, the downregulation of sia1 that normally takes place in the animal domain might underlie the disappearance of chrd.1 expression from the neural derivatives of the BCNE at the onset of gastrulation. While Nodal and sia converge in the positive regulation of chrd.1 in the marginal subdomain of the BCNE, it is not known if additional pathways, apart from the Wnt/nβ-cat/sia cascade, are necessary for triggering chrd.1 expression in the animal subdomain of the BCNE.

The default-state model of neural specification adopted for Xenopus and mouse embryos poses that the default fate of ectoderm (epiblast) is neural, but it is actively repressed by BMP signaling. This default-state is revealed during neural induction, when the ectoderm is exposed to BMP antagonists, like Chordin and Noggin (Levine and Brivanlou, 2007). In Xenopus, this initial step of neural induction occurs before gastrulation in the BCNE, as the BMP antagonists are directly expressed by neuroectodermal precursors fated to give rise to the forebrain (Kuroda et al., 2004). Like in Xenopus at the blastula stage (this work, Fig. 7A), the first neural tissue induced in mouse expresses the pan-neural marker Sox2 (Levine and Brivanlou, 2007). Moreover, this neural tissue initially has a forebrain character, but it is subsequently posteriorized during gastrulation to form the rest of the central nervous system (CNS) (Levine and Brivanlou, 2007) (Fig. 7I,J). However, unlike Xenopus, Chrd and Noggin are not expressed before gastrulation in the mouse. Chrd transcripts first appear in the GO at the mid-streak stage (Anderson et al., 2002; Kinder et al., 2001) (Fig. 7K). Noggin appears later in the GO, and both genes are expressed in the GO-derived AM (Fig. 7L) (McMahon et al., 1998; Bachiller et al., 2000), but they were not detected in the prospective forebrain, unlike in the Xenopus BCNE (Fig. 7A–C) (Kuroda et al., 2004). Therefore, the default model for anterior neural induction in mice implied that the source of BMP antagonists lies in the GO and its derived tissues (Levine and Brivanlou, 2007), i.e. outside the presumptive forebrain. However, we note that transcripts from a Chordin-like1 gene (Chrdl1) are first detected in the future neural plate at E7.0 in mice (Fig. 7K) and persist there during gastrulation (Fig. 7L) (Coffinier et al., 2001). Notably, Chrdl1 never appears in the node or the primitive streak (PS), thus establishing a complementary pattern in relation to Chrd1 (Fig. 7K,L) (Coffinier et al., 2001). Chrdl1 behaves as a BMP antagonist (Sakuta et al., 2001; Chandra et al., 2006) and mouse Chrdl1 was more potent than Xenopus chrd.1 in the induction of complete secondary axes in frogs (Coffinier et al., 2001). Thus, as early as at E7.0 [when a range of pre-streak to mid-streak embryos can be obtained, according to EMAP eMouse Atlas Project (http://www.emouseatlas.org)] the mouse future neural plate indeed expresses a BMP antagonist of the Chordin family. An ortholog chrdl1 gene was identified in Xenopus, but transcripts were not detected before tailbud stages (Pfirrmann et al., 2014). Therefore, it seems that cells in the presumptive brain territory already express BMP antagonists as early as at late blastula in Xenopus (chrd.1) or around the onset of gastrulation in mouse (Chrdl1), when the anterior AM just begins migrating from the early GO to completely underlie the future forebrain later in gastrulation.

Single knockout mice for Chrd or Noggin undergo normal neural induction and anterior CNS patterning (Bachiller et al., 2000; McMahon et al., 1998). Although double knockouts demonstrated that both are necessary for forebrain maintenance (Bachiller et al., 2000), it was not studied if forebrain specification, which normally takes place around the mid-streak stage (Levine and Brivanlou, 2007), effectively occurred. On the other hand, Chrdl1^−/− mice developed a CNS and survived to adulthood (Blanco-Suarez et al., 2018). Double knockouts of Chrd and Chrdl1 should be obtained to study if both genes are necessary for neural specification.

In Xenopus, chrd.1 expression in the chordal pre-organizer population requires induction by Nodal (this work, Fig. 7A). We notice that in the mid-streak stage mouse, while Nodal is strongly expressed in the posterior-proximal quadrant of the epiblast, where the PS forms and endomesoderm ingression takes place (Varlet et al., 1997) (Fig. 7K), Chrdl1 is oppositely expressed with the highest levels in the anterior quadrant (Coffinier et al., 2001) (Fig. 7K), where anterior neural specification is taking place (Fig. 7I). This pattern resembles the complementary expression of Xenopus chrd.1 and nodal genes in the BCNE/prospective brain and the NC (dorsal endoderm), respectively (Fig. 7A). However, it was not studied if Nodal normally controls mouse Chrdl1.

Remarkably, Nodal^−/− embryos showed an expanded and premature specification of neural ectoderm expressing forebrain markers in the epiblast (including the rostral forebrain regulator Hesx1), and Gsc expression persisted in the future GO region at pre-streak stages (Camus et al., 2006). When examined at streak-stages, Nodal^−/− embryos did not develop a morphologically distinguishable PS, and Gsc expression was always absent. Interestingly, Tbxt expression was also absent, except for some caudal cells in around 10% of mutant embryos. Moreover, some patches of other caudal mesoderm markers were present in 25% of mutant embryos (Conlon et al., 1994). FoxH1^−/− embryos neither expressed Gsc nor Foxa2 at the anterior tip of the PS during gastrulation and failed to form a definitive node and notochord (Yamamoto et al., 2001; Hoodless et al., 2001). Surprisingly, neither Chordin nor Noggin expression was analyzed in Nodal knockouts. However, conditional removal of Smad2 activity from the epiblast or deletion of the proximal epiblast enhancer (PEE) of the Nodal gene (which renders attenuated expression of Nodal in the PS) neither suppressed node formation nor anterior neural specification. However, Gsc expression corresponding to the prechordal plate was lost at the late-streak stage in both mutants, while Chordin and Noggin expression was aberrant and decreased along the axial midline in Smad2 mutants (it was not analyzed in Nodal-PEE mutants) (Vincent et al., 2003). Therefore, Gsc expression in the mouse AM region also appears to display two phases concerning Nodal responsiveness, like those we show here for Xenopus: an earlier one (before PS formation) in the future GO region, which does not depend on Nodal (Camus et al., 2006), and a later one, in the node/prechordal plate, which requires Nodal (Conlon et al., 1994; Yamamoto et al., 2001; Hoodless et al., 2001; Vincent et al., 2003). Overall, these results in mouse resemble our findings after blocking Nodal in Xenopus, indicating that this pathway is required for restricting forebrain induction before gastrulation (Camus et al., 2006), for the maintenance of PM cells during gastrulation, and for the development of notochordal cells (Conlon et al., 1994; Yamamoto et al., 2001; Hoodless et al., 2001; Vincent et al., 2003). However, the development of a subset of posterior mesodermal cells does not show a high dependency on Nodal in both, mouse (Conlon et al., 1994) and Xenopus (this work). These observations suggest that in amphibians and mammals, before overt signs of gastrulation, Nodal is not necessary for the initial specification of the prechordal pre-organizer, but as gastrulation progresses, Nodal is required for the maintenance of the GO, the PM, and the notochord.

In conclusion, while in Xenopus, the same gene (chrd.1) is expressed in both pre-brain and pre-organizer territories in the BCNE (Fig. 7A), separate mouse genes encoding BMP antagonists belonging to the Chordin family are expressed in the presumptive neural plate (Chrdl1) and the GO (Chrd1) (Fig. 7K,L). In Xenopus, Nodal differentially regulates chrd.1 among the BCNE cells, being necessary for the initiation of chrd.1 expression in the presumptive CM subpopulation and to restrict that of the pre-brain subpopulation. In mice, it is not known if Nodal analogously restricts Chrdl1 expression in the presumptive brain territory, although their complementary expression patterns suggest so. The SMAD2 conditional knockout indicates that Nodal is also necessary for Chrd1 expression during gastrulation in mice, although it is not known if it is required for its initial expression in the GO or its maintenance since the analysis was performed at headfold stage (Vincent et al., 2003).

In chick embryos, the posterior marginal zone (PMZ) was proposed as the NC equivalent (Joubin and Stern, 2001) (Fig. 7G,H). The Köller's sickle (KS) is a crescent-shaped region of the superficial portion of the PMZ (Stern, 1990) (Fig. 7H). Two cell populations that contribute to the chick GO (Hensen's node) were identified before the onset of gastrulation. One is initially located in the epiblast, just above the anterior face of KS at stage X, and moves anteriorly between stages XI and XIII to the center of the blastoderm. These are known as central cells (CC) (Fig. 7H). The second group, known as posterior cells (PC), is located within the middle layer and is initially associated with the inner face of KS (Fig. 7H). These PC start expressing Gsc before gastrulation and migrate later than the CC, within the middle layer, together with the tip of the PS. When both populations meet at the mid-streak stage, they establish a completely functional GO (Izpisúa-Belmonte et al., 1993; Streit et al., 2000; Joubin and Stern, 2001). As in Xenopus and unlike in mice, Chrd expression in chick embryos begins before the onset of gastrulation (Fig. 7G). Transcripts are initially found in a region just anterior to KS in the epiblast and the underlying cells of the middle layer. Therefore, it is expressed in both populations that contribute to the GO (Streit et al., 1998) (Fig. 7G). A rough fate map of the early blastula/pre-streak chick embryo showed that the prospective forebrain position lies in the epiblast, also immediately anterior to the KS (Stern et al., 2006) (Fig. 7H), suggesting that Chrd is also transiently expressed in the chick forebrain precursors, like in the Xenopus BCNE at equivalent stages (as suggested by Kuroda et al., 2004; Reversade et al., 2005) (Fig. 7G). Ectopic expression of Chrd in the non-neural ectoderm of the area pellucida in gastrulating chick embryos was unable to induce a secondary PS or neural markers. However, when Chrd was ectopically expressed earlier, in the anterior edge of the area pellucida before the onset of gastrulation, a secondary axis with PS, node, and neural ectoderm with the typical horseshoe shape of the anterior neural plate was induced (Streit et al., 1998). This suggests that forebrain precursors are competent to be recruited by Chrd before the onset of gastrulation in chick embryos.

In pre-streak chick embryos (stage XII/XIII), Nodal is expressed in a region confined to the middle two-thirds of KS, while Chrd expression is restricted to the midline region of the epiblast, just rostral to KS (Lawson et al., 2001) (Fig. 7G). Therefore, as in Xenopus, Nodal is expressed in the chick NC equivalent. Nodal alone was insufficient to induce ectopic Chrd in explants of anterior epiblast of pre-streak embryos but could induce it when combined with FGF8. On the other hand, in the absence of FGF8 signaling, Nodal expression remained unaffected in posterior blastoderm explants (containing KS) of pre-streak embryos, while Chrd expression was decreased (Matsui et al., 2008). These results indicate that FGF8 signaling from the nascent hypoblast is necessary for Chrd expression before the onset of avian gastrulation and that Nodal might cooperate but is insufficient to induce it. However, experiments blocking Nodal were not performed to address if it is indeed required for triggering Chrd expression in the posterior epiblast before gastrulation.

In Zebrafish, chrd is readily detected well before gastrulation in the dorsal region, including the future GO region and extending more or less towards the animal pole, depending on the bibliography (Miller-Bertoglio et al., 1997; Sidi et al., 2003; Branam et al., 2010) (Fig. 7D). In contrast to the other vertebrate models discussed here, chrd is additionally expressed during gastrulation in territories beyond the GO, including the prospective brain and other neuroectodermal regions (Miller-Bertoglio et al., 1997; Sidi et al., 2003; Branam et al., 2010) (Fig. 7E,F). Notably, in double mutants for nodal1 and nodal2 (squint/cyclops) or in maternal/zygotic mutants for the essential cofactor for Nodal signaling tdgf1 (MZoep mutants), the chrd domain in the dorsal center of the zebrafish blastula was reduced in size, but strong expression persisted in a considerable subdomain (Gritsman et al., 1999). Also, at blastula stages, tbxt expression was suppressed in the dorsal region but persisted in the remaining presumptive mesodermal ring, while gsc expression was suppressed in these mutants (Feldman et al., 1998; Gritsman et al., 1999). These observations indicate that the population of chrd+ cells in the zebrafish blastula is also heterogeneous in Nodal requirements, as we found in Xenopus (this work). They also show that the CM precursors (as we found in Xenopus) as well as the PM precursors (unlike what we observed in Xenopus) require Nodal signaling for their initial specification (Fig. 7D). As both PM and CM markers were suppressed at blastula stages, the chrd+ cells that remain in these mutants might as well represent future neural cells.

Interestingly, zebrafish has a chordin-like2 (chrdl2) gene which also behaves as a BMP antagonist. The zebrafish ortholog of the chrdl1 gene could not be identified in the genome (Branam et al., 2010). Chrdl2 is strongly expressed throughout the animal hemisphere before gastrulation (Fig. 7D), thus introducing an additional level of BMP antagonism to the dorsal region. Knockdown experiments showed that chrl2 is required together with chrd for dorsal development during the patterning of the embryonic dorsal-ventral axis (Branam et al., 2010).

The rostral forebrain is an evolutionary acquisition of vertebrates related to the appearance of the Hesx1 gene at the beginning of vertebrate evolution (Ermakova et al., 1999; Wilson and Houart, 2004; Ermakova et al., 2007; Bayramov et al., 2016) and derives entirely from the BCNE center (Kuroda et al., 2004). We have previously shown that expression of the rostral forebrain regulator hesx1 persists after blocking Nodal signaling in Xenopus (Murgan et al., 2014). Indeed, the expansion of the Hesx1 domain in Nodal^−/− mouse embryos occurred before gastrulation (Camus et al., 2006). Altogether, our findings and the comparison between different vertebrate models indicate that the establishment of the CNS during the development of vertebrates requires not only a very early induction of the brain territory but also its delimitation. This process involves Nodal signaling in the differential segregation of the cell populations that give rise to the dorsal structures (brain and AM) as early as at the onset of neural induction at the blastula stage.

Albino and wild-type Xenopus laevis embryos were obtained by natural mating or by in vitro fertilization using standard methods (Sive et al., 2010) and staged according to (Nieuwkoop and Faber, 1994). Parental animals were obtained from Nasco (Fort Atkinson, WI, USA). Protocols were approved by the Laboratory Animal Welfare and Research Committee (CICUAL) from Facultad de Medicina-UBA.

To obtain synthetic capped mRNAs, the following plasmids were employed as templates: Xcer-S pCS2+ (gift from Eddy de Robertis, University of California, LA, USA) (Bouwmeester et al., 1996), pCS2 MT foxH1-SID [gift from Uwe Strähle, European Zebrafish Resource Center (EZRC), Karlsruhe, Germany] (Chen et al., 1997), and pCS2-NLS-lacZ (gift from Tomas Pieler, Department of Developmental Biochemistry, University of Goettingen, Germany) (Bellefroid et al., 1996). These plasmids were linearized with NotI and transcribed with SP6 mRNA polymerase with the mMESSAGE mMACHINE SP6 Transcription Kit (Thermo Fisher Scientific, AM1340), following the manufacturer's instructions. Synthetic capped mRNAs were purified with the RNeasy Mini Kit (Qiagen, #74104).

Microinjection, culture, and fixation of embryos were performed as previously described (Murgan et al., 2014). Cer-S mRNA was injected in the vegetal region of the four cells at s3 (0.5 ng/cell). Foxh1-SID mRNA (0.25 to 1 ng) was unilaterally injected at s2. Microinjections included as tracer 30–40 ng/cell of Dextran Oregon Green 488, MW 10000, anionic, lysine fixable (DOG; Thermo Fisher Scientific, D7171) or of Dextran Alexa Fluor 594, MW 10000, anionic, fixable (Thermo Fisher Scientific, D22913); or 0.5 ng/cell of nuc-lacZ mRNA.

Plasmids for obtaining antisense RNA probes for wholemount ISH were linearized and transcribed as follows. Chrd1: pSB59-chrd1 (gift from Eddy de Robertis) (Sasai et al., 1994) was cut with EcoRI and transcribed with T7 RNA polymerase; Gsc: gsc pG500 (gift from Ken Cho, University of California, Irvine, CA, USA) (Cho et al., 1991), XbaI/T3; myod1: pSP73-XmyoD (gift from Cristof Niehrs, Institute of Molecular Biology, Mainz, Germany) (Hopwood et al., 1989), BamHI/SP6; not: pBS-KS-Xnot: HindIII/T7 (gift from David Kimelman, Department of Biochemistry, University of Washington, Seattle, WA, USA) (von Dassow et al., 1993); sia1: pBluescript RN3 Xsia ORF (gift from Patrick Lemaire, Institut de Biologie du Developpement de Marseille, Marseille, France) (Lemaire et al., 1995), HindIII/T7; sox2: pBS sox2 (gift from Yoshiki Sasai, RIKEN Center for Developmental Biology, Kobe, Japan) (Kishi et al., 2000), EcoRI/T7; tbxt: αpSP64T bra (gift from Abraham Fainsod, Faculty of Medicine, The Autism Centre Hebrew University, Jerusalem, Israel) (Smith et al., 1991), SalI/SP6. The preparation of digoxigenin-labeled antisense RNA probes and the wholemount ISH procedure was performed as previously described (Pizard et al., 2004), except that the proteinase K step was omitted. X-gal staining for revealing the nuc-lacZ tracer was performed as previously described (Franco et al., 1999).

Embryos injected with cer-S mRNA and uninjected siblings were allowed to grow until stages 9 or 10. For total RNA extraction, three embryos at s9 or eight embryos at s10 were placed in 1.5 ml tubes. After liquid withdrawal, tubes were placed on ice, and 200 µl (for s9 samples) or 400 µl (for s10 samples) of TRIreagent (Merck, cat. no. 93289) was added. Embryos were resuspended 10 times with a micropipette. Samples were stored at −80°C until RNA extraction, which was performed following the manufacturer's instructions. 1 μg of RNA was treated with DNAse I (Ambion, cat. no. AM2222) and used for first-strand cDNA synthesis, using High Capacity Reverse Transcription Kit with random primers (Applied Biosystems, cat no. 4368814). Amplification was performed in triplicate in an Applied Biosystems 7500 Real-Time PCR System machine using Power Up SYBR Green Master Mix (Applied Biosystems, cat. no. A25472). Melt curves were analyzed to confirm the specificity of PCR products. The efficiency of each PCR amplification was estimated using the slope of a standard curve. Relative gene expression was calculated using Pfaffl′s mathematical model, with Histone H4 expression levels as standard.

The following primers were used for RT-qPCR: chrd.1 F: ACTGCCAGGACTGGATGGT, chrd.1 R: GGCAGGATTTAGAGTTGCTTC (Leibovich et al., 2018); gsc F: TTCACCGATGAACAACTGGA, gsc R: TTCCACTTTTGGGCATTTTC (Leibovich et al., 2018); histone H4 F: GGCAAAGGAGGAAAAGGACTG, histone H4 R: GGTGATGCCCTGGATGTTGT (Cao et al., 2007); mix1 F: CAAAAGCCACCAAGCCCATT, mix1 R: TGCTGAAGGAAACATTGCCC (Sun et al., 2015); sox2 F: GAGGATGGACACTTATGCCCAC, sox2 R: GGACATGCTGTAGGTAGGCGA E.M. (De Robertis http://www.hhmi.ucla.edu/derobertis/).

Cer-S mRNA batches for ISH analysis were tested by the effect on the mesodermal marker myod1 at neurula stage and were used when myod1 expression was strongly attenuated or abolished in injected embryos in comparison to uninjected siblings (Fig. S1C,D; Table 1). Fox1-SID mRNA was tested by the effect on the pan-mesodermal marker tbxt at early gastrula stage by ISH analysis (Fig. S1E–G). For RT-qPCR analysis, cer-S mRNA-injected batches were tested at s9 and s10 by the effect on the expression of mix1, a direct target of Nodal signaling (Charney et al., 2017), and were used when mix1 expression was significantly reduced to less than 50% in relation to uninjected sibling controls (Fig. S1A,B).

The numbers of samples (n) and biological replicates (N) that were analyzed are indicated for each set of experiments in the figures and tables. Biological replicates represent batches of embryos from independent mating pairs, or from different groups of embryos from the same batch in the case of RT-qPCR assays at s9. Statistical tests applied for RT-qPCR analysis are described in the corresponding Materials and Methods section and the figures. Differences were considered significant when P<0.05.

We thank the following colleagues for providing us with DNAs: Ken Cho (gsc), Eddy De Robertis (chrd.1, cer-S), Abraham Fainsod (tbxt), Tomas Pieler (NLS-lacZ/nuc-lacZ), David Kimelman (not), Patrick Lemaire (sia1), Christof Niehrs (myod1), Yoshiki Sasai (sox2), and Uwe Strähle (foxh1-SID). We are grateful to Marianela Ceol Retamal, Andrea Pecile, Manuel Ponce, and Ezequiel Yamus for animal husbandry, and to María Belén Favarolo for help with experiments.