Nato3 plays an integral role in dorsoventral patterning of the spinal cord by segregating floor plate/p3 fates via Nkx2.2 suppression and Foxa2 maintenance

During the development of the central nervous system, neuroepithelial cells acquire distinct fates to give rise to a plethora of neuronal and glial cell types (Jessell, 2000; Placzek and Briscoe, 2005). Extrinsic signals, and consequently intrinsic factors, induce rostrocaudal and dorsoventral (DV) patterning that divides the neural tube into distinct structures. In the caudal neural tube, DV patterning is mainly coordinated by organizing centers comprising the roof plate and floor plate (FP). These organizers secrete morphogens that regulate the precise organization of the spinal cord and consequently determine the neural assembles that comprise its neuronal circuits.

An increased concentration of sonic hedgehog (Shh), secreted first from the notochord and later from the medial ventralmost cells, results in upregulation of the transcription factors Foxa2 and Nkx2.2 in the presumptive FP cells, which are therefore considered at this time point to have an FP/p3 shared identity (Balaskas et al., 2012; Lek et al., 2010; Matise, 2013; Ribes et al., 2010). Ultimately, medial cells destined to adopt a definitive FP identity must downregulate Nkx2.2, whereas more dorsal cells interpreting Shh signaling upregulate Nkx2.2 to become p3 progenitors. The detailed mechanism controlling the segregation between FP and p3 fates is not fully understood (Jessell, 2000; Matise, 2013; Placzek and Briscoe, 2005), although recently an elegant model for Shh gradient interpretation was formulated (Balaskas et al., 2012).

Shh dynamics was suggested to be a major determinant in the selection between FP and p3 fates, as the induction of FP identity requires a high concentration of Shh, whereas a lower concentration induces p3 fate (Ribes et al., 2010). As effectors of the Shh pathway, a reduction in Gli activity can explain the downregulation of Nkx2.2 in FP cells. This model is challenged by the fact that the amplitude of the Shh gradient increases in the developing spinal cord as development proceeds (Balaskas et al., 2012; Chamberlain et al., 2008); thus, prospective p3 cells are exposed to levels of ligand that are at least similar to those experienced by prospective FP cells (Lek et al., 2010). Furthermore, genetic manipulations of Gli genes show that the emergence or absence of FP and p3 cells coincides, suggesting adjoined induction and origin (Hui and Angers, 2011).

A non-graded mechanism was also suggested to contribute to FP/p3 spatial separation and diversification (Lek et al., 2010). Accordingly, a ‘temporal shift’ from non-neuronal to neuronal potential of the progenitors, accompanied by specific expression of neuronal factors, underlies the acquirement of the p3 fate. At the same time, Nkx2.2 is repressed and Shh/Gli-independent Foxa2 expression is maintained to assure the characteristics of the definitive FP fate. However, the mechanisms regulating these events are not fully understood.

Nato3 (also known as Ferd3l) is an evolutionarily conserved basic helix-loop-helix (bHLH) transcription factor that is dynamically expressed in the embryonic and adult nervous system (Mansour et al., 2011; Segev et al., 2001). In the developing neural tube, Nato3 expression is restricted to the FP in both the spinal cord and midbrain (Mansour et al., 2011). In the developing mesencephalon, loss of Nato3 does not affect FP differentiation, but suppresses proneural gene expression and induces cell cycle arrest. As a consequence, a loss of neurogenic activity in the mesencephalic FP causes a decrease in the number of midbrain dopaminergic neurons generated in Nato3 mutants (Ono et al., 2010). By contrast, FP cells in the spinal cord are correctly specified and display only a temporary downregulation of medial FP-specific genes. Thus, the function of Nato3 in the spinal FP is unclear (Ono et al., 2010).

Here, we aimed to unravel the cellular and molecular mechanisms regulating the establishment and maintenance of the unique identity of FP cells. Using in vivo analysis in the chick embryo, we have found that overexpression of Nato3 induces ectopic, non-dividing cells that express Foxa2, a key gene in FP induction and function. Nato3 is also shown to indirectly downregulate Nkx2.2 when it is expressed both ectopically and in presumptive FP cells. This ability was also exemplified in the Nato3 null mice that we generated, in which Nkx2.2 was not properly suppressed in FP cells lacking Nato3. Taken together, we provide evidence that Nato3 is involved in the segregation of the FP and p3 identities, which is an essential step towards the establishment of the definitive FP fate.

To explore the function of Nato3 during FP development, we first analyzed the spatiotemporal expression profile of Nato3 and related markers in the developing chick neural tube. At embryonic day (E) 2 [Hamburger-Hamilton (HH) stage 12-13], soon after neural tube closure, no detectable levels of Nato3 transcripts were evident (Fig. 1A) (Mansour et al., 2011). At this developmental stage, definitive FP identity has not yet been defined, as is evident by co-expression of Foxa2 and Nkx2.2 in the presumptive FP cells, sharing FP/p3 identity (Fig. 1B,C) (Dessaud et al., 2008; Jeong and McMahon, 2005; Lek et al., 2010; Ribes et al., 2010). By E3 (HH18-19), Nato3 and Foxa2 were co-expressed at high levels within the definitive FP cells (Fig. 1D,E), whereas Nkx2.2 was downregulated in FP cells and its expression domain was shifted dorsally (Fig. 1F). At E4.5 (HH24-25), Nato3 transcripts were still detectable in FP cells and colocalized with Foxa2 (Fig. 1G,H). These observations from chick are in agreement with findings from mouse, in which Nato3 expression was shown to be upregulated in parallel with the downregulation of Nkx2.2 in the FP (Ono et al., 2010; Ribes et al., 2010). Taken together, in both avian and rodent models, the dynamic expression of Nato3 is temporally correlated with the establishment of the definitive FP identity (Foxa2⁺/Nkx2.2^-) and the spatial separation of FP and p3 marker-expressing cells (Fig. 1I).

In mouse and chick Nato3 is specifically expressed at early stages of spinal cord development, around the time of FP differentiation, and we reasoned that Nato3 might participate in the specification/differentiation of FP cells. To address this hypothesis, we examined whether Nato3 is sufficient to induce an ectopic FP fate in the chick neural tube (Fig. 2). Overexpression of Nato3 by in ovo electroporation resulted in weak ectopic induction of the FP determinant factor Foxa2 at 24 hours post-electroporation (hpe) (∼5 ectopic Foxa2⁺ cells/section; Fig. 2A-Ac), but not in the non-transfected hemitube or upon transfection with the pCIG backbone vector (Fig. 2E). Moreover, Foxa2 induction was cell-autonomous, as indicated by the colocalization of GFP⁺ and Foxa2⁺ cells (Fig. 2Aa-c, arrows). Notably, Foxa2 was induced not only in ventral positions, where a high concentration of Shh is present, but in some cases also in more dorsal regions (data not shown), suggesting that Nato3 induces Foxa2 expression in an Shh-independent manner.

Nato3, like most class II bHLH family members, is a tissue-specific transcription factor (Segev et al., 2001; Verzi et al., 2002). A class II factor forms a heterodimer with E proteins, which are class I bHLH members, such as E12 and E47, which are the splice variants of the E2A gene (also known as Tcf3), and considered to be more ubiquitously expressed (Bertrand et al., 2002; Murre et al., 1994). We envisioned that co-expression of Nato3 and its heterodimerization partner E12 might enhance the induction of the ectopic Foxa2⁺ fate. Indeed, co-electroporation of Nato3 with E12 resulted in robust induction of the FP marker Foxa2 all along the DV axis of the transfected side (∼21 ectopic Foxa2⁺ cells/section; Fig. 2C,Ca-c). An induction was not detected upon electroporation of a control vector encoding E12 alone (Fig. 2F). The ectopic induction of Foxa2⁺ cells was cell-autonomous, as evident by the colocalization of GFP⁺ (marking Nato3⁺ cells) and Foxa2⁺ cells (100% of Foxa2⁺ cells were GFP⁺; Fig. 2C,Ca, arrows). Both the Foxa2 expression level and number of induced Foxa2⁺ cells were significantly higher upon co-electroporation of Nato3+E12 as compared with Nato3 by itself (Fig. 2A′,C). Taken together, these results demonstrate that ectopic expression of Nato3 is sufficient to induce Foxa2⁺ cells. Interestingly, Foxa2 induction by Nato3, or even Nato3 and E12, was not accompanied by subsequent expression of additional FP markers such as Arx and Shh (Fig. 2B,D,D′). Thus, ectopic expression of Nato3 did not endow the cells with a definitive FP fate but, rather, selectively induced Foxa2 expression.

Given that ectopic expression of Nato3 gave rise to Foxa2⁺ cells, we hypothesized that the normal identity of these cells was consequently converted. Indeed, ectopic expression of Nato3+E12 resulted in a reduction of the number of neuronal progenitors marked by Nkx2.2, Olig2, Nkx6.1, Nkx2.2, MNR2 (also known as Mnx1) and Pax7 (Fig. 3A-G), compared with the non-transfected side or with expression of the backbone control vector (Fig. 3H-M) or E12. Markedly, the identity change was confined to the area of Foxa2 induction (Fig. 3A-A′′,Aa). Taken together, these results demonstrate that misexpression of Nato3+E12 affected the expression of ventral and dorsal progenitor markers, thus supporting the idea that Nato3-induced upregulation of Foxa2 results in cell-autonomous fate conversion.

The Foxa2⁺ cells that were induced by Nato3 were situated in the mantle zone, lateral to the ventricular zone of the neural tube (Fig. 3A-G). During spinal neurogenesis, dividing progenitors in the ventricular zone start differentiating while upregulating neuronal genes such as Atoh1, Ngn1/2 and Ascl1, and then exit the cell cycle and migrate laterally to the mantle zone (Bertrand et al., 2002; Guillemot, 2007; Helms et al., 2005). Moreover, as young FP cells differentiate to become definitive spinal FP cells, they become postmitotic. Hence, we asked whether the induction of Foxa2⁺ cells by ectopic Nato3 was coupled to their cell cycle exit, as would be expected for FP cells.

We re-examined the distribution of transfected cells in neural tubes ectopically expressing Nato3 or Nato3+E12 (Fig. 3N,O). Cells transfected with Nato3, and to a greater extent Nato3+E12, were preferentially located at the lateralmost aspect of the neural tube (Fig. 3Nd-i). By contrast, control embryos electroporated with a backbone plasmid displayed GFP⁺ cells that were more evenly distributed (Fig. 3Na-c). This difference indicates that Nato3-induced Foxa2 expression had stimulated the transfected cells positioned around the lumen to migrate laterally out of the ventricular zone, indicating that they had probably undergone cell cycle exit.

To verify that the migration of the Nato3-expressing progenitors from the ventricular zone was concomitant with cell cycle exit, electroporated embryos were also pulse labeled with BrdU (Fig. 3N) and the number of double-positive cells (GFP⁺/BrdU⁺) counted (Fig. 3O). Whereas 92±6% of the proliferating cells were GFP⁺ when electroporated with the backbone vector, only 62±20% incorporated BrdU when transfected with Nato3. The combined expression of Nato3+E12 further and significantly reduced the proliferation of transfected cells to 16±8% as compared with transfection of E12 alone, which resulted in 79±7% double-positive cells.

Definitive spinal FP cells are Foxa2⁺ and postmitotic, and are considered glia-like as they do not express neuronal genes (Placzek and Briscoe, 2005). We investigated whether the Nato3-induced ectopic Foxa2⁺ cells also mimic this feature of definitive FP cells. Quantitative analysis of sections stained for neuron-specific class III β-tubulin (Tuj1, also known as Tubb3) revealed that there was no significant difference in the number of postmitotic neurons between the two sides of the neural tube. Moreover, cells positive for both Nato3 and Tuj1 were not detected upon merging the GFP (Nato3-expressing cells) and Tuj1 images (Fig. 3G).

In summary, our observations indicate that Nato3 (and to a greater extent Nato3+E12) is able to induce ectopic Foxa2⁺ cells in the embryonic avian spinal cord. These cells exit the cell cycle and become postmitotic while migrating to the mantle zone, but are not neural, as they do not express common pan-neuronal genes. These features of the induced Foxa2⁺ cells resemble some of the properties of normal FP cells, raising the hypothesis that Nato3 might function not only ectopically, but also endogenously, in the developing spinal FP.

During the specification of FP cells, before the definitive FP identity is demarcated, Foxa2 and the p3 determinant factor Nkx2.2 are co-expressed in the ventral midline (Fig. 1B,C). When FP progenitors differentiate to become definitive spinal FP cells, Nkx2.2 expression becomes restricted to the more dorsal p3 progenitors (Fig. 1E,F) (Jeong and McMahon, 2005; Lek et al., 2010; Ribes et al., 2010). The mechanism responsible for the downregulation of Nkx2.2, and consequently the temporal switch between the shared FP/p3 and the definitive FP identities, is not fully understood.

To delineate the effect of Nato3 on Nkx2.2 expression, rather than on patterning, we took advantage of the fact that forced expression of Nato3 induces fewer Foxa2⁺ cells than Nato3+E12 (Fig. 2A′,C). Ectopic expression of Nato3, but not of the backbone vector, was sufficient to repress the expression of Nkx2.2 in the p3 domain (Fig. 4A,Aa-d). The effect of Nkx2.2 repression was cell-autonomous, as judged by the absence of colocalization of Nato3-expressing cells (GFP⁺) and Nkx2.2⁺ cells, and the presence of adjacent GFP^-/Nkx2.2⁺ cells (Fig. 4A,Aa-d,G,G′). Strikingly, the repression of Nkx2.2 expression was specific, since Nato3⁺ cells were found to colocalize with Olig2⁺ (Fig. 4Ac,Ad), Nkx6.1⁺ (Fig. 4C,G,G′) and Pax7⁺ (Fig. 4D) cells.

Double immunostaining for Foxa2 and Nkx2.2 identified Nato3⁺ cells that downregulated Nkx2.2 but were Foxa2^- (Fig. 4B,B′), thus excluding the possibility that the repression of Nkx2.2 resulted from an Nkx2.2-to-Foxa2 fate conversion in this domain. The lack of fate conversion was also supported by the fact that the weak induction of Foxa2 was accompanied by robust repression of Nkx2.2 (Fig. 2A′, Fig. 4A) and that some cells, in which Nkx2.2 was downregulated, upregulated Olig2 (Fig. 4Ad, arrows). Collectively, these data demonstrate that Nato3 represses Nkx2.2 expression and that this effect is uncoupled from the induction of the Foxa2⁺ fate.

To manipulate the downregulation of Nkx2.2 in the FP we utilized a dominant-negative form of Foxa2 (dnFoxa2) that contains the forkhead DNA-binding domain but lacks the N- and C-terminal domains. Consequently, dnFoxa2 can bind target sites but is unable to activate transcription. FP expression of dnFoxa2 sustained Nkx2.2 and abolished Arx expression in the presumptive FP cells, indicating again the need for Foxa2 to acquire proper FP identity (Ribes et al., 2010).We next asked whether Nato3 could downregulate, in the FP itself, the sustained Nkx2.2 expression induced by dnFoxa2 (Fig. 4H-J′). As expected, the inhibition of Foxa2 activity resulted in abnormally elevated expression of Nkx2.2 in the FP (an average of 6.9 Nkx2.2⁺ cells/section; Fig. 4I-I′). By contrast, the backbone vector did not induce Nkx2.2⁺ cells (Fig. 4H-H′). Notably, when Nato3 was co-electroporated with dnFoxa2, Nkx2.2 was not upregulated in the FP cells (an average of 0.7 Nkx2.2⁺ cells/section; Fig. 4J-J′). Thus, Nato3 could block dnFoxa2-induced Nkx2.2 expression in FP cells and rescue the phenotype. Moreover, Nato3 also downregulated Nkx2.2 at its endogenous p3 domain (35% of Nkx2.2⁺ cells were also GFP⁺; Fig. 4K). Supporting this notion is the fact that, when electroporated alone, Nato3 had an even higher impact on Nkx2.2 downregulation (only 17% of Nkx2.2⁺ cells were also GFP⁺; Fig. 4A-A′′,G), whereas the control plasmid had no effect (80% of Nkx2.2⁺ cells were also GFP⁺; Fig. 4K).

Our studies using chick embryos indicate that Nato3 is capable of repressing Nxk2.2 ectopically, as well as in the FP per se, in a Foxa2-independent and cell-autonomous manner, suggesting that Nato3 has a role in the segregation of the shared FP/p3 fate in FP precursors.

bHLH transcription factors can function as transcriptional activators or repressors (Castro et al., 2011). To discriminate between the two mechanisms, we generated obligatory activator (Nato3-VP16) or repressor (Nato3-EnR) versions of Nato3 and tested their ability to mimic Nato3 functions, namely the induction of Foxa2⁺ cells and the inhibition of Nkx2.2 expression.

Overexpression of the transcriptional activator Nato3-VP16 induced weak ectopic Foxa2 expression (Fig. 5A). Foxa2 induction was further increased by co-expression of Nato3-VP16 and E12 (Fig. 5D), although less efficiently than when using unmodified Nato3+E12 (Fig. 2C,Ca-c). By contrast, the transcriptional repressor form Nato3-EnR, with or without E12, was not sufficient to induce ectopic expression of Foxa2 (Fig. 5E,E′,G). Misexpression of the functional activator Nato3-VP16 mimicked the capability of wild-type Nato3 to specifically repress Nkx2.2 and induce lateral migration of the transfected cells (Fig. 5B,Ba,H). Nearly complete repression of Nkx2.2 was observed, and GFP⁺ and Nkx2.2⁺ cells never overlapped upon transfection with Nato3-VP16 (Fig. 5B,Ba,H′), which was not the case with Nato3-EnR (Fig. 5H,H′). By contrast, double-positive cells were identified for GFP with other progenitor markers, such as Nkx6.1 (Fig. 5C), Olig2 and Pax7 (data not shown). Thus, induction of the Foxa2⁺ fate and the repression of Nkx2.2 were mediated by transcriptional activator activity of Nato3, suggesting that the repression of Nkx2.2 by Nato3 was indirect.

The lack of proliferation of Nato3-transfected cells and their lateral migration (Fig. 3B,I) indicated that Nato3 might have promoted premature neuronal differentiation and cell cycle exit. Also, Nato3-VP16-transfected cells were preferentially located in the mantle zone, lateral to the ventricular zone (Fig. 5Aa,I), instead of being almost uniformly distributed as in controls. Additionally, these cells rarely incorporated BrdU and only a few examples of transfected cells that also incorporated BrdU were observed (Fig. 5I). Unlike Nato3-VP16, cells electroporated with Nato3-EnR or the backbone vector were almost uniformly distributed (Fig. 5E-G,J). To further verify these observations, the localization of Nato3, Nato3-VP16 and Nato3-EnR transcripts was examined by in situ hybridization, and they showed a similar distribution to the corresponding proteins (data not shown). The induction of Foxa2 expression was mimicked by the obligatory activator Nato-VP16, but not the repressor Nato3-EnR. By contrast, the repression of Nkx2.2 was probably mediated indirectly, possibly through upregulation of an intermediate repressor protein.

We next examined whether Nato3 is also necessary for the differentiation of FP cells and downregulation of Nkx2.2 during spinal cord development in mammals. We utilized homologous recombination in embryonic stem cells to generate Nato3 null allele (Nato3^Null) mice, in which the entire coding region of Nato3 was removed (supplementary material Fig. S1). The knockout mice were viable, fertile and appeared normal, in agreement with published results (Ono et al., 2010).

We carefully analyzed the histology and development of the FP and the spinal cord for specification and differentiation defects, but no abnormalities were identified (data not shown). Shh expression in the notochord and the presumptive FP appeared normal in the wild-type and littermate null mouse embryos at E10.5 (Fig. 6A) and E11.5 (data not shown), implying proper FP induction and specification. Similarly, the expression of Foxa2 and Arx appeared similar in wild-type and null littermates.

Since in chick we identified a cell-autonomous repression of Nkx2.2 by Nato3 (Fig. 4), we examined Nkx2.2 expression in the spinal FP of wild-type and Nato3^Null mice (Fig. 6B). Neural tube sections at the hindlimb level at E10.5 displayed Foxa2 expression in the FP (Fig. 6Ba,i), while Nkx2.2 was already downregulated in the FP and its expression domain was shifted dorsally to the p3 region (Fig. 6Be,i). By contrast, in Nato3^Null cords, Nkx2.2 expression was maintained in the FP at the hindlimb level, as indicated by colocalization with Foxa2⁺ cells (Fig. 6Bf,j). At the forelimb level, which was developmentally more advanced, less Nkx2.2 expression was detected in the FP domain (Fig. 6Bh,l). This phenotype gradually decreased such that, at E11.5, no overlap between lacZ⁺ (marking Nato3) and Nkx2.2⁺ cells was observed in null embryos (Fig. 6Cf).

Next, we aimed to reinforce and prolong the effect of Nato3 loss of function to further reveal its role in FP differentiation. We assumed that the lack of Nato3 was compensated by another factor being upregulated by Foxa2. We took into consideration the reciprocal positive-feedback loop between Nato3 and Foxa2 shown here and in other studies (Mansour et al., 2011; Nissim-Eliraz et al., 2013) and decided to further block the Foxa2-Nato3 pathway by obtaining a bigenic mutant mouse. Since Foxa2 null (Foxa2^Null) embryos die at gastrulation (Ang and Rossant, 1994) and Nato3^Null mice are apparently normal, Foxa2 heterozygotes (Foxa2^Het) and Nato3^Null adults were mated to obtain Nato3^NullFoxa2^Het embryos.

To evaluate the outcome of the bigenic mutations, Nato3^NullFoxa2^Het, as well as wild-type, Foxa2^Het and Nato3^Null embryos were studied (Fig. 7). At E10.5, wild-type and Foxa2^Het embryos demonstrated a complete segregation of the FP/p3 fates, as seen by the lack of colocalization of Foxa2 and Nkx2.2 staining (Fig. 7Aa,b). As shown in Fig. 6, Nato3^Null embryos displayed partial overlap of Foxa2 and Nkx2.2 expression domains, mainly at the lateral FP (Fig. 7Ac). Notably, in Nato3^NullFoxa2^Het embryos, Nkx2.2 expression was detected not only in the p3 domain, but also in the FP (Fig. 7Ad). This is in agreement with our hypothesis that Nato3 deficiency, when combined with Foxa2 heterozygosity, would emphasize the phenotype due to a lack of compensation.

When analyzing the extended expression of Nkx2.2 in the FP of E10.5 Nato3^Null embryos, we identified Nkx2.2/Foxa2 colocalization at the hindlimb level (Fig. 6Bj), but not at the more developmentally advanced forelimb level (Fig. 6Bl). In addition, this defect was transient, and undetectable at E11.5 (data not shown). Thus, to further validate that the bigenic mutation enhanced the phenotype, we first examined Nkx2.2 expression in E10.5 Nato3^NullFoxa2^Het embryos along the rostrocaudal axis, from cervical to sacral levels. Colocalization of Nkx2.2 and Foxa2 in FP cells was evident at all the levels examined (data not shown). Next, to look for a sustained effect, we examined E11.5 embryos. Indeed, in Nato3^NullFoxa2^Het embryos (Fig. 7Bd), but not in wild-type, Foxa2^Het or Nato3^Null embryos (Fig. 7Ba-c), Nkx2.2 continued to be expressed in the FP cells, indicating that even at E11.5 a definitive FP fate was not acquired by these cells.

To investigate the specificity of the Nato3-Foxa2 signaling blockage on Nkx2.2 downregulation in FP cells, the embryos were further analyzed for additional patterning markers, including Nkx6.1 and Shh. The expression pattern of Shh was confined to FP cells at both E10.5 and E11.5, and appeared comparable in all four genotypes examined (Fig. 7Ae-h,Be-h). Similarly, no overt differences in Nkx6.1 staining were seen among these genotypes (Fig. 7Ai-l,Bi-l). Thus, FP cells were correctly specified but specifically failed to downregulate Nkx2.2 expression upon inhibition of the Nato3-Foxa2 pathway.

Altogether, we have shown by gain- and loss-of-function approaches in both mouse and chick that Nato3 is important for FP/p3 fate segregation, an essential step in the specification of the embryonic spinal cord.

The FP of the developing embryo has a crucial role in establishing cell type diversification and patterning of the neural tube across many species. In this study, we provide evidence that the bHLH transcription factor Nato3 plays a role during the specification and differentiation of spinal FP progenitors. Nato3, its transcriptional activator variant Nato3-VP16 and, to a greater extent, the Nato3+E12 heterodimer, are sufficient to endow neural progenitors with a Foxa2⁺ fate. Not only does Nato3 induce ectopic Foxa2 expression, but it also leads to cell cycle exit of the fate-converted cells. Nato3 is both necessary and sufficient to specifically inhibit the expression of Nkx2.2 in ventral midline cells. Taken together (Fig. 8), our findings imply that the reciprocal positive interactions between Nato3 and Foxa2 enhance and maintain their own expression, while downregulating Nkx2.2. Consequently, FP cells acquire their definitive identity and cease expressing the p3 factor Nkx2.2, and its expression domain shifts dorsally to delineate progenitors destined to become V3 neurons.

A plethora of studies in various organisms have demonstrated that Foxa2, which is a direct target of Shh signaling, is both necessary and sufficient for the specification and maintenance of FP cells (Balaskas et al., 2012; Lek et al., 2010; Ribes et al., 2010; Sasaki and Hogan, 1994). In both chick (this study) and mouse (Ono et al., 2010) embryos, Foxa2 expression precedes that of Nato3 (Fig. 8, t1-3), thus ruling out the possibility that Nato3 is involved in the induction of FP cells. Rather, it is suggested that Nato3 is involved in the maintenance of FP identity. In early stages of neural tube development, notochord-generated Shh induces Foxa2 expression in overlying FP precursors via Gli transcription factors, which are effectors of the Shh pathway (Sasaki et al., 1997). However, Gli was found to be downregulated soon after neural tube closure (Matise and Joyner, 1999; Sasaki et al., 1997), suggesting that, at this stage, FP expression of Foxa2 becomes independent of Shh signaling. FP cells become refractory to Shh signaling as a prerequisite for the elaboration and maintenance of their identity (Ribes et al., 2010). Despite extensive efforts addressing the Shh-independent expression of Foxa2, a positive autoregulatory feedback loop or another direct upstream gene has yet to be reported.

Our data imply a direct back-regulation of Foxa2 transcription by Nato3. The ability of Nato3 to induce Foxa2 in the dorsal neural tube, where the Shh concentration is low, and the identification of ectopic Foxa2⁺/Shh^- cells indicate that Nato3 acts independently of the Shh gradient. Nato3 induces Foxa2 through transcriptional activation, which is in contrast to most factors implicated in the specification of ventral progenitor domains, which mainly act as transcriptional repressors (Dessaud et al., 2008; Muhr et al., 2001; Novitch et al., 2001). Ectopic FP cells and Foxa2 expression can be induced by high levels of Shh protein or Shh signaling molecules (Allen et al., 2011; Placzek and Briscoe, 2005; Roelink et al., 1994) and also by the repressor activity of Nkx2.2 and Nkx6.1 (Lek et al., 2010). However, none of these factors was found to be induced upon misexpression of Nato3, further supporting direct regulation of Foxa2 by Nato3. We have recently shown that Foxa2 directly activates Nato3 transcription via two binding sites located within a Nato3 promoter (Mansour et al., 2011). Taken together, our results suggest a bidirectional positive regulatory loop connecting Nato3 and Foxa2 to guarantee persistent expression of Foxa2 (Fig. 8, t3-4). These reciprocal interactions relay Shh signaling and enhance Nato3 and Foxa2 expression at early stages, whereas afterwards, Nato3 confers the Shh-independent expression of Foxa2, hence maintaining FP identity in the ventral medial cells.

Cell proliferation, specification and differentiation must be precisely coordinated during nervous system development in order to ensure that diverse cell types are generated in the correct number, place and time (Guillemot, 2007; Pitto and Cremisi, 2010). In addition to the activity of bHLH transcription factors in determining distinct progenitor identities and regulating neurogenesis (e.g. Ben-Arie et al., 1997; Gazit et al., 2004; Guillemot, 1995), they have a role in the regulation of cell cycle withdrawal (Farah et al., 2000; Nakada et al., 2004; Nguyen et al., 2006). In agreement with this paradigm, our results reveal that Nato3, acting as a transcriptional activator, promotes cell cycle exit. Mechanistically, bHLH factors mediate cell cycle exit through upregulation of the cyclin-dependent kinase inhibitor p27Kip1 (Cdkn1b) (Farah et al., 2000; Guillemot, 2007; Nguyen et al., 2006). Additionally, Hes1, which is known to be required for non-dividing organizer cell development in the neural tube, is aberrantly upregulated in the FP of the mesencephalon in Nato3^Null mice and forced expression of Nato3 represses Hes1 expression and induces premature neurogenesis (Ono et al., 2010). Thus, one can envision that the promotion of cell cycle arrest in the spinal FP by Nato3 could be carried out by similar mechanisms.

We demonstrate that, in both chick and mouse, Nato3 is necessary and sufficient to specifically repress Nkx2.2, which raises the question of its biological importance. During the specification of the ventral cell lineage in the neural tube, different progenitor domains are generated in a sequential manner determined by different concentrations of Shh (Dessaud et al., 2008; Jeong and McMahon, 2005). Nkx2.2 is first activated in the ventral midline and is conjointly expressed with Foxa2 at the onset of FP specification. Later, Nkx2.2 is repressed in the prospective FP and its expression becomes confined to the p3 domain, contributing to the segregation of the FP and p3 identities (Dessaud et al., 2008; Jeong and McMahon, 2005; Lek et al., 2010; Nishi et al., 2009; Ribes et al., 2010).

What mediates this temporal switch of FP cells from a shared FP/p3 fate to the definitive FP fate? As our results demonstrate that Nato3 specifically represses Nkx2.2, independently of its ability to induce Foxa2, we conclude that Nato3 underlies a ‘temporal switch’ from the shared FP/p3 identity to the definitive FP fate. Indeed, the expression of Nkx2.2 precedes that of Nato3, the downregulation of Nkx2.2 was observed around the time of Nato3 induction, and the expression domains of Nato3 and Nkx2.2 in the FP do not appear to overlap. Moreover, the observation that, unlike Foxa2, Nato3 expression in the mouse neural tube is confined to the FP and not to the adjacent Nkx2.2 cells along the rostrocaudal axis provides further support for this view.

Since Nato3 functions by transcriptional activation, not repression, its downregulation of Nkx2.2 is probably indirect, similar to the repression of Nkx2.2 by Pax6, which acts indirectly via activation of gene expression (Lei et al., 2006). The identity of the mediator factor remains to be determined.

Foxa2 has been proposed to mediate the repression of Nkx2.2 in prospective FP cells (Ribes et al., 2010). Importantly, Nkx2.2 is persistently expressed in the FP of the Nato3 null mutant despite the expression of Foxa2, and Nato3 can rescue the upregulation of Nkx2.2 in FP cells following Foxa2 inhibition in the chick neural tube. Moreover, Foxa2 expression is not exclusively confined to the FP lineage in the developing spinal cord (Jeong and McMahon, 2005; Lek et al., 2010; Placzek and Briscoe, 2005; Ribes et al., 2010) and hindbrain (Jacob et al., 2007), where it is co-expressed with Nkx2.2. Overall, it is unlikely that Foxa2 is the sole mediator of this repression and fate separation.

The prolonged expression of Nkx2.2 in Nato3^NullFoxa2^Het embryos, which results in weakened Foxa2 signaling combined with a deficit of Nato3, emphasizes the importance of the positive-feedback loop between Foxa2 and Nato3, as shown here and in other studies (Mansour et al., 2011; Nissim-Eliraz et al., 2013). In these embryos, the FP cells are induced and start to differentiate, but the segregation of the FP/p3 fates is impaired. According to our model, in FP cells Nato3 acts downstream of Foxa2 to inhibit Nkx2.2 and, at the same time, feeds back to maintain Foxa2 expression, thus enhancing the establishment of the definitive FP fate of the cells located in the ventral midline. In the absence of Nato3, the FP/p3 fate segregation is temporarily delayed, but is completed at a later stage, suggesting a compensatory mechanism. In the bigenic Nato3^NullFoxa2^Het mutants, Nato3 cannot induce the expression of enough Foxa2, leading to an insufficient upregulation of the compensating factor(s), resulting in sustained Nkx2.2 expression in FP cells. It has been suggested that Foxj1 indirectly inhibits Nkx2.2 expression in the FP by reducing the sensitivity of FP cells to Shh signaling (Cruz et al., 2010); thus, one cannot exclude the possibility that Foxa2 and/or Foxj1 act with Nato3 to establish the definitive FP fate.

Overall, our findings suggest that Nato3 serves multiple functions during the development of FP cells. With the lack of Shh signaling and Gli expression, Nato3 maintains the expression of Foxa2, which, in turn, regulates many genes, including Nato3 itself, to maintain FP identity. Nato3 also mediates the segregation of FP and p3 fates through an indirect repression of Nkx2.2, as is evident in both chick and mouse embryos.

All procedures were performed according to the guidelines of the Institutional Animal Care and Use Committee (IACUC) of the Hebrew University, which is an AAALAC internationally accredited institute.

Electroporation was performed as previously described (Zisman et al., 2007). Embryos were harvested 24 or 48 hpe. Sections covering the length of the neural tube were inspected, but only sections from the thoracic region were used for the figures and for statistical analysis.

All expression plasmids were based on the pCIG3 vector containing a CMV promoter, IRES, and three nuclear localization sequences tagged with eGFP (Megason and McMahon, 2002; Niwa et al., 1991). Primers are described in supplementary material Table S1.

The coding region of chick Nato3 (Segev et al., 2001) and mouse E12 (Tcf3) were amplified from genomic DNA or pBS-mE12 (kindly provided by Michael Walker, Weizmann Institute of Science, Israel) and cloned into pCIG3. The VP16 activator domain was amplified from pCS2-VP16 and the Engrailed repressor domain (EnR) from pUAST-EnHTH (a gift from Adi Salzberg, Rappaport Faculty of Medicine, Technion, Israel). pCAGGS-Foxa2 and a dominant-negative form of mouse Foxa2 (dnFoxa2, pCAGGS-Fkh^a2) were a kind gift from James Briscoe (Jacob et al., 2007).

Immunofluorescent staining and DAPI (1 μg/ml) counterstaining were performed as previously reported (Mansour et al., 2011). Working dilutions for primary and secondary antibodies are given in supplementary material Table S2. RNA in situ hybridization was conducted as described (Krizhanovsky and Ben-Arie, 2006; Ma et al., 1997). For uniformity in statistical analysis, the comparison was performed in chick and mouse on transverse sections from the thoracic level, unless otherwise stated. The in ovo BrdU incorporation assay was performed as described (García-Campmany and Marti, 2007), except that BrdU was applied on the top of the embryo.

To generate Nato3 knockout mice, a targeting vector containing 4.4 kb and 6.2 kb of mouse genomic DNA, from 129SvEv, located 5′ and 3′ to the ORF encoding Nato3, respectively, was constructed using a pNZTK2 backbone (supplementary material Fig. S1). The construct was electroporated into a hybrid embryonic stem cell line (C57BL/6N and 129SvEv) at InGenious Targeting Laboratory (iTL). One recombinant clone surviving double selection was used to generate two chimeras that gave rise to fertile heterozygous mice (supplementary material Fig. S1). Heterozygotes mice were repeatedly backcrossed to C57BL/6 mice to obtain a pure background.

Foxa2 heterozygous mice (Ang and Rossant, 1994) were obtained from the European Mouse Mutant Archive (EMMA) as frozen embryos (EM:02526). Mouse contract rederivation services were provided by the Mary Lyon Centre at MRC Harwell.

We thank Naomi Book-Melamed for contribution to confocal imaging; Omri Schatz, Ben Jerry Gonzales, Or Avner and Amit Hefetz for assistance in the experimental work; Esther Golenser and Theodora Bar-El for carefully reading and editing the manuscript; Dalit Sela-Donenfeld for critical examination of the data; and James Briscoe, Adi Salzberg and Michael Walker for the generous contribution of expression constructs.