An Early Role for Wnt Signaling in Specifying Neural Patterns of Cdx and Hox Gene Expression and Motor Neuron Subtype Identity

The link between extrinsic signaling, progenitor cell specification and neuronal subtype identity is central to the developmental organization of the vertebrate central nervous system. In the hindbrain and spinal cord, distinctions in the rostrocaudal identity of progenitor cells are associated with the generation of different motor neuron subtypes. Two fundamental classes of motor neurons, those with dorsal (dMN) and ventral (vMN) exit points, are generated over largely non-overlapping rostrocaudal domains of the caudal neural tube. Cdx and Hox genes are important determinants of the rostrocaudal identity of neural progenitor cells, but the link between early patterning signals, neural Cdx and Hox gene expression, and the generation of dMN and vMN subtypes, is unclear. Using an in vitro assay of neural differentiation, we provide evidence that an early Wnt-based program is required to interact with a later retinoic acid- and fibroblast growth factor–mediated mechanism to generate a pattern of Cdx and Hox profiles characteristic of hindbrain and spinal cord progenitor cells that prefigure the generation of vMNs and dMNs.


Introduction
During the early development of the vertebrate central nervous system, the position of generation of post-mitotic neurons depends on the patterning of progenitor cells along the dorsoventral and rostrocaudal axes of the neural tube [1][2][3]. At many levels of the neuraxis, the dorsoventral pattern of progenitor cells, which later gives rise to motor, sensory, and local circuit neurons, is initiated by the opponent signaling activities of Sonic hedgehog (Shh) and bone morphogenetic proteins [2,4,5]. In contrast, the rostrocaudal pattern of neural progenitor cells that differentiate into distinct neuronal subtypes is imposed, in part, by opponent retinoid and fibroblast growth factor (FGF) signals [6][7][8][9]. Within the hindbrain and spinal cord, the rostrocaudal positional identity of neurons is reflected most clearly by the generation of different motor neuron (MN) subtypes. One fundamental distinction in MN subtype identity is the emergence of two major classes of MNs that exhibit distinctive axonal trajections, ventral exiting motor neurons (vMNs) and dorsal exiting motor neurons (dMNs) [10]. vMNs include most spinal MNs as well as hypoglossal and abducens MNs of the caudal hindbrain [1,10,11], whereas dMNs are found throughout the hindbrain and at cervical levels of the spinal cord [10]. Each of the many subsequent distinctions in MN subtype identity emerge through the diversification of these two basic neuronal classes [2].
Despite many advances in defining the mechanisms of MN diversification [7,8,[12][13][14], it remains unclear how neural progenitors in the hindbrain and spinal cord acquire a rostrocaudal positional character that results in the generation of dMN and vMN classes. At both hindbrain and spinal levels, Hox genes are informative markers of the rostrocaudal positional identity of progenitor cells. Within the hindbrain, distinct rhombomeres are delineated by the nested expression of 39 Hox genes [3,15], whereas the spinal expression of 59 Hox genes distinguishes progenitor cells and post-mitotic neurons at cervical, brachial, thoracic, and lumbar levels [1,[6][7][8]13,16]. Moreover, Hox genes are determinants of MN subtype identity in both hindbrain and spinal cord. In the hindbrain, for example, the restricted expression of Hoxb1 helps to determine the identity of facial MNs [1,[17][18][19][20], and in the spinal cord the restricted expression of Hox6, Hox9, and Hox10 proteins establishes MN columnar subtype [7,16]. In addition, a more complex Hox transcriptional regulatory network specifies spinal MN pool identity and connectivity [21]. The neural pattern of Hox expression is, in turn, regulated by members of the Cdx homeobox gene family [6,[22][23][24][25]. Cdx genes are transiently expressed in the caudalmost region of the neural plate prior to the onset of 59 Hox gene expression [26][27][28] and appear to be direct regulators of the expression of 59 Hoxb genes [6,23,24,27]. Thus, analysis of spatial profiles of Cdx and Hox gene expression may provide clues about the identity of signals that pattern MN subtypes in the hindbrain and spinal cord.
Several recent studies have provided insight into the signals that impose rostrocaudally-restricted patterns of neural Cdx and Hox expression. Retinoic acid (RA) and FGF signals appear to have opponent roles in the rostrocaudal patterning of Hox gene expression in the caudal hindbrain (cHB) and spinal cord [6,8]. Mesodermal-derived RA signals promote the expression of Hox genes characteristic of the cHB and rostral spinal cord (rSC) [11,29,30], whereas FGF signals pattern the expression of Hox genes at more caudal levels of the spinal cord. At an earlier developmental stage, neural progenitors have been shown to acquire caudal forebrain, midbrain, and rostral hindbrain positional identities in response to graded Wnt signaling at the gastrula stage [31,32]. It is unclear, however, whether an early phase of Wnt signaling is also required to establish Cdx and Hox gene expression profiles characteristic of the cHB and spinal cord, in turn specifying the generation of dMN and vMN subtypes.
This study uses in vitro assays of neural cell differentiation to obtain evidence that early Wnt signaling does indeed have a crucial role in specifying the identity of hindbrain and spinal cord progenitor cells as revealed by profiles of Cdx and Hox gene expression. This early influence of Wnt signaling is later refined by retinoid and FGF signals to impart additional rostrocaudal distinctions in Hox expression that correlate tightly with the generation of dMNs and vMNs. Our findings therefore define a crucial early role for Wnt signaling in inducing profiles of Cdx and Hox expression that prefigure the differentiation of dMN and vMN subtypes in the developing hindbrain and spinal cord.

Transcriptional Markers of Progenitor Cell Position and MN Subtype
To explore how progenitor cells of different rostrocaudal regional identity differentiate into dMNs and vMNs, we analyzed a panel of transcription factors that are expressed in different temporal and spatial patterns in the developing hindbrain and spinal cord.
Hindbrain and Spinal Cord Progenitor Cells Acquire Rostrocaudal Regional Identity at Early Somite Stages To study the patterning of Cdx and Hox genes by neural progenitor cells and its link to the differentiation of MN subtypes, we used in vitro differentiation assays that employed stage 4-8 prospective hindbrain and spinal cord explants, and stage 4 prospective forebrain (FB) explants. In stage 4 caudal (C) explants, cells have been exposed to caudalizing signals at the time of their isolation [31,32], and these explants were used to examine the signals that specify hindbrain and spinal cord character. Cells in stage 4 FB explants have not been exposed to caudalizing signals at the time of their isolation [31,32], and these explants were used in attempts to reconstitute more completely the events that direct the generation of dMNs and vMNs from ''naive'' neural cells.
Prior to stage 8, the caudal neural plate is either specified as rSC or cSC (HH 5-7), but no explants generating cells of cHB character can be isolated (unpublished data). By stage 8, however, cells in explants isolated at a position just rostral to the regressing Hensen's node (RN explants) did not express CdxB and CdxC ( Figure 2) and generated cells characteristic of the cHB that expressed Hoxb4 alone (Hoxb4 þ /b8 À /c9 À cells (85 6 10% of total cell number) after 40 h in culture ( Figure 3B). In contrast, cells in stage 8 explants isolated at the level of the node (NL explants) expressed CdxB and CdxC ( Figure 2). These explants also generated cells that expressed Hoxb4 and Hoxb8 in the absence of Hoxc9 (Hoxb4 þ /b8 þ /c9 À cells; 96 6 2% of total cell number), a marker profile characteristic of the rSC ( Figure 3C). Cells in stage 8 explants isolated caudal to the node (CN explants) also expressed CdxB and CdxC (Figure 2), and generated Hoxb4 þ /b8 þ /c9 þ cells (88 6 7% of total cell number), a profile characteristic of the cSC ( Figure 3D). Thus by stage 8, prospective hindbrain and spinal cord progenitor cells appear to have acquired a coarse rostrocaudal regional identity.
To determine whether the Hox gene profiles generated in stage 8 neural plate explants were correlated with the emergence of dMNs and vMNs, we exposed explants to the diffusible N-terminal fragment of the Shh protein (Shh-N) that exhibits MN-inducing activity [43,44]. In the presence of Shh-N (15 nM) for ;50 h, stage 8 RN explants generated many Tbx20 þ /Isl þ dMNs (22 6 7% of total cell number) and very few Hb9 þ /Isl þ vMNs (1 6 0.5% of total cell or number), a profile of MNs characteristic of the cHB (r7 and rostral r8) ( Figure 4B). Under these conditions, NL explants generated very few Tbx20 þ /Isl þ dMNs (1.5 6 1% of total cell number) and many Hb9 þ /Isl þ vMNs (25 6 4% of total cell number). Few, if any, of the induced Hb9 þ /Isl þ vMNs co-expressed Hoxc9 (0.3 6 0.3% of total cell number), indicative of an rSC positional character ( Figure 4C). CN explants did not generate Tbx20 þ /Isl þ dMNs but did generate Hb9 þ /Isl þ vMNs (27 6 7% of total cell number), most of which expressed Hoxc9 protein (80 6 6% of Hb9 þ cells)-a profile characteristic of the thoracic spinal cord ( Figure 4D). Thus, by stage 8, progenitor cells that occupy different rostrocaudal positions within the caudal neural plate, defined in part by their Cdx profiles, are specified as hindbrain and spinal cord cells of either rostral or caudal regional character, and have acquired sufficient positional information to differentiate into dMNs and vMNs in a position-appropriate manner.

Spinal Cord Cells of Caudal Character Are Specified at the Late Gastrula Stage
To examine when the early pattern of Cdx and Hox profiles characteristic of hindbrain and spinal cord progenitor cells is established, we first monitored the generation of cells expressing CdxB and CdxC by in situ hybridization in stage 4 C explants cultured for 15 h in vitro, corresponding to a stage 8 embryo. We tracked the rostrocaudal orientation of these explants by labeling cells at the caudal margin with DiI crystals. Under these conditions, CdxB þ and CdxC þ cells (36 6 11% of total cell number) were generated in a caudal domain of the explants ( Figure 5D), adjacent to the DiI labeling. To examine the rostrocaudal identity of the Cdx þ cells, we monitored the generation of Hoxb4 þ , Hoxb8 þ , and Hoxc9 þ cells by in situ hybridization in stage 4 C explants cultured for 44 (C) In stage 20 (42-somite) chick embryos, the rostral boundaries of expression of Hoxb4, Hoxb8, and Hoxc9 in the neural tube are maintained as in a stage 17 embryo (6 1-somite). Tbx20 þ /Isl þ dMNs are present at high numbers in the cHB and at lower numbers in the rSC. No Tbx20 þ /Isl þ dMNs are found in the cSC. In contrast, Hb9 þ /Isl þ vMNs are present at high numbers in both rSC and cSC, and at lower numbers in r8 of the cHB. Hoxc9 protein is expressed in a subset of Hb9 þ /Isl þ vMNs in the cSC and thus, distinguishes vMNs in the cSC from vMNs in the rSC. (D) Horizontal bars represent rostrocaudal restrictions (applied to Figure 1A h. Stage 4 C explants cultured for 44 h generated Krox20 þ cells (45 6 15% of total cell number), characteristic of the rHB away from the DiI label and, in a separate caudal domain (adjacent to the DiI label) Hoxb4 þ /b8 þ /c9 þ cells (52 6 12% of total cell number) characteristic of the cSC ( Figures 7B and  5E). Thus, the generation of CdxB þ /C þ cells and Hoxb4 þ /b8 þ /c9 þ cells in a similar caudal domain of the explants supports the view that the expression of Cdx genes in neural plate cells is restricted to prospective spinal cord cells. In contrast, only a small domain of the explants generated Hoxb4 þ /b8 þ /c9 À cells (12 6 8% of total cell number), a marker profile characteristic of the rSC. Moreover, these explants lacked cells that expressed Hoxb4 alone (Hoxb4 þ /b8 À /c9 À cells) ( Figures 5E and  7B), characteristic of the cHB. These results provide evidence that, at stage 4, prospective caudal neural plate cells are specified as cells of rHB and cSC character, and only later acquire cHB or rSC character.
We next examined whether prospective rHB and cSC cells have acquired sufficient rostrocaudal positional information by stage 4 to permit them to differentiate into dMNs or vMNs when exposed to Shh-N. To test this possibility, we cultured stage 4 C explants for 28 h to allow cells to acquire a stable rostrocaudal positional identity, and then for an additional 38 h in the presence of Shh-N (15 nM) (combined culture time corresponding to a stage 20, ;40-somite embryo). In the presence of Shh-N, Tbx20 þ /Isl þ dMNs (1 6 1% of total cell number) and Hb9 þ /Isl þ vMNs (16 6 2% of total cell number) were generated in separate domains of the explants ( Figure  6B). A majority of the vMNs expressed Hoxc9 (10 6 2% of total cell number)-a profile characteristic of vMNs at thoracic levels of the spinal cord ( Figure 6B). These results indicate that by stage 4, cells in the prospective caudal neural plate have acquired a positional character that permits them to differentiate into dMNs and vMNs.

Joint Wnt and FGF Signaling at Late Gastrula Stages Specifies Spinal Cord Character
We next addressed the identity of the secreted signals that might be involved in the early specification of cells of spinal cord character. Several Wnt and Fgf genes are expressed in and around the prospective caudal neural plate [45][46][47], and the specification of rHB cells at stage 4 requires convergent Wnt and FGF signaling [32,48]. To examine whether combined Wnt and FGF signaling is required for the specification of cells of spinal cord positional character, we cultured stage 4 C explants in the presence of a soluble fragment of the Frizzled receptor 8 protein (Frz8CRD-IgG), an antagonist of Wnt signals [32,49,50], or with SU5402, an antagonist of FGF receptor signaling [49,51,52] and monitored the expression of Cdx and Hox genes.
In the presence of Frz8CRD-IgG or SU5402 (5 lM) for 15 h, the expression of both CdxB and CdxC was blocked ( Figure 5F and 5H). After 44 h under these conditions, the generation of cells of both rHB (Krox20 þ ) and of spinal cord (Hoxb4 þ /b8 þ /c9 À and Hoxb4 þ /b8 þ /c9 þ ) character was almost completely blocked (3 6 3% caudal cells remaining versus 64 6 10% in the controls) ( Figure 5G and 5I). Instead, Otx2 þ forebrain-like cells were generated (79 6 9% of total cell number versus 0% in the controls) ( Figure 5G and 5I). These results support the idea that the specification of cells of spinal cord positional character also involves convergent Wnt and FGF signaling.
To test whether Wnt signaling in prospective rHB and cSC cells is required for the generation of vMNs and dMNs, stage 4 C explants were cultured in the presence of mFrz8CRD-IgG, and after 28 h Shh-N (15 nM) was added for a further 38 h. Under these conditions, the generation of Tbx20 þ /Isl þ dMNs and Hb9 þ /Isl þ and Hoxc9 þ /Hb9 þ vMNs was blocked (0% of total cell number), and instead Isl þ /Tbx20 À /Hb9 À /Hoxc9 À neurons, characteristic of the ventral forebrain [53,54], were generated (18 6 3% of total cell number) ( Figure 6C). Thus, exposure of prospective caudal neural plate cells to Wnt signals is required for the generation of vMNs and dMNs.

High-Level Wnt Signaling Promotes Spinal Cord Character
We next examined whether differences in the level or duration of exposure to Wnt and FGF signals contribute to the early distinction in hindbrain and spinal cord character. To test this possibility, we exposed stage 4 C explants to exogenous Wnt and FGF signals for 15 h or 44 h in vitro. In explants exposed to Wnt3A (75 ng/ml) and FGF4 (60 ng/ml) simultaneously for 15 h, CdxB þ and CdxC þ cells were generated throughout the entire explant ( Figure 5J). After 44 h of culture under these conditions, the generation of Krox20 þ cells was largely suppressed (3 6 2% of total cell number versus 45 6 15% in the controls) and most cells acquired a Hoxb4 þ /b8 þ /c9 þ cSC character (96 6 2% of total cell number versus 52 6 12% in the controls) ( Figure 5K). To examine whether mesodermal cells were generated under these conditions, we monitored the expression of Mox1 [55], which is expressed in caudal paraxial mesoderm and of Brachyury (Bra) [56], which at caudal levels is expressed in both the mesoderm and in cells of the forming caudal neural plate Exposure to FGF4 (60-120 ng/ml), or Wnt3A (150 ng/ml) alone did not increase the number of cells expressing Cdx genes, nor did it change the ratio of Krox20 þ and Hoxb4 þ /b8 þ /c9 þ cells ( Figure S1 and unpublished data). These results are consistent with the view that exposure of cells to prolonged or higher level Wnt and FGF signaling promotes the specification of cells of spinal cord rather than midbrain or hindbrain character.

RA Imparts Caudal Character to Hindbrain Cells and Rostral Character to Spinal Cord Cells
How then, are rostral and caudal sub-domains of the hindbrain and spinal cord established? Since Wnt signaling contributes to the distinction in specification of prospective rHB and cSC cells at gastrulation stages, we examined first whether exposure of prospective hindbrain and spinal cord cells to Wnt signals beyond stage 8 was required for the later specification of cHB and rSC cells. Stage 8 caudal neural plate explants exposed to Frz8CRD-IgG still generated cells of cHB and spinal cord character (unpublished data); indicating that prolonged exposure to Wnt signals is not required for the generation of these two sub-domains of the caudal neural tube.
By stage 8, the retinoid acid synthesizing enzyme retinal-dehyde dehydrogenase 2 (RALDH2) is expressed in the paraxial mesoderm adjacent to the prospective cHB and rSC [57][58][59]. RA signaling might therefore promote the generation of cells of cHB and rSC character by acting on neural cells that have already acquired an initial caudal character, through convergent Wnt and FGF signaling. To test this idea, we exposed stage 4 C explants to RA, and used the combinatorial expression of Hoxb4, Hoxb8, and Hoxc9 to distinguish cells of cHB, and rSC or cSC character. Since Cdx expression does not distinguish between cells of rSC and cSC character, we did not monitor Cdx expression in these experiments. To map prospective rHB and cSC cells in these explants, we tracked their rostrocaudal orientation by labeling cells at the caudal margin with DiI crystals. We found that Hoxb4 þ /b8 þ /c9 þ cells derive from the caudal-most region of the explants, whereas Krox20 þ cells derive from a rostral domain of the explants ( Figure 7B).

FGF Signaling Contributes to the Distinction between Cells of cHB and rSC Character
We next examined how the distinction between cHB and rSC cells is established. RA promotes the generation of cHB cells, and FGFs promote the expression of Hox genes characteristic of the caudal region of the spinal cord [8,46]. Moreover, RA and FGF signals act in an opponent manner during rostrocaudal patterning of Hox gene expression and MN progenitor cell specification [6,8]. These observations led us to examine the possibility that FGF and RA signals converge during the initial assignment of cHB and rSC positional character.
We exposed stage 4 C explants to both RA (10 nM) and FGF4 (30 ng/ml) and assayed their Hox profile after 44 h. The generation of Hoxb4 þ /b8 À /c9 À cHB cells was suppressed (1 6 1% of total cell number versus 47 6 10% in explants cultured with RA alone), and most cells (92 6 3% of total cell number versus 55 6 10% in explants cultured with RA alone) expressed Hoxb4 and Hoxb8, a profile indicative of rSC character. A few Hoxb4 þ /b8 þ /c9 þ cells (7 6 4% of total cell number versus 0% in explants cultured with RA alone), characteristic of cSC character were also detected ( Figure  7D). Thus, in the presence of RA, prolonged exposure to FGF signals promotes the generation of cells of rSC at the expense of cHB character. This finding supports the idea that the status of FGF signaling biases whether RA exposure induces Hox gene profiles characteristic of cHB or rSC.

Combinatorial Wnt, RA, and FGF Signals Impose Hindbrain and Spinal Cord Character in Naive Neural Cells
As a further test of the sufficiency of Wnt, FGF, and RA signals in establishing hindbrain and spinal cord pattern, we examined whether a combination of these factors can establish appropriate Cdx and Hox gene profiles in naive rostral forebrain cells that appear not to have been exposed to caudalizing signals in ovo.
Collectively, these findings provide evidence that Wnt signals in combination with RA and/or FGF exposure, induce Hox profiles in neural cells that predict the later positionspecific emergence of dMNs and vMNs characteristic of the hindbrain and spinal cord. These observations support the idea that early exposure to Wnt signals, together with later RA and FGF signals imposes Hox profiles that anticipate the patterned generation of dMN and vMN subclasses in the developing hindbrain and spinal cord.

Discussion
This study has examined the link between extrinsic patterning signals, regionally restricted profiles of transcription factor expression in neural progenitor cells, and the specification of MN subtype along the rostrocaudal axis of the hindbrain and spinal cord. Our results support four main conclusions: (i) Wnt signaling is required to specify cells of spinal cord character, (ii) the initial specification of spinal cord progenitor cells appears to require prolonged, or higher level Wnt signaling than does the specification of cells of hindbrain character, (iii) early Wnt signaling provides a positional context for the later actions of RA and FGF signals in specifying the rostrocaudal regional identity of hindbrain and spinal cord cells, and (iv) the interplay of Wnt, retinoid, and FGF signals establish distinction in progenitor cell Cdx and Hox profiles that anticipate the rostrocaudal position of generation of dMNs and vMNs in the hindbrain and spinal cord. Below, we discuss the evidence that supports each of these conclusions.

Interplay between Early Wnt and Later FGF and RA Signals in the Assignment of Caudal and Neural Fates
The generation of different subclasses of MNs along the rostrocaudal axis of the hindbrain and spinal cord depends on two crucial early steps of caudal neural development: first, the early specification of cells of hindbrain and spinal cord character, and second, the subsequent refinement of rostrocaudal regional character of hindbrain and spinal cord progenitor cells. Wnt signaling has been implicated in the generation of caudal neural cells [48,[60][61][62][63][64][65][66][67][68][69], and results in chick have provided evidence that FGF and graded Wnt signaling in neural cells specify cells of caudal forebrain, midbrain, and rostral hindbrain character [32,66]. The present study provides evidence that early Wnt signaling is also essential to impose caudal hindbrain and spinal cord character on neural progenitor cells. The results also support the view that the specification of spinal cord progenitor cells (E) 4 FB explants cultivated in the presence of Wnt3A (;150 ng/ml) and RA (10 nM) did not generate CdxC þ /CdxB þ presumptive caudal neural cells (n ¼14 explants). DOI: 10.1371/journal.pbio.0040252.g008 requires prolonged, or higher level Wnt signaling than is required for the specification of cells of hindbrain character, and they are consistent with previous findings that Cdx genes respond to Wnt signals and act upstream of 59 Hox genes in neural progenitor cells [23,[70][71][72][73][74][75]. Thus, taken together, our results support the idea that in the presence of FGF signals, graded Wnt signaling imposes midbrain, hindbrain, and spinal cord character on prospective caudal neural plate cells.
The signals and mechanisms that act in the subsequent step to impose rostrocaudal regional identity on hindbrain and spinal cord progenitor cells have been examined previously. From early somite stages, RA supplied by the paraxial mesoderm and newly formed somites, promotes the expression of Hox genes characteristic of the cHB and rostral levels of the spinal cord [6,8,29,31,67,76,77]. FGF signals derived from the regressing primitive streak promote the expression of progressively more caudal Hox-c proteins in a concentration-dependent manner [7,8]. Thus, RA and FGF signals act in an opponent manner to impose rostrocaudal regional identity on hindbrain and spinal cord progenitor cells [6,78]. Our findings extend these results by showing that early Wnt signaling establishes a positional context for the later actions of RA and FGF signals in specifying hindbrain and spinal cord cells of rostral and caudal regional identity.
Collectively, our results suggest a model of how hindbrain and spinal cord cells of early rostrocaudal regional identity are generated (Figure 11). At gastrula stages, prospective caudal neural plate cells are exposed to Wnt signals derived (B-F) Sox1 was used as a general neural marker. Bars represent mean 6 s.e.m. number of cells in Otx2 þ , Hoxb4 þ /b8 À /c9 À , Hoxb4 þ /b8 þ /c9 À , and Hoxb4 þ / b8 þ /c9 þ domains, respectively, as percentage of total cell number. Each row represents consecutive sections from a single explant. (B) Control stage 4 FB explants generated Sox1 þ /Otx2 þ but no caudal neural cells (n ¼ 24 explants). (C) Stage 4 FB explants cultured in the presence of Wnt (;150 ng/ml) and FGF (60 ng/ml) generated Hoxb4 þ /b8 þ /c9 þ cells and only a few Krox20 þ cells (n ¼ 24 explants). (D) Cultivation in the presence of Wnt3A (;150 ng/ml), RA (10 nM), and FGF (30 ng/ml) generated Hoxb4 þ /b8 þ /c9 À cells and a few Hoxb4 þ /b8 þ /c9 þ cells (n ¼ 18 explants). (E) Cultivation in the presence of Wnt3A (;150 ng/ml) and RA (10 nM) generated Hoxb4 þ /b8 À /c9 À cells and no, or only a few, Hoxb4 þ /b8 þ /c9 À cells (n ¼ 28 explants). (F) Exposure to Wnt3A (;150 ng/ml) and RA (10 nM) in the presence of SU5402 (3 lM), an inhibitor of FGF signaling, generated Hoxb4 þ /b8 À /c9 À cells (n ¼ 12 explants from the emerging caudal paraxial mesoderm and from epiblast cells; and to FGF signals derived from the primitive streak [32,46,47,79,80]. In response to convergent Wnt and FGF signaling, prospective caudal neural plate cells are initially specified either as cells of rHB character or as Hoxb4 þ /b8 þ /c9 þ cells characteristic of caudal/thoracic spinal cord. At early somite stages, caudal paraxial mesoderm and newly formed somites located adjacent to the prospective cHB and rSC, express high levels of Raldh2, providing a local source of RA [59,81]. Our results suggest that RA specifies   cells of r7/r8 cHB character by inducing Hoxb4 expression in prospective rHB cells and cells of rSC identity by preventing the expression of Hoxc9 in prospective cSC cells. At these stages, several Fgfs are expressed in the regressing Hensen's node and primitive streak adjacent to the developing spinal cord [8,46,79]. We find that FGF signals maintain the specification of cSC cells, and in the presence of RA, promote the generation of rSC cells. This model is strengthened by our data providing evidence that distinct combinations of Wnt, RA, and/or FGF signals can reconstitute rostral and caudal hindbrain and spinal cord character in naive prospective FB cells in a predictable manner. Genetic analyses have provided evidence that inactivation of Wnt genes, expressed in the caudal regions of gastrula stage mouse and zebrafish embryos, perturbs the development of the caudal neural plate [60,62,63,69,82]. However, the formation of paraxial mesoderm, which serves as a local source of neural caudalizing signals [63,[83][84][85], is also impaired in these mutant embryos [45,69,86]. Thus, these genetic studies left unresolved the issue whether the effect of perturbed Wnt signaling on caudal neural development reflects the impaired formation of paraxial mesoderm or reflects direct Wnt signaling in neural cells. Our in vitro studies establish direct effects of Wnt signals on neural tissue in the absence of other tissues and clarify the integrative mechanisms that control the early development of the hindbrain and spinal cord. FGF and retinoid signals also regulate the temporal pattern of differentiation of caudal neural progenitor cells. FGF has been shown to keep cells in a stem zone-like state [14,78,87], whereas RA promotes the differentiation of neural cells [14,78,88]. Consistent with these roles of FGF and RA, exposure of naive neural explants, cultured with Wnt and FGF to low levels of RA (2 nM), does not change the rostrocaudal identity of neural progenitor cells but results in an increased number of differentiated MNs (unpublished data), whereas under these conditions, increased levels of FGFs greatly reduce MN differentiation (unpublished data). These findings fit well with the suggested opponent activities for FGF and RA in deciding the balance between neural cell proliferation and differentiation [87].

Hox Gene Profiles of Hindbrain and Spinal Cord Progenitor Cells Predict the Pattern of dMNs and vMNs
The patterned expression of Hox genes in neural progenitor cells appears to be a major determinant of the identity of different MN populations [1,7,11,16,18]. Earlier studies have provided evidence that RA and FGF act in an opponent manner on caudal neural cells to establish the rostrocaudal pattern of distinct subclasses of differentiated MNs [7][8][9]. dMNs and vMNs represent two major MN subclasses that are generated in distinct rostrocaudal patterns in the hindbrain and spinal cord [10,89]. Our findings provide evidence that Wnt signaling in neural progenitor cells is required for the generation of both vMNs and dMNs. We also show that Wnt, RA, and/or FGF signals can induce cells with Hox gene profiles characteristic of cHB, rSC, and cSC progenitor cells that differentiate into corresponding dMNs and vMNs when exposed to Shh-N. These findings therefore reveal a tight link between Wnt, RA, and FGF signals, profiles of progenitor cell Hox expression, and the rostrocaudal pattern of dMN and vMN generation in the hindbrain and spinal cord.
Other recent studies have revealed a determinative role of Hox genes in MN subtype specification. In the hindbrain, Hoxb1 is expressed throughout r4, and has been shown to be required for the specification of facial branchiomotor neurons [17][18][19]. Similarly, targeted expression of Hoxa3 in the rHB leads to the generation of ectopic somatic MNs [11]. In the spinal cord, the rostrocaudal profile of genes of the Hox6 to Hox9 paralog group have been shown to establish distinctions in MN columnar and pool subtype [7,16]. It seems likely, therefore, that the initial profiles of Hox expression, shown here to depend on early Wnt signaling, are involved in establishing domains of dMN and vMN formation. Furthermore, Hox genes appear to represent a common regulatory target for the three classes of signaling factors-Wnts, FGFs, and RA [90]-that conspire to regulate the position of generation of dMNs and vMNs.
Thus, our results reveal that an early Wnt-based program is required to interact with a later RA-and FGF-mediated mechanism to generate a pattern of neural progenitor cells with Cdx and Hox profiles that prefigures the generation of two major subclasses of MNs in the developing hindbrain and spinal cord. Further studies will reveal how these three signals are integrated at the molecular level to regulate Cdx and Hox gene profiles, leading to the subsequent differentiation of dMN and vMN classes.

Materials and Methods
Embryos. Fertilized white leghorn chicken eggs were obtained from Agrisera AB, Umeå , Sweden. Chick embryos were staged according to Hamburger and Hamilton (HH) [91].
Culture of tissue explants. Explants were cultured in vitro as previously described [32]. To enable tracing the rostrocaudal orientation of the dissected explants, rostrocaudally asymmetrical explants were isolated and then placed in a defined orientation in collagen matrix, where their orientation was maintained during cultivation, fixation, and cryo-sectioning (see Figure 5A); or the caudal margin of the explants was labeled with DiI crystals (Molecular Probes) during the dissection process ( Figures 5D, 7B, 7C, and 7D). Recombinant human FGF4 (R & D Systems, Minneapolis, Minnesota, United States) was used at 30 and 60 ng/ml. The FGF receptor inhibitor SU5402 (Calbiochem, EMD Biosciences, San Diego, California, United States) was used at 3 and 5 lM. All-trans retinoic acid (RA) (Sigma-Aldrich, St. Louis, Missouri, United States) was used at 10-40 nM. Purified recombinant mouse Wnt3A (R & D Systems) was used at 150 ng/ml. Soluble Wnt3A and control-conditioned media [92] were obtained as described [32] and used at 50-100 ll/ml, which for Wnt3A conditioned medium, mimicked the activity 75-150 ng/ml of Wnt3A protein (R & D Systems). Soluble mouse Frizzled 8 (mFrz8CRD-IgG) [50] and control-conditioned medium were generated as described [32] and used at 300-500 ll/ml of culture medium. Explants cultured with control-conditioned medium behaved like explants cultured alone. In MN differentiation studies, Shh-N (R & D Systems) was used at 15 nM, and cultivated explants were washed to remove Wnt, RA, and/or FGF4 from the medium before addition of Shh-N.