Dlx proteins position the neural plate border and determine adjacent cell fates

The juxtaposition of presumptive neural plate and epidermis forms a signaling center responsible for the stereotypic pattern of dorsal neurons in the neural plate (including lateral primary neurons in amphibians and fish),roofplate, neural crest cells and cranial placodes(Baker and Bronner-Fraser,2001; Moury and Jacobson,1990; Selleck and Bronner-Fraser, 1995). In addition, signals from this region influence mesodermal structures such as somites(Lassar and Munsterberg, 1996;Pourquie, 2000) and heart(Raffin et al., 2000;Sarasa and Climent, 1987). Although considerable progress has been made recently in the molecular characterization of neural-inducing factors, relatively little is known about the molecules that determine the site of the border between neural plate and epidermis.

Diffusible proteins such as BMP, Wnt and FGF isoforms play an important role in patterning neural and non-neural ectoderm along the mediolateral axis in Xenopus and zebrafish. BMP antagonists, such as chordin and noggin, initiate neural induction in the dorsal ectoderm while BMP signaling in the ventral ectoderm represses neural fates and promotes epidermal differentiation (Brewster et al.,1998; Hemmati-Brivanlou and Melton, 1997; Kuo et al.,1998; Mizuseki et al.,1998; Nakata et al.,1997; Sasai et al.,1994; Smith et al.,1993; Wilson and Hemmati-Brivanlou, 1995). One model for neural plate border formation suggests that the ectoderm is differentially patterned by threshold levels of BMP signaling: high levels induce epidermal fates, low or absent signaling permits neural differentiation while intermediate levels induce border fates (Marchant et al.,1998; Morgan and Sargent,1997; Nguyen et al.,1998; Wilson et al.,1997). However, studies showing that not all aspects of neural and neural crest induction are recapitulated by modulating BMP levels in non-neural tissues (e.g. LaBonne and Bronner-Fraser, 1998) have prompted an examination of additional factors that might synergize with BMP to control cell fate at the border region of late blastula embryos (Bachiller et al., 2000; Klingensmith et al., 1999; Streit and Stern,1999). In particular, Wnt and FGF isoforms appear critical for induction of neural crest and cranial placodal cells(Adamska et al., 2000;Chang and Hemmati-Brivanlou,1998; LaBonne and Bronner-Fraser, 1998; Mayor et al., 1997; Phillips et al.,2001; Saint-Jeannet et al.,1997; Streit and Stern,1999; Vallin et al.,2001; Wilson et al.,2001).

Several transcription factors have been proposed to influence neural/non-neural ectodermal patterning. For example, Xiro, Xash-3 and Zic family members are induced by pre- to early gastrula stage dorsalizing and neuralizing signals, such as noggin, and with time their expression becomes localized to the future neural plate(Gomez-Skarmeta et al., 1998;Kuo et al., 1998;Mizuseki et al., 1998;Nakata et al., 1997;Turner and Weintraub, 1994). Overexpression of these factors in whole embryos expands the neural plate. Where examined, however, either an abnormally broad domain of neural crest marker expression overlaps the ectopic neural plate region or the neural crest markers are lost (Gomez-Skarmeta et al.,1998; Kuo et al.,1998; Mizuseki et al.,1998; Nakata et al.,1997; Turner and Weintraub,1994), indicating that normal border signaling has not occurred. In contrast, the Dlx family of transcription factors represent a class of proteins that, along with Msx-1 and PV.1, inhibit neural plate differentiation when overexpressed in Xenopus(Ault et al., 1997;Feledy et al., 1999;Pera and Kessel, 1999;Suzuki et al., 1997). Prior to gastrulation, transcripts encoding these factors are expressed diffusely in the ectoderm but subsequently become excluded from the developing neural plate(Luo et al., 2001a). At least one Dlx family member, Dlx3, has been shown to be regulated by BMP and canonical Wnt/β-catenin signaling that provide ventralizing signals in pre-gastrula stage Xenopus embryos(Beanan et al., 2000). These data suggest that Dlx genes might reinforce or refine the early ectodermal pattern established by BMP and Wnt signaling. Interestingly, the medial expression borders, adjacent to the neural plate, differ among Dlx family members, leading Luo et al., to propose that distinct cell fates might arise through differential action of Dlx family members(Luo et al., 2001b). These studies suggest a model in which Dlx activity may regulate the position of the neural plate border. A direct test of this model by loss-of-function has not been done nor have the consequences of shifting the endogenous spatial expression pattern of Dlx expression on adjacent cell fates been examined. Moreover, it is not certain whether the normal role of Dlx proteins is solely inhibitory. We originally postulated a positive role based on our observation that Dlx3 transcripts were downregulated in narrowminded, a zebrafish mutant that is deficient in Rohon-Beard neurons and that exhibits delayed appearance of neural crest cells (Artinger et al., 1999).

In this paper, we describe loss- and gain-of-function studies to investigate the role of Dlx genes in positioning the lateral edge of the neural plate and specifying adjacent cell fates. Dlx3 or Dlx5 homeodomains,which are highly conserved among Dlx family members, were fused to the Engrailed transcriptional repressor or the VP16 transcriptional activator domains to modulate transcription of genes regulated by Dlx. These fusion proteins were misexpressed in localized regions within Xenopusectoderm to modify endogenous Dlx function and subsequent alterations in cell fates were analyzed. Whereas we envisaged that pre-gastrula stage BMP, Wnt and FGF signaling biases ectodermal cells towards neural or epidermal fates, our results indicate that Dlx-dependent transcription positions of the lateral border of the neural plate and, importantly, is required in non-neural ectoderm to induce signals that specify the stereotypic pattern of neural crest cells and cranial placodes.

The VP16-Dlx3hd, EnR-Dlx3hd, VP16-Dlx3hd-GR and EnR-Dlx3hd-GR constructs were made by subcloning the homeodomain coding region of the zebrafishDlx3 by PCR using the following oligonucleotide primers:5′CCCCATCGATATGGTAAACGGAAAACCC (for the EnR constructs) or 5′CCCCATGTCTATGGTAAACGGAAAACCC (for the VP16 constructs) and 5′CCCCCGGCGCGCCCTTCTTGAACTTGGATCTC. The amplified fragments were then digested with ClaI and AscI (for the EnR constructs) orBglII and AscI (for the VP16 constructs) and ligated into the polylinker of vectors containing either the EnR repressing domain or VP16 activating domain and the GR ligand binding domain described by Kolm and Sive(Kolm and Sive, 1995),separated by a polylinker in a pCS2+(Turner and Weintraub, 1994)backbone that contains an SP6 promoter used to generate synthetic mRNA for microinjection. Enr-Dlx5hd-GR and VP16-Dlx5hd-GR plasmids were constructed in an identical manner by subcloning the Xenopus Dlx5 homeodomain.

Xenopus laevis embryos were fertilized in vitro, dejellied in 2%cysteine-HCl (pH 7.8), and maintained in 0.1×MMR. Embryos were reared at 14-22°C and staged according to Nieuwkoop and Faber(Nieuwkoop and Faber, 1994). Capped mRNA was synthesized using the mMessage Machine kit (Ambion). mRNA was injected into one dorsoanimal blastomere of four to eight cell stage embryos in 3% Ficoll in 1× MMR. For all VP16 constructs, 100-200 pg of mRNA was injected while for the EnR constructs, 50-80 pg of mRNA was injected. All injections included 250-300 pg of β-galactosidase mRNA to provide a lineage tracer. Dexamethasone (10 μM) was added at either stage 5/6, stage 8/9, stage 10 or stage 11.5-12 to induce the GR fusion proteins.

EnR-Dlx3hd mRNA along with β-galactosidase mRNA and GFP mRNA as lineage tracers were injected into two ventral animal blastomeres of four to eight cell stage host embryos. Control host embryos were uninjected or were injected with the lineage tracers alone. Donor embryos were first injected with rhodamine dextran (10×10³M_r;Molecular Probes) into both blastomeres of a two-cell stage embryos. Embryos were then cultured until stage 12 when a portion of the neural plate and subjacent mesoderm was transplanted into the ventral ectoderm of recipient stage 12 embryos. The fluorescent signal was acquired in monochrome and pseudo-colored green for clarity.

Embryos that were co-stained for β-galactosidase were fixed for 40 minutes at room temperature in MEMFA (0.1 M Mops pH 7.4, 2 mM EGTA, 1 mM MgSO₄, 3.7% formaldehyde), rinsed in 1× PBS with 2 mM MgCl₂ and incubated in staining solution at 37°C with Magenta Gal (Biosynth) before processing for in situ hybridization as described previously (Harland, 1991). The following plasmids were used to generate digoxigenin-labeled probes (RNA polymerase was used for linearization): pSp72-Xslug (BglII,SP6) (Mayor et al., 1995),pSp72-XSnail (BglII, Sp6) (Essex et al., 1993), pKS-XSox2 (XbaI, T7)(Mizuseki et al., 1998),pKS-Msx-1 (Hox7.1) (EcoRI, T7)(Su et al., 1991),pKS-Dll2 (Dlx3) (BglII, T7)(Papalopulu and Kintner,1993), pKS-XHairy2a (BamHI, T7) (J. W.),pGEM3-keratin (BamHI, SP6)(Jonas et al., 1989),pBSII-Xsix1 (NotI, T7)(Pandur and Moody, 2000),sp70-NCAM (EcoRV, SP6) and N-tubulin (BamHI, T3)(Oschwald et al., 1991;Richter et al., 1988). Stained embryos were postfixed in MEMFA and embedded in JB4 according to the manufacturer's directions (Polysciences) for histological examination.

Embryos were fixed in MEMFA and processed for immunohistochemistry(Hemmati Brivanlou and Harland,1989) using an EpA antibody(Jones, 1985) detected with an alkaline-phosphatase-conjugated secondary antibody.

The progressive refinement of the Dlx3 expression pattern in late cleavage to early gastrula stage Xenopus embryos(Fig. 1A-C) suggests that it mediates or responds to cues involved in neural plate border formation. During late blastula stages, Dlx3 mRNA is detectable throughout the ectoderm(Fig. 1A). By the onset of gastrulation, Dlx3 mRNA is reduced in the dorsal ectoderm and becomes localized ventrally to non-neural ectoderm as gastrulation proceeds(Fig. 1B). Expression is restricted exclusively to the non-neural ectoderm by the beginning of neurulation (Fig. 1C). OtherDlx family members have overlapping expression patterns at these stages (Luo et al., 2001a). To examine the role played by endogenous Dlx, we engineered transcriptional activating and repressing constructs using the homeodomain of Dlx3 or Dlx5(Fig. 1D). The homeodomains were fused to either the VP16 activation domain (VP16-Dlx3hd) or the engrailed repressor (EnR) domain (EnR-Dlx3hd). The amino acid sequences of the homeodomains are highly conserved among Dlx family members, sharing 80-88%positional identity. Critically, all 4 of the residues in helices three to four responsible for basepair-specific contacts between homeodomain proteins and the DNA major groove are identical(Dave et al., 2000;Laughon, 1991;Mathias et al., 2001). Therefore, these constructs are predicted to modulate target genes of all known Dlx family members, which is advantageous given their overlapping expression. Conditionally inducible versions of these constructs were made by fusion to the glucocorticoid receptor ligand binding domain. Cytoplasmic and nuclear factors are commonly expressed as GR fusion proteins to maintain transcription factors in an inactive heat shock complex until the addition of the dexamethasone (Kolm and Sive,1995). By manipulating the timing of dexamethasone addition, it is possible to investigate the temporal requirements of Dlx activity.

Ectopic activation of target genes by injection of VP16-Dlx3hd mRNA resulted in a loss of the early neural plate marker, Xsox2(Kishi et al., 2000) (91% of embryos showed a loss, n=69; Fig. 2B). This confirms that the anti-neural plate potential of the intact Dlx3 protein (Feledy et al.,1999) (Fig. 3C)depends on transcriptional activator function. In this and all subsequent experiments, β-galactosidase mRNA was co-injected to mark the progeny of injected blastomeres (magenta stain). We then investigated whether repression of downstream targets of endogenous Dlx proteins would have a reciprocal effect on neural plate formation. As shown inFig. 2C, injection of EnR-Dlx3hd mRNA expanded Xsox2 expression laterally on the injected side of the embryo (92% expanded, n=79). Control embryos injected with β-galactosidase mRNA alone showed no change in Xsox2expression (0% expanded, n=136;Fig. 2A). Similar results were seen when embryos were injected with VP16-Dlx5hd-GR (83% showed a loss,n=30) (not shown) and EnR-Dlx5hd-Gr (61% expanded Xsox2,n=31) (Fig. 2D) when dexamethasone was added immediately (stage 5/6).

To ensure that the effects on the neural plate were not limited toXsox2 or to early stages of neural induction, we examined the effects of the Dlx constructs on NCAM expression. NCAM was lost after VP16-Dlx3hd injection (93% lost, n=42) but was expanded laterally by EnR-Dlx3hd(89% expanded, n=57; Fig. 2E-G). Transverse sections of older stage (stage 27-30) embryos revealed that the neural tube had grossly normal morphology although the region marked by β-galacotosidase had expanded(Fig. 2I) as compared to the contralateral side or comparable region inFig. 2H.

Co-injection with full-length Dlx3 mRNA rescued the EnR-Dlx3hd effects in a dose-dependent manner (Fig. 3)indicating that the EnR-Dlx3hd phenotype reflects repression of normal targets of Dlx proteins. As would be epxected, the highest doses of full-length Dlx3 caused the overexpression phenotype (Fig. 3C). Thus, we conclude that the phenotypic effects of the Dlx fusion proteins reflect a modulation of natural Dlx target genes. Taken together, these data suggests that Dlx activity delimits the neural plate and prevents its expansion, consistent with conclusions drawn from overexpression studies (Feledy et al., 1999;Luo et al., 2001b).

We then asked if suppression or expansion of the neural plate is accompanied by compensatory changes in epidermal specification, which is visualized by expression of epidermal keratin and the epidermal-specific antibody, EpA (Jones,1985). VP16-Dlx3hd did not affect keratin or EpA levels when expressed in the neural plate region (keratin: 76% of embryos exhibited no change, n=84, Fig. 2K; EpA: 86% no change, n=59Fig. 2O) despite its marked ability to inhibit neural plate differentiation. In contrast, EnR-Dlx3hd suppressed keratin and EpA expression where injected(keratin: 95% suppressed, n=138Fig. 2L,M; EpA: 85%,n=73, Fig. 2P),consistent with expansion of the neural plate(Fig. 2C,G). Control embryos injected with β-galactosidase mRNA showed no change in keratinor EpA expression (keratin: 0% affected, n=129Fig. 2J; EpA: 4%,n=81, Fig. 2N). We conclude that although Dlx activity inhibits neural plate differentiation, it is not sufficient to redirect presumptive neural plate cells to adopt an epidermal cell fate.

The preceding experiments show that EnR-Dlx3hd can expand the neural plate. Expansion of the neural plate is also seen upon overexpression of Zic, Xash3 and Xiro family members, but the neural crest domain is either expanded diffusely or missing and lateral neurons occur ectopically(Gomez-Skarmeta et al., 1998;Kuo et al., 1998;Mizuseki et al., 1998;Nakata et al., 1997;Turner and Weintraub, 1994),most likely reflecting a disturbance of the normal patterning mechanism. We therefore examined a panel of markers to ask whether patterning of these cell lineages were similarly disordered or occurred normally in EnR-Dlx3hd-injected tissues. Xh2a and Xmsx-1 mark the neural plate border at the end of gastrulation (stage 12). Xsnail and Xslug are expressed in the neural crest, which appears during late gastrula/early neural plate stages (stages 12-14). N-tubulin marks the medial, intermediate and lateral rows of primary neurons in neurula and neural tube stage (stages 13-16) embryos (Chitnis et al.,1995). Cranial placodes arise from thickenings in the ectoderm immediately lateral to the neural plate and give rise to a wide variety of derivatives, including paired sense organs (reviewed byBaker and Bronner-Fraser,2001). Xsix1, a homeobox-containing transcription factor,is expressed in the cranial placodal thickenings present in an anterior and lateral band of early neurulae, marking the prospective olfactory anlagen and persists in the late neurulae, where it marks the olfactory, otic, trigeminal and dorsolateral placodes (Pandur and Moody, 2000).

We frequently observed that each marker was displaced to the lateral margin of the areas that contained the injected EnR-Dlx3hd and β-galactosidase mRNAs (Fig. 4;Table 1). Notably, the size of the Xh2a and Xmsx expression domains were not changed when displaced laterally (Fig. 4C,F)nor was the premigratory neural crest field, marked by Xslug andXsnail, altered (compare the injected and uninjected sides inFig. 4I,N,P). Similarly, the characteristic three rows of primary neurons, marked by N-tubulin,arose in the expanded neural plate in EnR-Dlx3hd-injected embryos,but their pattern was shifted outwards such that the lateral neurons were now positioned along the new neural plate border(Fig. 4S). An identical result was observed in EnR-Dlx3hd-injected zebrafish as well(Fig. 4T,U) suggesting that a Dlx-dependent mechanism positions lateral neurons in both species. The neurons remained tightly organized in rows and ectopic neurons were not observed. The placodal marker Xsix1 was also shifted outwards in embryos expressing EnR-Dlx3hd (compare Fig. 4V with 4X). Interestingly, Xsix1 was displaced to a region just lateral to the domain of EnR-Dlx3 expression, consistent with expression outside the expanded neural plate. Lateral displacement of trigeminal placodes was also observed in embryos stained with N-tubulin, which also marks these placodes (not shown). The ability of EnR-Dlx3hd to displace the stereotypic pattern of marker expression laterally by late gastrula stages suggests that endogenous Dlx activity is upstream of neural crest, lateral primary neuron and cranial placode precursor specification.

While the cell fate markers were often displaced laterally to the border of the EnR-Dlx3hd region, we also observed a loss of markers in some embryos(Table 1,Fig. 4Y and see below). Loss was correlated with large domains of injected EnR-Dlx3hd that extended to the ventral side of the embryo. Thus, depletion of endogenous Dlx activity permits the neural plate to expand maximally to about twice its normal size,regardless of whether or not a larger area expresses EnR-Dlx3hd. The loss of markers, however, raised the possibility that Dlx proteins are required in the non-neural ectoderm to provide neural crest inducing signals. This hypothesis is tested by the transplant experiments below.

In contrast to the loss-of-function studies, injection of VP16-Dlx3hd or wild-type Dlx3 did not shift marker expression medially, but rather abolished expression of both border and neural crest markers(Table 1 andFig. 4B,E,H,M,R). The sole exception was Xsix1, which lies outside the neural plate normally(Fig. 4W). Medial displacement occurred when only a small area of the neural plate contained the injected VP16-Dlx3hd mRNA; large patches of VP16-Dlx3hd-injected tissue resulted in a loss of Xsix1. The general inability to shift these markers medially suggests that ectopic Dlx activity in presumptive neural plate cannot induce factors needed for correct patterning of the cell lineages that arise at the neural plate border. However, the medial displacement of Xsix1 that occurs when only a minimal gap separates the neural plate from non-neural ectoderm suggests that short-range communication between neural and non-neural ectoderm is critical for placode formation.

To investigate the temporal requirement for Dlx function in neural plate border formation, embryos were injected as above but with mRNAs encoding the conditional proteins EnR-Dlxh3-GR or VP16-Dlx3hd-GR. Dexamethasone was added at either stage 6, 8, 10 or stage 11.5-12 and embryos were examined forXsox2 and Xslug expression. Injected embryos cultured in the absence of dexamethasone exhibited no change in any of the markers tested (not shown). When dexamethasone was added at stage 6, embryos injected with EnR-Dlx3hd-GR (70% expanded, n=34) or VP16-Dlx3hd-GR (80% repressed,n=25) showed the expansion or repression of Xsox2,respectively (Fig. 5A,B),typical of the preceding results with the constitutive constructs. Similarly,the neural crest marker Xslug was either lost (48% lost,n=39) or shifted laterally (52% shifted laterally, n=39;Fig. 5C) when EnR-Dlx3hd-GR-injected embryos were treated with dexamethasone at stage 6. The incidence of expanded Xsox2 expression and laterally shiftedXslug (Fig. 5I)decreased as dexamethsone was added at later stages. Addition at stage 11.5-12 did not affect Xsox2 expression (wild-type pattern was observed in 86% of EnR-Dlx3-GR embryos, n=66 and in 90% of the VP16-Dlx3hd-GR-injected embryos, n=93;Fig. 5E,F). Similarly,EnR-Dlx3-GR did not alter Xslug expression when dexamethasone was added at stage 11.5-12 (65% were unaffected, n=126;Fig. 5G). Therefore,specification of the neural plate border and induction of neural crest requires Dlx function before the end of gastrulation. We did notice, however,that VP16-Dlx3hd-GR eliminated or reduced Xslug expression regardless of whether dexamethasone was added at stage 6 (88% eliminated or reduced,n=20; Fig. 5D) or stage 11.5-12 (72% eliminated or reduced, n=114;Fig. 5H). Thus, neural crest,but not neural plate, remains sensitive to Dlx inhibition after late gastrulation.

The preceding results showed that local depletion of Dlx function during gastrulation causes an expansion of the neural plate and the lateral displacement of the normal pattern of lateral primary neuron, neural crest and cranial placode cell fates. This could occur by a repressive mechanism whereby Dlx genes simply prevent neural plate formation. However, when the area of EnR-Dlx3hd expression extended to the most ventral ectoderm, we noticed that the neural plate expanded maximally to only about twice its normal distance from the dorsal midline and that all of the cell lineage markers examined were absent (Fig. 6A-D; identical results obtained with EnR-Dlx5hd, not shown). This suggests that Dlx proteins not only repress cell fates, but might also be necessary in non-neural ectoderm for the production of factors that are critical for specification of lateral primary neuron, neural crest, and cranial placode cells.

To determine if Dlx function is required in non-neural ectoderm, we took advantage of previous studies showing that grafting neural plate tissue to the non-neural ectoderm causes neural crest cells to be induced in both host and donor tissues where juxtaposed (Mancilla and Mayor, 1996; Moury and Jacobson, 1990; Selleck and Bronner-Fraser, 2000). Regions of neural plate from fluorescent dextran-injected donors were grafted to the ventral ectoderm of host embryos that had been injected with nuclear-localized β-galactosidase mRNA,either alone or with EnR-Dlx3hd mRNA (Fig. 6E). We then assayed the graft region for the induction ofXslug and Xsix1 to determine whether the induction of neural crest and cranial placodes is inhibited when Dlx activity is downregulated. As expected, control grafts showed Xslug expression induced at the graft-host tissue interface (blue stain,Fig. 6E inset). Higher magnification views (Fig. 6J-N)show induction of both Xslug and Xsix1 (blue stain) at the interface of the graft (green fluorescent label) and host (magentaβ-galactosidase stain) tissues, confirming that juxtaposition of competent neural plate and non-neural ectoderm induces neural crest and cranial placodes. In contrast, Xslug and Xsix1 were not induced when neural plate was transplanted into ventral ectoderm expressing EnR-Dlx3hd (Fig. 6O-S;Table 2). In cases where donor neural ectoderm spanned EnR-Dlx3hd-expressing and -non-expressing host tissue(Fig. 6P), Xslug was induced only at the site where the host tissue lacked injected mRNAs (black arrowhead) but not in or near cells with the injected mRNA (red arrowhead). No or minimal Xslug and Xsix1 transcripts were detected in donor neural ectoderm cultured alone (Fig. 6F,G) whereas similar explants taken from a more lateral position that spanned the border region expressed Xslug and Xsix1(Fig. 6H,I), indicating that expression in the grafts reflects induction and not contamination with donor border region tissue. Taken together, these studies demonstrate that Dlx-dependent transcription is required in non-neural ectoderm to produce factors that act at over short range to induce neural crest and cranial placode fates.

Gain-of-function experiments showed that Dlx genes position the interface between the neural plate and non-neural ectoderm by locally inhibiting neural plate differentiation (Fig. 2),confirming previous observations that full-length Dlx3 and Dlx5 antagonizes neural plate formation (Feledy et al.,1999; Luo et al.,2001a). Loss-of-function experiments also support this conclusion(Fig. 2) and can be rescued with full-length Dlx3 (Fig. 3),suggesting that our constructs specifically affect natural targets of the full-length Dlx family of proteins. Because of the high degree of conservation among Dlx homeodomains and overlapping expression during late blastula and gastrula stages, our results reflect perturbation of endogenous targets of Dlx proteins in aggregate and we cannot ascribe activities to individual family members or address questions of functional redundancies. Further studies will be required to determine the precise function of individual family members.

A central finding of our study is that local inhibition of Dlx function shifts the stereotypic pattern of lateral primary neurons, neural crest and cranial placodes laterally (Figs2,3,4). Thus, Dlx family members act upstream of the specification of cell fates that arise near the lateral border of the neural plate. Furthermore, Dlx genes are required for the induction of these cell fates, as revealed by the lack of marker expression when neural plate is grafted to EnR-Dlx3hd-expressing ectoderm (Fig. 6).Fig. 7 shows a schematic model of Dlx function at the lateral margin of the neural plate. Normally, Dlx activity in non-neural ectoderm (yellow) delimits the mediolateral position of the neural plate/non-neural ectoderm border and lateral neurons, neural crest and cranial placode cells (Fig. 7B). Inhibition of Dlx function in a localized region results in both an expansion of the neural plate and a lateral displacement of border region cell fates to a position abutting Dlx-positive ectoderm(Fig. 7C). Inhibition of a broader region of Dlx activity, however, causes the neural plate to expand maximally to about twice its normal distance from the dorsal midline, but normal border region cell lineages do not arise(Fig. 7D). The absence of stereotypic border region markers in these cases suggests that Dlx activity functions in the prospective epidermis to induce short range signals that specify border cell fates. In this model, the short-range signals are unable to traverse the interposed tissue (Fig. 7D). Ectopic Dlx activity generally fails to displace border cell fates medially (Fig. 7E). In these cases, as with broad regions of Dlx underexpression, we presume that signals needed to induce and pattern lateral neurons, neural crest and cranial placode cells cannot traverse the interposing tissue, which does not express neural or epidermal markers (Fig. 2).

Stereotypic displacement of border fates was not observed in previous studies in which the neural plate was expanded by overexpression of Zic family members, XBF-2, and Xash-3 (Kuo et al.,1998; Mariani and Harland,1998; Mizuseki et al.,1998; Nakata et al.,1997; Turner and Weintraub,1994). In these cases, the domains of neural crest and lateral neurons were either expanded or eliminated, indicating that these genes control differentiation of neural plate cells and their derivatives rather than regulate formation of a normal border region per se.

We did not see any consistent effects on the forebrain marker Otx2when embryos were injected with either the activating or inhibiting constructs(not shown). We also did not see a shift in the anterior portion of theXsix1 domain (not shown). Previous studies have shown that overexpression of Dlx3 does not affect Otx2 expression(Feledy et al., 1999). These results suggest that Dlx genes function primarily posterior to the forebrain. Positioning the mediolateral and anteroposterior positions of forebrain neurectoderm might rely on a separate mechanism, possibly involving Wnt, FGF or retinoic acid signaling (Villanueva et al., 2002).

The modulation of late gastrula stage markers(Fig. 4A-I) and the timecourse experiment of EnR-Dlx3hd-GR activation by dexamethasone addition(Fig. 5) both indicated that Dlx activity patterns cell fates at the medial neural/non-neural ectodermal interface by stages 11.5-12. Because this time precedes the morphological appearance and expression of molecular markers of neural folds and lateral primary neurons, neural crest cells and cranial placodes, Dlx activity is needed upstream of the process that induces these cell fates. Classical graft and extirpation experiments suggested that the anterior neural plate border is no longer susceptible to signals that reposition the border by late blastula stage (stage 9+) (Zhang and Jacobson,1993). While the mediolateral border was not examined specifically, it is possible that these studies revealed the action of diffusible signals, such as BMP and Wnt, that bias ectodermal fates. Our data indicates that endogenous Dlx activity is required somewhat later, through stage 11.5-12 (Fig. 5),consistent with Dlx genes being regulated by BMP and Wnt signaling. Interestingly, the stages when Dlx activity is required corresponds to the time when the transcripts encoding Dlx, Xmsx1 and Xiro become spatially localized to respect the neural plate border. The potential involvement of Dlx proteins with these and other proteins to regulate their own expression and influence neural, epidermal and border region cell fates is discussed below.

Current models for partitioning ectoderm into neural plate and epidermis involve initiating Wnt, BMP and FGF signals(Baker et al., 1999;Barth et al., 1999;Ikeya et al., 1997;Launay et al., 1996;Streit and Stern, 1999;Wilson and Hemmati-Brivanlou,1995; Wilson et al.,2001). Wnt signaling through the canonical β-catenin pathway provides dorsal cues during cleavage stages (e.g.Heasman et al., 1994;Larabell et al., 1997). By pre-gastrula stages, ectopic activation of BMP and Wnt/β-catenin signaling pathways elicit a ventralizing effect on embryos suggesting that these proteins might normally antagonize induction of the neural plate at these stages (Baker et al.,1999; Christian and Moon,1993; Hawley et al.,1995; Sasai et al.,1995; Wilson and Hemmati-Brivanlou, 1995;Wilson et al., 2001). At leastDlx3 is negatively regulated by the pre-MBT Wnt/β-catenin signaling (Beanan et al.,2000), and induced by pre-gastrula BMP signaling(Feledy et al., 1999); thus,Dlx genes probably act downstream of these proteins. Whether Dlx activity feeds back to influence gastrula-stage BMP and/or Wnt signaling is unclear.

Dlx proteins might regulate border cell fates by cooperating with other transcription factors that are induced by early BMP, Wnt and/or FGF signaling. Injection of VP-Dlx3hd prevents induction of markers of lateral primary neurons, neural crest and cranial placodes(Fig. 4). Moreover, epidermal markers were not induced (Fig. 2), suggesting that Dlx activity alone is insufficient to respecify neural plate to non-neural ectoderm (see diagram inFig. 7E) and implicating an additional factor. One candidate is Msx1, which is an immediate downstream target of BMP signaling localized to the ventral ectoderm in early gastrula stage Xenopus embryos (Feledy et al., 1999; Suzuki et al.,1997). Overexpression of Msx1 represses neural plate and border cell fates, but unlike Dlx3, also converts prospective neural plate into epidermis (Suzuki et al.,1997). Neural inhibition by Msx-1, unlike Dlx factors, appears to act via transcriptional repression(Yamamoto et al., 2000),suggesting that Msx-1 and Dlx factors regulate distinct target genes. Since Msx and Dlx proteins can heterodimerize(Zhang et al., 1997),regulatory interaction between the factors is also likely. Msx-1 loss-of-function experiments are needed to address its role in cell fate specification at the neural plate border. Defects of Msx1-deficient mice are not known to involve an expanded neural plate, but potential redundancy or compensation complicates interpretation(Jumlongras et al., 2001;Satokata and Maas, 1994). Nonetheless, it seems possible that induction of neural crest, lateral neurons and cranial placodes might require both Dlx activity and the epidermal promoting activity of Msx1.

The lateral displacement of border cell fates by Dlx genes could indicate a genetic interaction with prospective neural plate factors (see diagram inFig. 7A). One candidate isXiro1, which is initially expressed broadly in the dorsal ectoderm at the onset of gastrulation but becomes restricted to the prospective anterior neural plate as gastrulation proceeds(Gomez-Skarmeta et al., 1998). Overexpression of Xiro1 expands the neural plate and, in some instances, can displace Xslug expression outward(Gomez-Skarmeta et al., 1998). Furthermore, ectopic Xiro1 downregulates endogenous Bmp4expression whereas ectopic Xmsx1 downregulates endogenousXiro1 (Gomez-Skarmeta et al.,2001). These data combined with our results raise the intriguing possibility that Dlx genes, like Xmsx1, are involved in a mutual repression circuit with Xiro1.

In summary, our results support a model in which diffusible BMP, Wnt, and FGF signaling establishes graded expression patterns of transcription factors involved in positioning the neural plate border and determining adjacent cell fates. The neural plate antagonizing activities of Dlx proteins probably function in a reciprocal (inhibitory) circuit with transcription factors that promote neural plate, such as the Xiro-1 protein. The consequence of this interaction is to sharpen the border between the neural and non-neural ectoderm. In addition, we showed that Dlx genes also play a positive role in non-neural ectoderm for the production of factors that are required for the induction and stereotypic mediolateral patterning of lateral neurons, neural crest and placode cells.

The authors thank Nancy Papalopulu (MRC, Cambridge) for providingXenopus Dlx cDNAs, Marie-Andree Akimenko (Loeb Research Institute,Ottawa) for providing zebrafish Dlx3 cDNA, Sally Moody (George Washington University, Washington D.C) for providing SixI cDNA, Michael Sargent (NIMR,London) for providing Xslug and Xsnail cDNA and Yoshiki Sasai for providing Xsox2 (Kyoto University, Kyoto). We also would like to thank Ruchika Gupta for constructing the pCS2+BA activator and repressor plasmids and Kim Bettano,Stacey Coleman, Chrissy Evola and Chris Simpson for technical assistance. We gratefully acknowledge support from the American Cancer Society(RPG97-121-01-DDC) and NIH (RO1HL59502) to M. M. and a Medical Foundation postdoctoral fellowship to K. B. A.