LSox5 regulates RhoB expression in the neural tube and promotes generation of the neural crest

Since the discovery of Sry (Gubbay et al., 1990), many transcription factors have been characterised and included in the Sox family (Sry-related HMG box), based on their homology to the HMG DNA-binding domain. These proteins are involved in a number of developmental processes, making the Sox family one of the most prolific of its kind with more than 30 vertebrate members recognised to date. Through the phylogenetic analysis of the HMG box, Sox family members have been classified into 11 groups (Bowles et al.,2000). Together with Sox6 and Sox13, Sox5 belongs to group D.

SoxD genes contain more exons than do other Sox genes(Schepers et al., 2002). Additional complexity is introduced through tightly regulated mRNA processing,the functional significance of alternative splicing being highlighted by the differential expression of the distinct isoforms. Indeed, a short variant of Sox5 (SSox5) is found only in adult testis, while longer isoforms (LSox5) are expressed in other tissues during mammalian development(Denny et al., 1992; Lefebvre et al., 1998; Hiraoka et al., 1998). LSox5 variants share common structural features with other group D members,including a leucine zipper and glutamine-rich regions thought to be involved in dimerisation (Hiraoka et al.,1998; Lefebvre et al.,1998). In the mouse, LSox5/Sox6 heterodimers cooperate with Sox9 to activate the Col2a1 gene, which encodes type II collagen, an extracellular matrix component that is essential for chondrogenesis (Lefebvre et al.,1998). Indeed, it is in this process that LSox5 has been most thoroughly studied. Mutations in SOX9 produce campomelic dysplasia, a syndrome that is often associated with autosomal XY sex reversal and involves the severe malformation of most cartilage-derived structures(Meyer et al., 1997). Furthermore, mouse chimeras containing Sox9^–/–embryonic stem cells inactivate early cartilage markers, including the Col2a1 gene (Bi et al.,1999), a phenomenon that is also observed in Sox5, Sox6double mutant embryos (Smits et al.,2001). This scheme is further clarified by the severe downregulation of LSox5 and Sox6 expression produced after Sox9 inactivation using the Cre/LoxP system in chondrogenic cell lineages, thus demonstrating that LSox5 and Sox6 are genetically downstream of Sox9 in chondrocytes(Akiyama et al., 2002). Interestingly, in the chick, Sox9 has been implicated in the differentiation of the neural crest (Cheung and Briscoe, 2002) and here we characterise chick LSox5 and show its participation in cranial neural crest development.

The neural crest is a cell population that originates at the boundary between the neural plate and the prospective epidermis. Once specified, the cells of the neural crest delaminate from the neural folds/neural tube by undergoing a process of epithelium to mesenchyme transition (EMT). These cells then migrate along characteristic pathways and differentiate into a wide variety of derivatives upon reaching their destination, including neurons and glia of the peripheral nervous system, pigment cells, craniofacial cartilage and bone (LeDouarin and Kalcheim,1999). Diffusible factors of the BMP, BMP antagonists, FGF and WNT families, together with retinoic acid, seem to direct the first steps of neural crest induction (Liem et al.,1995; Selleck et al.,1998; Garcia-Castro et al.,2002; Villanueva et al.,2002), although different transcription factors are thought to interpret these extracellular signals. Indeed, the winged-helix family member FoxD3 and Sox9 have been implicated in the induction of the neural crest(Kos et al., 2001; Dottori et al., 2001; Cheung and Briscoe, 2003); and members of the Snail family of transcription factors, Snail and Slug, are required to trigger the EMT (reviewed in Nieto, 2002). The neural crest derivatives become committed at different stages of their development and some transcription factors seem to direct this lineage commitment. At low concentrations, Sox10 apparently maintains the multipotency of the crest cells and at higher doses it inhibits neuronal differentiation, favouring the generation of peripheral glia and melanoblasts(Kim et al., 2003; Paratore et al., 2001). by contrast, FoxD3 represses melanogenesis(Kos et al., 2001; Dottori et al., 2001) and it has been demonstrated that Sox9 is necessary for the determination of the chondrogenic lineage in cranial neural crest cells(Mori-Akiyama et al.,2003).

We have characterised chick LSox5, which shows a high degree of similarity to its mammalian counterpart and an identical HMG box. Two splice variants were found that contain specific structural motifs susceptible to post-translational modification. We show that LSox5 is expressed in premigratory cranial neural crest cells and that following delamination, LSox5 expression coincides with a characteristic pattern of crest migration. High expression levels are maintained in the crest-derived components of cranial glia, including Schwann cells and satellite glia. Misexpression of LSox5 in the midbrain and the hindbrain provoked the rapid, cell autonomous upregulation of RhoB. This misexpression of LSox5 also led to an extension of the dorsal territory and of the developmental window in which the neural crest is produced, and augmented the generation of cephalic neural crest. The less immediate effects of LSox5 include the upregulation of other neural crest markers, such as Slug, FoxD3 and Sox10. Thus,we propose that LSox5 participates in the generation of the cranial neural crest and in the subsequent differentiation of the cranial glia lineage.

Fertilised chicken eggs were purchased from Granja Santa Isabel(Córdoba, Spain). Eggs were incubated, opened and staged according to Hamburger and Hamilton (Hamburger and Hamilton, 1951).

Chicken LSox5 was isolated in a subtractive screen to isolate genes involved in the early stages of brain development. A cDNA subtraction library was generated with the `Clontech PCR select-cDNA subtraction kit'(Clontech, catalog number K1804-1), using the anterior region of stage 8 and stage 11 chick embryos as the `driver' and `tester' populations, respectively. One of the clones isolated in the screen was a partial cDNA for LSox5, and the full-length cDNAs of LSox5-I and LSox5-II (2.3 and 2.4 kb, respectively) were obtained by PCR walking using a stage14 lambda zap cDNA library. The nucleotide sequence of LSox5 appears in the EMBL, GenBank and DDBJ Nucleotide Sequence Databases under the Accession Numbers AJ626988 and AJ626989. A putative short isoform (SSox5) results from the deletion of the first 1248 coding nucleotides of LSox5-II. Hence, we analysed the expression of the alternatively spliced LSox5 isoforms by RT-PCR using the flanking oligonucleotides: sense 5′-CTCCCAGCCTTTCACCTTCCC-3′; and antisense 5′-GCTTTCTGGAGTCCCTTTTAT-3′.

Full-length LSox5-II (772 amino acids) and SSox5 (356 amino acids) cDNAs were cloned into a pCX vector(Niwa et al., 1991) to be used in electroporation assays. Each plasmid (4 mg/ml) was injected between the neural folds of stage 8 chicken embryos together with pCX-EGFP (0.5 mg/ml), and the embryos were immediately electroporated as described previously (Itasaki et al.,1999) with some modifications. Electroporation was performed with a TSS20 Intracel square pulse generator, programmed to deliver five 50 mseconds, 15 V pulses at 4 Hz, through custom-made platinum electrodes. The eggs were sealed and allowed to develop for another 6-7, 12 or 20-24 hours, at which point the embryos were removed and processed for in situ hybridisation or immunohistochemistry. Control electroporation assays were performed by injecting pCX-EGFP (4.5 mg/ml) alone.

Single or double whole-mount in situ hybridisation was carried out as described (Nieto et al.,1996). For double hybridisation, one probe was labelled with digoxigenin-UTP and the second with fluorescein-UTP. The probes were detected with anti-digoxigenin or anti-fluorescein antibodies coupled to alkaline phosphatase, and visualised with NBT-BCIP (digoxigenin) or INT-BCIP(fluorescein; all reagents and antibodies supplied by Roche). The LSox5 probes were prepared from the full-length LSox5-IIcDNA isoform. The fluorescein-labelled LSox5 probe was often used to control for electroporation without masking the blue staining of the probe under analysis. Embryos were then photographed prior to embedding in gelatin or agarose for vibratome sectioning.

The LSox5-II cDNA was cloned into the pRSET plasmid (Invitrogen),from which a fusion protein was generated to immunise rabbits and obtain antisera. Protein extracts from embryonic fibroblasts stably infected with RCAS vectors expressing LSox5-I or LSox5-II were analysed by western blotting to test the affinity and specificity of the sera. One serum,32A-III, specifically recognised LSox5 in western blots and produced a pattern of immunostaining comparable with that obtained by in situ hybridisation (not shown). This serum was used throughout this work. For immunohystochemistry, 10μm cryostat or 40 μm vibratome agarose sections were permeabilised with 0.5% Triton X-100 (USB), blocked with 10% FBS and incubated overnight at 4°C with the primary antibody. After washing, the cryostat sections were incubated for 1 hour with secondary antibodies or overnight in the case of agarose vibratome sections. The primary antibodies were used at the following concentrations: LSox5, 1:4000; EGFP, 1:1000 (Molecular Probes); Laminin 1,1:1000 (Sigma); Pax7, 1:1000 (DSHB)(Ericson et al., 1996); Slug,1:1000 (62.1E6, DSHB) (Liem et al.,1995); P₀, 1:1000 (1E8, DSHB) (Bhattacharyya et al.,1991); HNK-1, 1:4000 (prepared from a cell line obtained from ATCC);Islet-1/2, 1:1000 (40.2D6, DSHB) (Ericson et al., 1992). Cy2- or Cy3-conjugated secondary antibodies were used (Jackson; 1:1000 dilution). HNK-1 whole-mount immunohistochemistry was performed as described (Nieto et al.,1996).

In an attempt to isolate genes involved in the early stages of chick brain development, we have identified clones corresponding to chick LSox5in a subtractive screen where cDNAs from stage 8 chick embryos were compared with those isolated from stage 11. cDNA clones that showed a strong similarity to mammalian LSox5 and that were more intensely expressed at stage 11 were identified. Then, a full-length cDNA was generated by PCR cDNA walking,which rendered two distinct isoforms. The longer LSox5-II differed from the shorter LSox5-I sequence because of a 105 bp insertion that encoded additional amino acids at position 56 (the amino acid sequence of chicken LSox5 and its alignment with the human orthologue is shown in Fig. S1). By comparing the LSox5 cDNA sequence with public database entries, we mapped the chicken Sox5 gene to chromosome 1 and established an exon/intron organisation very similar to that described for its mammalian orthologues (Ikeda et al.,2002) (the location and sequence of exon/intron boundaries is shown in Table S1).

The two cDNA isoforms isolated in the chick encode proteins of 737 and 772 amino acids, predicted to be of 80.99 and 84.95 kDa, respectively. The amino acid sequences are highly similar to their mammalian counterparts, LSox5-II sharing 95.7% and 93% of its residues with its human and mouse orthologues,respectively. The HMG box is identical in all three species, but out of this region, the chick LSox5-II contains 113 serine/threonine residues, 10% more than its mammalian orthologues. Strikingly, the majority of the additional S/T residues are concentrated between amino acids 426-456, and within this region,four serines and one threonine are concentrated in a short nine amino acid stretch. In silico screening of EST databases, together with the analysis of the corresponding genomic entries has shown that the inclusion of this serine-rich domain (SRD) varies across species. It is constitutively incorporated in zebrafish and fugu LSox5 polypeptides, absent in human, mouse and rat, and it may be alternatively spliced in the chicken (see Fig. S2shows the DNA sequence at the intron-exon boundary of different species,providing a molecular basis for this diversity). These phylogenetic variations may reflect the evolution of mechanisms that regulate LSox5 activity which involve differential phosphorylation.

We analysed the expression of the two distinct LSox5 isoforms by RT-PCR on cDNA samples from stages 9-23, with a pair of primers flanking the inserted domain. Two bands corresponding in size to that predicted for each isoform were detected at all stages and at the same intensity. Furthermore, in western blots of whole embryo protein extracts, a serum generated against LSox5 also detected two bands of similar intensity and with apparent molecular masses slightly higher than those predicted from the primary structure (the expression of LSox5 as obtained by both RT-PCR and western blot is shown in Fig. S1). Thus, we detected LSox5 in embryos at all the stages examined and no differences in expression could be detected between the two isoforms during development, either by RT-PCR or immunoblotting. Nevertheless, as these studies were performed on whole embryos, we cannot exclude the possibility that local differences do indeed exist.

The long splice variants derived from the Sox5 gene were first identified in the mouse. They have been shown to play a crucial role in chondrogenesis (Lefebvre et al.,1998) and to be expressed in the developing pancreas, as well as in a variety of adult human tissues (Ikeda et al., 2002; Lioubinski et al., 2003). At early stages of chick development, LSox5mRNA appeared to be distributed in the neural crest and its derivatives(Fig. 1). The expression of LSox5 was first detected at stage 7 in the neural folds(Fig. 1A), closely following the onset of the expression of other neural crest markers such as Slug (data not shown) and FoxD3(Kos et al., 2001). At all the stages analysed, LSox5 was also expressed at high levels in the notochord (Fig. 1C).

Within the neural plate, the expression of LSox5 was dependent on its anteroposterior location, as transcripts were found both dorsally and ventrally at prosencephalic levels (Fig. 1B), whereas at more caudal levels they were restricted to the most dorsal region where the neural crest forms(Fig. 1C). As the neural folds extended caudally, so did the dorsal restriction of LSox5 expression. At stages 8 and 9, LSox5 continued to be expressed in the anterior prosencephalon, although this region does not produce neural crest(Fig. 1D,F) and nor were neural crest markers such as Slug and FoxD3 expressed here(Fig. 1E)(Kos et al., 2001). At later stages, LSox5-expressing cells became migratory(Fig. 1G) and were found among the population of mesencephalic and rhombencephalic migratory crest cells(Fig. 1I). LSox5expression then progressively diminished in the dorsal cephalic neural tube caudal to the forebrain, coincident with its apparition in cells that were migrating along well-defined neural crest pathways(Anderson and Meier, 1981; Tosney, 1982; Lumsden et al., 1991).

To confirm whether the migratory cells that express LSox5 were neural crest cells, we examined the expression of Slug, a marker of the premigratory and the migratory neural crest population(Fig. 1J)(Nieto et al., 1994; del Barrio and Nieto, 2004). Within the neural tube, all cells that expressed LSox5 also expressed Slug(Fig. 1J-L), confirming that LSox5 is expressed in premigratory neural crest cells. Nevertheless, a subpopulation of cells only expressed Slug, indicating that LSox5 is not common to all premigratory crest cells. In the early stages of migration, many cells expressed both these proteins, although as migration proceeded, LSox5 expression augmented while the expression of Slug diminished(del Barrio and Nieto, 2004). The onset of Slug expression prior to that of LSox5indicated that Slug precedes LSox5 in the genetic cascade at play during neural crest development.

We then compared the expression of LSox5 with that of other neural crest markers and its distribution in neural crest derivatives. RhoB is a small GTPase implicated in the delamination of the neural crest that lies downstream of Slug (Liu and Jessell,1998; del Barrio and Nieto,2002). RhoB mRNA is expressed in a small population of premigratory neural crest cells and in the early migratory cells. Its expression pattern is very similar to that of LSox5 in the premigatory neural crest cells, suggesting that at a particular stage of development they may be co-expressed in neural crest cells. The glycoprotein epitope HNK1 (Tucker et al.,1984) is found in the majority of migratory neural crest cells (Le Dourain and Kalcheim, 1999; del Barrio and Nieto, 2004). Although most migratory neural crest cells were labelled for both HNK1 and LSox5 (Fig. 1M), an early migratory cell population contained LSox5 alone.

In the head, all the peripheral glia and most neurons in both the sensory and autonomic ganglia are lineages derived from the neural crest (reviewed by LeDourain and Kalcheim, 1999). The population of peripheral glia includes satellite glia of the cranial ganglia and the Schwann cells associated with the cranial nerves. Our immunohistochemical analysis revealed that LSox5 expression was maintained in both these glial cell types. In Schwann cell precursors that ensheath the cranial nerves, LSox5 is co-expressed with P₀, an early marker of both myelinating and nonmyelinating cells(Bhattacharyya et al., 1991) (Fig. 1N). In the cranial ganglia, anti-LSox5 labelled small rod-shaped nuclei whose morphology indicated that they might correspond to satellite glia. This population of LSox5-positive cells was clearly different from the neuroblasts that express Lim-domain factors of the Islet class(Fedtsova et al., 2003)(Fig. 1O).

To investigate whether LSox5 plays a role in the migration and/or differentiation of the neural crest, we ectopically induced expression of the longest LSox5-II isoform on the right-hand side of the neural tube. This was achieved by electroporating pCX-LSox5 (together with pCX-EGFP as a marker of transformation) into the cephalic region of stage 8 embryos. After 20-24 hours, uniform EGFP labelling was detected on one side of the neural tube and in the characteristic streams of migratory cranial neural crest (Fig. 2A). A similar distribution of ectopic LSox5 was observed by immunohistochemistry(Fig. 2B,C). Under these conditions, the expression of HNK1 increased on the electroporated side,particularly in the post-otic regions, although a similar effect was also observed at different cranial levels (Fig. 2D,E). Moreover, many of the additional HNK1-labelled cells were located at more retarded positions along the migratory pathway, and ectopic epithelial cells were also labelled by HNK1 throughout the electroporated side of the neuroepithelium (Fig. 2F-H). None of these effects was observed when pCX-EGFPalone was electroporated (not shown).

The additional population of migratory cells were very probably bona fide neural crest cells, as they also expressed premigratory neural crest markers. Indeed, we found a dramatic increase in the number of Pax7-, RhoB-and Slug-expressing cells on the electroporated side of the dorsal neural tube (Fig. 2I-K, L and M, respectively). In addition to neuroepithelial cells in the dorsal neural tube, Pax7 is also expressed in a population of cranial neural crest migrating from the mesencephalon and rhombomeres 1, 3 and 5(Kawakami et al., 1997). Coincidently, N-cadherin, the expression of which disappears in delaminating crest cells (Nakagawa and Takeichi,1995), was also diminished in this region(Fig. 2N). Finally, the basement membrane was disrupted, indicating that these crest cells could undergo the epithelial-mesenchyme transition and delaminate(Fig. 2K,L). The misexpression of LSox5 also generated some scattered RhoB-expressing cells at more ventral locations within the electroporated tube, and some of these cells appeared to delaminate from this ectopic site(Fig. 2O). However, ectopic expression of Slug was never detected at intermediate or ventral regions of the tube (Fig. 2M). In summary, ectopic overexpression of LSox5 in the neural tube upregulated Pax7, Slug and RhoB expression dorsally,expanding the delaminating domain and increasing the migratory neural crest population. More ventrally, some ectopic RhoB-positive cells were generated that did not express Slug, although some of them seemed capable of exiting the neural tube.

Under normal conditions, the onset of Slug expression occurs before that of LSox5. Considering that the onset of LSox5and RhoB expression occurs within a similar timescale, and that LSox5 is able to induce RhoB in areas where Slugwas not induced (Fig. 2O), we examined whether LSox5 could induce RhoB within short periods of time and whether this was independent of Slug induction. On the transformed side of the neural tube, we could identify a number of cells that expressed EGFP and LSox5 6-7 hours after electroporation(Fig. 3A,B). In these embryos,ectopic RhoB expression was also detected in the intermediate and ventral regions of the neural tube (Fig. 3C), but no abnormal Slug expression could be observed(not shown). Moreover, no ectopic cells expressing RhoB were observed in the contralateral side of the tube or in control embryos electroporated with pCX-EGFP alone (not shown). The rapid appearance of ectopic cells expressing both LSox5 and RhoB suggested that the forced expression of LSox5 directly induced RhoB expression in the neural tube (Fig. 3D-F). Interestingly, this induction was transitory in intermediate and ventral regions of the neural tube because 12 hours after electroporation only a fraction of the EGFP-labelled cells here expressed RhoB. Similarly,after 20 hours, the dorsal domain of RhoB expression had expanded and a few scattered ectopic RhoB-expressing cells were detected out of this region.

The transcription factors Sox10 and FoxD3 have both been implicated in the generation of the neural crest and in the posterior selection of crest-derived non-neural phenotypes (Aoki et al.,2003; Britsch et al.,2001; Dottori et al.,2001; Kos et al.,2001). Given the expression pattern of LSox5 in the peripheral glia, we examined whether LSox5 misexpression influenced either Sox10 or FoxD3 activity. The expression pattern of these two factors did not appear to have been altered by the misexpression of LSox5 within 6-7 hours of electroporation (data not shown). However,12 hours after electroporation, the number of cells expressing both Sox10 and FoxD3 increased on the electroporated side of the embryo (Fig. 4A-D). Under these conditions, the embryos had developed to stage 11-11+ when the generation of the neural crest at mesencephalic and rostral hindbrain levels has normally terminated, and the neural crest cells are already at a distance from the dorsal neural tube. Indeed, this distancing of the neural crest population was apparent on the control side of the embryos. By contrast, cells continued to delaminate, and Sox10- and FoxD3-expressing cells were still detected in the neural tube on the electroporated side(Fig. 4B,D). This result indicates that the capacity to produce neural crest cells was temporally extended following LSox5 misexpression.

Similar, yet more dramatic, phenotypes were observed 20-24 hours after electroporation, when the embryos had reached stages 13-14(Fig. 4E-K). In these embryos,retarded migratory cells expressing Sox10 and FoxD3 were found in the midbrain and rostral hindbrain, where neural crest production normally ceases around stage 11 (Fig. 4E,G,K). In more caudal regions, Sox10 and FoxD3expression was significantly augmented in the different migratory streams of crest cells, particularly in the circumpharyngeal crest(Fig. 4H-K). Cells exiting the neural tube at aberrant, more ventral locations were also observed in these embryos (Fig. 4L,M). Moreover,as for Slug, the ectopic cells expressing LSox5 did not express Sox10 or FoxD3(Fig. 4L,M).

The persistence of LSox5 in the cephalic neural crest derived glial lineage, and the increased expression of both Sox10 and FoxD3 upon LSox5 misexpression, prompted us to analyse the fate of electroporated cells. In embryos that were allowed to develop for 48 hours after LSox5 electroporation, many EGFP-labelled cells reached the branchial arches or were spread across the most rostral mesenchyme (not shown). Thus, it seemed likely that these cells might be related to the non-neural crest derivatives. In the cranial ganglia, electroporated cells were preferentially detected at sites where Schwann cell precursors accumulate, both at the entrance and exit of the corresponding cranial nerve(Fig. 5). We also found labelled cells interspersed with neuronal precursors of placodal origin at the distal region of the trigeminal ganglion(Fig. 5A,B), the geniculate ganglion (Fig. 5C,D) and the petrosal ganglion (not shown). By contrast, labelled cells were very rarely found in the proximal region of the trigeminal ganglion where neural crest-derived neuronal precursors accumulate. In the ciliary and superior ganglia, the neuronal precursors are all of neural crest origin, but they did not coincide with LSox5 overexpressing cells(Fig. 5E-H). These data suggest that LSox5 overexpression in neural crest cells drives the differentiation of non-neuronal phenotypes.

As well as LSox5, a shorter isoform is also generated from the SOX5 genes, SSox5 (Fig. 6A), which contains only the HMG box for DNA binding and a unique coiled-coil domain (Denny et al.,1992; Wunderle et al.,1996). Although SSox5 is expressed in adult testis in the mouse(Hiraoka et al., 1998), it was not detected in embryonic chick extracts. However, as a first approach to analyse the functional domains responsible for the activity of LSox5, we analysed the effects of expressing the putative chick orthologue of mammalian SSox5 in the neural tube. With respect to the expression of RhoB,Sox10 and FoxD3 after electroporation, this short form behaved very similarly to the long isoform (Fig. 6B,C). These results indicate that the structural information contained in SSox5 is sufficient to influence neural crest development in the manner that we have observed with LSox5.

We have cloned the chicken orthologue of the Sox5 gene, which appears to generate at least two long isoforms (LSox5-I and -II) during early embryogenesis. Sox5 is a member of the D subgroup of the Sox gene family (Schepers et al.,2002). A short isoform of Sox5 (SSox5), containing an HMG DNA-binding domain and a small coiled-coil region, was originally described in adult mouse testis (Denny et al.,1992). Long isoforms were later found in mammals(Hiraoka et al., 1998; Lefebvre et al., 1998; Ikeda et al., 2002) that, like other members of the D subgroup (Sox6, Sox13), contain an extra coiled-coil domain and glutamine-rich stretches. In spite of the tight regulation that restricts SSox5 expression to specific adult tissues, we have shown that SSox5 is capable of producing the effects characterised in this study. If LSox5 does not harbour a transactivation domain, as suggested in some in vitro studies(Lefebvre et al., 1998), the formation of heterodimers with other transcription factors should be considered and the unique coiled-coil structure present in SSox5 may represent an interactive domain.

Not only does the overall sequence of chick LSox5 show a high degree of similarity to that of its mammalian orthologue but the HMG box is identical. Nevertheless, one structural feature that distinguishes chick LSox5 derives from the fact that 16 out of the 33 amino acid substitutions between the mouse and the predicted chicken protein lie within residues 426-456. Strikingly,nine of these changes imply the gain of serine or threonine residues in the chick protein, including a specific serine rich domain (SRD). Indeed, when scanned for putative phosphorylation sites(Blom et al., 1999), they mainly lie in this region of LSox5 and some of the substitutions in the chick sequence generate new sites. This suggests that in the chicken, specific mechanisms may exist to regulate LSox5 activity by phosphorylation. Indeed,the regulation of Sox function by serine phosphorylation has been demonstrated: LSox5, Sox6 and Sox9 cooperate in the transactivation of a Col2a1 chondrocyte-specific enhancer during cartilage formation, and the activity of Sox9 in this process is controlled by cyclic AMP-dependent protein kinase A phosphorylation (Huang et al., 2000).

We isolated chicken LSox5 while searching for genes that are induced or upregulated between embryonic stages 8-11, a period during which the cephalic neural crest forms and begins to migrate. LSox5satisfied the criteria of the screen because it is expressed in the cephalic neural crest, both premigratory and migratory. However, we also detected LSox5 expression in the trunk neural crest, albeit at much lower levels (not shown). At early developmental stages, LSox5 is observed in the neural folds, but while it is expressed in the whole of the prospective forebrain epithelium, in more caudal regions it is restricted to the most dorsal part of the neural tube where the neural crest form. When compared with other neural crest markers, this expression pattern is unique to LSox5. The two genes Slug and Sox9 are restricted to the dorsal region of the neural plate and the onset of their expression coincides with the generation of the neural crest. By contrast, FoxD3is first expressed in the most rostral region of the neural folds and only becomes restricted to the territories where the neural crest forms prior to delamination (Kos et al., 2002). Thus, the initial expression of FoxD3 and the persistence of LSox5 in the anterior prosencephalon are likely to be related to developmental processes other than the generation of the neural crest.

In the head, the onset of LSox5 expression follows the well-established anteroposterior developmental gradient in the dorsal neural folds/tube, as do most other neural crest markers. However, Sox9,Slug and FoxD3 all precede LSox5, which approximately coincides with RhoB and precedes Sox10 expression. It is interesting to note that LSox5, as Slug, is continuously expressed at low levels throughout rhombomere (r) 3, while other markers such as Sox9, FoxD3, RhoB and Sox10 are never expressed in this rhombomere (Liu and Jessell,1998; Cheng et al.,2000; Kos et al., 2002) (S.P.-A., M.A.N. and J.A.B., unpublished). Although we do not know the functional significance of this at present, the absence of many neural crest markers may be related to the fact that the neural crest produced in r3 undergoes massive apoptosis(Graham et al., 1993). Those crest cells that survive in r3 are deviated rostrally or caudally to contribute to the major migratory streams adjacent to r2 and r4(Birgbauer et al., 1995). Thus,based on their relative expression patterns, we would locate LSox5downstream of Sox9, Slug and FoxD3, and at the same hierarchical level or slightly upstream of RhoB and Sox10.

Taking into account the gene hierarchy that emerged by analysing the endogenous expression in the neural crest, the upregulation of RhoBappears to be an immediate consequence of LSox5 misexpression in the neural epithelium. This upregulation is a cell-autonomous phenomenon and it initially presents near complete penetrance, i.e. most transformed cells become RhoB-positive regardless of their location in the dorsoventral axis of the neural tube. With time, and in parallel with the number of LSox5-expressing cells within the neural epithelium, the number of RhoB-expressing cells increases dramatically at the dorsal aspect of the neural tube, concomitant with a massive delamination of cells expressing different neural crest markers. By contrast, the number of RhoB-expressing cells remains low at intermediate and ventral positions, suggesting that these environments are hostile to its induction. RhoB has been proposed to contribute to the changes in cell shape and adhesion that occur during the delamination of neural crest cells by regulating actin polymerisation, the formation of focal adhesions and of stress fibres(Liu and Jessell, 1998). Accordingly, it should be noted that some of the cells that are induced to express RhoB but that lie outside of the dorsal domain of neural crest competence, remain in the neural tube and maintain an epithelial aspect. Only a few cells located outside of, but close to, the domain of competence for crest formation seem to be able to delaminate and enter the adjacent mesenchyme. However, even when they are able to delaminate, these cells do not express any neural crest markers other than the ectopic LSox5. These results indicate that LSox5 is likely to induce RhoB expression directly. Moreover, although this makes the cells competent to undergo changes in cell shape, it is not compatible with them undergoing complete EMT and fully acquiring the neural crest phenotype, unless they are located in the crest-competent domain. We also describe the appearance of HNK1-positive cells within the neural epithelium as a consequence of LSox5 misexpression,a phenotype similar to that described following FoxD3 or Sox9 misexpression (Kos et al., 2002; Cheung and Briscoe, 2003). Although we have not checked if this aberrant HNK1 expression coincides with or follows RhoB ectopic expression, it seems plausible that this phenomenon mimics the normal development of migratory neural crest cells,which acquire the HNK1 epitope after expressing LSox5 and RhoB. In addition, although we do not know the molecular mechanisms involved in this particular behaviour, they may also operate under physiological conditions. Whereas physiological expression of Slug, Sox9 and FoxD3 is restricted to the neural crest competent region in the dorsal neural tube, both RhoB- and HNK1-positive cells can be found at more ventral locations later in development (S.P.-A., M.A.N. and J.A.B., unpublished).

We have shown that LSox5 misexpression in the neuroepithelium produces a dramatic increase in the generation of neural crest cells, which express several neural crest markers, including Pax7, Slug, FoxD3, RhoB, Sox10 and HNK1. Within the head region, this effect can be detected at all levels of the AP axis. Quantitative variation was found between embryos that might reflect the precise timing of the electroporation and analysis, and the efficiency of transfection. Under our standard electroporation conditions, the extension of the `temporal window' of neural crest generation was best visualised in the mesencephalic and rostral hindbrain regions. Other aspects such as the increase in the number of migratory cells were more evident in the circumpharyngeal streams. Whereas the induction of RhoB expression can be detected soon after the appearance of LSox5, the other markers are only induced after longer periods of time, suggesting that their induction is indirect. Before delamination, the ectopic crest cells are located within a crest competent domain as defined by overexpressing Slug(del Barrio and Nieto, 2002). The expansion of this presumptive neural crest territory is paralleled by the extension of the region in which the basement membrane disassembles, allowing the delamination of the neural crest cells that follow the normal migratory pathways. In addition, the `temporal window' of neural crest generation is also extended, as has been observed after the misexpression of other neural crest markers, such as Noelin 1 (Barenbaum et al., 2000), FoxD3 (Kos et al.,2001; Dottori et al.,2001), Slug (del Barrio and Nieto, 2002), Sox10(Aoki et al., 2003) and Sox9 (Cheung and Briscoe,2003). As a matter of fact, many features related to the increase of neural crest production after LSox5 overexpression are similar to those described in the aforementioned studies. It is thought that neural crest overproduction occurs at the expense of other CNS cell types(Dottori et al., 2001; Cheung and Briscoe, 2003). Consequently, the overproduction of neural crest cells frequently appears to deplete the cells from the most dorsal region of the neural tube, as evident morphologically following LSox5 misexpression.

Regardless of the generalised phenotype of neural crest overproduction reported here, it is of interest to place LSox5 in the genetic cascade of neural crest development. The endogenous expression patterns tentatively locate it downstream of Sox9, Slug and FoxD3,implying it acts after neural crest induction and the segregation of the crest from the other neuroepithelial cells. The results obtained after LSox5 overexpression indicate that it lies upstream of RhoBand Sox10, compatible with studies indicating that RhoB is involved in the delamination process and Sox10 in the survival and maintenance of the stem cell properties in the migratory population(Paratore et al., 2001; Kim et al., 2003; Mollaaghababa and Pavan,2003). However, earlier markers were also upregulated when LSox5 was ectopically expressed in neural crest cells at more extended times, suggesting that a feedback loop may be at work in vivo. A precedent comes from in vitro studies of EMT induction where a balanced cross-modulation of cell-cell adhesion molecules, cell-ECM adhesion molecules and cytoskeletal molecules can trigger and orchestrate EMT(Newgreen and Minichiello,1995; Somasiri et al.,2001). Moreover, in a quail neural epithelial cell system in vitro, the pharmacological inhibition of protein kinase C immediately affects the cytoskeleton, provoking transformation into crest-like cells(Minichiello et al., 1999) and the upregulation of Slug and of Sox10 expression (D. F. Newgreen, personal communication). In addition, a non-autonomous cell response may be induced by extracellular molecules (membrane proteins, soluble factors or matrix molecules). Indeed, many neural crest cells in the migratory streams on the transfected side did not express the EGFP marker, raising the possibility of secondary or indirect induction.

With respect to signalling molecules, previous studies have shown that signals emanating from the non-neural ectoderm and paraxial mesoderm, such as the BMPs, WNTs and FGFs, induce neural crest differentiation(Liem et al., 1995; Ikeya et al., 1997; LaBonne and Bronner-Fraser,1998). However, the overproduction of neural crest caused by the misexpression of both FoxD3 and Sox9 has been placed downstream of these dorsalising pathways(Dottori et al., 2001; Cheung and Briscoe, 2003). The same seems to hold true for LSox5 misexpression. We have verified that Bmp4, Bmp7 and Wnt1 expression remains unaltered in treated embryos, although it was sometimes diminished as a consequence of the depletion of neuroepithelial cells in the dorsal tube (S.P.-A., M.A.N. and J.A.B., unpublished).

During chondrocyte differentiation, Sox9 is required for the expression of LSox5 (Akiyama et al., 2002),which subsequently interacts cooperatively with Sox6 and Sox9 to activate the type II collagen gene and promote chondrogenesis(Lefebvre et al., 1998). We have analysed chick Sox6 expression by in situ hybridisation,verifying that it is not expressed in either the dorsal neural tube or in early migratory crest (S.P.-A., M.A.N. and J.A.B., unpublished). In the dorsal neural tube, Sox9 activates a pathway leading to neural crest induction, in which Sox10 is later induced (Cheung and Briscoe, 2003). Sox10 expression is then maintained in the peripheral glial lineage (Britsch et al.,2001), where it regulates the expression of P₀(Peirano et al., 2000). Thus,it is tempting to speculate that in neural crest development a similar interaction between SoxE and SoxD factors to that observed during chondroblast differentiation occurs. In this case, LSox5 would be the downstream a member of the Sox E group, Sox9, and would interact with Sox10, another E-group member during the premigratory stages and/or later on during the differentiation of the glial lineage. In addition, as our misexpression experiments suggest LSox5 may lie upstream of Sox10, the expression of which is induced later to cooperate at stages of differentiation. Alternatively,Sox10 expression could be independent of LSox5, but may be upregulated or induced after forced LSox5 expression, in which case it would enter in the pathway of neural crest generation by mimicking the role of Sox9.

We are grateful to M. Okabe and J. Miyazaki for kindly providing us with the pCX-EGFP vector. The monoclonal antibodies used in this work were obtained from the Developmental Studies Hybridoma Bank developed under the auspices of the NICHD and maintained by the University of Iowa. This work was supported by a grant to J.A.B. from the Spanish Ministry of Science and Technology (BMC2002-00600). S.P.-A. is the recipient of a predoctoral fellowship from the Spanish Ministry of Science and Technology.