Neural crest cells are induced by interactions between the adjacent neural and non-neural ectoderm (Dickinson et al., 1995; Selleck and Bronner-Fraser, 1995) by a combination of signals, including Wnt proteins, bone morphogenetic proteins (BMPs), fibroblast growth factors (FGFs), retinoic acid, and the Notch/Delta system (Liem et al., 1995; LaBonne and Bronner-Fraser, 1998; Christiansen et al., 2000; Aybar et al., 2002; Aybar and Mayor, 2002; Endo et al., 2002; Garcia-Castro et al., 2002; Villanueva et al., 2002; Wu et al., 2003). Neural crest cells emerge from the dorsal neural tube, migrate along defined pathways and differentiate into many cell types, including neurons and glia of the peripheral sensory and autonomic nervous ganglia, Schwann cells, and melanocytes, as well as chromaffin cells of the adrenal medulla, and much of the skeletal and connective tissue of the craniofacial complex (Le Douarin, 1982; Hall and Horstadius, 1988).
Induction of neural crest cells leads to expression of numerous down-stream target genes (Gammill and Bronner-Fraser, 2002), including transcription factors such as Sox8, Sox9, and Sox10 of the SoxE subgroup (Bowles et al., 2000). These genes are expressed in premigratory neural crest cells and later in distinct neural crest-derived subpopulations. Sox9 is the first to be expressed, at a similar time to FoxD3, overlapping with expression of BMP4, cadherin 6b, Slug, and RhoB (Liu and Jessell, 1998; Cheung and Briscoe, 2003). Sox10 is expressed shortly after Slug, and Sox8 expression begins soon after Sox10, before migration from the neural tube (Cheng et al., 2000; Cheung and Briscoe, 2003). As migration commences at trunk levels, Slug, RhoB, N-cadherin, and cadherin 6b are down-regulated (Akitaya and Bronner-Fraser, 1992; Monier-Gavelle and Duband, 1995; Nakagawa and Takeichi, 1995; Liu and Jessell, 1998), while Sox10 and FoxD3 continue to be expressed in migratory neural crest cells (Cheng et al., 2000; Dottori et al., 2001). All migrating neural crest cells in avian embryos then up-regulate the carbohydrate epitope HNK-1 (Vincent et al., 1983; Tucker et al., 1984; Bronner-Fraser, 1986) and a subset up-regulate cadherin-7 (Nakagawa and Takeichi, 1995).
Sox8 is expressed in enteric neurons and at nerve entry/exit points and, like Sox9, in the branchial arches (Wright et al., 1995; Bell et al., 2000; Schepers et al., 2000; Li et al., 2002; Spokony et al., 2002). Sox9 has known functions in cartilage development and sex determination, and mutations in human Sox9 are responsible for campomelic dysplasia (Foster et al., 1994; Wagner et al., 1994; Bell et al., 1997; Healy et al., 1999; Bi et al., 2001; Akiyama et al., 2002; Mori-Akiyama et al., 2003). Furthermore, early overexpression of chick Sox9 results in generation of ectopic neural crest cells (Cheung and Briscoe, 2003) and knockdown in Xenopus causes a loss of both neural crest precursors and later craniofacial skeletal elements (Spokony et al., 2002). Sox10 is required for the early survival and migration of neural crest cells and the later differentiation of Schwann cells and melanocytes in many vertebrates. Sox10 expression continues in developing glia, but is down-regulated in other neural crest lineages (Bondurand et al., 1998; Kuhlbrodt et al., 1998; Pusch et al., 1998; Southard-Smith et al., 1998; Cheng et al., 2000; Dutton et al., 2001; Aoki et al., 2003; Young et al., 2003). In humans, mutations in Sox10 cause Waardenburg-Shah type IV (WS4) syndrome, which combines the phenotype of Hirschsprung's disease and Waardenburg syndrome, i.e., aganglionosis of the distal colon, cochlear deafness, and pigmentation defects (Pingault et al., 1998). In the absence of Sox10, neural crest cells can form but fail to differentiate and ultimately die, leading to a loss of melanocytes and peripheral glia (Kapur, 1999; Britsch et al., 2001; Dutton et al., 2001; Paratore et al., 2001; Sonnenberg-Riethmacher et al., 2001; Honore et al., 2003). Overexpression or knockdown of Sox10 in Xenopus from the two-cell stage increased or decreased, respectively, the number of neural crest precursors and expression of neural crest markers (Aoki et al., 2003; Honore et al., 2003). Of interest, when Sox10 expression remains elevated in vitro, neural crest cells are maintained in a multipotent state (Kim et al., 2003).
Previous studies in the chick have focussed on examining the function of Sox9 in the neurulating neural tube, shortly after its initial expression. However, SoxE genes are maintained in the neural tube and subpopulations of differentiating neural crest cells. Here, we examine their later role in the neural crest and neural tube development by overexpressing Sox10 in the closed neural tube of E2 embryos and comparing the effects with that of another neural crest marker, FoxD3, as well as the other SoxE genes. Of interest, we find that all SoxE genes up-regulate neural crest markers in the neural tube and subsequently lead to emigration of transfected cells from all dorsoventral levels, including the floor plate. Furthermore, maintaining SoxE expression in neural tube-derived cells appeared to stall the cells in an undifferentiated state. These results suggest that normal expression of SoxE genes affects the balance of cells that remain in the neuroepithelium vs. those that emigrate and that subsequent levels of expression are critical for allowing multiple lines of differentiation.
Distribution of SoxE Genes During Trunk Neural Crest Migration
We compared the distribution pattern of Sox8, 9, and 10 in embryos from E2 to E4 to establish the onset and duration of their expression during the course of neural crest formation and migration. Although all three SoxE genes are expressed in neural crest cells at some time in development, their individual patterns and timing of appearance differ (Fig. 1). As previously reported (Cheung and Briscoe, 2003), Sox9 is the first to appear in premigratory neural crest cells in the dorsal neural tube. However, Sox9 is down-regulated after neural crest cells initiate their migration. Sox9 expression is observed in prechondrogenic tissue of the somitic sclerotome and notochord, from early stages of its formation. Expression of Sox10, followed by Sox8, appears just before neural crest cells exit the neural tube. Sox8 is expressed transiently and down-regulated shortly after emigration as neural crest cells enter the somitic sclerotome. In contrast, Sox10 is expressed as migration proceeds and is seen in neural crest cells as they extend into the somites. It is retained on the dorsal-most neural crest cells but appears to be down-regulated in the ventrally migrating neural crest around the dorsal aorta.
These results show that Sox10 is the primary SoxE gene expressed by migrating neural crest cells. HNK-1 staining reveals that most but not all migrating neural crest cells in the rostral portion of the sclerotome are double-labeled with Sox10 (Fig. 1). Later, however, strong expression of Sox10 is observed in neural crest cells condensing to form the dorsal root and sympathetic ganglia. After differentiation, Sox10 expression is maintained in glial cells (Kuhlbrodt et al., 1998; Cheng et al., 2000; Britsch et al., 2001; Dutton et al., 2001; Aoki et al., 2003).
Ectopic Expression of Sox10 in Trunk Neural Tube Causes an Epithelial–Mesenchymal Transition
The above results demonstrate that Sox10 is the primary SoxE family member expressed in migrating neural crest cells, whereas Sox9 is primarily expressed in the dorsal neural tube, sclerotome, and notochord, and Sox8 is only transiently expressed in the migrating neural crest population. Therefore, we focused on Sox10 as the best candidate for a SoxE gene functioning during neural crest migration. Sox10 was ectopically expressed in the neural tube of E2 embryos, which were characterized during and at the end of neural crest migration, 16–48 hours postelectroporation (hpe). On the transfected side of the neural tube in embryos electroporated with Sox10, most transfected cells had migrated away from all aspects of the neural tube, including the ventrolateral region such that the neural tube was reduced in size. The cells appeared in the periphery as loose aggregates (Fig. 2b, n = 16 for 16–48 hpe). Numerous Sox10 transfected cells were also observed along the dorsolateral region populated by melanocyte precursors (Loring and Erickson, 1987), albeit deeper in the dermatome than those in control embryos (Fig. 2b).
In sharp contrast, the neural crest of embryos electroporated with control plasmid encoding green fluorescent protein (GFP) underwent a normal epithelial–mesenchymal transition (EMT) from the dorsal neural tube and formed a migratory population of neural crest cells, many of which were GFP-positive. Control transfected neural crest appeared indistinguishable from HNK-1–positive cells on the nonelectroporated side of the same embryo. They were found in normal neural crest sites, including the dorsal root and sympathetic ganglia, and nerve bundles, while a small number were observed in the dorsolateral melanocyte pathway, in and close to the epidermis (Fig. 2a, n = 10 for 16–48 hpe). Other electroporated cells in control embryos remained in the neural tube, and those in the ventrolateral region projected axons along the ventral root. No differences in the segmental pattern of neural crest through the rostral half of the sclerotome were observed in either Sox10 or control embryos (Fig. 2c,d), but in transverse sections, the migratory Sox10 transfected cells were more scattered than control neural crest cells.
Analysis at various time points after electroporation showed that the Sox10 transfected cells in intermediate and ventral regions of the neural tube underwent ectopic EMT and commenced migration around 36 hpe, at least 24 hr later than the normal neural crest EMT in GFP control and unelectroporated embryos. The neural crest marker HNK-1 (Fig. 2e), normally expressed after neural crest cells have emigrated from the neural tube, was expressed by Sox10 electroporated cells before their departure from the neural tube (Fig. 2f). This finding is similar to the results observed after overexpression of Sox9 or FoxD3 (Dottori et al., 2001; Kos et al., 2001; Cheung and Briscoe, 2003). In contrast to observations in Sox9 overexpressing embryos reported previously (Cheung and Briscoe, 2003), Sox10 transfected cells migrated not only from the dorsal neural tube, but also from mid- and ventral levels (Fig. 2b,h,i). Of interest, as Sox10 transfected cells detached en masse from the neural tube, they continued to express N-cadherin (Fig. 2h,i), which is normally down-regulated from the cell surface of neural crest cells as they delaminate (Fig. 2g; Akitaya and Bronner-Fraser, 1992; Nakagawa and Takeichi, 1998).
We examined the expression of several markers of premigratory and migratory neural crest cells after ectopic expression of Sox10 in the neural tube. RhoB, which is normally transiently expressed in premigratory neural crest (Fig. 2j; Liu and Jessell, 1998), was down-regulated on the electroporated side of the neural tube (Fig. 2k,l). Pax3 is transiently expressed in neural crest cells (Stark et al., 1997) but was not up-regulated in Sox10 transfected cells from 16–48 hpe (not shown). Cadherin-7, normally expressed by some migratory neural crest (Nakagawa and Takeichi, 1995), was up-regulated in a proportion of Sox10 transfected cells in the neural tube (not shown), similar to HNK-1.
Sox10 Overexpression Maintains Cells in an Undifferentiated State
Absence of Sox10 leads to defects in differentiation of both melanocytes and glia (reviewed in Mollaaghababa and Pavan, 2003). To determine whether overexpression of Sox10 had long-term effects on cell differentiation, we assessed the phenotype of electroporated embryos at E6 (4 days postelectroporation, n = 7). Control transfected cells were observed in normal neural crest sites and within the neural tube (Fig. 3a, n = 5). In ovo electroporation results in transient transfection; however, Sox10 expression assessed 6 days postelectroporation (E8) was detectable in the majority of GFP-positive cells (not shown). Of interest, the electroporated side of the neural tube in Sox10 experimental embryos, which had been severely depleted of cells at E4, had reformed significantly by E6, although not attaining the size of the nonelectroporated side (Fig. 3b,d,g). However, few Sox10-transfected cells were found within the neural tube by this time in contrast to the initial distribution of these cells. In addition, few electroporated cells were found within the dorsal root and sympathetic ganglia (Fig. 3b). No Sox10 transfected cells, including the few within ganglia, expressed neuronal markers such as HuC/HuD (Fig. 3b). Many Sox10 transfected cells were located adjacent to the neural tube, around the dorsal root ganglia, in the dorsal and ventral roots, and dermis. Others were present in ectopic sites such as within the myotome among skeletal muscle, a few within cartilage, and in small clusters immediately dorsal to the neural tube. However, none displayed overt characteristics of differentiation, such as skeletal muscle markers (Fig. 3c). At E6, many Sox10 transfected cells retained expression of HNK-1 (Fig. 3e), which is normally expressed in dorsal root and sympathetic ganglia, Schwann cells and the neural tube (Fig. 3d; Tucker et al., 1984). Although many Sox10 transfected cells were in the appropriate location to differentiate into melanocytes or Schwann cells, none were observed expressing MEBL-1, a melanocyte precursor marker (Fig. 3f; Kitamura et al., 1992; Wakamatsu et al., 1998), and few expressed P0, a marker of Schwann cell differentiation (Fig. 3g,h; Peirano et al., 2000).
Ectopic Expression of SoxE Genes in the Ventral Neural Tube
The massive increase in emigrating cells at the expense of neural tube cells after Sox10 expression suggested that all dorsoventral levels of the neural tube produced migratory cells displaying some neural crest characteristics under these circumstances. All of these electroporations included the endogenous dorsal neural tube, raising two possible explanations: (1) that this result may reflect an expansion of the dorsal domain; and/or (2) that all dorsoventral levels of the neural tube may be capable of producing neural crest-like cells given appropriate signals.
To investigate if Sox10-induced delamination must be initiated from the dorsal neural tube, we targeted electroporations to the ventral neural tube alone. The results showed that ectopic expression of any of the SoxE genes, including Sox9, in this location resulted in extensive emigration of ventral neural tube cells away from the neural tube by 2 days postelectroporation (Fig. 4b–d, n = 7 for each gene). Expression of HNF3β, normally found in the floor plate at E3 (Fig. 4e), was abolished cell-autonomously in SoxE transfected floor plate cells, but its expression was maintained in untransfected floor plate cells in SoxE transfected embryos (Fig. 4f,g). The SoxE transfected cells left the neural tube by means of the ventral root and through the basal lamina of the ventral neural tube. Many of the migrating transfected cells expressed the HNK-1 epitope (Fig. 4h) and retained expression of N-cadherin, similar to observations at other dorsoventral axial levels (Fig. 4i,j). Whereas Sox8, 9, and 10 each induced substantial ventral migrating cells, there were differences in the migration patterns. Sox10 transfected cells were chiefly found along the ventral roots and axon pathways albeit in greater numbers than control cells, and they largely respected the perinotochordal exclusion zone (Fig. 4b,h). In contrast, Sox9 transfected cells were much more widely dispersed dorsoventrally, much less confined to axon pathways, and invaded the perinotochordal zone (Fig. 4d). It is interesting to note that they mingle with Sox9 expressing sclerotomal cells in this location (see Fig. 1). Some Sox8 transfected cells were observed along axon pathways, while others were dispersed dorsoventrally and located adjacent to the notochord (Fig. 4c,j). The motor axon trajectories, assessed by neurofilament staining (not shown), were not disrupted by ectopic ventral expression of SoxE genes.
In contrast to experimental embryos, most control transfected ventral neural tube cells remained in the neural tube, in some cases extending axons along the ventral roots (Fig. 4a, n = 4). As previously observed, a small number of neural tube cells migrate and form Schwann cells in the proximal ventral root but do not migrate further than this (Weston, 1963; Rickmann et al., 1985; Lunn et al., 1987).
Electroporation of Sox9 and Sox8 Into the Neural Tube Maintain Transfected Cells in an Undifferentiated State
Longer-term effects of overexpression of Sox8 and Sox9 have not been described. We, therefore, examined the location and expression of markers of differentiation in Sox8 and Sox9 expressing cells 4 days after electroporation into the lateral aspect of the neural tube (n = 4 for each gene) revealing both common and unique phenotypes for each gene. In all cases, transfected cells surrounded but typically were not incorporated into neural crest locations, e.g., they were found around the periphery of rather than within the dorsal root ganglia. In terms of differentiation, no electroporated cells outside the neural tube expressed the neuronal marker HuC/HuD (Fig. 5a). Transfected cells also were noted in the ventral root, along axons, and throughout the dermis (Fig. 5a,b). Many transfected cells were observed in locations usually avoided by neural crest cells, including the region adjacent to the notochord and within the developing myotome. HNK-1 expression was maintained in many Sox8 and Sox9 transfected cells outside the neural tube (not shown). However, no Sox8 or Sox9 transfected cells in the dermis or elsewhere expressed the melanocyte marker MEBL-1, and, similar to Sox10 transfected cells, they were located deeper in the dermis than control melanocytes (not shown). A proportion of Sox8 or Sox9 transfected cells located in the ventral root and along axon pathways expressed the Schwann cell marker P0; however, many transfected cells along or near axons were P0-negative (Fig. 5b–g). Sox8 and Sox9 transfected cells located in sites not normally occupied by neural crest cells rarely expressed P0 (Fig. 5b).
Of interest, differences were noted between the different SoxE genes. For example, more Sox8 transfected cells remained in the neural tube than those transfected with Sox9 or Sox10, and some Sox8 transfected cells in the neural tube expressed the neuronal marker HuC/HuD (Fig. 5a). More Sox8 transfected cells in the ventral root and along axons appeared to express P0 than did Sox9 transfected cells (Fig. 5c–g). These results suggest that there are some commonalities in function within the SoxE family, but also subtle differences in their function, particularly at later stages.
Comparison of Gene Expression Changes After Overexpression of SoxE Genes and FoxD3
Common phenotypes after overexpression of Sox8, Sox9, Sox10, or FoxD3 are as follows: expression of HNK-1 in the neural tube, ability of transfected cells to undergo EMT, similar migration patterns, and for SoxE genes, maintenance in an undifferentiated state (Dottori et al., 2001; Kos et al., 2001; Cheung and Briscoe, 2003).
As a first step toward identifying the epistatic relationships between these genes, we compared gene expression changes after overexpression of each of these genes. The results are summarized in Table 1, using data from our experiments together with those of previously published studies (Cheung and Briscoe, 2003; n = 3 for each gene and time point). Overexpression of SoxE genes or FoxD3 caused a down-regulation of the premigratory neural crest marker RhoB on the electroporated side of the neural tube. Another premigratory neural crest marker, Slug, was reported to be briefly up-regulated after Sox9 overexpression at 6 hpe and then down-regulated by 12 hpe (Cheung and Briscoe, 2003). However, we did not observe any increase in Slug expression in Sox9 transfected cells at any stage examined in the present study. Nor did overexpression of Sox8, Sox10, or FoxD3 induce expression of Slug at any time (our observation; Dottori et al., 2001).
Table 1. Gene Expression Changes in SoxE and FoxD3 Transfected Neural Tube Cellsa
A dash indicates there was no difference between the control and electroporated side of the neural tube. Increases or decreases in gene expression are indicated by ↑ and ↓, respectively. The asterisk indicates data taken from Cheung and Briscoe (2003).
6h: -24h: slight↑
6h: -24h: slight↑
6h: -24h:slight ↓
6h: -24h:slight ↓
6h: -24h: -
6h: -24h: -
6h: ↑24h: slight↑
6h: -24h: ↑
6h: -24h: -
6h: ↓/-* 24h: ↑
6h: ↓/-16-24h: ↓
6h: -24h: -
6h: -24h: -
6h: -24h: -
6h: -24h: -
After overexpression of FoxD3, Sox10 was up-regulated in transfected cells at 24 and 48 hpe (Fig. 6a,b) and Sox8 was weakly up-regulated at 24 hr; however, Sox9 expression did not appear to be altered in FoxD3 transfected cells. In contrast, Sox10 overexpression caused a decrease in the expression of FoxD3 in dorsal neural cells at 16 and 24 hpe (Fig. 6c,d), similar to the decrease in RhoB. Sox9 overexpression caused an up-regulation of Sox10 in transfected cells at 6 hpe and a weak up-regulation of Sox10 at 24 hpe. At 24 hpe, the level of Sox10 in Sox9 transfected cells appeared to be far less than the endogenous level of Sox10 expression, assessed by the density of staining. Sox8 also appeared to be slightly up-regulated in Sox9 transfected cells at 24 hpe. FoxD3 expression after Sox9 overexpression initially decreased at 6 and 12 hpe and then increased in transfected neural tube cells at 24 hpe (Cheung and Briscoe, 2003). Sox8 and Sox10 overexpression both appeared to cause a slight reduction in the level of Sox9 on the electroporated side of the neural tube at 24 hpe but had little or no effect on each other. These results suggest that Sox9 and FoxD3 are likely to precede and induce Sox10, with later cross-regulation among the genes.
Sox10 Is Expressed in Migratory Trunk Neural Crest Cells
Up-regulation of SoxE genes appears to be an early response to neural crest induction, with Sox9 expressed in premigratory neural crest and Sox10 expressed as migration begins. After emigration from the neural tube, Sox9 is down-regulated in the migrating population. Its primary expression pattern at times of neural crest migration is not in the neural crest but in the notochord and somitic sclerotome. In contrast, Sox10 is expressed on migrating trunk neural crest cells as they move through the rostral portion of the sclerotome. It is then down-regulated in the neural crest cells localized ventrally around the dorsal aorta where sympathetic neuroblasts congregate and differentiate. As neural crest derivatives form, Sox10 remains on in condensing dorsal root ganglia and is re-expressed on the sympathetic ganglia as well as melanoblast precursors. In contrast to Sox9 and 10, Sox8 is expressed only transiently in neural crest cells at their earliest times of migration.
Ectopic expression of Sox10, as well as other SoxE family members, causes neural tube cells to undergo an EMT and leave the neural tube at all dorsoventral levels. Previous studies have also shown that Sox9 and FoxD3 cause an excess of emigration from intermediate levels of the neural tube (Dottori et al., 2001; Cheung and Briscoe, 2003), but it was not clear if this finding represented a dorsal expansion or if ventral neural tube cells could respond independently by undergoing an EMT. To address this question, we targeted the ventral-most neural tube, including the region that normally forms the floor plate. Although intermediate neural tube can form neural crest, the ventral-most portion adjacent to the floor plate is normally refractory to neural crest induction by ectoderm in vitro (Dickinson et al., 1995). In contrast, ectopic expression of SoxE genes appears sufficient to allow these cells to emigrate from the neural tube and express some characteristics of neural crest cells. SoxE transfected floor plate cells did not express characteristic markers of floor plate cells such as HFN3β; however, adjacent untransfected floor plate cells did express such markers. A small number of Schwann cells in the proximal ventral root arise from the ventral neural tube during normal development (Weston, 1963; Rickmann et al., 1985; Lunn et al., 1987). However, the neural crest-like cells formed after ectopic SoxE expression were present in large numbers and migrated far beyond the proximal ventral root.
These results suggest that expression of any SoxE gene is sufficient to drive neuroepithelial cells to express some neural crest markers, undergo EMT, and migrate from any dorsoventral level of the neural tube. The normally coordinated and dynamic expression of BMP4 and its inhibitor Noggin have been shown to initiate the EMT of neural crest cells from the neural tube (Sela-Donenfeld and Kalcheim, 1999). When added to neural tube explants, BMP4 can induce migratory neural crest-like cells from the intermediate neural tube (Liem et al., 1995). However, ectopic expression of SoxE genes did not result in up-regulation of BMP4 (Cheung and Briscoe, 2003). Other growth factors such as Wnt proteins have also been implicated in induction of the neural crest (Garcia-Castro et al., 2002; Wu et al., 2003). Wnt3a is up-regulated for 12 hr after Sox9 overexpression; however, overexpression of Wnt3a alone in neural tube cells does not induce EMT or HNK-1 expression (Cheung and Briscoe, 2003).
Sonic hedgehog (Shh) is expressed by the notochord and induces the development of the floor plate (Patten and Placzek, 2000). Shh is distributed in a ventral–dorsal gradient throughout the neural tube and is required for patterning the ventral neural tube. In vitro studies have shown that Shh can inhibit induction of the neural crest at an early stage and inhibit neural crest cell migration (Selleck et al., 1998; Testaz et al., 2001). However, SoxE transfected ventral neural tube cells, despite being exposed to high levels of Shh, underwent migration comparable to that of transfected lateral neural tube cells at some distance from the source of Shh. This observation is consistent with the finding that notochord grafts, in the presence of normal dorsalizing signals do not prevent development or migration of neural crest cells (Artinger and Bronner-Fraser, 1992).
BMP4 up-regulates expression of RhoB, a marker of premigratory neural crest cells in the dorsal neural tube. Inhibition of Rho activity using a C3 exotoxin prevented the EMT of HNK-1–expressing neural crest cells (Liu and Jessell, 1998). However, SoxE or FoxD3 transfection results in an expanded region of delamination and repression of endogenous RhoB (Dottori et al., 2001; Cheung and Briscoe, 2003). Furthermore, RhoB is not expressed in the ventral neural tube, yet expression of SoxE genes induces substantial levels of EMT, HNK-1 expression, and cell migration. Thus, the relationship of RhoB to EMT is unclear. It is clearly not involved in migration, because it is down-regulated at migratory stages of neural crest development (Liu and Jessell, 1998), suggesting it might function transiently to prevent premature EMT.
Slug has been demonstrated to be involved in induction of neural crest and may also be important in EMT (Nieto et al., 1994; LaBonne and Bronner-Fraser, 2000; del Barrio and Nieto, 2002), although Slug expression cannot promote neural crest formation in the presence of Noggin at trunk levels (Sela-Donenfeld and Kalcheim, 1999). Whereas overexpression of SoxE genes and FoxD3 resulted in neural tube cells undergoing an EMT, this finding was not associated with up-regulation of Slug. Cheung and Briscoe (2003) observed an up-regulation of Slug after Sox9 overexpression at 6 hpe. This difference may be due to differences in the axial level, stage electroporated, and time of analysis. Slug is more highly expressed at cranial levels in premigratory and migratory neural crest, and overexpression of Slug increases the amount of cranial but not trunk neural crest (del Barrio and Nieto, 2002). Together, these findings provide support for the proposal that ectopic SoxE and FoxD3 expression can induce EMT of neural tube cells independent of RhoB and Slug, possibly by driving expression of downstream effectors of EMT and migration that are separate to the Slug and RhoB pathway (Dottori et al., 2001). Thus, whereas RhoB and Slug may be involved normally in the process of neural crest generation, they may not be essential for EMT of neural tube-derived cells.
SoxE Overexpression Affects Migration Pathways
The ectopic emigration of neural tube-derived cells caused by overexpression of SoxE genes nevertheless resulted in production of cells with migratory behaviors that only partially replicated those observed in endogenous neural crest. The neural crest-like cells migrated segmentally through the rostral but not caudal hemisomites similar to endogenous neural crest cells. However, many SoxE transfected cells were also dispersed in the dorsoventral plane and located in sites inappropriate for neural crest cells. SoxE transfected cells in the dermis did not colocalize exactly with endogenous melanocytes but rather were found in deeper layers.
Differences were noted between different SoxE family members. Ventrally transfected Sox10 cells migrating from the ventral neural tube mainly followed axon pathways, such as the ventral root, and maintained a distance from the notochord. In contrast, Sox9 transfected cells were far more dispersed in the dorsoventral plane, and many Sox9 and Sox8 transfected cells were observed adjacent to the notochord, a zone normally strictly avoided by neural crest cells (Newgreen et al., 1986).
Sox10 is normally expressed by migrating neural crest cells, whereas expression of Sox8 and Sox9 is down-regulated in migrating trunk crest cells. Sox10 and Pax3 promote expression of c-Ret (Lang et al., 2000), a receptor for glial cell-derived neurotrophic factor (GDNF), which is involved in enteric neural crest chemotaxis (Young et al., 2001). It is possible that overexpression of Sox10 allows expression of cell surface receptor molecules required to detect some normal cues involved in neural crest pathfinding, whereas Sox8 and Sox9 do not allow normal expression or function of such molecules. Consistent with this hypothesis, many more Sox8 or Sox9 compared with Sox10 transfected cells migrated into the region surrounding the notochord, a region known to be rich in repulsive molecules such as proteoglycans and Shh (Newgreen et al., 1986; Testaz et al., 2001).
One observed cell surface change induced by SoxE gene expression is the maintenance of N-cadherin expression. N-cadherin is normally strongly expressed by neuroepithelial cells and removed from cell junctions when neural crest cells migrate from the neural tube (Akitaya and Bronner-Fraser, 1992; Monier-Gavelle and Duband, 1995; Nakagawa and Takeichi, 1998). SoxE transfected cells are clearly mesenchymal and highly migratory in nature, raising the question of the role of N-cadherin in maintaining neuroepithelial cellular morphology. Overexpression of N-cadherin causes failure of premigratory neural crest cells to migrate from the neural tube (Nakagawa and Takeichi, 1998). Whereas N-cadherin is normally associated with neuroepithelial cells, its expression is not sufficient to maintain them in an epithelial state in the presence of high levels of SoxE gene expression.
Overexpression of SoxE Genes Maintains Cells in an Undifferentiated State
All three SoxE genes failed to drive differentiation of the transfected population into any particular lineage. A small proportion of transfected cells differentiated into Schwann cells as detected by the marker P0. The location of transfected cells was in many cases appropriate for Schwann cells or melanocytes, yet only some axon-associated cells expressed the Schwann cell marker P0, and no cells in the dermis expressed the melanocyte marker MEBL-1. None of the very few SoxE transfected cells observed within ganglia expressed the neuronal marker HuC/HuD. These findings are in accord with and extend the in vitro observation that overexpression of Sox10 in neural crest cells inhibited differentiation into neuronal lineages and preserved multipotency (Kim et al., 2003). These results contrast with previous observations of neural crest differentiation after Sox9 overexpression (Cheung and Briscoe, 2003), possibly resulting from axial level differences.
Our data indicate high-level expression of all SoxE genes can maintain neural tube-derived cells in an undifferentiated state in vivo and inhibit transfected cells from localizing in sites of neurogenesis. However, the reason for lack of differentiation remains unclear. High-level expression of Sox9 is associated with and necessary for cartilage formation by both cranial neural crest and mesodermally derived cells (Bi et al., 2001; Spokony et al., 2002; Mori-Akiyama et al., 2003), and Sox10 promotes expression of melanocyte and Schwann cell genes (Bondurand et al., 2000, 2001; Peirano et al., 2000; Potterf et al., 2000; Britsch et al., 2001). This finding indicates that, expressed at normal levels, SoxE genes are not in themselves inhibitory to differentiation. Of interest, Sox10 overexpression from the two-cell stage in Xenopus greatly increased the number of pigment cells expressing Trp-2 (Aoki et al., 2003).
Sox family members are believed to often pair with a cofactor to activate expression of downstream targets (Kamachi et al., 2000; Wilson and Koopman, 2002). It is possible that overexpression of SoxE genes at an early stage in neural crest development gives an imbalance in relation to the expression of cofactors required for differentiation. Sox10 and Pax3 interact to promote expression of the melanocyte gene Mitf (Bondurand et al., 2000; Potterf et al., 2000); however, we did not observe Pax3 expression in Sox10 transfected cells, so this imbalance may have been sufficient to prevent melanocyte differentiation. Furthermore, Sox10 is expressed only transiently in melanoblasts, whereas Sox10 transfected cells maintained expression until at least E8. Similarly, high-level expression driven by the β-actin promoter used in our expression constructs may provide an excess of SoxE protein, overriding the effects of differentiation-promoting cofactors. Promoters of several SoxE gene targets such as P0 contain dimeric and monomeric binding sites (Peirano et al., 2000; Peirano and Wegner, 2000; Schlierf et al., 2002; Bernard et al., 2003). The functional consequences of monomeric and dimeric binding are unclear but expression of high levels of SoxE genes may interfere with the balance between these binding modes, resulting in failure to activate normally SoxE responsive promoters. Alternatively, overexpression of SoxE genes within the neural tube results in extensive migration of neural crest-like cells but at a considerably later stage than the endogenous neural crest. It is possible that a proportion of SoxE induced migratory cells are unable to access the normal differentiation microenvironments and signals due to the late stage at which they migrate and, therefore, are inhibited from differentiating.
Interactions Between SoxE Genes and FoxD3
The similarities between overexpression of SoxE genes and FoxD3 suggest that these genes act in a similar manner and raise the possibility that they may interact with each other at some level. FoxD3 and Sox9 overexpression induced Sox10 expression within the neural tube, indicating that they are both upstream of Sox10. In addition, Sox8 was weakly up-regulated by Sox9. This finding is consistent with the finding that Sox9 and FoxD3 expression begins earlier in the prospective neural crest than either Sox8 or Sox10 (Cheung and Briscoe, 2003). The level of up-regulation of Sox8 and Sox10 after Sox9 overexpression was slight in comparison to the endogenous expression level of these genes, suggesting that additional factors are required to drive high-level expression. Sox10 overexpression repressed expression of FoxD3, indicating that there is a feedback mechanism linking these transcription factors during neural crest development.
Similarities between overexpression of SoxE genes suggest that SoxE genes may be able to compensate for each other in the production of migratory neural crest cells, but are less competent to do so when required for later development in separate sublineages in which SoxE expression patterns do not overlap. Compound knockout or knockdown of SoxE genes at different stages of neural crest development are required to help further define the function and interrelationships of SoxE genes in neural crest development.
Full-length chick Sox10 (cSox10, supplied by Y.-C. Cheng) and Sox9 (cSox9, supplied by P. Sharpe) were subcloned into pMES (internal ribosomal entry site [IRES] -GFP construct with a chick β-actin promoter, donated by C. Krull (Swartz et al., 2001). Empty vector pCMS-EGFP (Clontech, CA) or pMES were used as controls. Sox8pCAGG and Sox9pCAGG were supplied by Y.-C. Cheng. pCA-cFoxD3-IRES-EGFP was supplied by M. Dottori (Dottori et al., 2001). DNA was prepared using Qiagen midiprep columns and concentrated to 4 mg/ml.
Hamburger and Hamilton stage (HH; Hamburger and Hamilton, 1951) 12–14 (E2) chick embryos (Research Poultry, Research, Australia) were used for trunk electroporations. The shell was opened, and India ink (Pelikan no. 17 Black; Pol Equipment, Sydney, Australia) diluted 1:20 in phosphate buffered saline (PBS) was injected under the blastoderm using a 30-gauge needle and 1-ml syringe to visualize the embryo. A small hole was created in the vitelline membrane above the most caudal somite. A total of 1 μl of DNA was mixed with Fast Green (Sigma, St. Louis, MO) for visualisation, and drawn into a finely pulled glass pipette. Approximately 0.05 μl of DNA was injected into the neural tube caudal to the last somite of each embryo using a mouth pipette. A total of 100 μl of PBS was then placed on the embryo, and 0.5-mm gold electrodes placed either side of the embryo, with the anode to the right side. Embryos were electroporated with three pulses of 25 V, 50 msec using a BTX Electro Square Porator ECM 830 (Fisher Biotec, West Perth, Australia). When electrodes were oriented dorsoventrally to electroporate the floor plate, the anode was placed beneath the embryo, and 30 V pulses were used. Approximately 5 ml of albumen was removed to lower the embryo and the egg was resealed with adhesive tape and reincubated at 38°C for 6 hr to 6 days. Embryos were then removed and placed in PBS and viewed as whole-mounts using a Leica MZ FL III fluorescence stereomicroscope (Leica Microsystems, Switzerland). Images were captured by using a Leica DC200 digital camera III and processed with Leica IM1000 software. The GFP-positive region of the embryo was dissected out and fixed in 4% paraformaldehyde in PBS (PFA) for 30 min at room temperature, then washed in PBS. Embryos used for in situ hybridization were fixed overnight in PFA at 4°C. For cryostat sectioning, embryos were placed in 30% sucrose in PBS overnight, embedded in TBS Tissue Freezing Medium in Tissue Tek cryomolds (both from ProSciTech, Thuringowa, Australia), and frozen in isopentane cooled with dry ice. Then, 10- to 20-μm sections were cut transversely using a Leica CM1900 cryostat and collected on Superfrost microscope slides (Biolab Scientific, Auckland, New Zealand) coated with poly-L-lysine. Adjacent sections were collected on separate slides to allow the use of different antibodies on adjacent sections.
In Situ Hybridization
Section in situ hybridization was performed essentially following the methods detailed in Strahle et al. (1994) and Myat et al. (1996) with the exception that Tris-buffered saline was used in place of maleic acid buffer. Antisense digoxigenin-labeled probes were prepared by in vitro transcription from plasmid templates as described by the manufacturers (Ambion, TX). Plasmids used were chick Sox10 (Cheng et al., 2000), chick Sox8 and Sox9 (Bell et al., 2000; Cheng et al., 2001), chick FoxD3 (Dottori et al., 2001), chick Slug (del Barrio and Nieto, 2002), and chick Pax3 (Stark et al., 1997).
Sections were permeabilized with 0.1% Triton X-100 (Tx) in 1% bovine serum albumin (BSA) in PBS for 15 min at room temperature. Primary antibodies were diluted in 1% BSA/PBS, and 100 μl was applied as a standing drop on slides followed by a coverslip, and incubated overnight at 4°C. Primary antibodies used were mouse IgG anti-MF20 (skeletal muscle marker, 1/50), mouse anti-RhoB (56-4H7, 1/100), mouse anti-HNF3β (4C7, 1/10), mouse anti-P0 (EC8, 1/100; all supplied by the Developmental Studies Hybridoma Bank, University of Iowa, Iowa City, IA), mouse IgM anti–HNK-1 supernatant (1/50, hybridoma maintained at MCRI), goat anti-GFP (1/400, Rockland, PA), mouse IgG anti-ACAM (N-cadherin, 1/200, clone GC-4, Sigma), mouse anti-HuC/HuD (1/50, Ab 16A11, Molecular Probes), rabbit anti-neurofilament (1/200, Ab 1987, Chemicon), melanocyte-specific mouse anti-melanosome matrix protein 115 (MEBL-1, 1/30, developed by K. Kitamura [Kitamura et al., 1992], kindly provided by Y. Wakamatsu, Sendai, Japan), rabbit anti-SoxE (1 μg/ml, kindly provided by C. Smith, MCRI, Australia), and mouse anti–cadherin-7 (1/100, kindly provided by M. Takeichi, RIKEN Centre, Japan). Sections were washed three times in PBS, and secondary antibodies were diluted in 1% BSA/PBS applied for 2 to 3 hr at room temperature, then washed three times in PBS. Secondary antibodies used were donkey anti-sheep Alexa 488 (1/400), donkey anti-mouse and anti-rabbit Alexa 594 (1/1,500, all from Molecular Probes), and biotinylated donkey anti-rabbit and anti-mouse IgM (1/100) followed by PBS washes and Streptavidin:AMCA (1/500, all from Jackson ImmunoResearch, PA). Sections were viewed by using an Olympus IX70 inverted fluorescence microscope, and images were captured using a Spot RT monochrome camera and Spot software (version 3.5, Diagnostic Instruments, Inc., Scitech, Australia) or MTI 3CCD camera and Oculus TciPro software (version 3.1, Coreco, Inc., Scitech). Images were compiled using Adobe Photoshop 6.0 (Adobe Systems, Inc., CA).
We thank Yoshio Wakamatsu for the MEBL-1 antibody; Yi-Chuan Cheng for chick Sox8, Sox9, and Sox10 cDNA; Mirella Dottori for FoxD3 constructs; Craig Smith for SoxE antibody, Sox8 and Sox9 in situ probes; Masatoshi Takeichi for cadherin-7 antibody; Paul Sharpe for chick Sox9 cDNA; Angela Nieto for the Slug probe; and Cathy Krull for the pMES construct. We also thank the Developmental Studies Hybridoma Bank, developed under the auspices of the NIHCD and maintained by the University of Iowa, Department of Biological Sciences (Iowa City, IA), for the following antibodies: MF20, developed by D.A. Fischmann; 1E8 (P0), developed by E. Frank; 4C7 (HNF3β), developed by T. Jessell and S. Brenner-Morton; and 56-4H7 (RhoB) developed by T. Jessell. S.J.M. is supported by an Australian Postgraduate Award. V.M.L. is supported by an American Heart Association Postdoctoral Fellowship.