Temporal control of neural crest lineage generation by Wnt/β-catenin signaling

Many neural and non-neural cell types in vertebrates are produced by the neural crest, a transient embryonic structure that emerges during neurulation in the dorsal part of the neural tube (Le Douarin et al., 2008). Among the neural crest derivatives are the pigment cells in the skin that are reported to originate either directly from neural crest cells migrating from the neural tube on a dorsolateral pathway or from nerves innervating the skin (Sommer, 2011). Melanocyte lineage specification of dorsolaterally migrating cells is thought to occur at an early stage of neural crest development. Cells fated for the melanocyte lineage stall in the migration staging area (MSA) adjacent to the neural tube before continuing their dorsolateral migration (Weston, 1991). In the MSA, melanocyte specification is indicated by expression of the microphthalmia-associated transcription factor (Mitf), which controls melanoblast development and survival (Opdecamp et al., 1997). Both in cell culture and in vivo, expression of Mitf is regulated by canonical Wnt signaling, indicating a central role of this signal transduction pathway in melanocyte development (Dorsky et al., 2000; Takeda et al., 2000; Hari et al., 2002; Widlund et al., 2002).

Canonical Wnt signaling involves the intracellular signaling component β-catenin, which, upon activation, translocates to the nucleus and induces specific transcriptional responses (Gordon and Nusse, 2006). In cultures of mouse and quail neural crest cells, treatment with Wnt or activation of β-catenin enhances melanoblast proliferation and differentiation (Dunn et al., 2000; Jin et al., 2001). More strikingly, constitutive activation of β-catenin in zebrafish in vivo promotes the formation of melanocytes while repressing neural cell lineages (Dorsky et al., 1998). By contrast, upon inhibition of Wnt signaling, neural crest cells adopt a neural rather than a pigment cell fate. Similarly, in the mouse, conditional deletion of β-catenin (Ctnnb1) in premigratory neural crest cells also prevents the generation of melanocytes (Hari et al., 2002). However, unlike in zebrafish, inactivation of Ctnnb1 in the mouse neural crest not only prevents melanocyte formation but also sensory neurogenesis (Hari et al., 2002). Moreover, in contrast to zebrafish, Ctnnb1 signal activation in the mouse neural crest in vivo does not enhance melanocyte formation, but rather promotes the widespread formation of sensory neurons at the expense of virtually all other possible neural crest cell fates, including melanocytes (Lee et al., 2004).

In the present study we offer an explanation for these inconsistent findings, suggesting that in mouse embryos Wnt/β-catenin signaling controls sensory fate acquisition before regulating melanocyte lineage formation. To analyze the temporal control of neural crest lineage generation by Wnt/β-catenin signaling, we genetically manipulated Ctnnb1 at different stages of neural crest and early melanocyte development. Our results are in agreement with the idea that sensory neuronal and melanocyte lineages are successively generated from neural crest cells by sequential β-catenin signaling.

The Cre-loxP system was used to conditionally express a stabilized form of β-catenin as described (Lee et al., 2004). Mutant embryos were heterozygous for Cre and for Ctnnb1^Δex3. Littermates inheriting an incomplete combination of alleles, i.e. either Cre or Ctnnb1^Δex3, served as controls. For in vivo fate mapping experiments, either ROSA26 reporter (R26R) (Soriano, 1999) or Z/EG (Novak et al., 2000) mouse lines were used. Embryos carrying one allele for lineage tracing are indicated as R26R or Z/EG. The following Cre lines were used: Wnt1-Cre (Danielian et al., 1998), Sox10-Cre (Matsuoka et al., 2005), Dhh-Cre (Jaegle et al., 2003; Joseph et al., 2004), Plp-CreERT2 (Leone et al., 2003) and Tyr-CreERT2 (Bosenberg et al., 2006). Genotyping for all Cre alleles was performed by PCR with primers Cre-up (5′-ACCAGGTTCGTTCACTCATGG-3′) and Cre-lo (5′-AGGCTAAGTGCCTTCTCTACAC-3′) and 35 cycles of 94°C for 45 seconds, 58°C for 30 seconds and 72°C for 1 minute, to amplify a 217 bp fragment within the Cre ORF.

For activation of the CreERT2 protein, tamoxifen (Sigma T5648) was dissolved in 1:10 ethanol:sunflower oil at 10 mg/ml. Tamoxifen (1 mg) was injected intraperitoneally into the pregnant mother on 1 day for Plp-CreERT2 mice at the indicated stages and on 2 consecutive days at E11.5 and E12.5 for Tyr-CreERT2 mice.

In situ hybridization on cryosections with digoxigenin-labeled riboprobes was performed as described (Hari et al., 2002). Mitf- and Dct-specific riboprobes were used as described (Hari et al., 2002). The 1.2 kb Mitf riboprobe corresponded to position 177-1399 of the published mouse Mitf sequence (GenBank accession no. NM_008601) (Bondurand et al., 2000).

lacZ reporter gene expression was detected using X-Gal staining. Cryosections were fixed for 5 minutes on ice in PBS containing 2% formaldehyde, 0.2% glutaraldehyde, 0.02% NP40 and stained for 4-16 hours at 37°C in PBS containing 1 mg/ml X-Gal, 10 mM K₃Fe(CN)₆, 10 mM K₄Fe(CN)₆, 2 mM MgCl₂, 0.02% NP40.

For immunohistochemistry, cryosections were fixed for 5 minutes in 4% formaldehyde at room temperature and treated with blocking buffer (1% BSA, 0.3% Triton X-100, in PBS) for 30 minutes. Primary antibodies were used as follows: mouse anti-Neurog2 (1:10; gift from D. J. Anderson, California Institute of Technology, Pasadena, CA, USA), goat anti-Dct (1:200; Santa Cruz sc-10451), mouse anti-Mitf (1:200; Abcam ab12039, C5 clone), rat anti-c-Kit (1:200; eBioscience 14-1171), mouse anti-PHH3 (1:200; Cell Signaling 9706), goat anti-NeuroD (1:200; Santa Cruz sc-1084), mouse anti-Cre (1:150; Abcam ab24607), chicken anti-β-galactosidase (1:2000; Abcam ab9361), mouse anti-TH (1:200; Sigma T1299), goat anti-Sox10 (1:200; Santa Cruz sc-17342), rabbit anti-GFP (1:500; Abcam ab290) and rabbit anti-Brn3a (1:2000; gift from E. E. Turner, Seattle Children’s Research Institute, Seattle, WA, USA). Secondary antibodies were from Jackson ImmunoResearch or Invitrogen.

All quantifications were performed on 18-μm transverse sections. Recombination efficiency of the Wnt1-Cre line was assessed by counting β-galactosidase/Sox10 double-positive cells among total Sox10-positive cells in three different Wnt1-Cre/R26R E10.5 embryos at the level of the hindlimbs, counting five sections per embryo. Recombination efficiency of Sox10-Cre was quantified by counting Cre/Sox10 double-positive cells among total Sox10-positive cells in three different Sox10-Cre embryos at the level of the forelimbs, counting 11 sections per embryo. Fig. 3U represents the average numbers of Dct-positive cells per MSA at E11.5 and per mm length of skin at E16.5. At E11.5, Dct-positive cells were counted in 19 MSAs on ten sections per embryo. At E16.5, Dct-positive cells were counted in the skin of six entire lateral sides (dorsal to ventral) on six sections. Quantification for both stages was at the level of the forelimbs of three control and three Sox10-Cre/Ctnnb1^Δex3 embryos. Proliferation for Fig. 3V was analyzed by counting PHH3/Sox10 double-positive cells among total Sox10-positive cells. At E9.5, quantification was performed on every fourth section between the forelimbs and the hindlimbs; at E10.5, quantification was on six sections at the level of the forelimbs. Three control and three Sox10-Cre/Ctnnb1^Δex3 embryos were quantified for each stage. Brn3a- and Mitf-positive cells for Fig. 5K were counted on at least seven sections of the sympathetic ganglia at the level of the forelimbs. Quantifications were performed on four Wnt1-Cre/Ctnnb1^Δex3 embryos and on four Sox10-Cre/Ctnnb1^Δex3 embryos. Quantification of Mitf- and Brn3a-positive cells was on adjacent or near-adjacent sections of the same embryos. All results are shown as mean ± s.d. Statistical analyses (two-tailed unpaired Student’s t-test; calculation of s.d.) were performed with Microsoft Excel.

Although, in mice, neural crest cells concomitantly engage in the ventral and dorsolateral migratory pathways (Serbedzija et al., 1990), neural and melanocyte cell type specification might occur sequentially. This is reflected by the sequential expression of transcription factors associated with lineage specification. Neurogenin 2 (Neurog2), a basic helix-loop-helix (bHLH) transcription factor required for sensory neurogenesis (Ma et al., 1996), is already expressed in neural crest cells as they begin to delaminate from the neural tube (Sommer et al., 1996). By contrast, the bHLH leucine zipper transcription factor Mitf, which promotes melanocyte development, is only expressed from E10.5 onwards (Opdecamp et al., 1997). Accordingly, Neurog2 was prominently expressed in a subset of migrating neural crest cells and in forming dorsal root ganglia (DRG) in caudal regions of mouse embryos at embryonic day (E) 10.5. By contrast, dopachrome tautomerase (Dct), an enzyme of the pigment synthesis pathway and an early marker of the melanocyte lineage, was not expressed in these cells (Fig. 1A). Similarly, at this early stage, neural crest cells marked by the transcription factor Sox10 (Kuhlbrodt et al., 1998; Paratore et al., 2001; Kleber et al., 2005) expressed neither Mitf protein, Mitf mRNA nor c-Kit, another early marker of the melanocyte lineage (Fig. 1B-D). At E11.5, however, Neurog2 expression was weak, whereas Mitf was strongly expressed in Sox10-positive cells found in the MSA (Fig. 1E,F). Expression of Dct, c-Kit and Mitf mRNA at this stage further demonstrates that traits of the melanocyte lineage become apparent after expression of the sensory lineage marker Neurog2 (Fig. 1A-H).

To address whether the sequential appearance of sensory and melanocyte lineage markers might involve reiterative β-catenin signaling, we aimed to genetically manipulate β-catenin activation in a temporally controlled manner using the Cre-loxP system. In Wnt1-Cre transgenic mice, Cre is expressed in the dorsal neural tube at stages before neural crest delamination (Danielian et al., 1998; Jiang et al., 2000; Hari et al., 2002). Therefore, Cre recombinase in these mice is active at least in a subset of the premigratory neural crest population and its derivatives, as demonstrated for the stages E8.0 and later by Cre-induced β-galactosidase expression in Wnt1-Cre embryos crossed to the ROSA26 (R26R) Cre reporter line (Soriano, 1999) (Fig. 2A,C,E,G). Immunohistochemistry for Cre and Sox10 in Wnt1-Cre mice at E9.5 confirmed prominent expression of Cre already at premigratory stages of neural crest development in the dorsal neural tube, whereas Sox10 was only expressed in Cre-positive cells after their delamination from the neural tube (Fig. 2I-L). Importantly, quantification of Cre-dependent β-galactosidase expression in Wnt1-Cre/R26R mice at E10.5 (Fig. 2M-P) demonstrated that 96±0.4% of all Sox10-expressing migratory neural crest cells had been subject to Wnt1-Cre-mediated recombination. Given that Wnt activity drops after neural crest cells have emigrated from the neural tube (Kleber et al., 2005), our data indicate that the vast majority of neural crest cells are very efficiently targeted by Wnt1-Cre activity before, or at the time of, their emigration from the neural tube.

As in Wnt1-Cre/R26R mice, Cre reporter activity was shown to occur in all neural crest-derived tissues in Sox10-Cre/R26R animals (Matsuoka et al., 2005) (Fig. 2D). Unlike Wnt1-Cre, Sox10-Cre marked neural crest cells only after their emigration from the neural tube, as shown by staining for Cre reporter-driven β-galactosidase expression both on whole-mount embryos and on transverse sections at E8.0, E9.5 and E10.5 of Sox10-Cre/R26R embryos (Fig. 2B,D,F,H). Furthermore, Cre protein was detected in 78±1.7% of all Sox10-positive migratory neural crest cells, but not in the neural tube of E9.5 Sox10-Cre/R26R embryos (Fig. 2Q-T, arrowhead). The combined data demonstrate that Wnt1-Cre and Sox10-Cre are expressed in a large population of lineage-related neural crest cells, albeit at different stages of their development.

The sequential Cre expression in the neural crest population of Wnt1-Cre and Sox10-Cre animals allowed us to address a potential stage-dependent role of canonical Wnt signaling during neural crest development. Previously, we have demonstrated that expression of a constitutively active stabilized form of β-catenin (Ctnnb1^Δex) in the premigratory neural crest of Wnt1-Cre/Ctnnb1^Δex3 embryos promotes sensory neurogenesis at the expense of virtually all other neural crest derivatives, including melanocytes (Lee et al., 2004). Accordingly, presumptive melanoblasts expressing Mitf or Dct were absent in the MSA and around the otic vesicle (OV) of Wnt1-Cre/Ctnnb1^Δex3 embryos at E11.5 (Fig. 3A-D). By contrast, Mitf- and Dct-positive cells were present next to the neural tube, in the MSA, and around the OV of Sox10-Cre/Ctnnb1^Δex3 embryos at E11.5 (Fig. 3E-L). Dct-positive cells were able to colonize the skin and hair follicle primordia of Sox10-Cre/Ctnnb1^Δex3 embryos at E16.5 (Fig. 3M,O). The number of recombined cells found in the hair follicle primordia was comparable in Sox10-Cre/Ctnnb1^Δex3/R26R and control embryos (Fig. 3N,P, arrowheads). Quantification of Dct-positive cells at E11.5 in the MSA and at E16.5 in the skin further demonstrated the presence of similar numbers of Dct-positive melanoblasts in Sox10-Cre/Ctnnb1^Δex3 and control embryos (Fig. 3U).

These data are in agreement with an earlier study (Delmas et al., 2007) reporting that constitutive activation of the Wnt/β-catenin pathway in the melanocyte lineage neither increased proliferation nor the number of melanoblasts. Similarly, proliferation of neural crest cells was not altered in Sox10-Cre/Ctnnb1^Δex3 animals, as demonstrated by quantifying the frequency of mitotic phospho-histone H3 (PHH3)/Sox10 double-positive cells among neural crest cells at E9.5 and E10.5 (Fig. 3Q-T,V). Likewise, BrdU incorporation experiments performed previously in our laboratory revealed no change in proliferation in cultures of neural crest cells isolated from Wnt1-Cre/Ctnnb1^Δex3 embryos (Lee et al., 2004). Thus, although β-catenin signal activation in premigratory neural crest using the Wnt1-Cre transgenic mouse line prevents formation of Dct- and Mitf-positive melanoblasts, it does not interfere with melanocyte lineage specification and localization of appropriate numbers of Dct-positive cells to the OV and the skin, if activated after emigration of neural crest cells from the neural tube using the Sox10-Cre mouse line.

The normal generation of orthotopic Dct- and Mitf-positive melanoblasts in Sox10-Cre/Ctnnb1^Δex3 embryos raised the question of whether constitutive β-catenin activity affects neural crest cell fates only at developmental stages of premigratory, but not migratory, neural crest cells. To address this issue, we further analyzed the phenotype of Sox10-Cre/Ctnnb1^Δex3 embryos, focusing on various neural crest derivatives that fail to properly form in Wnt1-Cre/Ctnnb1^Δex3 animals (Lee et al., 2004). In the head, Sox10-Cre/Ctnnb1^Δex3 displayed gross morphological anomalies from E13.5 onwards, with varying degrees of malformations and reduced population of craniofacial structures by neural crest-derived cells, as shown by in vivo fate mapping at E14.5 (Fig. 4A-C). Fate mapping experiments using Z/EG reporter mice (Novak et al., 2000) revealed a strong reduction of peripheral nerves and of the enteric nervous system in Sox10-Cre/Ctnnb1^Δex3/Z/EG embryos (Fig. 4D-I). Furthermore, the number of autonomic neurons expressing tyrosine hydroxylase (TH) was drastically reduced in the sympathetic ganglia of Sox10-Cre/Ctnnb1^Δex3 animals (Fig. 4J-M). These data demonstrate that β-catenin signaling interferes with the generation of several neural crest-derived cells types, even if activated only after neural crest cell emigration.

In Wnt1-Cre/Ctnnb1^Δex3 mice, activation of β-catenin signaling in premigratory neural crest cells results in the generation of ectopic, often enlarged sensory ganglia (Lee et al., 2004). At E10.5, such oversized ganglia can be found, for instance, at the location of normal sympathetic ganglia and contain undifferentiated Sox10-positive cells, Neurog2-positive sensory progenitors, and Brn3a-expressing (Pou4f1 – Mouse Genome Informatics) differentiated sensory neurons (Fig. 5G,H) (Lee et al., 2004). By contrast, although ectopic Brn3a-positive cells were still present at the sites of sympathetic ganglia, the ganglia were smaller in Sox10-Cre/Ctnnb1^Δex3 than in Wnt1-Cre/Ctnnb1^Δex3 embryos (^*P<0.05) (Fig. 5G-J). Intriguingly, whereas in Wnt1-Cre/Ctnnb1^Δex3 mice only a few Mitf-positive cells were present at sites where sympathetic ganglia usually reside, Sox10-Cre/Ctnnb1^Δex3 embryos exhibited significantly more Mitf-positive cells in these structures at E10.5 (Fig. 5B-E,K). The relative number of Brn3a-expressing cells was reduced accordingly in ganglia of Sox10-Cre/Ctnnb1^Δex3 embryos (Fig. 5K). Neither Brn3a-nor Mitf-positive cells were detected in control sympathetic ganglia (Fig. 5A,F). In addition to Mitf, some of the ectopic cells found in Sox10-Cre/Ctnnb1^Δex3 embryos already expressed Dct (data not shown). Double immunohistochemistry for Mitf and Brn3a, Mitf and NeuroD, and c-Kit and Brn3a, revealed exclusive expression of ectopic melanocyte and sensory neuronal markers in Sox10-Cre/Ctnnb1^Δex3 embryos at E10.5 (Fig. 5L-W). These experiments indicate that migratory neural crest cells in Sox10-Cre/Ctnnb1^Δex3 embryos undergo proper sensory and melanocyte lineage formation and lineage segregation, but that a considerable fraction of these lineages is found at an ectopic site. Moreover, the concurrent suppression of autonomic neurogenesis and other neural crest cell lineages in Sox10-Cre/Ctnnb1^Δex3 embryos (Fig. 4) suggests that β-catenin signaling influences neural crest cell fate decisions after their emigration from the neural tube.

To investigate whether ectopic Dct- and Mitf-positive cells in Sox10-Cre/Ctnnb1^Δex3 embryos are restricted to sympathetic ganglia or more broadly found, we systematically analyzed various mutant embryo structures for the presence of cells expressing Mitf and Dct. At E10.5, Mitf-positive cells were found not just within sympathetic ganglionic anlagen, but also dispersed around these structures (Fig. 6A,B). At subsequent developmental stages, Dct-positive cells were detected at several ectopic places in Sox10-Cre/Ctnnb1^Δex3 embryos, including in many tissues that usually do not harbor neural crest derivatives. For instance, at E12.5, the kidney primordium comprised Dct-expressing cells, which were shown by in vivo fate mapping to be of neural crest origin (Fig. 5C-F). Additionally, neural crest-derived Dct-positive cells were detected in the diaphragm, in the spleen and in the urogenital tract (Fig. 6G-R). Ectopic Dct-positive cells were also found at further locations, for example in reproductive organs (testes and ovaries) and around the dorsal aorta (data not shown), and were sometimes even integrated into epithelial structures. As confirmed by Cre-mediated genetic cell tracking, all these organs were devoid of neural crest-derived cells in the control. Of note, many cells marked by Dct expression or by Cre-induced β-galactosidase expression in different organs of Sox10-Cre/Ctnnb1^Δex3/R26R embryos also displayed pigmentation (Fig. 6S-X). Thus, expression of a constitutively active form of β-catenin in migratory neural crest cells in Sox10-Cre/Ctnnb1^Δex embryos leads to the ectopic localization of neural crest-derived Dct- and Mitf-positive melanoblasts and differentiated bone fide melanocytes in various embryonic organs.

The emergence of extra melanoblasts, concomitant with the loss of other neural crest derivatives in Sox10-Cre/Ctnnb1^Δex3 embryos, is consistent with the idea that β-catenin signal activation regulates cell fate choices in multipotent migratory neural crest cells, promoting melanoblast formation at the expense of other fates. Alternatively, β-catenin signal activation might cause an expansion of melanoblasts on their way from the MSA to the skin, leading to the ectopic localization of excessively produced cells. To exclude this alternative explanation for the presence of ectopic melanoblasts, we made use of the inducible Tyr-CreERT2 allele that drives Cre recombinase in neural crest cells only after dorsolateral migration and melanocyte fate specification (Bosenberg et al., 2006) (Fig. 7A). Induction of Ctnnb1^Δex3 expression by treatment of the pregnant females with tamoxifen at E11.5 and E12.5 resulted in recombination in melanoblasts in the skin and in forming hair follicle primordia at E16.5 (Fig. 7C). Consistent with a previous report (Delmas et al., 2007), β-catenin signal activation did not cause an overt increase in melanoblast cell numbers in the skin of Tyr-CreERT2/Ctnnb1^Δex3 embryos (data not shown), disfavoring a role of β-catenin signaling in promoting melanoblast proliferation. Of note, we did not observe any ectopic melanoblasts in Tyr-CreERT2/Ctnnb1^Δex3/R26R embryos, as demonstrated by in vivo fate mapping and in situ hybridization analysis of Dct expression (Fig. 7D,E, the spleen is shown as an example). These data are consistent with the idea that ectopic melanoblasts in Sox10-Cre/Ctnnb1^Δex3 animals are not a consequence of increased orthotopic melanoblast proliferation.

Recently, it has been reported that melanocytes not only arise from neural crest cells migrating along the dorsolateral pathway, but also from Schwann cell precursors (SCPs) located in nerves innervating the skin (Adameyko et al., 2009). To address whether melanocyte fate acquisition at ectopic sites results from elevated β-catenin signaling in nerves, we expressed Ctnnb1^Δex3 specifically in peripheral glia by means of Dhh-Cre-mediated recombination (Jaegle et al., 2003) (Fig. 7A). As previously shown, Dhh-Cre is active in neural crest-derived cells present along peripheral nerves (Jaegle et al., 2003; Joseph et al., 2004). In agreement with this, immunostaining for Cre protein confirmed that Dhh-Cre is not yet active in migratory neural crest cells at E9.5 (data not shown). Although recombination in nerves appeared to be efficient in Dhh-Cre/Ctnnb1^Δex3/R26R embryos from E12 onwards (Jaegle et al., 2003) (data not shown), we observed neither recombined orthotopic melanoblasts (Fig. 7G, arrowhead) nor melanoblasts at ectopic locations in these animals (Fig. 7H,I; data not shown). These findings do not support a role of β-catenin in promoting the generation of melanocytes from nerve cells.

Based on fate mapping experiments with an inducible Plp-CreERT2 line, it has been suggested that the SCP-to-melanocyte transition occurs within a narrow time window during murine development at ∼E11 (Ernfors, 2010). To exclude the possibility that Dhh-Cre activity might be too late to track and influence a SCP-to-melanocyte lineage switch, we induced expression of Ctnnb1^Δex3 at various early and late time points using the Plp-CreERT2 line (Leone et al., 2003). At E9.5, Cre is broadly expressed in migratory Sox10-positive neural crest cells of Plp-CreERT2/R26R embryos (Fig. 8B-E). Accordingly, tamoxifen treatment at this early stage led to Cre-mediated recombination and β-galactosidase expression in the majority of neural crest-derived cells at E10.5 (Fig. 8F-I) and in multiple neural crest derivatives at E16.5, such as neuronal cells in the DRG and in the enteric nervous system, nerves and presumptive melanoblasts in the skin (Fig. 8J,K; data not shown). This indicates that at this early developmental stage, Plp-CreERT2 expression occurs in unspecified neural crest cells and is not restricted to SCPs. In agreement with the results obtained with Sox10-Cre/Ctnnb1^Δex3 animals, β-catenin signal activation in Plp-CreERT2/Ctnnb1^Δex3 embryos at E9.5 led to the appearance of ectopic Dct-positive melanoblasts at E16.5 (Fig. 8L,M, the spleen is shown as an example). Cre activity in Plp-CreERT2/Ctnnb1^Δex3 embryos became more restricted, however, upon induction at E11.5 and later stages (Fig. 8N-U). In contrast to Dhh-Cre, Plp-CreERT2 efficiently marked nerves as well as melanoblasts in the skin when induced at E11.5, E12.5, E14.5, or at postnatal stages (Fig. 8N,O,R,S; data not shown). When activated from E11.5 onwards, β-catenin signaling failed to promote ectopic melanoblast formation in Plp-CreERT2/Ctnnb1^Δex3 animals (Fig. 8P,Q,T,U). These data indicate that β-catenin-mediated generation of melanoblasts at ectopic sites is due to its activation in neural crest cells before they have localized to nerves or to the skin.

Canonical Wnt/β-catenin signaling has been shown to play reiterative roles during neural crest development, sequentially controlling neural crest induction, neural crest cell delamination, lineage specification, and differentiation of neural crest derivatives (Garcia-Castro et al., 2002; Burstyn-Cohen et al., 2004; Lewis et al., 2004; Kleber and Sommer, 2004; Steventon et al., 2009; Li et al., 2011). In the present study, we add a new twist to this theme, demonstrating that the generation of distinct neural crest cell fates in the mouse is also subject to sequential Wnt/β-catenin signaling. Activation of β-catenin in the premigratory neural crest promotes sensory neurogenesis while suppressing virtually all other possible lineages, including melanocytes (Lee et al., 2004). If β-catenin signal activation occurs in migratory rather than premigratory neural crest cells, however, melanocytes represent a major cell type generated, while most other neural crest cell lineages still fail to develop. These data suggest that neural crest cell fate decisions are regulated by stage-dependent differential responsiveness to Wnt/β-catenin signal activation.

Wnt/β-catenin signal activation in neural crest cells has previously been associated either with expansion of pigment cell numbers or with increased sensory neurogenesis (Dorsky et al., 1998; Jin et al., 2001; Lee et al., 2004). Differences in the outcomes of these studies have been attributed, among others, to possible species-specific mechanisms controlling neural crest lineage decisions (Sommer, 2011). Our work provides an additional explanation for how to reconcile the divergent reports: the exact timing of Wnt/β-catenin signal manipulation appears to have an influence on whether neural crest cells choose to produce sensory neurons or melanoblasts. The use of mouse lines expressing Cre recombinase at different stages of neural crest development allowed us to temporally control conditional Ctnnb1 activation. As demonstrated by in vivo fate mapping, the use of Wnt1-Cre results in recombination in the dorsal neural tube at stages when this structure comprises the premigratory neural crest cell population. By contrast, Sox10-Cre is not active in premigratory neural crest cells in the dorsal neural tube. Hence, genetic tracking of Sox10-Cre-mediated recombination fails to mark cells of the dorsal neural tube at any developmental stage. However, Sox10-Cre efficiently labels neural crest cells after their emigration from the neural tube. As shown by immunostaining for Cre expression and in vivo fate mapping by means of Cre-mediated β-galactosidase expression, Wnt1-Cre and Sox10-Cre are sequentially active in a large fraction of lineage-related neural crest cells. Therefore, the differential outcome of Wnt1-Cre-driven versus Sox10-Cre-driven Ctnnb1 activation is consistent with the idea that canonical Wnt signaling regulates sensory neurogenesis in neural crest cells before or during their emigration, when the cells are still localized in or close to the dorsal neural tube. By contrast, according to this idea, the melanocyte lineage is promoted after neural crest cell emigration. This view is also compatible with our previous finding that conditional inactivation of Ctnnb1 in the premigratory neural crest reveals a requirement of the Wnt pathway for both sensory neuron as well as melanocyte formation (Hari et al., 2002).

Although we show that β-catenin signal activation is sufficient to influence neural crest cell lineage decisions in a stage-dependent manner, we were unable for technical reasons to demonstrate a stage-dependent requirement for β-catenin. Using Cre lines active in specified melanoblasts and/or in glial progenitors, we have defined a relatively narrow time window during which neural crest cells respond to β-catenin activation by increased melanocyte formation (see below for further discussion): after neural crest cell emigration, but before the cells have reached the dorsolateral migratory pathway or the nerves. Since loss of protein lags behind conditional gene deletion, Sox10-Cre-dependent ablation of Ctnnb1 did not abolish canonical Wnt signaling during this narrow time window of signal responsiveness, and neural crest cells were found, therefore, to adopt a melanocytic fate in Sox10-Cre/Ctnnb1^flox/flox animals (L.H. and L.S., unpublished). However, the combined in vivo loss- and gain-of-function data, together with the observation that the Wnt/β-catenin pathway is active in a fraction of migratory neural crest cells in vivo (data not shown), point to a role of canonical Wnt signaling in controlling melanocyte fate in migratory neural crest cells after their delamination from the neural tube.

The virtual loss of several neural crest derivatives in Sox10-Cre/Ctnnb1^Δex3 animals, concomitant with the generation of surplus melanoblasts in these mice, is consistent with the hypothesis that migratory neural crest cells can adopt multiple fates in vivo and that neural crest lineage segregations in these cells can be influenced after their emigration from the neural tube. Alternatively, Wnt signaling might induce expansion of committed melanoblasts, as has been suggested previously based on cell culture experiments (Dunn et al., 2000). However, constitutive activation of β-catenin in melanoblasts using Tyr-CreERT2-driven gene expression led neither to cell expansion in the skin, confirming work by others (Delmas et al., 2007), nor to the appearance of melanoblasts at ectopic sites. Therefore, altered proliferation of melanoblasts cannot explain the phenotype described in the present study.

Apart from dorsolaterally migrating neural crest cells, neural crest-derived cells along nerves innervating the skin have been reported to contribute to pigment cell formation (Adameyko et al., 2009). We used two different Cre-expressing mouse lines to exclude the possibility that β-catenin activation promotes melanoblast generation from nerve cells rather than from migratory neural crest cells. In a first mouse line, Cre is expressed from Dhh promoter elements, which leads to Cre recombinase activity specifically in nerves from early stages onwards (Jaegle et al., 2003). Notably, we found no history of recombination in melanoblasts and melanocytes in these mice, and Dhh-Cre-driven β-catenin activation neither promoted extra melanoblast generation nor the suppression of other fates. In a second mouse line, inducible Plp-CreERT2 expression has been used by others to investigate a transition from SCPs to melanocytes (Leone et al., 2003; Adameyko et al., 2009). In the present study, we show that Cre-dependent recombination in these mice marks multiple neural crest derivatives, including neurons, glia and melanocytes, when induced at early stages of neural crest development. Induction at later stages revealed a more restricted expression pattern of Cre, with recombination being detectable in both glia and melanocytes at all stages examined, including postnatally. Given the reported SCP-melanocyte transition observed in Plp-CreERT2 mice, it is unclear why Dhh-Cre-mediated recombination does not label melanoblasts. Possibly, some melanocytic cells are marked independently of glia in Plp-CreERT2 animals. Alternatively, Dhh-Cre and Plp-CreERT2 might be expressed in distinct nerve cell subpopulations (Sommer, 2011). In any case, the dynamic activity of Plp-CreERT2 during development allowed us to confirm that ectopic melanoblasts found in Sox10-Cre/Ctnnb1^Δex3 animals are due neither to enhanced SCP-to-melanoblast transition nor to expansion of melanoblasts: surplus melanoblasts were only seen in Plp-CreERT2/Ctnnb1^Δex3 when Cre recombinase was induced at early stages to mark multiple neural crest derivatives, but not when recombination in these mice was restricted to nerve cells and melanocytes at stages E11.5 or later. These data demonstrate that β-catenin activation can promote the generation of melanoblasts only during a narrow time window of neural crest development, after emigration of neural crest cells and before their association with nerves or engagement in the dorsolateral pathway.

The proposed relatively narrow time window of β-catenin responsiveness in migratory neural crest cells might also explain why surplus melanoblasts are found at ectopic sites within the embryo, rather than at their normal location in the skin. Normally, neural crest cells destined for a melanocytic fate localize to the MSA and then migrate dorsolaterally to populate the skin. Conceivably, β-catenin activation in Sox10-Cre/Ctnnb1^Δex3 embryos imposes a melanocytic fate on neural crest cells before they have reached their destination in the skin. This causes ventrally migrating neural crest cells to generate melanoblasts at the expense of other derivatives. Intriguingly, in Sox10-Cre/Ctnnb1^Δex3 mice, ectopic melanoblasts and melanocytes are not only found in neural crest target structures, such as the autonomic ganglia, but also in tissues that are normally devoid of neural crest derivatives, such as the spleen, urogenital tract and diaphragm. Although this needs to be addressed, the extensive migration of β-catenin-overexpressing melanocytic cells might reflect processes that are also relevant to the high migratory capacity of cells present in aggressive melanoma.

Earlier clonal assays in avian embryos indicated that at least some neural crest cells are multipotent in vivo, generating clones composed of both pigment and neural cell types (Bronner-Fraser and Fraser, 1988; Bronner-Fraser and Fraser, 1989; Frank and Sanes, 1991). A more recent study suggested, however, that neural crest cells are already lineage restricted prior to their emigration from the neural tube: at early stages, cells migrate ventrally to form neural structures only, whereas at later stages cells migrate dorsolaterally to exclusively produce melanocytes (Krispin et al., 2010a; Krispin et al., 2010b). According to these data, multifated neural crest cells (i.e. cells adopting multiple fates) would be rare in avian embryos, and the majority of neural crest cells would consist of discrete cell subpopulations generating single rather than multiple cell types (Krispin et al., 2010b). In contrast to avian embryos, there is no evidence for a temporal switch from ventral to dorsolateral migration in mouse embryos, as mouse neural crest cells migrate dorsolaterally at the onset of neural crest migration (Serbedzija et al., 1990). In addition, single-cell labeling in mouse embryos identified many clones composed of multiple neural crest derivatives, including neural cells and presumptive melanocytic cells in the dorsolateral pathway (Serbedzija et al., 1994).

Despite these data, our findings support the idea that the generation of distinct neural crest lineages in the mouse is also subject to temporal control mechanisms, as in avian embryos. Because most, but not all, neural crest cells are targeted by Cre in the mouse lines used in this study, we cannot exclude a certain heterogeneity in the neural crest cell population with respect to Wnt/β-catenin responsiveness at a given time point. Nonetheless, a large fraction of neural crest cells appears to be multipotent rather than lineage determined at premigratory as well as migratory stages, and their fate can be influenced in vivo by modulation of cues controlling lineage choices. However, our findings do not exclude the possibility that cells with the potential to respond to β-catenin signaling and to adopt multiple fates upon signal manipulation might actually be lineage restricted during normal development. Thus, whether multipotent cells are also multifated at different stages of neural crest development in vivo needs to be elucidated in future studies.

We thank Dies Meijer, A. McMahon, C. Lobe and P. Soriano for providing transgenic animals and E. Turner and D. J. Anderson for antibodies.