To identify functional zebrafish sox10 regulatory regions we generated a series of plasmid constructs using 5' genomic DNA flanking the zebrafish Sox10 coding region to drive expression of GFP. Plasmid constructs psox10^-4725:GFP (containing sox10 regulatory regions used previously 1367) and psox10^-1252:GFP contain the sox10 gene DNA respectively 4725 bp and 1252 bp immediately 5' of the transcript start site. The latter is defined from our sequencing of zebrafish sox10 cDNAs in which the characteristic signature of the 5' 7-methyl-guanosine cap 68 was found immediately 5' to sequences of exon 1 as defined in Fig. 1 (Dutton et al., 2001; TJC, PhD Thesis, University of Bath). Both constructs also contain the zebrafish sox10 exon 1, intron 1 and the start of exon 2 (Fig. 1). The GFP sequence was placed so the ATG sequence of GFP is in the same relative position as the Sox10 ATG (Fig. 1C). The constructs were then used to test if these regions of DNA contain sufficient regulatory elements to drive detectable GFP expression in a pattern reflecting the endogenous sox10 pattern in embryonic zebrafish. When injected into 1-cell zebrafish embryos approximately one third of injected embryos subsequently expressed GFP in one or more discrete cell types. Examples of cells expressing GFP in transient transgenic sox10:GFP embryos are shown in Fig. 2. In embryos injected with constructs psox10^-4725:GFP or psox10^-1252:GFP, GFP expression was initially seen at early somite stages in areas corresponding to premigratory neural crest (Fig. 2A, B). By 24 hpf expression was seen in migrating neural crest cells in the head and body. These included migrating cells in all major pathways, including the streams forming the branchial arches, cells populating the peripheral head, and cells on both medial and lateral pathways in the trunk and tail (data not shown). In these transient transgenic embryos some of these cells had the morphology and characteristic pigment of melanophores and xanthophores. GFP expression was also prominent in the olfactory neurons (Fig. 2C), otic epithelium (Fig. 2D) and CNS neurons in both the brain and spinal cord. At 48 hpf GFP expression in xanthophores and melanophores was still visible (Fig. 2E) and Rohon-Beard neurons were sometimes labeled. Cells of the developing inner ear also continued to express GFP strongly. GFP expression at 48 hpf became visible in differentiating cartilage cells of the pectoral fin and the jaw and persisted until at least 7 dpf (Fig. 2H). GFP expression persisted in the interneurons of the CNS for 4 days (Fig. 2F). At 72 hpf GFP expression became visible in developing oligodendrocytes, initially in the ventral neural tube and then in a more widespread pattern as they disperse throughout the CNS (Fig. 2G). GFP expression in lateral line Schwann cells was not seen in transient transgenic fish expressing GFP from the psox10^-1252:GFP construct but was visible, rarely, in transient transgenic embryos injected with the psox10^-4725:GFP construct. In addition, GFP expression was seen in notochord and muscle cells, common sites of ectopic reporter expression in transgenic zebrafish 62. These patterns are fully consistent with those described for constructs containing intermediate length promoter fragments (TJC, PhD Thesis, University of Bath). Together, these data suggested that the genomic sox10 DNA region present in all these constructs, but importantly even in the smallest one containing only 1252 bp of upstream sequence, contained regulatory sequences able to drive mosaic GFP expression in transient transgenic zebrafish with an expression profile and timing consistent with that previously described for the zebrafish sox10 gene by in situ hybridisation 8.

Before extending the promoter analysis, we checked whether our transient sox10:GFP transgenics data would be consistent with the pattern of GFP expression seen in integrated germ line sox10:GFP transgenics. We have previously published a similar line made from the p4.9:GFP plasmid, Tg(-4.9sox10:egfp)^ba2 19, but since this was only a single line it was not clear exactly how representative that pattern might be. Therefore, we generated a number of sox10:GFP germ line transgenic zebrafish lines using the psox10^-4725:GFP and psox10^-1252:GFP constructs. Multiple transgenic founders were identified for each construct and representative germline transgenic lines were established. Initially 10 founders for sox10^-4725:GFP and 6 for sox10^-1252:GFP were identified. F2 embryos from all 10 sox10^-4725:GFP founders had similar GFP patterns with no obvious expression domain differences. Lines Tg(-4725sox10:GFP)^ba3 , Tg(-4725sox10:GFP)^ba4were subsequently maintained in our fish facility as representative lines although Tg(-4725sox10:GFP)^ba4 embryos have a slightly greater GFP expression level. The Tg(-1252sox10:GFP)^ba5 line was the only one maintained from the shorter construct founders, since it showed no obvious ectopic GFP expression domains; offspring from the other -1252sox10:GFP founders sometimes displayed ubiquitous ectopic GFP expression suggesting they were less insulated against insertion position effects.

Transgenic lines Tg(-4725sox10:GFP)^ba3, Tg(-4725sox10:GFP)^ba4 and Tg(-1252sox10:GFP)^ba5 showed a pattern of GFP expression consistent with the transient transgenic studies utilising the same DNA constructs. GFP expression in transgenic embryos of lines Tg(-4725sox10:GFP)^ba4 and Tg(-1252sox10:GFP)^ba5 are compared in Table 1 alongside the GFP expression profile of transiently transgenic embryos injected with construct psox10^-1252:GFP. Examples of GFP expression patterns in Tg(-4725sox10:GFP)^ba4and Tg(-1252sox10:GFP)^ba5 are shown in Fig. 3. The earliest GFP expression seen in the Tg(-4725sox10:GFP)^ba4 was in the cranial premigratory neural crest at the 1-somite stage. This expression was not visible in Tg(-1252sox10:GFP)^ba5. At 16 hpf GFP expression was visible throughout the premigratory neural crest (Fig. 3A) and by 24 hpf in migrating neural crest cells, branchial arches, ear epithelium and olfactory neurons (Fig. 3B, C). Weak GFP expression was visible in the neural tube. At 48 hpf GFP expression was observed in the ear epithelium, and both melanophores (Tg(-4725sox10:GFP)^ba4 only) and xanthophores were labelled (Fig. 3D), along with jaw and pectoral fin cartilage. GFP expression was seen in both brain and neural tube. Additional sites of reporter expression were seen at 96 hpf when labelled Schwann cells and oligodendrocytes were observed (Fig. 3E, F). GFP expression in pigment cells was largely absent by this time but the jaw and pectoral fin expression remained (Fig. 3G, H). These GFP expression patterns show strong similarities to those we have characterised for intermediate length promoter fragments, although Tg(-4.9sox10:eGFP)^ba2 is unique in showing reduced expression in otic vesicle 19 (TJC, PhD Thesis, University of Bath).

Although transgenic offspring from founders injected with the psox10^-1252:GFP sequence expressed much weaker GFP levels than individual transient transgenics containing the same sequence, the GFP patterns in the transgenic lines were consistent with that deduced from transiently transgenic cohorts injected with psox10^-1252:GFP. Lines Tg(-4725sox10:GFP)^ba3 and Tg(-4725sox10:GFP)^ba4 expressed GFP at a higher level than Tg(-1252sox10:GFP)^ba5 but the pattern and timing of GFP expression were also comparable in these lines. Some differences in GFP expression between the transgenic lines and the transient transgenic cohorts could be attributed to weak levels of expression in the Tg(-1252sox10:GFP)^ba5 lines. For example, no melanophores expressing GFP were seen in Tg(-1252sox10:GFP)^ba5 embryos although this was seen in the corresponding transient transgenic embryos. Conversely, both glia of the lateral line and DRG associated glia were seen to express GFP in the germline transgenics where this expression had not been seen in the corresponding transient transgenic cohort. This difference presumably reflects the mosaic nature of transgene expression in transients and the consequent low probability of transgene expression in cells derived from uncommon progenitors.

Analysis of the GFP expression pattern in lines Tg(-4725sox10:GFP)^ba4 and Tg(-1252sox10:GFP)^ba5, together with our previous transgenic line 1969, confirmed our conclusion from the transient transgenic analysis that the sox10 genomic sequences described contained regulatory sequences able to recapitulate the temporal and spatial pattern of early sox10 expression in zebrafish. In addition, however, these data considerably reduce the genomic DNA interval containing these regulatory regions compared with the published Tg(-4.9sox10:egfp)^ba2 line. Consequently, we decided to extend our analysis of zebrafish sox10 regulatory regions using a transient transgenic approach.

In order to identify more precisely where key sox10 regulatory elements were located, we created a nested series of 5' zebrafish sox10 promoter deletions (Fig. 4). Cohorts of transient transgenic fish were examined for patterns of GFP expression at 48 hpf and 96 hpf points and scored under a fluorescent dissecting microscope at 600× for both numbers of fish expressing any GFP signal and those with GFP signal representing sox10 expression (Fig. 4, Table 2). Significantly, the representative sox10-like pattern of GFP characterised above was maintained in constructs following increasingly greater 5' deletions. Although the overall level of GFP expression was reduced with larger 5' deletions (as deduced from the lower percentage of fish expressing GFP in any pattern; due to the selection for visible expression the level of GFP expression in individual cells where seen was generally comparable), deletion of the transcript start was required for complete loss of the sox10-like pattern of GFP expression. Although expression in fish injected with psox10^-100:GFP, a construct containing only 100 bp of sequence 5' to the transcript start, resulted in a very low percentage of GFP-expressing fish, the spatial pattern of GFP was generally maintained with all sox10-like GFP-positive cell-types represented. This suggested that although control of the level of the sox10:GFP expression was mediated by sequences 5' of the transcript start, critical elements regulating the spatial pattern of sox10 expression were likely to be located elsewhere within the construct.

To attempt to identify these elements we next constructed plasmids from a zebrafish sox10 cDNA (i.e. lacking intron 1), but with the addition of flanking genomic sequences from exon 1 (Fig. 4). The first of these were psox10-1000cDNA:GFP and psox10^{-1252Δ(+133–+2220)}:GFP. These both deleted the whole of intron 1 and differed only in 252 bp of DNA 5' of the transcript start. Transgenic fish containing either of these two constructs contained only extremely rare GFP expressing cells and all sox10-like expression was absent. This immediately suggested that regulatory elements controlling the sox10:GFP expression pattern were present within intron 1. Restriction sites present in intron 1 enabled analysis of constructs containing either 5' or 3' portions of intron 1. Construct psox10^{-1252Δ(+1862–+2220)}:GFP deleted 376 bp from the 3' part of intron 1. Transgenics made with this construct showed no GFP expression in a sox10-like pattern. In contrast, psox10^{-1252Δ(+159–+1862)}:GFP transgenics retained the 3' part of intron 1 and expressed GFP in a sox10-like pattern (Fig. 4B). These data suggested that regulatory elements controlling sox10:GFP expression were likely to be contained within the 376 bp at the 3' end of intron 1.

We used the FirstEF promoter prediction algorithm 70 with a view to beginning to identify the location of the sox10 promoter in nine species. This software takes a genomic sequence and identifies putative promoters by evaluating genomic regions for CpG islands, promoter regions and first exon donor sites. This algorithm has been shown to appropriately identify about 80% of first exons with a false prediction rate of approximately 15%. In addition to the position of the predictions, FirstEF output includes the probability of finding a true promoter at the predicted location. These analyses revealed that all species have a predicted promoter within intron 1 of the sox10 gene (Fig. 5), with results from only two of the eight species (human and chicken) having a probability significantly lower than 1. Interestingly, this software did not predict the transcription start site used in the identified zebrafish sox10 cDNAs (AA, data not shown).

Multi-species comparative sequence analysis is emerging as a powerful tool for identifying transcriptional regulatory elements. Indeed, this approach has proven useful for sox10 using data from multiple mammalian species 276061. However, this technique presents unique problems at greater phylogenetic distances and detailed analysis using zebrafish as a reference sequence has not revealed any significant non-coding nucleotide conservation compared to multiple mammalian species or chicken at sox10 (data not shown; see also 61). However, because functionally conserved elements can be missed by these in silico approaches 66, we explored the possibility that smaller motifs (e. g. transcription factor binding sites) are conserved between these species. Specifically, we identified regions 100% identical and at least six basepairs long in seven mammalian species at sox10 intron 1 using the ExactPlus software 27. The conserved fragments were then submitted to TRANSFAC to identify putative transcription factor binding sites. These analyses revealed six biologically-relevant predictions in mammalian species (Fig. 5B; See Additional file 1). To determine if these predictions were conserved in chicken and zebrafish, we submitted the entire intron one sequence for each species to TRANSFAC. Interestingly, chicken and zebrafish sequences have the same predictions as mammals in the 3' most region of intron one. Furthermore, not only are these sequences conserved but so is their arrangement. We hypothesised that these conserved transcription factor binding sites might contribute to an enhancer regulating sox10 expression in zebrafish.

The critical regions of the sox10 promoter highlighted in this study contained numerous binding sites for transcription factors previously implicated in both neural crest and sox10 regulation, including Notch, β-catenin/Tcf, FoxD3 and SoxE proteins. Before focusing our attention on the highlighted regions we wished to confirm whether these proteins were indeed individually competent to induce sox10 expression in zebrafish. Embryos at the single cell stage were injected with constructs containing notch^ΔEmv, a constitutively active Notch, Δβ-catenin (an activated βcatenin) or sox9b under the control of a heatshock promoter. The embryos were incubated at 28.5°C for 6 hours before being subjected to a 60 min, 37°C heatshock, then left for 1 hour before fixing and processing to detect sox10 transcripts by in situ hybridization. These embryos, equivalent to approximately 8-9 hpf embryos, would be expected to be too young to show endogenous sox10 expression at this stage. Consistent with this, uninjected embryos subjected to heatshock did not express detectable levels of sox10. In contrast, embryos expressing Sox9b, Notch or β-catenin showed induced sox10 transcripts in the expected mosaic pattern (Fig. 6A,B; data not shown). Quantitation indicated that all three transcription factors were able to induce sox10 expression in these early embryo tests (Table 3). In contrast, heatshock promoter driven FoxD3 was unable to induce ectopic sox10 expression in this experiment (data not shown).

To support further the idea that regulatory regions within the promoter region analysed here might be responsive to these transcription factors, we repeated the heatshock experiments in the Tg(-4725sox10:GFP)^ba3 reporter line. Embryos were injected with 400 pg of either hs:notch^ΔEmvΔGFP (as hs:notch^ΔEmv but with GFP coding region deleted) or hs:zsox9b, heat-shocked for 1 hour at 6 hpf, then left for one hour, before being fixed and examined by whole-mount mRNA in situ hybridisation for gfp expression. Uninjected embryos showed no gfp expression confirming that at the stage examined endogenous gfp expression is lacking; in contrast a large proportion of embryos injected with either the sox9b or notch constructs showed gfp expression (Table 4; Figure 6C). Interestingly, the sox9b construct showed a significantly increased proportion of embryos with ectopic gfp⁺ in injected embryos after heat shock when compared with non-heat-shocked controls, whereas the notch construct did not. Given the known leakiness of this heat shock promoter 71, this result indicates that the transgene is very sensitive to even low levels of Notch activity. Together, these data confirm the responsiveness of the sox10 promoter to Sox9b and Notch.

Our combined bioinformatics and physical promoter deletion studies had independently converged to indicate that the 3' end of zebrafish sox10 intron 1 might contain regulatory sequences critical for sox10 expression. To characterise functionally the 376 bp region at the 3'end of intron 1 it was initially divided into 3 overlapping regions (Fig. 7). DNA pieces corresponding to these were generated by PCR and ligated into the psox10^{-1252Δ(+1862–+2220)}:GFP vector. This vector lacks the entire 376 bp piece and both transient and germline transgenic fish made with this vector do not express GFP in a sox10-like pattern. Re-introduction of fragment A into this vector did not restore any GFP expression. Insertion of fragment B restored the GFP expression, albeit only weakly (in 12% of scored fish; n = 33). Notably, the distribution of these GFP positive cells was consistent with the sox10-like expression pattern (see +B in Table 2). Re-introduction of fragment C into the inactive vector also restored GFP to 23% of the fish scored and again the distribution of GFP positive cells largely corresponded to that expected for sox10 expression (see +C, Table 2). Significantly, a construct (psox10^{-1252Δ(+1862–+2220)+BC}) containing a fragment (BC) spanning both the B and C regions was able to restore GFP expression in appropriate cell types and with a distribution not significantly different (p > 0.05) to the parent psox10^-1252:GFP construct (see +BC, Table 2). Thus, adding the BC fragment, which spans all six conserved, biologically relevant sequence predictions identified bioinformatically, to the psox10^{-1252Δ(+1862–+2220)} vector is sufficient to restore most aspects of the sox10-like pattern. The positions of these putative transcription factor binding sites in the B and C regions are shown in Fig. 7.

To assay the functional relevance of these sites in sox10:GFP expression another intronic deletion sequence containing these sites and including the minimal required region was generated. The 2–5 fragment contained multiple sites with invariant CAAA cores of consensus Lef/Tcf/Sox sites(A/T A/T CAA A/T G/T) and includes the evolutionarily conserved Tcf and NFKappaB sites but not the FoxD3/Sox sites identified in the TRANSFAC analysis of sox10 promoters. Reintroduction of this piece into an inactive vector (psox10-^{1252Δ(+1862–+2220)}) could restore GFP expression in 22% of fish counted and gave a sox10-like pattern (see +2–5, Table 2), although the robustness of the sox10-like profile was reduced. We concluded that this region with multiple transcription binding sites was sufficient to restore the correct sox10-like expression of the heterologous reporter.

Given the robust induction of sox10 transcription by β-catenin, and given the prominent role of Wnt signalling in neural crest induction 395472, we wanted to test if the Tcf/Lef binding sites identified in the B and 2–5 fragments could act as functional Tcf/Lef binding sites. Electrophoretic mobility shift assays (EMSA) were performed using radioactively labelled fragments and myc-tagged Lef1^MT protein (Fig. 8A). A specific DNA band with reduced mobility was present in reactions containing Fragment B and Lef1^MT protein but not in those with GFP protein made in vitro. To ensure that the band was not due to the presence of rabbit reticulocyte extract a supershift assay was performed with anti-Myc antibody. In the presence of the anti-Myc tag antibody the specific retarded DNA complex was abolished and replaced by a supershifted band (*, Fig. 8A). The specific complex was however unaffected by incubation with an unrelated anti-WT1 antibody. Competition assays with unlabelled double stranded oligonucleotides showed that an oligonucleotide containing a wild-type Lef1 binding site (GTCAAAG) could abolish the specific binding of Lef1^MT to fragment B but oligonucleotides containing only a mutated site (AGCTGAG) were unable to compete for specific binding (Fig. 8B).

To directly address whether the Lef1 binding sites might be crucial for sox10 regulation, we next used site directed mutagenesis (Table 5) to disrupt the CAA core of the conserved Lef1 binding sites B1, B2 and site 5 (see Figure 7). Mutation of site B1 alone had little effect on Lef1 binding in vitro (Frag B1, Fig 8C) but additional mutation of site B2 reduced Lef1 binding to a fragment carrying both mutations (Fragment B1B2, Fig 8C). Fragment 2–5, previously shown to restore sox10:GFP spatial expression in transient transgenics, also efficiently bound Lef1 in mobility shift assays (Fig 8D) and mutagenesis of the sites B1, B2 and site 5 in this fragment also reduced but did not eliminate Lef1 binding (Frag 2–5^mut, Fig 8D). To eliminate residual Lef1 binding to the B1 site we made fragment 3–5. This fragment contains only two of the identified evolutionarily conserved Lef1 binding sites. Fragment 3–5 bound in vitro made Lef1 less efficiently than fragment 2–5 (Fig 8E) but insertion of this fragment into the inactive vector was still able to restore sox10:GFP expression (18/70 embryos screened at 48 hpf). While this result further delineates the regulatory region required for sox10-like reporter expression to that surrounding the evolutionarily conserved transcription factor binding sites identified in our sequence comparison, mutagenesis of the conserved Lef1 target sites in the 3–5 sequence could not eliminate Lef1 binding (Fig. 8E) or sox10:GFP reporter expression (not shown), indicating that additional target sequences in this region contribute to embryonic sox10 expression; these additional sequences may include non-canonical Lef-binding sites, as well as sites binding other transcription factors. Taken together, our data provides evidence that regulation of early sox10 expression is complex, and is likely to depend on a combination of both Lef1 and other transcription factors binding to a defined region of intron 1 of sox10.

Embryos were obtained through natural mating and maintained at 28.5°C. Embryonic stages are as described by Kimmel et al. 79. Embryo ages are described in hours post-fertilization (hpf).

A PAC clone (BUSMP706I16137Q2) with an 84 Kb insert encompassing the zebrafish Sox10 coding region was partially sequenced 5' and 3' of the previously identified sox10 coding region (GenBank RefSeq ID NM_131875; TJC, PhD Thesis University of Bath) (Fig. 1A and 1B). This sequence enabled both sub-cloning of potential zebrafish sox10 promoter sequences and a bioinformatics comparison of the zebrafish sox10 5' UTR with potential sox10 promoter sequences from other vertebrates. Plasmid psox10^-4725:GFP (Figs 1 and 4) was constructed as follows. A 7 kb EcoR1-SpeI fragment terminating at the 3' end just prior to the ATG of the zebrafish sox10 gene was isolated from BUSMP706I16137Q2 (TJC, PhD. Thesis University of Bath) and cloned into EcoRI-SpeI digested pBluescript KS+ (Stratagene). The fragment was subsequently released as a SalI-XbaI fragment and cloned into SalI-XbaI digested pCS2XLTGFP to form psox10^-4725:GFP. Note that in this manuscript we number all constructs with respect to the transcription start site (+1); the psox10^-4725 promoter fragment corresponds to the promoter fragment used in previous publications 1367. Plasmids psox10^-1252:GFP and psox10^-521:GFP were created by double-digesting psox10^-4725:GFP with EcoRI and MluI, and with EcoRI and NdeI, respectively. Sticky ends were rendered blunt with T4 DNA polymerase and plasmids recircularised using T4 DNA ligase. Plasmid p+20:GFP was described by TJC (PhD. Thesis, University of Bath). Constructs psox10^-1000cDNA and psox10^{-1252Δ(+133–2238)} were constructed by PCR using annealed overlapping fragments generated from a sox10 5' RACE clone 8 and psox10^-4725:GFP. Site directed mutagenesis (Quik Change XL, Stratagene) was used to insert restriction sites for further deletions: NheI at position -100, EcoRV at +159, SmaI at +2220. Deletions between engineered and naturally occurring sites were used for further deletion constructs: psox10^-100:GFP was formed by re-ligation between NheI at -100 and EcoRI at -4725. Other deletions were: Δ+159 – +1862 re-ligated between EcoRV and SmaI, Δ+1862 – +2220 re-ligated between SmaI sites, Δ+159 – +468 re-ligated beween EcoRV and BmgB1 and Δ+468 – +1862 re-ligated between BmgBI and SmaI. Restriction analysis and sequencing were used to confirm vector construction. Prior to embryo injection all plasmids were isolated using a High Purity Maxi-prep kit (Marlingen), DNA concentration was determined by spectrophotometry and confirmed by agarose gel electrophoresis.

The hs:notch^ΔEmv-hs:GFP plasmid encodes mNotchΔE, a Notch mutant which undergoes ligand-independent proteolytic cleavage, under the control of the heatshock promoter. The plasmid also encodes GFP under the control of a second HS promoter and was a gift from Dr Caroline Beck. To test this construct in the Tg(-4725sox10:GFP)^ba3 transgenic line, we deleted the GFP coding region by double digestion with NdeI/SpeI followed by religation, to make plasmid hs:notch^ΔEmvΔGFP. pCH85 encoding hs:Δβ-catenin was provided by Arne Lekven. Hs:Fkd6 plasmid was constructed by cloning the zebrafish Fkd6 coding region as a BamHI-XhoI fragment into pCsHsp. The zebrafish Sox9b coding region was cloned as an XbaI fragment into pCsHsp from a plasmid supplied by John Postlethwait (University of Oregon, USA) to generate plasmid hs:zsox9b. Myc-tagged Lef1 was made from pCs2Lef1^Myc37, kindly provided by Richard Dorsky (University of Utah, USA).

10 nl purified, undigested plasmid DNA at a concentration of 40 ng/μl in water containing 0.1% phenol red was injected into the yolk of 1 cell stage fertilized, wild-type zebrafish embryos. Morphologically abnormal embryos were identified and discarded at 24 hpf. GFP expression was initially observed in live embryos using a MZFL fluorescent dissecting microscope (Leica) before confirmation and documentation using an Eclipse E800 compound microscope (Nikon). GFP expressing cells were identified by their location and morphology and cohorts of at least 35 GFP-positive embryos were examined in detail to compile an overall expression profile for each construct, paying special attention to known sox10 expression domains. Photographic documentation utilised a C4880 CCD camera (Hamamatsu) and compiled using Adobe Photoshop.

sox10 promoter:GFP germline transgenic zebrafish lines were generated by co-injection of plasmid constructs containing I-SceI restriction sites and I-SceI restriction enzyme (Roche). Oligonucleotides containing I-SceI restriction sites (TAGGGATAACAGGGTAAT) and appropriate flanking sequences were inserted into sox10 promoter constructs at the HindIII and Asp718 sites flanking the transgene. 10 nl purified, undigested plasmid DNA at a concentration of 10 ng/μl was injected into the cell cytoplasm of 1-cell stage fertilized embryos in injection mix containing 0.5× Mg free I-SceI buffer and 0.2 units/μl I-SceI (Roche, aliquoted and stored at -80°C). Injected embryos with mosaic GFP expression were identified after 24 hpf and the individuals with the highest contribution of GFP-positive cells (approximately 30% of GFP positive embryos) were raised to adulthood. Subsequent crossing with wild-type fish enabled identification of founder germline transgenic fish for individual lines following observation of non-mosaic GFP expression in F1 embryos. Temporal and spatial expression of GFP was monitored and documented as for the transient transgenic embryos. Additional images were taken using a LSM 510 (Zeiss) confocal microscope.

Genomic sequences were obtained by sequencing bacterial artificial chromosomes harboring the sox10 locus or from the University of California at Santa Cruz (UCSC) Genome Browser http://genome.ucsc.edu as previously described (Antonellis et al. 2006): human (chr22:36606845–36725000, July 2003 UCSC assembly), cat (GenBank AC137542), dog (GenBank AC137537), cow (GenBank AC137534), pig (GenBank AC137657), rat (GenBank AC137528), mouse (chr15:79203055–79300747, March 2005 UCSC assembly), chicken (GenBank AC147863) and zebrafish (see above). Genomic sequences corresponding to the 5' region of sox10 from each species were submitted to the FirstEF web-based software 70 to predict promoters and adjacent first exons. Default FirstEF parameters http://rulai.cshl.edu/tools/FirstEF/ were employed. Sequences from intron 1 that were identical in 7 mammalian species 27 were submitted to TRANSFAC 80 version 8.1 using the Matrix Search for Transcription Factor Binding Sites (MATCH) and Pattern Search for Transcription Factor Binding Sites (PATCH) interfaces. PATCH parameters were set to identify TRANSFAC entries: (1) in vertebrate genes and for vertebrate transcription factors: (2) six basepairs or greater; and (3) with the maximum number of mismatches set at zero. MATCH parameters were set to identify TRANSFAC entries using the "minimize false negatives" setting. The entire zebrafish and chicken intron 1 sequences were analyzed in the exact same manner.

Zebrafish embryos were injected with 10 nl of a heatshock plasmid construct at 40 ng/μl concentration at the 1-cell stage before incubation for 6 hours at 28.5°C. The embryos were then transferred to pre-warmed embryo media for 60 min at 37°C. After a further 1 hour incubation at 28.5°C embryos were fixed and processed for ectopic sox10 or egfp expression by in situ hybridization 8.

PCR generated target probes were end labeled with γ³²P ATP using T4 polynucleotide kinase (New England Biolabs) and purified by native PAGE. In vitro generated protein was made from purified plasmid vectors using TNT coupled transcription/translation systems (Promega) according to the manufacturers instructions. Protein production was confirmed by Western blotting. For electrophoretic mobility assays a 20 μl reaction mixture containing in vitro made protein, 2000 cpm labeled probe, 1 μg poly(dI-dC)-poly(dI-dC) (Sigma), 50 mM NaCl, 10 mM Hepes pH7.9, 5 mM MgCl₂, 0.5 mM EDTA, 3% Ficoll, 0.1 mM DTT and 1 mg/ml BSA was incubated on ice for 20 minutes before separation on a 5%^w/_v polyacrylamide (37:1), 0.5% TBE gel at 200 V at 4°C. Dried gels were exposed to Biomax MS X-Ray film (Kodak). In supershift experiments 20 μg antibody (anti-WT1 C19 (SantaCruz, USA), anti-Myc (gift of Jonathan Slack, University of Bath, UK) was added to the reaction tube 5 minutes prior to addition of the labelled probe.

Chi squared tests were employed using Yates' correction. Bonferroni correction was used to allow for multiple comparisons.