Hypomorphic Sox10 alleles reveal novel protein functions and unravel developmental differences in glial lineages

Sox10, like other transcription factors of the Sox family, is an important regulator of embryonic development (Hong and Saint-Jeannet, 2005; Kelsh, 2006; Wegner, 1999; Wegner and Stolt, 2005). Neural crest development, in particular, is heavily dependent on Sox10 in all model vertebrate species analyzed so far(Aoki et al., 2003; Britsch et al., 2001; Cheung and Briscoe, 2003; Dutton et al., 2001; Herbarth et al., 1998; Southard-Smith et al., 1998). Survival of neural crest stem cells and the maintenance of their multipotency require Sox10 (Kim et al.,2003). In rodents, several lineage decisions of neural crest stem cells additionally depend on Sox10 so that melanocytes, peripheral glia and cells of the enteric nervous system (ENS) are all missing in its absence(Britsch et al., 2001; Herbarth et al., 1998; Southard-Smith et al., 1998). The continued presence of Sox10 in peripheral glia furthermore suggests additional functions during later developmental stages and terminal differentiation of neural crest derivatives(Ito et al., 2006; Peirano et al., 2000).

Such a late function has already been verified in oligodendrocytes(Stolt et al., 2002). These myelin-forming cells of the central nervous system represent the second main site of Sox10 expression and rely, for terminal differentiation and activation of several key myelin genes, on Sox10(Bondurand et al., 2001; Schlierf et al., 2006; Stolt et al., 2002).

Sox10 structure-function studies have identified several domains that are characteristic of a transcription factor and necessary for its specific cellular actions. These include the DNA-binding high mobility group (HMG)domain (amino acids 101-180) and a transactivation domain at the extreme C-terminus (amino acids 400-466)(Kuhlbrodt et al., 1998b). These domains are highly similar to the corresponding regions in Sox8 and Sox9, to which Sox10 is closely related and with which it forms subgroup E of the Sox protein family.

Further regions conserved among these three SoxE proteins include amino acids 61-101 and 233-306. Amino acids 61-101 have been defined as a DNA-dependent dimerization domain (Peirano and Wegner, 2000; Schlierf et al., 2002). Known Sox10 target genes usually contain multiple response elements in their promoter, some of which bind one Sox10 molecule,whereas others bind two Sox10 molecules in a cooperative manner (for a review,see Wegner, 2005). This cooperative binding requires the dimerization domain. Although the dimerization domain is needed for full activation of target gene promoters in vitro (Schlierf et al., 2002),its physiological relevance in vivo remains to be proven.

No function has yet been attributed experimentally to the region between amino acids 233-306. This region, which we named the K2 domain, has been shown to possess transactivation potential in Sox8(Schepers et al., 2000). It might thus also be involved in mediating Sox10-dependent transcriptional activation. To analyze the function of the dimerization domain and the K2 domain in mice, we replaced the wild-type Sox10 allele by mutant versions that either lacked the K2 domain or carried a triple alanine substitution in the dimerization domain that rendered it inactive in vitro.

The K2 domain mutation was generated by overlap-PCR, deleting bp 1279-1500 in the rat Sox10 cDNA (accession NM019193), corresponding to amino acids 233-306 in the Sox10 protein. The Sox10 aa1 mutant carries a substitution of amino acids Cys71, Ile72 and Arg73 by alanines(Schlierf et al., 2002). Sequences corresponding to the open reading frame for one of the mutant Sox10 proteins and a neomycin resistance cassette with flanking loxP sites were placed between 5′ and 3′ Sox10 genomic regions as homology arms (4.3 kb and 1.5 kb, respectively) in the context of a pPNT vector backbone as described (Britsch et al., 2001; Kellerer et al.,2006; Ludwig et al.,2004b). The targeting vector thereby replaced the Sox10coding exons with a continuous Sox10 reading frame carrying the deletion of the K2 domain or the aa1 mutation(Fig. 2A,B). Sequencing confirmed that no additional unintended mutation had been introduced during cloning.

Both constructs were linearized with XbaI and electroporated into E14.1 ES cells, which were then selected with G418 (400 μg/ml) and gancyclovir (2 μM). Selected ES cell clones were screened by Southern blotting with a 0.6 kb 5′ probe, which recognized a 6.6 kb fragment of the wild-type allele, a 5.9 kb fragment of the Sox10^aa1allele and a 10.9 kb fragment of the Sox10^ΔK2 allele in genomic DNA digested with NcoI (Fig. 2A-C). Appropriate integration of the 3′ end of the targeting construct was verified using a 0.6 kb 3′ probe on ES cell DNA digested with BamHI and ScaI. This probe hybridized to a 10.4 kb fragment in the Sox10^aa1 allele and to a 10.2 kb allele in the Sox10^ΔK2 allele,as opposed to a 4.6 kb fragment in the wild-type allele(Fig. 2A-C). Two targeted ES cell lines for each mutant were injected into C57Bl/6J blastocysts to generate chimeras. Chimeric males from two independent clones transmitted the targeted allele to their offspring. No differences were detected between mice derived from the two different ES cell lines. The neomycin resistance cassette was removed by Cre-mediated recombination, and removal was verified as described(Kellerer et al., 2006). Heterozygous mice were backcrossed on a C3H background. Homozygous mutant embryos (F2-F5) were generated by heterozygote intercrosses.

Genotyping was routinely performed on DNA from tail tips or, in the case of embryos from yolk sacs, by PCR analysis using a common primer located 81 bp upstream of the start codon (5′-CAGGTGGGCGTTGGGCTCTT-3′) and two primers located 487 bp downstream of the start codon in intron 3 of the Sox10 gene (5′-TCCCAGGCTAGCCCTAGTG-3′) or 617 bp downstream of the start codon in mutant Sox10 reading frames(5′-GTGAGCCTGGATAGCAGCAG-3′). Fragments of 568 and 698 bp were indicative of the wild-type and targeted alleles, respectively(Fig. 2D).

Expression plasmids for Sox10, Sox10 ΔK2, and Sox10 Q377X were based on pCMV5 (Kuhlbrodt et al.,1998a; Ludwig et al.,2004a). Luciferase reporter carried the 1 kb Mpz promoter(P0-luc), 3.7 kb of the dopachrome tautomerase promoter (3.7 Trp2-luc) and 0.7 kb of the Mbp promoter (0.7 Mbp-luc)(Ludwig et al., 2004a; Peirano et al., 2000; Stolt et al., 2004).

For luciferase assays Neuro2a neuroblastoma cells were cultivated in 24 well plates, transfected with 0.5 μg reporter and 0.5 μg effector plasmids in the case of P0-luc and 3.7 Trp2-luc or with 0.025 μg reporter and 0.5 μg effector in the case of 0.7 Mbp-luc using Superfect reagent(Qiagen). Cells were transfected in triplicate and harvested 48 hours after transfection.

Immunocytochemistry was performed on Neuro2a cells seeded on coverslips in 3-cm plates and transfected with 2 μg pCMV5-Sox10 or pCMV5-Sox10 ΔK2. After treatment with 4% paraformaldehyde (PFA) for 30 minutes, coverslips were incubated successively with anti-Sox10 rabbit antiserum (1:4000)(Stolt et al., 2003) and goat anti-rabbit Cy3 (1:200, Dianova).

For analysis of protein stability, Neuro2a cells were transfected with wild-type and mutant Sox10 constructs using calcium phosphate precipitates and treated 48 hours after transfection with 25 μg/ml cycloheximide for up to 24 hours. Extracts were prepared after various times of cycloheximide treatment, and Sox10 detected by western blot as described(Stolt et al., 2006). Sox10-specific bands on western blots were quantified using the NIH ImageJ software.

Embryos were isolated at 10.5-18.5 dpc from staged pregnancies and processed for extract preparation, RNA isolation, immunohistochemistry or in situ hybridization (Kellerer et al.,2006; Stolt et al.,2004; Stolt et al.,2003).

Immunoprecipitations were performed on whole embryo extracts using the anti-Sox10 rabbit antiserum(Wißmüller et al.,2006). After reverse transcription of 2 μg whole embryo RNA in a 20 μl reaction, 1 μl of the obtained cDNA was amplified by PCR on a Roche Lightcycler according to the manufacturer's instructions using the LightCycler-FastStart DNA Master SYBR Green Kit with the following primer pairs: primers 1 and 2, 5′-TCAGTCTCGGCTGTCCAGCC-3′ and 5′-GAAGAGCCCAACGCCACCT-3′; primers 3 and 4,5′-TCTACAGCCCCATCTCTGAC-3′ and 5′-CAGCCTCCTCCACTGCCA-3′ (Fig. 2A). Rpl8 transcripts were used for normalization(Kellerer et al., 2006). Annealing was at 64°C.

In situ hybridizations (except probe hybridization and final colorimetric detection) were performed automatically on a Biolane HTI (Hölle and Hüttner AG) using DIG-labeled antisense riboprobes for Mbp, Plp, Mpz,c-Kit, Dct and Sox10 (Britsch et al., 2001; Stolt et al.,2002).

Primary antibodies for immunohistochemistry included: anti-Sox10 guinea pig antiserum (1:2000) (Maka et al.,2005), anti-Oct6 rabbit antiserum (1:2000)(Friedrich et al., 2005),anti-Phox2b rabbit antiserum (1:2000, gift of C. Goridis, Ecole Normale Superieure, Paris, France), anti-Olig2 rabbit antiserum (1:10,000, gift of D. Rowitch, DFCI, Boston, MA), anti-B-FABP rabbit antiserum (1:10,000, gift of C. Birchmeier and T. Müller, MDC, Berlin, Germany), anti-SoxB1 (Sox1/2/3)rabbit antiserum (1:500) (Tanaka et al.,2004), anti-Th rabbit antiserum (1:1000, Biomol), anti-NF165 mouse monoclonal (1:200, Developmental Studies Hybridoma Bank) and anti-NeuN mouse monoclonal (1:500, Chemicon). Secondary antibodies conjugated to Cy3 immunofluorescent dyes (Dianova) were used for detection. Dissected gastrointestinal tracts were also used for NADPH diaphorase staining(Scherer-Singler et al.,1983). Samples were analyzed and documented as described(Kellerer et al., 2006).

To analyze the functional consequences of Sox10 mutations in vivo,we replaced the wild-type Sox10 gene sequences by mutant versions in which amino acids 71-73 of Sox10 were changed from CIR to AAA (aa1 mutant) or in which amino acids 233-306 were deleted (ΔK2 mutant)(Fig. 1A). The aa1 mutant has previously been found to selectively abolish the ability of Sox10 to form DNA-dependent dimers (Schlierf et al.,2002). Taking into account that the ΔK2 mutant affects neither the HMG-domain of Sox10 nor its dimerization domain, its DNA-binding ability is likely to be unaffected. The ΔK2 mutant furthermore localized as efficiently as wild-type Sox10 to nuclei in transfected Neuro2a cells(Fig. 1B). There was also no obvious difference in the stability of the ΔK2 mutant versus wild-type Sox10 in transfected cycloheximide-treated cells(Fig. 1C). Despite these normal characteristics, Sox10 ΔK2 exhibited altered transactivation abilities. In luciferase reporter gene assays, activation of the Mpz and dopachrome tautomerase (Dct) promoters by the ΔK2 mutant was 3.6-fold lower than by wild-type Sox10(Fig. 1D). The residual activity of the ΔK2 mutant still elicited a 20- to 23-fold promoter activation, whereas a truncated Sox10 protein without the previously identified C-terminal transactivation domain(Kuhlbrodt et al., 1998b) was completely inactive (Sox10 Q377X in Fig. 1D). In contrast to Mpz and Dct promoters, the Mbp gene promoter, as a third Sox10 target, was activated almost normally by the ΔK2 mutant (Fig. 1D). We conclude that the ΔK2 mutation selectively affects the transactivation capacity of Sox10 on a subset of its target genes and thus functions biochemically as a weak context-dependent transactivation domain, in contrast to the constitutive transactivation domain in the C-terminal region.

To introduce the aa1 and ΔK2 mutations into the Sox10genomic locus, all coding regions present in exons 3-5, as well as introns 3 and 4, were removed by homologous recombination and replaced by an uninterrupted Sox10 reading frame carrying either the aa1 or theΔK2 mutation followed by a neo selection cassette(Fig. 2A,B). Before analysis,the neo selection cassette was removed by Cre-mediated recombination.

A similar strategy was previously employed to replace Sox10 coding sequences by lacZ sequences (Sox10^lacZ), rtTA sequences (Sox10^rtTA) or Sox8coding sequences (Sox10^Sox8ki)(Britsch et al., 2001; Kellerer et al., 2006; Ludwig et al., 2004b). These studies showed that no essential regulatory sequences were removed during recombination and that the inserted sequences were expressed in a Sox10-specific pattern throughout embryonic development. Detailed expression studies on the Sox10^Sox8ki allele had also shown that expression levels after recombination do not differ significantly from expression levels of the wild-type allele(Kellerer et al., 2006). To ensure that expression levels of the Sox10^aa1 and Sox10^ΔK2 alleles were also comparable to that of the wild type, RNA was isolated from homozygous mutant and wild-type embryos at either 10.5 dpc or 11.5 dpc. Sox10expression levels were analyzed by quantitative RT-PCR with two different primer pairs (primers 1,2 and 3,4 in Fig. 2A) and found to be indistinguishable between wild-type and mutant genotypes (Fig. 3A). Immunoprecipitation from embryo extracts additionally revealed that wild-type Sox10 and aa1 mutant protein were present in similar amounts(Fig. 3B). Although the similar molecular mass of the ΔK2 mutant and the immunoglobulin heavy chains prevented an analysis of this mutant on the protein level, our results indicate that neither transcript nor protein amounts were significantly altered in the two Sox10 mutants.

Additionally, it deserves to be mentioned that replacement of the wild-type Sox10 allele, which has three coding exons, by one in which a continuous reading frame is present, leads to no obvious phenotypic abnormalities (S.S. and M.W., unpublished). The phenotypic changes observed in mice carrying the Sox10^aa1 or the Sox10^ΔK2 allele are therefore caused by the mutations in the Sox10 protein and not by inadvertently introduced changes in expression levels.

Mice with a single Sox10^aa1 or Sox10^ΔK2 allele were viable and fertile. In contrast to what has been observed in Sox10^+/lacZ mice, so far we have not observed colonic aganglionosis in the heterozygotes. However, from the second generation onwards, some Sox10^+/aa1 mice developed a white belly spot on a C3H background, similar to that observed in Sox10^+/lacZ mice, indicating that the Sox10 aa1 mutation affects melanocyte development (data not shown).

Homozygous Sox10^aa1/aa1 and Sox10^ΔK2/ΔK2mice were also born alive and at expected frequencies and did not exhibit the characteristic body posture of Sox10-deficient mice at birth (data not shown). Whole-mount in situ hybridization studies at 10.5 and 11.5 dpc confirmed that both alleles were expressed in pattern and intensity indistinguishable from the wild type (Fig. 3C). This again confirms that transcript levels were not significantly altered, but also argues that early development of neural crest derivatives, especially cranial ganglia, dorsal root and sympathetic ganglia, was normal. Sox10^aa1/aa1 and Sox10^ΔK2/ΔK2mice thus differ from Sox10^lacZ/lacZ mice, which at this age already show severely hypoplastic cranial ganglia as well as the first signs of degeneration in dorsal root ganglia (DRG)(Britsch et al., 2001; Herbarth et al., 1998; Kapur, 1999; Southard-Smith et al.,1998).

Although born at expected frequencies, all Sox10^ΔK2/ΔK2mice and many Sox10^aa1/aa1 died within the first hours after birth. Some Sox10^aa1/aa1 mice, however, survived for up to 3 days. These Sox10^aa1/aa1 survivors failed to thrive and soon exhibited the typical signs of a megacolon. We conclude that both alleles are hypomorphic.

Because of the obvious gastrointestinal innervation defects in Sox10^aa1/aa1 mice, we first studied the ENS in both Sox10 mouse mutants. At 18.5 dpc in the wild type, the gastrointestinal tract was completely colonized by enteric neural crest cells and cytodifferentiation had progressed significantly. As a consequence, enteric neurons (marked by PGP9.5 or NOS, also known as Uchl1 and Nos1, respectively - Mouse Genome Informatics)as well as enteric glia (marked by B-FABP, also known as Fabp7) were present in high numbers in the stomach and in all parts of the small and the large intestine, such as duodenum and caecum. Because all markers yielded similar results, only those for NOS-positive neurons obtained by NADPH diaphorase staining are shown (Fig. 4).

Whereas Sox10-deficient embryos lack an ENS throughout the complete gastrointestinal tract (Herbarth et al.,1998; Kapur, 1999; Southard-Smith et al., 1998),the stomach wall of Sox10^ΔK2/ΔK2embryos contained a fully developed ENS at 18.5 dpc(Fig. 4A). Their large intestine was, however, completely devoid of enteric neural crest, enteric neurons or glia (Fig. 4D). Colonization of the gut by ENS cells had stopped in this mouse mutant in the proximal region of the small intestine(Fig. 4C). The region immediately adjacent to the stomach was little affected(Fig. 4B), and although the exact limit of colonization varied between individual embryos of this genotype, it never went beyond the duodenum. By comparison, Sox10^aa1/aa1 embryos exhibited a more severe phenotype. Already the stomach wall contained fewer neurons and glia than either the age-matched wild-type or Sox10^ΔK2/ΔK2embryos (Fig. 4A). Hardly any cells with characteristics of enteric neural crest, neurons or glia could be detected in small and large intestine (Fig. 4B-D). Both the aa1 and ΔK2 mutations thus affect ENS development, but with different severity.

The observation of a white belly spot in some Sox10^+/aa1 mice indicated that melanocyte development is disturbed in the Sox10 aa1 mouse mutant. To confirm this and extend the analysis to the Sox10 ΔK2 mutant, we investigated melanocyte marker gene expression by whole-mount in situ hybridization in homozygous mutant embryos at 12.5 dpc and compared it with wild-type littermates. Whereas streams of migrating melanoblasts were detected around the eye and in the hindlimb region of wild-type embryos, melanoblast numbers were strongly reduced in both mouse mutants, as assessed using either Dct or kit oncogene(c-Kit) as a marker (Fig. 5A). Quantification in the hindlimb region of Sox10^aa1/aa1 mice revealed that Dct-positive cells were nearly absent and c-Kit-positive cells were reduced by two-thirds (Fig. 5C). Taking into account that c-Kit, in contrast to Dct, also labels cell types other than melanoblasts and that the number of residual c-Kit-positive cells was comparable to that in Sox10-deficient mice,which lack melanoblasts altogether (Britsch et al., 2001; Southard-Smith et al., 1998), the melanocyte lineage is likely to be absent in Sox10^aa1/aa1 mice. This was also corroborated by analysis of the skin of the few Sox10^aa1/aa1 survivors at postnatal day 3 (data not shown).

Sox10^ΔK2/ΔK2embryos, by contrast, had melanoblasts at 12.5 dpc(Fig. 5A), although their number corresponded to only 20% of wild-type levels as determined by in situ hybridization for Dct (Fig. 5B). c-Kit in situ hybridization also revealed a strong reduction, by approximately 50%. This and the absence of a belly spot in adult Sox10^+/ΔK2 mice indicate that although both the dimerization domain and the K2 domain participate in melanocyte development, an intact dimerization function appears to be the more important.

Sox10-deficient mice also lack glial cells throughout the peripheral nervous system (PNS), and this secondarily leads to nerve defasciculation and degeneration of DRG neurons (Britsch et al., 2001). To study the importance of dimerization and the K2 domain for these processes, we analyzed PNS development in Sox10^aa1/aa1 and Sox10^ΔK2/ΔK2mice.

Comparison of the DRG in Sox10^aa1/aa1 and Sox10^ΔK2/ΔK2embryos with wild-type embryos at 11.5 and 12.5 dpc did not reveal any obvious alterations in the shape or size of DRG. NeuN-positive neurons (NeuN is also known as Neuna60 - Mouse Genome Informatics) and B-FABP-positive glia were present in similar numbers in the two mutants and wild type(Fig. 6). There was also no obvious difference in the number of cells expressing wild-type and mutant Sox10 proteins.

Ganglia still appeared normal in their morphology in both mutants at 14.5 dpc and expressed NeuN as well as B-FABP. However, the mutant Sox10 protein had almost completely disappeared in the DRG of Sox10^ΔK2/ΔK2embryos (Fig. 6), which is likely to be due to reduced expression levels as the abundance of Sox10 transcripts was also severely reduced (data not shown). By contrast, other Sox10-expressing tissues continued to express the mutant Sox10 protein in the Sox10^ΔK2/ΔK2embryos. The loss of Sox10 expression led to a substantial increase in apoptosis in DRG of Sox10^ΔK2/ΔK2mice at 14.5 dpc that not only affected the normally Sox10-expressing satellite glia but also the sensory neurons (data not shown). As a consequence, DRG were severely reduced in size at 18.5 dpc in Sox10^ΔK2/ΔK2embryos. Sox10^aa1/aa1 embryos continued to express the mutant Sox10 protein in their DRG. In size and marker gene expression, DRG in Sox10^aa1/aa1 embryos were not affected even at 18.5 dpc. Only the slight changes in DRG shape indicated that minor defects also existed in this genotype. In summary, these results indicate that generation of DRG is dependent neither on an intact dimerization function, nor on the presence of the K2 domain. Further development and maintenance of DRG on the other hand appears to specifically require the activity of the K2 domain of Sox10. At those times when the defect became apparent, Sox10 is predominantly expressed in satellite glia of DRG, arguing that the K2 domain might be particularly important in this cell type.

We also analyzed development of Schwann cells in Sox10^aa1/aa1 and Sox10^ΔK2/ΔK2mice as the second main type of glial cell in the PNS. Schwann cells along peripheral nerves, visualized by immunohistochemistry for NF165 (also known as Nefm - Mouse Genome Informatics), contained Sox10 protein at all times of embryonic development in both mutants and appeared similar to the wild type(Fig. 7 and data not shown). At 18.5 dpc, many of the Schwann cells along wild-type spinal nerves had entered the promyelinating stage in which they express the transcription factor Oct6(also known as Pou3f1 - Mouse Genome Informatics)(Bermingham et al., 1996; Jaegle et al., 1996). Oct6 expression was present in spinal nerves of Sox10^ΔK2/ΔK2embryos at 18.5 dpc. Peripheral nerves from age-matched Sox10^aa1/aa1 embryos, by contrast, lacked significant Oct6 expression (Fig. 7). Instead,Schwann cells along the nerve of Sox10^aa1/aa1 embryos continued to express the SoxB1 transcription factor Sox2 which, in the majority of wild-type Schwann cells, is already downregulated at this age and is a marker for immature Schwann cells (Le et al., 2005). Schwann cells in Sox10^aa1/aa1mice thus remained in the immature stage and did not enter the promyelinating stage. As a consequence, expression of the peripheral myelin genes Mbp and Mpz was not detected along the nerves of Sox10^aa1/aa1 mice at 18.5 dpc, in contrast to the wild type (Fig. 7).

Although Schwann cells from Sox10^ΔK2/ΔK2mice expressed Oct6 and were thus likely to have entered the promyelinating stage, they continued to express Sox2 and therefore also behaved aberrantly(Fig. 7). Similar to Schwann cells from Sox10^aa1/aa1 mice, Schwann cells from Sox10^ΔK2/ΔK2mice did not express myelin genes indicating that they also did not reach the myelinating stage.

In contrast to DRG and peripheral nerves, sympathetic ganglia contain comparatively few glial cells. Neither at 12.5 dpc nor at 18.5 dpc did we detect significant changes in the development of rostral sympathetic chain ganglia of Sox10^aa1/aa1 and Sox10^ΔK2/ΔK2embryos. Immunohistochemistry with antibodies directed against Sox10, Phox2b and tyrosine hydroxylase (Th) yielded numbers of positive cells and signal intensities in the mutants that were comparable to the wild type(Fig. 8 and data not shown).

Finally, we analyzed oligodendrocyte development in the spinal cord of both mouse mutants. Because the mutant Sox10 proteins were expressed throughout oligodendrocyte development at levels comparable to the wild type (data not shown), we used Sox10 expression to follow oligodendrocyte development. Oligodendrocyte precursors were specified in the correct spatiotemporal pattern at 12.5 dpc in the ventral part of the spinal cord of both mouse mutants (Fig. 9A). In the following days, oligodendrocyte precursors spread throughout the mantle zone of the spinal cord and finally started to accumulate in the marginal zone at 18.5 dpc in preparation for terminal differentiation. Both migration pattern and marginal zone accumulation were normal in Sox10^aa1/aa1and Sox10^ΔK2/ΔK2embryos, arguing that there is no defect in oligodendrocyte precursors. Comparing Olig2 expression in control and mutant mice, we also detected no changes in the distribution or number of oligodendroglial cells(Fig. 9B and data not shown).

In situ hybridization with antisense riboprobes for the myelin genes Mbp and Plp (also known as Plp1 - Mouse Genome Informatics) at 18.5 dpc revealed that in contrast to Sox10-deficient embryos(Stolt et al., 2002), terminal differentiation took place at significant rates in both Sox10^aa1/aa1 and Sox10^ΔK2/ΔK2embryos (Fig. 10A). Quantification revealed a slight reduction, to ∼60-70% of wild-type levels, in the number of myelinating oligodendrocytes in Sox10^ΔK2/ΔK2embryos (Fig. 10B). By contrast, terminal oligodendrocyte differentiation and myelination proceeded at wild-type rates in Sox10^aa1/aa1 embryos(Fig. 10C). Oligodendrocyte development is thus surprisingly little affected in the two mutants.

The importance of the DNA-binding and transactivation domains for a transcription factor is almost self-evident and has been confirmed for Sox10 in vitro (for a review, see Wegner,2005) and by the existence of mutations in human patients that either lead to a loss of DNA-binding owing to amino acid changes in the HMG-domain, or to a loss of the constitutive C-terminal transactivation domain(see Inoue et al., 2004).

Transcription factors also contain regions with more subtle functions. For Sox10, these include a domain that is in direct apposition to the beginning of the HMG-domain and which has been found in vitro to mediate DNA-dependent dimerization (Peirano et al.,2000; Peirano and Wegner,2000). There is also a region in the middle of the protein that is fully conserved between Sox10 and its relatives Sox8 and Sox9, and which has transactivation potential in Sox8(Schepers et al., 2000; Wegner, 1999).

Here, we replaced wild-type Sox10 in the mouse by mutant versions that interfered with the function of these domains. For the aa1 variant, three consecutive amino acid residues were substituted. This had previously been shown to interfere with DNA-dependent dimerization(Schlierf et al., 2002). In the ΔK2 mutant, the complete conserved region was removed. Although we replaced the three coding exons of Sox10 by a single continuous open reading frame during the process, expression rates were not significantly altered. The defects observed in mice homozygous for the Sox10^aa1 or the Sox10^ΔK2 allele are therefore primarily due to the changes in the Sox10 sequence.

Compared with a mouse with Sox10 deficiency, Sox10^aa1/aa1 and Sox10^ΔK2/ΔK2mice have a milder hypomorphic phenotype. Furthermore, there are significant differences between the two mouse mutants and these differences, as well as the phenotypic similarities between them, offer interesting insights into the function of Sox10.

We detected no change in neural crest formation and emigration in either mouse mutant, indicating that early neural crest development is not particularly sensitive to either mutation. If we take into account that Sox8 can replace Sox10 during this phase in the mouse(Kellerer et al., 2006) and in chicken (Cheung and Briscoe,2003), neither process appears to require the presence of a particular SoxE protein. Accordingly, the temporal pattern with which SoxE proteins are switched on and function during early neural crest development varies among vertebrate species (O'Donnell et al., 2006).

Melanocyte and ENS development, on the other hand, are significantly affected in both mouse mutants. Because these two processes were also defective in Sox10^{Sox8ki/Sox8ki} mice, they appear to be exquisitely sensitive to changes in Sox10 and reliant on an intact Sox10 protein. Such an assumption is also supported by the fact that the vast majority of heterozygous SOX10 mutations identified to date in human patients lead to defects in melanocyte and ENS development and manifest as a combination of Waardenburg syndrome and Hirschsprung disease(Inoue et al., 2004; Pingault et al., 1998).

ENS and melanocyte development were more affected in the Sox10^aa1/aa1 than in the Sox10^ΔK2/ΔK2mutant, pointing to clear differences between the two hypomorphic alleles. In fact, the melanocyte defect in the Sox10^aa1/aa1 mutant was comparable to the Sox10-deficient situation, arguing that DNA-dependent dimerization plays an essential role during melanocyte development.

Two target genes, Mitf and Dct, have so far been identified for Sox10 in the melanocyte lineage(Bondurand et al., 2000; Britsch et al., 2001; Lee et al., 2000; Potterf et al., 2000; Potterf et al., 2001; Verastegui et al., 2000). In zebrafish, Sox10 function in melanocytes is mainly mediated through mitf activation (Elworthy et al.,2003). Taking into account that the Mitf promoter, in contrast to the Dct promoter, does not contain dimeric Sox10 binding sites (Bondurand et al., 2000; Ludwig et al., 2004a), a strong effect on melanocyte development was not necessarily expected for the Sox10^aa1/aa1 mutant.

Several possibilities exist to explain the severity of the melanocyte defect in the Sox10^aa1/aa1 mutant. In addition to the Mitf promoter, Sox10 might require additional regulatory regions to activate Mitf expression in vivo and the activity of these regulatory regions might depend on dimeric Sox10 binding. The existence of a distal enhancer with Sox10-binding capability has indeed been reported for the Mitf gene (Watanabe et al.,2002). Alternatively, and in contrast to zebrafish, Sox10 might also work through pathways other than Mitf activation in mouse melanocytes. Evidence for such interspecies differences exist(Hou et al., 2006). Instead of Mitf, additional melanocytic target genes might then require dimeric Sox10 binding. Finally, it cannot be ruled out that the amino acid substitutions in the aa1 mutant affect other as yet unidentified functions in addition to DNA-dependent dimerization.

Compared with the defects in ENS and melanocyte development, developmental alterations in peripheral ganglia and nerves had a later onset and were milder. Interestingly, they were more pronounced in both Sox10^ΔK2/ΔK2and Sox10^aa1/aa1 mutants than in the Sox10^{Sox8ki/Sox8ki} mutant, arguing that a fully functional Sox8 protein with intact DNA-dependent dimerization function and K2 domain is better able to substitute for Sox10 in PNS development than the Sox10 mutants. As the PNS defects in the Sox10^ΔK2/ΔK2and Sox10^aa1/aa1 embryos appeared no earlier than 14.5 dpc, they also uncover novel functions for Sox10 in peripheral glia that were impossible to see in the Sox10-deficient mutant, in which peripheral glia are not specified and thus completely missing(Britsch et al., 2001).

In Sox10^ΔK2/ΔK2embryos, DRG develop normally and first contain a near normal complement of neurons and glial cells. However, from 14.5 dpc onwards, Sox10 expression is extinguished. Despite the fact that cells specified to a glial fate remain detectable for some time after the loss of Sox10 expression, DRG begin to degenerate and lose glial cells as well as neurons.

Taking into account that the stability of the ΔK2 protein was similar to that of wild-type Sox10 in tissue culture and that ΔK2 transcripts in DRG were also reduced, we assume that Sox10 expression is selectively downregulated in DRG of Sox10^ΔK2/ΔK2embryos. The K2 domain might in fact be needed to maintain Sox10 expression in DRG glia by a positive autoregulatory feedback mechanism.

Because of the dearth of specific markers for glial cells in the DRG, it is impossible at present to exactly define at which stage after specification satellite glia development is disrupted in the Sox10^ΔK2/ΔK2mutant. It nevertheless shows that under normal conditions, Sox10 is not only important during the specification event, but also has additional roles during later phases of differentiation and maturation of satellite glia. Considering the preferential expression of Sox10 in glia rather than DRG neurons, the simplest explanation for the observed phenotype in the Sox10^ΔK2/ΔK2mutants would thus be a primary failure of satellite glia to differentiate and to produce the signaling molecules needed for neuronal survival. According to this model, the neuronal loss would be secondary to the glial defect. Interestingly, DRG are one of the few tissues in which loss of the K2 domain has a stronger effect than loss of the DNA-dependent dimerization function.

Glial development was also disturbed along peripheral nerves. However, no significant loss of Schwann cells was observed in either mouse mutant and the mutant Sox10 proteins remained strongly expressed throughout embryonic development. The glial cells along the nerve also developed normally into Sox2-expressing immature Schwann cells. In the case of the Sox10^aa1/aa1 mutant, Schwann cells failed to enter the promyelinating stage and did not turn on Oct6 expression. Schwann cells in the Sox10^ΔK2/ΔK2mutant, by contrast, progressed into an Oct6-positive stage, but then also failed to turn on myelin genes. Our analysis therefore also proves for the first time that a fully functional Sox10 is needed for the final two stages of the development of myelinating Schwann cells. This is consistent with biochemical evidence that Sox10 directly activates expression of Krox20 (also known as Egr2 - Mouse Genome Informatics) as the master regulator of peripheral myelination and of several peripheral myelin genes(Bondurand et al., 2001; Ghislain et al., 2002; Peirano et al., 2000). However, our study provides the first genetic evidence and points to the fact that the peripheral neuropathy observed in some patients with heterozygous SOX10 mutations might be due to defective terminal differentiation of Schwann cells rather than to early Schwann cell specification defects(Inoue et al., 2004). The fact that the two Sox10 mutations affect development of Schwann cells and satellite glia differently furthermore corroborates the distinct nature of these two glial populations in the PNS.

It is also interesting that Schwann cell development is more sensitive to alterations to the functional domains of Sox10 than to an exchange of Sox10 for the related Sox8 (Kellerer et al.,2006). This observation on Schwann cells furthermore contrasts with that on oligodendrocytes, which are more severely affected in the Sox10^{Sox8ki/Sox8ki} mutant than in either the Sox10^ΔK2/ΔK2or the Sox10^aa1/aa1 mutant. Surprisingly, the onset of terminal oligodendrocyte differentiation even appears to be completely normal in the Sox10^aa1/aa1 mutant.

Our study has yielded important information regarding the cell types and developmental steps at which DNA-dependent dimerization or the K2 domain becomes relevant for Sox10 function. The differential effect of the ΔK2 mutation on several tissues furthermore supports the data from biochemical studies that define this region as a promoter-specific transactivation domain. In future experiments, it will be interesting to determine which Sox10 target genes might explain the specific defects in the Sox10^aa1/aa1 mutant, and identify the Sox10 interaction partners that mediate K2 domain function in select tissues.

We thank C. Goridis, T. Müller and C. Birchmeier for the gift of antibodies. This work was supported by a grant from the Deutsche Forschungsgemeinschaft to M.W. (We1326/8-1).