Developmental Biology 281 (2005) 309 – 317 www.elsevier.com/locate/ydbio Genomes & Developmental Control Impact of Sox8 on oligodendrocyte specification in the mouse embryonic spinal cord

C. Claus Stolt, Simone Schmitt, Petra Lommes, Elisabeth Sock, Michael Wegner*

Institut fu¨r Biochemie, Universita¨t Erlangen-Nu¨rnberg, Fahrstrasse 17, 91054 Erlangen, Germany

Received for publication 10 December 2004, revised 23 February 2005, accepted 10 March 2005

Abstract

The myelin-forming oligodendrocytes of the mouse embryonic spinal cord express the three group E Sox Sox8, Sox9, and Sox10. They require Sox9 for their specification from neuroepithelial cells of the ventricular zone and Sox10 for their terminal differentiation and myelination. Here, we show that during oligodendrocyte development, Sox8 is expressed after Sox9, but before Sox10. Loss of Sox8 did not impair oligodendrocyte specification by itself, but enhanced the Sox9-dependent defect. Oligodendrocyte progenitors were still generated in the Sox9-deficient spinal cord, albeit at 20-fold lower rates than in the wildtype. Combined loss of Sox8 and Sox9, in contrast, led to a near complete loss of oligodendrocytes. Other cell types such as ventricular zone cells and radial glia remained unaffected in their numbers as well as their rates of proliferation and apoptosis. Oligodendrocyte development thus relies on the differential contribution of all three group E Sox proteins at various phases. D 2005 Elsevier Inc. All rights reserved.

Keywords: Sry; High-mobility-group; Sox9; Sox10; Glia; Oligodendrogenesis; Redundancy; Knockout

Introduction their high-mobility-group domain, they bind into the minor groove of DNA in a sequence-specific manner, thereby Oligodendrocytes are the myelin-forming glial cells of widening the minor groove and introducing a strong bend the central nervous system. In the spinal cord, they are into the DNA (Bowles et al., 2000; Wegner, 1999). Of the primarily generated from neuroepithelial cells in a defined twenty mammalian Sox proteins, Sox10 and its relatives ventral region of the ventricular zone (Rowitch, 2004). Sox8 and Sox9 (also referred to as group E Sox proteins) are Neuroepithelial cells in this pMN domain give also rise to expressed in spinal cord oligodendrocytes in a complex and motoneurons. However, generation of motoneurons and overlapping manner. We have previously shown that Sox9 is oligodendrocytes is temporally separated with motoneuron already present in neuroepithelial cells of the pMN domain, production preceding oligodendrogenesis. Several transcrip- whereas the onset of Sox10 expression coincides with tion factors with importance for generation, development, specification to the oligodendrocyte lineage (Stolt et al., and differentiation of oligodendrocytes have been charac- 2003). After specification, both proteins are co-expressed in terized over the last years, including the bHLH proteins developing oligodendrocytes until terminal differentiation Olig2 and Olig1 and homeodomain proteins of the Nkx when Sox9 expression is extinguished. Sox10, in contrast, family (Rowitch et al., 2002; Wegner, 2001). continues to be expressed and is involved in activating Sox proteins represent another group of transcription myelin expression as part of the terminal differ- factors that figure prominently in oligodendrogenesis. With entiation program in oligodendrocytes (Stolt et al., 2002). Defects in oligodendrocyte development were observed in the absence of Sox9 and Sox10 with Sox9-specific * Corresponding author. Fax: +49 9131 85 22484. defects preceding those occurring in the absence of Sox10. E-mail address: [email protected] (M. Wegner). In the Sox9-deficient spinal cord, oligodendrocyte specifi-

0012-1606/$ - see front matter D 2005 Elsevier Inc. All rights reserved. doi:10.1016/j.ydbio.2005.03.010 310 C.C. Stolt et al. / Developmental Biology 281 (2005) 309–317 cation at mid-embryogenesis was severely affected (Stolt et sections from the forelimb level of genotyped, age-matched al., 2003), whereas oligodendrocytes in the Sox10-deficient mouse embryos were used with the following primary spinal cord failed to undergo terminal differentiation at the antibodies: anti-PCNA mouse monoclonal (1:100 dilution, end of embryogenesis (Stolt et al., 2002). However, neither Roche Biochemicals), anti-Hb9 mouse monoclonal (1:50 the early specification defect in the Sox9-deficient spinal dilution, Developmental Studies Hybridoma Bank), anti- cord nor the terminal differentiation defect in Sox10- Nkx2.2 mouse monoclonal (1:50 dilution, Developmental deficient spinal cord was absolute. In case of the terminal Studies Hybridoma Bank), anti-SoxB1 rabbit antiserum differentiation defect, residual myelin gene expression was (1:500 dilution, Tanaka et al., 2004), anti-B-FABP rabbit attributed to Sox8 occurrence in oligodendrocytes (Stolt et antiserum (1:10,000 dilution, gift of C. Birchmeier and T. al., 2004). Expression levels at this time were considerably Mu¨ller, MDC, Berlin), anti-Olig2 rabbit antiserum (1:2000 lower for Sox8 than for Sox10. This may explain why Sox8 dilution, gift of H. Takebayashi, Kyoto University), affinity- can be deleted from oligodendrocytes without any obvious purified anti-Sox9 rabbit antiserum (1:2000 dilution, Stolt et developmental consequences and why it compensates loss al., 2003), anti-Sox10 guinea pig antiserum (1:2000 of Sox10 so inefficiently (Sock et al., 2001; Stolt et al., dilution, Maka et al., 2005), anti-Sox8 guinea-pig antiserum 2004). (1:2000 dilution), or anti-Nkx6.1 guinea-pig antiserum Here, we analyze Sox8 expression during early phases of (1:1000, gift of M. Sander, UCI, Irvine). Secondary oligodendrocyte development. We show that Sox8 is already antibodies conjugated to Cy2 and Cy3 immunofluorescent expressed around the time of oligodendrocyte specification dyes (Dianova) were used for detection. Incorporated BrdU in neuroepithelial cells of the ventricular zone and that its was detected on tissue sections using an Alexa-488 coupled presence is responsible for remaining oligodendrocyte mouse monoclonal antibody directed against BrdU (Molec- progenitors in the Sox9-deficient spinal cord. Accordingly, ular Probes) at a 1:20 dilution. TUNEL assays were oligodendrogenesis is virtually absent in spinal cords performed according to the manufacturer’s protocol (Chem- deficient for Sox8 and Sox9. icon). Samples were analyzed and documented using a Leica inverted microscope (DMIRB) equipped with a cooled MicroMax CCD camera (Princeton Instruments, Materials and methods Trenton, NJ).

Animal husbandry, genotyping, BrdU labeling, tissue preparation, and in situ hybridization Results

Mice with Sox9loxP alleles (Akiyama et al., 2002) and Sox8 appears in the ventral spinal cord after Sox9, but additional Sox8lacZ (Sock et al., 2001) alleles were kept as before Sox10 double homozygous females, or asdouble heterozygous males in the presence of a nestin Cre transgene (Tronche et al., Using a lacZ marker inserted into the Sox8 genomic 1999) and crossed to generate CNS-specific Sox8lacZ/lacZ, locus, we have previously shown that cells of the Sox9D/D embryos. Genotyping was performed by PCR (Sock oligodendrocyte lineage in the spinal cord of Sox8+/lacZ or et al., 2001; Stolt et al., 2003). Sox8lacZ/lacZ mice express Sox8 (Sock et al., 2001; Stolt et Embryos (from 10.5 dpc to 16.5 dpc) were obtained from al., 2004). A newly developed antibody against Sox8 now staged pregnancies. After fixation in 4% paraformaldehyde, allowed us to directly study Sox8 occurrence in the wildtype 10-Am sections of genotyped, age-matched mouse embryos spinal cord in comparison to its close relatives Sox9 and were generated on a Leica cryotome, and used for Sox10. At 10.5 dpc, Sox8 was not yet expressed at immunohistochemistry. Alternatively, 20-Am sections were sufficiently high levels to be detected by the antibody in used for in situ hybridization with DIG-labeled probes the forelimb region of the spinal cord (Fig. 1b). Similarly, specific for Olig2, PDGF a, and Myelin Basic Sox10 was absent from the spinal cord and occurred only in (MBP) as described (Stolt et al., 2002). For BrdU the developing peripheral nervous system (Fig. 1c). At the labeling, pregnant mice were injected intraperitoneally with same time, Sox9 was already strongly expressed in neuro- 100 Ag BrdU (Sigma) per gram body weight 2 h before epithelial cells throughout the ventricular zone of the spinal embryo preparation (Stolt et al., 2003). cord (Fig. 1a). Whereas there was still no Sox10 expression at 11.0 dpc (Fig. 1f ), weak immunoreactivity against Sox8 Antibody production, immunohistochemistry, and TUNEL became visible at this stage in the ventral part of the assays ventricular zone (Fig. 1e). By 11.5 dpc, intensity of this signal had increased so that Sox8 protein was now easily Antibodies against Sox8 were generated in guinea pig detected in nuclei of ventricular zone cells (Fig. 1h). Sox8- against a purified bacterially expressed protein consisting of positive cells were restricted to the ventral region of the amino acids 2–60 of mouse Sox8 fused to glutathione-S- spinal cord, including the pMN domain from which the first transferase. For immunohistochemistry, 10-Am cryotome oligodendrocyte progenitors emigrated at this time as C.C. Stolt et al. / Developmental Biology 281 (2005) 309–317 311

(Fig. 2a) and Olig2 (Fig. 2b), which at this time are specific markers for the p3 and the pMN domain, respectively (Jessell, 2000), pointed to strong Sox8 expression in the two ventralmost domains of the ventricular zone. Lower amounts of Sox8 were additionally present dorsally adjacent to the pMN domain (Fig. 2). Expression of Sox8 extended farther from the floorplate than that of Nkx6.1 (Fig. 2c) which ends in the p2 domain (Jessell, 2000). Thus, the region of Sox8 expression in the ventral spinal cord reaches at least into the p1 domain. Sox8 was expressed in ventricular zone cells together with Sox9 (Figs. 3a–c) as well as in oligodendrocyte progenitors in combination with both Sox9 (Figs. 3a–c) and Sox10 (Figs. 3d–f). We conclude that, within the pMN domain, Sox8 appears in cells that already contain Sox9. These cells give rise to oligodendrocyte progenitors that additionally turn on Sox10 expression.

Loss of Sox8 does not alter ventricular zone and radial glia development even in the absence of Sox9

Previous studies have already revealed that spinal cord development in the absence of Sox8 proceeds normally (Sock et al., 2001). Similarly, no developmental defects were detected in the spinal cord of Sox8-deficient mice that additionally lacked one Sox10 allele until late in embryo- Fig. 1. Comparative expression of group E Sox proteins in the embryonic genesis (Stolt et al., 2004). Therefore, we asked whether spinal cord. Immunohistochemistry with antibodies specific for Sox9 spinal cord development would be affected in the combined (a,d,g,j), Sox8 (b,e,h,k), and Sox10 (c,f,i,l) were performed on transverse sections from the forelimb region of wildtype embryos at 10.5 dpc (a–c), absence of Sox8 and Sox9. For these studies, Sox9 had to be 11.0 dpc (d–f), 11.5 dpc (g–i), and 12.5 dpc (j–l). deleted in a tissue-specific manner using a loxP-containing Sox9 allele (Akiyama et al., 2002; Stolt et al., 2003)in evident from Sox10 immunohistochemistry (Fig. 1i). At combination with a CRE transgene under the control of 12.5 dpc, Sox8 was still strongly expressed in the same regulatory sequences from the nestin gene (Tronche et al., region of the ventral spinal cord (Fig. 1k). Additionally, 1999). Using this strategy, Sox9 was largely ablated from weak Sox8 staining was detected in ventricular zone cells of the spinal cord in the forelimb area by 12.5 dpc. Only few the dorsal spinal cord. At the same time, the number of Sox9 expressing cells remained as previously shown (Stolt Sox10-positive oligodendrocyte progenitors had increased et al., 2003). When development of ventricular zone cells significantly (Fig. 1l). Sox9 expression persisted throughout and radial glia was analyzed in these Sox8lacZ/lacZ, Sox9D/D the ventricular zone (Figs. 1g,j). To determine which ventricular zone cells exactly express Sox8 at 11.5 dpc, we performed co-immunohisto- chemistry (Fig. 2). Co-expression of Sox8 with Nkx2.2

Fig. 2. Expression of Sox8 protein in the ventral ventricular zone. Co- Fig. 3. Co-expression of group E Sox proteins in the embryonic spinal cord. immunohistochemistry on the ventral spinal cord of 11.5 dpc-old wildtype Immunohistochemistry on the pMN domain of 12.5 dpc-old wildtype embryos with antibodies against Sox8 (red) and the domain markers (all in embryos with antibodies against Sox8 (green in a,c,d,f), Sox9 (red in b,c), green) Nkx2.2 (a), Olig2 (b), and Nkx6.1 (c). Identified domains of the or Sox10 (red in e,f) is presented as single (a,b,d,e) or merged (c,f) pictures. ventral ventricular zone are listed on the right. Several oligodendrocyte precursors are marked with arrowheads. 312 C.C. Stolt et al. / Developmental Biology 281 (2005) 309–317 spinal cords using antibodies directed against group B1 Sox from the wildtype situation (Figs. 5a,b,c,d,g) arguing that proteins, nestin, or brain fatty acid binding protein (B- overall proliferation rates at this early stage are normal. We FABP), no differences were detected relative to the wildtype equally failed to detect differences in apoptosis rates using both regarding distribution (compare Figs. 4a,c,e to 4b,d,f) TUNEL assays (Figs. 5e,f,h). and quantity (Fig. 4k and data not shown). Supporting the Motoneuron numbers, in contrast, were higher in the notion that ventricular zone cells were present in normal Sox8lacZ/lacZ,Sox9D/D spinal cords at 12.5 dpc than in the numbers and had retained their identity, Olig2-positive wildtype (Figs. 4g,h). The overall number of Hb9-positive neuroepithelial cells were present at the expected dorsoven- cells was increased by 44% (1 + 2 in Fig. 4l). The tral position (Figs. 4i,j) and were found in wildtype numbers subpopulation of Hb9-positive cells, that were still at a in Sox8lacZ/lacZ,Sox9D/D spinal cords (VZ in Figs. 4m). medial position close to the ventricular zone, was threefold Neither BrdU labeling experiments on Sox8lacZ/lacZ,Sox9D/D larger (1 in Fig. 4l), presumably indicating that significant spinal cords nor immunohistochemistry for proliferating cell numbers of motoneurons were still born in the Sox8lacZ/lacZ, nuclear antigen (PCNA) revealed significant deviations Sox9D/D spinal cords at 12.5 dpc. The increase in moto-

Fig. 4. Ventricular zone, radial glia, and motoneuron development in embryonic spinal cord deficient for both Sox8 and Sox9. (a–j) Immunohistochemistry with antibodies directed against SoxB1 proteins (SoxB1 in a,b), B-FABP (c,d, as well as higher magnification of the ventral spinal cord in e,f), Hb9 (g,h), and Olig2 (i,j) were performed on transverse sections from the forelimb region of wildtype (a,c,e,g,i) and Sox8lacZ/lacZ,Sox9D/D embryos (b,d,f,h,j) at 12.5 dpc. (k– m) Quantification of B-FABP-positive cells (k), Hb9-positive cells (l), and Olig2-positive cells (m) in ventral spinal cords of wildtype (open bars) and Sox8lacZ/lacZ,Sox9D/D embryos (filled bars) at 12.5 dpc. At least 15 separate 10-Am sections from the forelimb region of 2 independent embryos were counted for each genotype. The number of Hb9-positive cells in the medial region (region 1, inset) was separately determined from the total number of Hb9-positive cells including those in the ventral horn (region 2, inset) (l), as was the number of Olig2-positive cells within (VZ) and outside (MZ) the ventricular zone (m). Data are presented as mean T SEM. Differences to the wildtype (set to 100%) were statistically significant for the mutant genotype as determined by the Student’s t test ( P  0.001) in case of medial Hb9-positive cells (300% T 89%), total Hb9-positive cells (144% T 28%), and Olig2-positive cells outside the ventricular zone (7% T 7%). C.C. Stolt et al. / Developmental Biology 281 (2005) 309–317 313

Fig. 5. Proliferation and apoptosis in embryonic spinal cord deficient for both Sox8 and Sox9. (a–f) Immunohistochemistry with antibodies specific for the S phase marker PCNA (a,b) or BrdU (c,d) was performed on transverse sections from the forelimb region of 12.5 dpc-old embryos previously labeled for 2 h with BrdU. Adjacent sections were used for TUNEL assays (e,f). (a,c,e) Wildtype embryos; (b,d,f) Sox8lacZ/lacZ,Sox9D/D embryos. (g,h) Quantification of BrdU- labeled cells in the ventral spinal cord (g), and TUNEL-positive cells in the total spinal cord (h) of wildtype (open bars) and Sox8lacZ/lacZ,Sox9D/D embryos (filled bars) at 12.5 dpc. At least 15 separate 10-Am sections from the forelimb region of 2 independent embryos were counted for each genotype. Data are presented as mean T SEM. No statistically significant differences to the wildtype were detected for the mutant genotype as determined by the Student’s t test. neurons was only transient as already reported for Sox9D/D mantle zone (Figs. 4j,m). Labeling of oligodendrocyte spinal cords (Stolt et al., 2003). No surplus motoneurons progenitors and neuroepithelial cells in the pMN domain were generated at later phases of development, indicating by Olig2 makes it difficult to distinguish between both that even in the absence of Sox8 and Sox9, neuroepithelial populations at early times of development with this marker. cells in the pMN domain at some point lose their Sox10, in contrast, is selectively expressed in the oligoden- competence to develop into motoneurons. At these later drocyte lineage (Figs. 3e and 6a). Immunolabeling failed to stages, pMN domain cells in the Sox8lacZ/lacZ,Sox9D/D spinal detect Sox10-positive cells in the Sox8lacZ/lacZ,Sox9D/D cords rather appear to persist as radial glia. spinal cord at 12.5 dpc (Fig. 6e). Even at 14.5 and 16.5 dpc, Sox10-positive oligodendrocyte progenitors were not Combined loss of Sox8 and Sox9 prevents detected in the Sox8lacZ/lacZ,Sox9D/D spinal cord (Figs. 6f,g). oligodendrogenesis In contrast, oligodendrocyte progenitors had proliferated and spread throughout the parenchyma of the wildtype In Sox9D/D spinal cords, the transient increase in spinal cord (Figs. 6b,c). Our failure to detect Sox10 even at motoneuron numbers was accompanied by a simultaneous 16.5 dpc, when oligodendrocyte numbers in the Sox9D/D decrease in oligodendrocyte progenitors (Stolt et al., 2003). spinal cord had expanded to 70% of wildtype levels, However, this specification defect was not complete. impressively confirmed the increased severity of the Oligodendrocyte progenitors were born at a twenty-fold oligodendrocyte specification defect in the Sox8/Sox9- lower number than in the wildtype, but were able to double deficient mouse. replenish the oligodendrocyte progenitor pool as embryonic To exclude the possibility that significant numbers of development proceeded so that oligodendrocyte numbers oligodendrocyte progenitors were still present in the absence had recovered to 70% by 16.5 dpc (Stolt et al., 2003). of Sox10 expression, we employed additional markers of In wildtype embryos, Olig2-positive cells are found at the oligodendrocyte lineage. Immunohistochemistry 12.5 dpc both in the ventricular zone and in the mantle zone revealed only few, weakly Olig2-positive cells in some with the latter population corresponding to oligodendrocyte Sox8lacZ/lacZ,Sox9D/D spinal cords at 14.5 dpc and 16.5 dpc progenitors (Figs. 4i,m). In Sox8lacZ/lacZ,Sox9D/D spinal (Figs. 6d,h,i). No Olig2 signal was detected by in situ cords, Olig2-positive cells were selectively lost from the hybridization (Figs. 7a,b). PDGF receptor a-positive cells 314 C.C. Stolt et al. / Developmental Biology 281 (2005) 309–317

Fig. 6. Oligodendrocyte development in the embryonic Sox8/Sox9-deficient spinal cord. (a–h) Immunohistochemistry with antibodies specific for the oligodendrocyte marker Sox10 (a,b,c,e,f,g) and Olig2 (d,h) was performed on transverse sections of wildtype spinal cords (a–d) and Sox8lacZ/lacZ,Sox9D/D spinal cords (e–h) from the forelimb region. (a,e) 12.5 dpc; (b,f) 14.5 dpc; (c,d,g,h) 16.5 dpc. (i) Quantification of Olig2-positive cells in wildtype (open bars) and Sox8lacZ/lacZ,Sox9D/D (filled bars) spinal cords at 14.5 dpc and 16.5 dpc. At least 15 separate 10-Am sections from the forelimb region of 2 independent embryos were counted for each genotype. Data are presented as mean T SEM. Differences to the wildtype (set to 100%) were statistically significant for the mutant genotype as determined by the Student’s t test ( P  0.001) both at 14.5 dpc (2% T 2%) and at 16.5 dpc (5% T 3%). or MBP-positive cells were equally absent at 16.5 dpc or Sox9D/D spinal cords at 16.5 dpc. The strong reduction of all earlier times in the Sox8lacZ/lacZ,Sox9D/D spinal cord (com- analyzed markers of the oligodendrocyte lineage argues that pare Figs. 7d,f to 7c,e and data not shown). Additionally, there are indeed very few remaining oligodendrocyte less than one proliferating and therefore oligodendroglial progenitors in the developing spinal cord of Sox8/Sox9- Nkx2.2-expressing cell (Zhou et al., 2001) was on average double deficient mice. observed per section in the mantle zone of Sox8lacZ/lacZ,

Discussion

All three group E Sox proteins are expressed in the oligodendrocyte lineage

Of the three group E Sox proteins, Sox9 is strongly expressed in chondrocytes, whereas neural crest cells express high levels of Sox10 (for a review, see Wegner, 1999). Dramatic phenotypes are observed in many tissues with high levels of Sox protein expression after loss of the respective Sox protein, leading to the skeletal mal- formation syndrome Campomelic Dysplasia following Sox9 inactivation (Foster et al., 1994; Wagner et al., 1994), and Waardenburg–Hirschsprung syndrome after Sox10 loss (Pingault et al., 1998). Oligodendrocytes, the myelin-forming glia of the central nervous system, express significant amounts of all three group E Sox proteins. Whereas Sox9 and Sox10 have been detected with specific antibodies, Sox8 expression has so far only been analyzed via a lacZ reporter that was integrated into the Sox8 locus during gene deletion and so far appeared to faithfully recapitulate Sox8 expression (Sock et al., 2001; Stolt et al., 2004). Here, we directly detected Sox8 protein in Fig. 7. Oligodendrocyte marker gene expression in the embryonic Sox8/ cells of the oligodendrocyte lineage. This is important as Sox9-deficient spinal cord. In situ hybridization with antisense probes minor discrepancies between expression of Sox8 and the specific for the oligodendrocyte markers Olig2 (a,b), PDGF receptor a (PDGF-Ra in c,d), and MBP (e,f) was performed on transverse sections lacZ reporter still cannot be excluded and actually seem to from the forelimb region of wildtype (a,c,e) and Sox8lacZ/lacZ,Sox9D/D exist. Expression of the lacZ reporter in the ventricular zone (b,d,f) embryos at 16.5 dpc. of the spinal cord was, for instance, only detected in an area C.C. Stolt et al. / Developmental Biology 281 (2005) 309–317 315 surrounding the pMN domain and appeared thus more can indeed cope with the loss of Sox9 and still undergo restricted than Sox8 expression (Stolt et al., 2003). specification to oligodendrocytes. Similarly, lacZ expression in the spinal cord outside the This accessory role in early specification is paralleled by ventricular zone was mostly restricted to cells of the a similar role of Sox8 in late oligodendrocyte development. oligodendrocyte lineage, whereas Sox8 is additionally found During terminal differentiation, Sox8 cooperates with in a significant fraction of radial glia. Sox10 in the activation of the myelination program (Stolt In the spinal cord, the three group E Sox proteins are et al., 2004). Compared to Sox10, Sox8 is expressed at turned on successively with Sox9 preceding Sox8 which in lower levels and again appears to play a relatively minor turn appears before Sox10. Due to the different onset of role. expression, Sox9 and Sox8 are already expressed in In the absence of Sox9 and therefore also in the absence neuroepithelial cells, including those of the pMN domain of Sox9 and Sox8, surplus motoneurons are generated for a whereas the onset of Sox10 expression is concomitant with prolonged period from the pMN domain. We have specification to oligodendrocyte progenitors. previously argued that the most plausible explanation for In addition to this sequential induction of Sox gene these motoneurons is that they are produced instead of expression during early oligodendrocyte development, there oligodendrocyte progenitors because their generation coin- is a selective down-regulation at later phases. Terminally cides with the normal time of oligodendrocyte specification differentiating oligodendrocytes turn off Sox9, but continue and correlates with a reciprocal decrease in oligodendrocyte to express Sox8 and Sox10 (Stolt et al., 2003, 2004). There progenitors (Stolt et al., 2003). The increase was only is also evidence for different levels of Sox gene expression transient because of the apoptotic adjustment of motoneuron in the oligodendrocyte lineage. Thus, during late oligoden- numbers to target tissue innervation (Yamamoto and drocyte development, Sox10 appears to be expressed at Henderson, 1999) and the failure to produce further higher levels than Sox8 (Stolt et al., 2004). Taking these motoneurons at later phases of development. Even in the qualitative and quantitative differences into account, it is absence of Sox9 and Sox8, neuroepithelial cells in the pMN possible to explain oligodendrocyte development in genet- domain at some point lose their competence to develop into ically altered mice with functional loss of group E Sox motoneurons, possibly through loss of neurogenic tran- . scription factors such as Neurogenin2 (Mizuguchi et al., 2001). Sox8 has accessory roles both in early and late oligodendrocyte development Implications for group E Sox protein function

In agreement with its high expression levels in the pMN The joint regulation of developmental processes by domain at the time of oligodendrocyte specification, loss of multiple group E Sox proteins might be fairly common. Sox9 leads to specification defects of the oligodendrocyte Thus, functional cooperation of Sox9 and Sox8 has been lineage (Stolt et al., 2003). This defect, however, is not observed during development of the male gonad (Chaboiss- complete and a few residual oligodendrocyte progenitors ier et al., 2004), and might also be at work during muscle were detected that later expanded and replenished the development (Schmidt et al., 2003). oligodendrocyte progenitor pool. Here, we show that Of the three group E Sox proteins, Sox8 appears to have additional removal of Sox8 strongly enhances the severity the most widespread expression during embryogenesis of this oligodendrocyte specification defect, arguing that (Pfeifer et al., 2000; Schepers et al., 2000; Sock et al., Sox9 and Sox8 cooperate in their ability to specify 2001). At the same time, it also appears to be the least oligodendrocytes. Taking into account that loss of Sox8 important as it is the only group E Sox protein that can be on its own has no influence on oligodendrocyte specifica- deleted in mammals without affecting viability (Bi et al., tion (Sock et al., 2001; Stolt et al., 2004), Sox8 clearly has a 1999; Britsch et al., 2001; Sock et al., 2001). Many Sox8- smaller impact than Sox9. expressing tissues tolerate deletion of the gene, whereas It is conceivable that the few oligodendrocyte progenitors their development is strongly impaired when the co- that originally become specified in the Sox9-deficient spinal expressed Sox9 or Sox10 is deleted. So far, osteoblasts are cord owe their existence to the mode of Cre-dependent Sox9 the only cell type shown to depend on Sox8 for proper ablation. Mosaic and variable Cre activity as well as the development (Schmidt et al., 2005). latency between deletion of gene and protein allow Sox9 Co-expression of group E Sox proteins likely is a persistence in a small number of neuroepithelial cells at the consequence of their evolutionary history. Invertebrates right time and place for oligodendrogenesis. Residual Sox9 possess a single group E Sox protein, whereas vertebrates might in turn be sufficient to drive oligodendrocyte have three group E Sox proteins or – in case of teleosts – specification on the background of endogenous Sox8 in three duplicated sets (Bowles et al., 2000; Koopman et al., these cells. Alternatively, neuroepithelial cells within the 2004). Therefore, group E Sox proteins must have arisen pMN domain may differ in their amount of endogenous early in the vertebrate lineage from a single common Sox8 so that a subpopulation with high amounts of Sox8 ancestor. Their overlapping expression pattern furthermore 316 C.C. Stolt et al. / Developmental Biology 281 (2005) 309–317 indicates that each of the three group E Sox proteins has References inherited most of its expression domains from this common ancestor. Whereas Sox9 and Sox10 have lost Akiyama, H., Chaboissier, M.-C., Martin, J.F., Schedl, A., de some of the original expression domains and retained (or Crombrugghe, B., 2002. The transcription factor Sox9 has essential gained) strong expression levels in the remaining ones, roles in successive steps of the chondrocyte differentiation pathway Sox8 may have best preserved the ancestoral expression and is required for expression of Sox5 and Sox6. Genes Dev. 16, 2813–2828. pattern. The main function of Sox8 may now be in Bi, W., Deng, J.M., Zhang, Z., Behringer, R.R., de Crombrugghe, B., 1999. osteoblasts and other cell types that rely on Sox8 for their Sox9 is required for cartilage formation. Nat. Genet. 22, 85–89. development. Low expression levels in several other Bowles, J., Schepers, G., Koopman, P., 2000. Phylogeny of the SOX family tissues, including oligodendrocytes, may rather be an of developmental transcription factors based on sequence and structural evolutionary relic than a physiologically relevant back-up indicators. Dev. Biol. 227, 239–255. Britsch, S., Goerich, D.E., Riethmacher, D., Peirano, R.I., Rossner, M., mechanism, because Sox8 cannot effectively compensate Nave, K.A., Birchmeier, C., Wegner, M., 2001. The transcription factor loss of Sox9 or Sox10. Sox10 is a key regulator of peripheral glial development. Genes Dev. In such a model, function of all three Sox proteins is 15, 66–78. considered equivalent at first approximation so that the Cai, J., Qi, Y., Hu, X., Tan, M., Liu, Z., Zhang, J., Li, Q., Sander, M., Qiu, differential contribution of each family member to a M., 2005. Generation of oligodendrocyte precursor cells from mouse dorsal spinal cord independent of Nkx6 regulation and Shh signaling. phenotype is primarily a function of its amount in the Neuron 45, 41–53. respective tissue or cell type. In support of such a model, Chaboissier, M.-C., Kobayashi, A., Vidal, V.I.P., Lu¨tzkendorf, S., van de comparative biochemical studies have so far failed to reveal Kant, H.J.G., Wegner, M., de Rooij, D.G., Behringer, R.R., Schedl, A., significant functional differences between different group E 2004. Functional analysis of Sox8 and Sox9 during sex determination in Sox proteins (Schepers et al., 2003; Stolt et al., 2004). the mouse. Development 131, 1891–1901. Cheung, M., Briscoe, J., 2003. Neural crest development is regulated by the Ectopic expression studies in the chick embryo furthermore transcription factor Sox9. Development 130, 5681–5693. indicate that all three group E Sox proteins influence early Foster, J.W., Dominguez-Steglich, M.A., Guioli, S., Kwok, C., Weller, P.A., neural crest development identically (Cheung and Briscoe, Stevanovic, M., Weissenbach, J., Mansour, S., Young, I.D., Goodfellow, 2003). P.N., Brook, J.D., Schafer, A.J., 1994. Campomelic dysplasia and Nevertheless, functional differences between the three autosomal sex reversal caused by mutations in an SRY-related gene. Nature 372, 525–530. group E Sox proteins may so far have gone unnoticed. Jessell, T.M., 2000. Neuronal specification in the spinal cord: inductive The existence of unique functions for Sox8 would signals and transcriptional codes. Nat. Rev., Genet. 1, 20–29. explain why at least low levels of Sox8 are maintained Koopman, P., Schepers, G., Brenner, S., Venkatesh, B., 2004. Origin and in tissues or cell lineages that mostly rely for their diversity of the SOX transcription factor gene family: genome-wide development on the other group E Sox proteins. analysis in Fugu rubripes. Gene 328, 177–186. Maka, M., Stolt, C.C., Wegner, M., 2005. Identification of Sox8 as a Furthermore, these functions would have to be subtle, modifier gene in a mouse model of Hirschsprung disease reveals as their loss is not phenotypically evident under standard underlying molecular defect. Dev. Biol. 277, 155–169. housing conditions in Sox8-deficient mice. Replacing one Mizuguchi, R., Sugimori, M., Takebayashi, H., Kosako, H., Nagao, M., group E Sox gene by another in the mouse will be Yoshida, S., Nabeshima, Y., Shimamura, K., Nakafuku, M., 2001. helpful in solving this issue. Combinatorial roles of and neurogenin2 in the coordinated in- duction of panneuronal and subtype-specific properties of motoneurons. Neuron 31, 757–771. Pfeifer, D., Poulat, F., Holinski-Feder, E., Kooy, F., Scherer, G., 2000. The Note added in revision SOX8 gene is located within 700 kb of the tip of 16p and is deleted in a patient with ATR-16 syndrome. Genomics 63, 108–116. While this manuscript was under revision, oligodendro- Pingault, V., Bondurand, N., Kuhlbrodt, K., Goerich, D.E., Prehu, M.-O., Puliti, A., Herbarth, B., Hermans-Borgmeyer, I., Legius, E., Matthijs, cytes were shown to be generated in the spinal cord not only G., Amiel, J., Lyonnet, S., Ceccherini, I., Romeo, G., Smith, J.C., Read, from the pMN domain, but in addition from a dorsal domain A.P., Wegner, M., Goossens, M., 1998. Sox10 mutations in patients at a later time (Cai et al., 2005; Vallstedt et al., 2005). On the with Waardenburg–Hirschsprung disease. Nat. Genet. 18, 171–173. basis of these reports and our findings, we have to conclude Rowitch, D.H., 2004. Glial specification in the vertebrate neural tube. Nat. that combined loss of Sox8 and Sox9 also interferes with Rev., Neurosci. 5, 409–419. Rowitch, D.H., Lu, Q.R., Kessaris, N., Richardson, W.D., 2002. An oligodendrogenesis from this second domain. Foligarchy_ rules neural development. Trends Neurosci. 25, 417–422. Schepers, G.E., Bullejos, M., Hosking, B.M., Koopman, P., 2000. Cloning and characterisation of the sry-related transcription factor gene sox8. Acknowledgments Nucleic Acids Res. 28, 1473–1480. Schepers, G., Wilson, M., Wilhelm, D., Koopman, P., 2003. Sox8 is We thank H. Kondoh and M. Sander for antibodies, G. expressed during testis differentiation in mice and synergizes with Schu¨tz for the nestin Cre mice, and A. Schedl for Sox9loxP SF1 to activate the AMH promoter in vitro. J. Biol. Chem. 278, 28101–28108. mice. Supported by grants from the Deutsche Forschungs- Schmidt, K., Glaser, G., Wernig, A., Wegner, M., Rosorius, O., 2003. Sox8 gemeinschaft (SFB473) and the Schram-Stiftung (T287/ is a specific marker for muscle satellite cells and inhibits myogenesis. 14172/2004) to M.W. J. Biol. Chem. 278, 29769–29775. C.C. Stolt et al. / Developmental Biology 281 (2005) 309–317 317

Schmidt, K., Schinke, T., Haberland, M., Priemel, M., Schilling, A.F., Tronche, F., Kellendonk, C., Kretz, O., Gass, P., Anlag, K., Orban, P.C., Mueldner, C., Rueger, J.M., Sock, E., Wegner, M., Amling, M., 2005. Bock, R., Klein, R., Schu¨tz, G., 1999. Disruption of the glucocorticoid The high-mobility-group transcription factor Sox8 is a negative receptor gene in the nervous system results in reduced anxiety. Nat. regulator of osteoblast differentiation. J. Cell. Biol. 168, 899–910. Genet. 23, 99–103. Sock, E., Schmidt, K., Hermanns-Borgmeyer, I., Bo¨sl, M.R., Wegner, Vallstedt, A., Klos, J.M., Ericson, J., 2005. Multiple dorsoventral origins of M., 2001. Idiopathic weight reduction in mice deficient in the oligodendrocyte generation in the spinal cord and hindbrain. Neuron 45, high-mobility-group transcription factor Sox8. Mol. Cell. Biol. 21, 55–67. 6951–6959. Wagner, T., Wirth, J., Meyer, J., Zabel, B., Held, M., Zimmer, J., Pasantes, Stolt, C.C., Rehberg, S., Ader, M., Lommes, P., Riethmacher, D., J., Bricarelli, F.D., Keutel, J., Hustert, E., Wolf, U., Tommerup, N., Schachner, M., Bartsch, U., Wegner, M., 2002. Terminal differentiation Schempp, W., Scherer, G., 1994. Autosomal sex reversal and of myelin-forming oligodendrocytes depends on the transcription factor campomelic dysplasia are caused by mutations in and around the Sox10. Genes Dev. 16, 165–170. SRY-related gene Sox9. Cell 79, 1111–1120. Stolt, C.C., Lommes, P., Sock, E., Chaboissier, M.-C., Schedl, A., Wegner, Wegner, M., 1999. From head to toes: the multiple facets of Sox proteins. M., 2003. The Sox9 transcription factor determines glial fate choice in Nucleic Acids Res. 27, 1409–1420. the developing spinal cord. Genes Dev. 17, 1677–1689. Wegner, M., 2001. Expression of transcription factors during oligoden- Stolt, C.C., Lommes, P., Friedrich, R.P., Wegner, M., 2004. Transcription droglial development. Microsc. Res. Tech. 52, 746–752. factors Sox8 and Sox10 perform non-equivalent roles during oligoden- Yamamoto, Y., Henderson, C.E., 1999. Patterns of programmed cell death drocyte development despite functional redundancy. Development 131, in populations of developing spinal motoneurons in chicken, mouse, 2349–2358. and rat. Dev. Biol. 214, 60–71. Tanaka, S., Kamachi, Y., Tanouchi, A., Hamada, H., Jing, N., Kondoh, H., Zhou, Q., Choi, G., Anderson, D.J., 2001. The bHLH transcription factor 2004. Interplay of SOX and POU factors in regulation of the nestin gene Olig2 promotes oligodendrocyte differentiation in collaboration with in neural primordial cells. Mol. Cell. Biol. 24, 8834–8846. Nkx2.2. Neuron 31, 791–807.