Differential activity of transcription factor Sox9 in early and adult oligodendroglial progenitor cells

The high‐mobility‐group domain‐containing transcription factor Sox9 confers glial competence to neuroepithelial precursors in the developing central nervous system and is an important determinant of astroglial and oligodendroglial specification. In oligodendroglial cells, it remains expressed in oligodendrocyte progenitor cells (OPCs) of the developing nervous system, but is shut off in differentiating oligodendrocytes as well as in OPCs that persist in the adult nervous system. To better understand the role of Sox9 in OPCs, we generated mouse models that allowed oligodendroglial expression of a Sox9 transgene during development or in the adult. With transgene expression beginning in the last trimester of pregnancy, the number of OPCs increased dramatically, followed by comparable gains in the number of pre‐myelinating and myelinating oligodendrocytes as assessed by marker gene expression. This argues that Sox9 boosts oligodendrogenesis during ontogenetic development at all stages, including terminal oligodendrocyte differentiation. When Sox9 transgene expression started in the adult, many transgene‐expressing OPCs failed to maintain their progenitor cell identity and instead converted into myelinating oligodendrocytes. As infrequent and inefficient differentiation of adult OPCs is one of the main obstacles to effective remyelination in demyelinating diseases such as Multiple Sclerosis, increased Sox9 levels in adult OPCs may substantially increase their remyelination capacity.


| INTRODUCTION
Oligodendrocytes in the vertebrate central nervous system (CNS) arise from neuroepithelial precursor cells of the ventricular zone and progress through oligodendrocyte progenitor cell (OPC) and pre-myelinating oligodendrocyte stages before reaching their functional state as myelinating oligodendrocytes. The Olig2 and Sox10 transcription factors determine the identity of the oligodendroglial lineage and drive development in cooperation with a set of temporally restricted stage-specific factors such as the differentiation promoting Myrf (Sock & Wegner, 2021).
Another stage-specific factor is Sox9. In contrast to Sox10, it is already expressed in neuroepithelial precursor cells where it conveys glial competence and is an essential determinant of the glial specification process (Scott et al., 2010;Stolt et al., 2003). Once OPCs are specified, they turn on Sox10 and jointly express Sox9 and Sox10.
Considering the close relation and structural similarity of both Sox proteins, Sox9 and Sox10 are believed to have very similar functions in OPCs during development. Indeed there is in vivo evidence for such functional redundancy. Whereas Sox9 deletion in neuroepithelial cells severely disturbs OPC generation because of its essential role in glial specification (Stolt et al., 2003), its deletion in already specified OPCs during ontogenetic development remains without consequence (Finzsch et al., 2008). Similarly, there are only mild OPC defects after loss of Sox10 (Stolt et al., 2002). However, following deletion of both Sox9 and Sox10, Pdgfra expression is dramatically reduced, resulting in severely compromised OPC survival and migration (Finzsch et al., 2008). In line with the concept of functional redundancy, Pdgfra appears to be a direct target of both Sox9 and Sox10 in OPCs during ontogenetic development (Baroti et al., 2016;Finzsch et al., 2008).
More detailed studies have shown that Sox9 and Sox10 expression levels are inversely correlated both in vivo and in vitro during specific stages of ontogenetic oligodendroglial development (Reiprich et al., 2017). In vivo, a good example is the early phase of terminal differentiation where a surge of Sox10 coincides with the disappearance of Sox9. There is also evidence that oligodendroglial Sox10 protein reduces Sox9 protein levels by inducing microRNAs that lower Sox9 translation (Reiprich et al., 2017). Together with the differential interaction of both Sox factors with Nfia and reports of antagonistic functions in developing astrocytes, cranial neural crest cells, and melanoma (Glasgow et al., 2014;John et al., 2011;Shakhova et al., 2015), functional redundancy between Sox9 and Sox10 may not fully explain their relationship in oligodendroglial cells. Unique and separable functions may exist as well. Additionally, Sox9 has not yet been studied in OPCs of the adult CNS.
Here we use mouse models that allow temporally controlled Sox9 transgene expression to determine its functions in OPCs both during development and in the adult. Our findings argue that Sox9 has a strong positive effect on multiple stages of oligodendrogenesis during development and that its absence in adult OPCs may be one major cause for their inability to differentiate efficiently into myelinating oligodendrocytes.

| Mice
Mice with a controllable oligodendroglial expression of a Sox9 transgene were generated by combining a bidirectional tdTomato-TRE-Sox9 transgene (TRE::Sox9) and a Rosa26 stopflox-tTA allele (Vogel et al., 2020) with a Cre recombinase under the control of the Olig2 promoter (Olig2-Cre) (Dessaud et al., 2007) or a tamoxifen inducible CreERT2 recombinase under the control of Cspg4 regulatory sequences (Cspg4-CreERT2, also known as NG2-CreERT2) . For experiments with the Olig2-Cre, littermates without the Cre recombinase were routinely used as controls. They yielded similar results as age-matched mice carrying the Olig2-Cre but lacking the Sox9 transgene. For experiments with the Cspg4-CreERT2, control mice were generated by combining a Rosa26 stopflox-tdTomato allele (Madisen et al., 2010) with the Cspg4-CreERT2 transgene. Genotyping was performed as described (Dessaud et al., 2007;Huang et al., 2014;Madisen et al., 2010;Vogel et al., 2020). Mice were kept under standard housing conditions with 12:12 h light-dark cycles with continuous access to food and water in accordance with animal welfare laws and ethical regulations. For a defined onset of Sox9 transgene expression in mice carrying the Olig2-Cre, 0.2 mg/mL tetracycline was added to sucrose-containing (0.05 mg/mL) drinking water and given to pregnant mice from the day of conception until embryonic day 13.5 (E13.5). For inducing Sox9 transgene expression in adult mice carrying the Cspg4-CreERT2, 100 mg tamoxifen per kg body weight was administered to mice at P60 for five consecutive days. The cuprizone intoxication model was adapted from previous studies (Kaddatz et al., 2021;Kang et al., 2012). Mice were fed with daily refreshed powder chow (Cat. V1120, SSNIFF, Germany) supplemented with 0.3% cuprizone (Cat. 14690, Sigma-Aldrich) for the first week and 0.2% cuprizone for the subsequent 2 weeks. Loss of oligodendrocytes in the middle part of the corpus callosum was $90%.
Tissues were dissected from embryos (E15.5), early postnatal mice (P0, P7) and 4% paraformaldehyde-perfused adult mice. For immunohistochemistry and in situ hybridization, tissues underwent overnight fixation in 4% paraformaldehyde and incubation in 30% sucrose. For embryonic and early postnatal tissues, embedding and freezing at À80 C was performed before cryotome sectioning at 10 μm thickness. Tissues from adult mice did not undergo embedding.
For in situ hybridization, 10 μm cryotome sections were incubated with DIG-labeled antisense riboprobes specific for Plp1 as described (Stolt et al., 2002). Samples were analyzed and documented with a Leica MZFLIII stereomicroscope equipped with an Axiocam (Zeiss).

| Electron microscopy
After perfusion of mice at P7 with a 4% paraformaldehyde, 2.5% glutaraldehyde solution, brains were dissected, sagittally cut and underwent fixation in Ito-buffer. Sample processing followed established protocols and included post-fixation in 2% osmium tetroxide, dehydration in ascending steps of ethanol and acetone and embedding in epon (Foggetti et al., 2019). After trimming, ultrathin sectioning and staining with uranyl-acetate, images were acquired at a nominal magnification as indicated in the figure legend with a JEOL 1400 Plus transmission electron microscope at 100 kV acceleration voltage.

| RNA sequencing
For RNA sequencing, brains were obtained from newborn mice expressing either Sox9 and tdTomato or tdTomato alone. Meninges were removed, and brain tissue was homogenized in HBSS using a gentleMACS Dissociator. The material was passed through a 70 μm Gene signatures consequent to Sox9 overexpression were identified using the Gene Set Enrichment Analysis (GSEA) tool from the Broad Institute (http://www.gsea-msigdb.org/gsea/index.jsp).

| Quantifications and statistical analysis
Results from independent animals, experiments, or separately generated samples were treated as biological replicates (n ≥ 3). Quantifications were performed using Fiji ImageJ software. Three standard size fields in the corpus callosum (next to the midline, above the ventricle) and cortex (directly above chosen corpus callosum region) were defined using Fiji ImageJ and analyzed for cell numbers for each animal. To determine whether differences in cell numbers were statistically significant, a two-tailed Student's t test was performed (*p ≤ .05; **p ≤ .01, ***p ≤ .001) using GraphPad Prism6 software.

| RESULTS
3.1 | Characterization of the mouse model for early onset oligodendroglial expression of the Sox9 transgene We have previously generated a transgene that carries Sox9 coding sequences under control of a bidirectional tetracycline responsive element (TRE) (Vogel et al., 2020). An aminoterminal Flag tag has been added at the beginning of the coding sequences so that Sox9 transgene expression can be directly detected via the Flag tag as well as by the tdTomato reporter that is transcribed from the TRE in the opposite direction (Figure 1a). To induce transgene expression, we combined it with a Rosa26 stopflox-tTA allele and a Cre transgene with spatiotemporally controlled expression. Cre expression then leads to tTA induction and subsequently, via tTA-dependent activation of the TRE, to Sox9 expression. Tetracycline-dependent tTA inactivation allows Sox9 transgene expression furthermore to be switched on and off in a reversible manner. Unfortunately, the system does not allow a reproducible fine-tuning of expression levels, which would be desirable as the exact Sox9 levels present in a cell may have an impact on its function within the gene regulatory network.
During the first postnatal week, the Flag tag and tdTomato reporter were detected in spinal cord and forebrain by immunohistochemistry in Olig2-and Sox10-positive oligodendroglial cells (Figure 1e,f). In fact, the majority of transgene expressing cells in these regions were positive F I G U R E 1 Mouse model for early onset expression of transgenic Sox9 in developing oligodendroglial cells. (a) Schematic representation of alleles and transgenes used to achieve early oligodendroglial expression of Sox9 (tgSox9). Cre, Cre recombinase inserted into the Olig2 gene locus; stop, stopflox cassette flanked by loxP sites (arrowheads); tTA, tetracycline-dependent transactivator inserted into the Rosa26 locus; TRE, bidirectional tetracycline responsive promoter; tdTomato, tdTomato reporter; pA, polyadenylation signal; SA, splice acceptor. (b, c) Immunohistochemically stained transverse spinal cord section (thoracic level, b) and coronal forebrain section (c) at E15.5 from transgenic mice treated with tetracycline until E13.5. Scale bar: 50 μm (b), 1 mm (c). (d) Cells expressing the Sox9 transgene (tgSox9) quantified by Flag-tag (white bars) and tdTomato (gray bars) at P0 and P7 in white (WM) and gray (GM) matter of spinal cord, or per mm 2 in CC and CTX of transgenic mice. (e, f) Co-immunohistochemical staining for tdTomato (red, e) and Olig2 (green, e) or Flag epitope (red, f) and Sox10 (green, f) in CC and CTX of transgenic mice at P0 (scale bar: 50 μm). (g, h) Quantification showing the percentage of transgene-positive cells co-expressing Olig2 (g) or Sox10 (h) at P0 and P7 in the CC and CTX. Data are presented as mean ± SEM for n = 3.
for Olig2 as well as Sox10 both at P0 (95 ± 1% in the CC and 91 ± 1% in the CTX for Olig2; 94 ± 1% in the CC and 94 ± 1% in the CTX for Sox10) and at P7 (92 ± 1% in the CC and 90 ± 1% in the CTX for Olig2; 90 ± 4% in the CC and 85 ± 1% in the CTX for Sox10) (Figure 1g,h). In contrast, there was virtually no overlap between the Flag or tdTomato signal and NeuN or Tbr1 as neuronal, Gfap or Aldh1l1 as astrocytic and Iba1 as microglial markers in co-immunohistochemical stainings ( Figure   S1b-f). We therefore conclude that expression of the Sox9 transgene in spinal cord as well as CC and CTX of the dorsal forebrain is restricted to and widespread throughout the oligodendroglial population during the first postnatal week.

| Consequences of early onset Sox9 transgene expression on oligodendroglial development
Mice expressing the Sox9 transgene in oligodendroglial cells from E15.5 onwards were born in expected Mendelian ratios and analyzed at P0 and P7 to study oligodendrocyte development by looking at the CC and adjacent CTX. Intriguingly, we detected a higher density of Dapi-stained nuclei and thus more cells in both CC and CTX at P0 and P7 in Sox9 transgenic mice as compared to controls (Figure 2a,b). The observed increases corresponded well to higher numbers of oligodendroglial cells, independent of whether we used Sox10 or Olig2 as a pan-lineage marker for these cells (Figure 2c-f). The rise in oligodendroglial cell numbers was particularly impressive in the CC at P0 where numbers were two-to threefold higher than in controls depending on the use of Sox10 or Olig2 as oligodendroglial lineage markers. In the CTX of Sox9 transgene expressing mice, oligodendroglial cell numbers were 1.5-fold increased over controls at P0 and P7.
At P0, many oligodendroglial cells are still in the OPC stage and express Pdgfra in control brains. Accordingly, OPC numbers were only slightly lower than the total number of oligodendroglial cells and exhibited a comparable increase in the CC and CTX of transgenic mice as observed for all oligodendroglial cells (Figure 2g,h). This argues that Sox9 transgene expression increased OPC numbers substantially.
By P7, the numbers of Pdgfra-positive OPCs had decreased especially in the CC in line with the beginning oligodendroglial differentiation. Differences between the two genotypes had furthermore disappeared indicating that the surplus oligodendroglial cells do not persist in the OPC stage and that Sox9 transgene expression does not prevent OPCs from progressing into a more differentiated state.
The beginning differentiation was also evident from Bcas1 expression as a marker that is turned on early during the differentiation process in pre-myelinating oligodendrocytes (Figure 3a,b) (Fard et al., 2017). Especially in the CC, Bcas1 numbers increased substantially from P0 to P7 in both genotypes. Relative to controls, the number of Bcas1-positive cells was four to fivefold higher in the CC and CTX of Sox9 transgenic mice at P0 and three to fourfold higher in the CC and CTX at P7. The dramatic increase in Bcas1-positive cells at P7 furthermore indicated that a substantial fraction of the surplus OPCs in the Sox9 transgenic mice had progressed into the pre-myelinating stage. Similar results were also obtained for Myrf. As Myrf induction lags behind Bcas1 appearance, we could not detect Myrf expression in either genotype at P0, but reproduced the dramatic increase seen for Bcas1 in Sox9 transgenic mice with Myrf at P7 (Figure 3c,d). By P7, the first myelinating oligodendrocytes were also visible by Plp1 in situ hybridization (Figure 3e,f). Again, significantly more myelinating oligodendrocytes were observed in the Sox9 transgenic mice than in controls. We conclude from these data that surplus OPCs in Sox9 transgenic mice progress without hindrance into the pre-myelinating and further into the myelinating stage. Considering the fact that the relative contribution of Pdgfra-positive OPCs to the total oligodendroglial population declines faster in Sox9 transgenic mice than in controls and that the fraction of differentiating oligodendrocytes increases more rapidly (Figure 3g), we even have to assume that the Sox9 transgene promotes differentiation.
In line with such an assumption, we reproducibly detected by electron microscopy low numbers of axonal segments that were already surrounded by several layers of myelin in the CC of Sox9 Additionally, we performed RNA sequencing on oligodendroglial cells that expressed the Sox9 transgene and were isolated from brains of newborn pups following dissociation and cell sorting for tdTomatoevoked fluorescence. Oligodendroglial cells from mice that expressed a tdTomato reporter from the Rosa26 locus after Olig2-Cre mediated removal of a stop-flox cassette served as controls. As evident from PCA plots and clustering by Euclidian distance, samples from mice expressing the Sox9 transgene were much more similar to each other than to control samples and clustered separately (Figure 6a,b). Using a log2-fold change in expression levels ≥ ±2 and adjusted p-value ≤.05 as filters, we found 426 genes altered in their expression upon Sox9  (h) Co-immunohistochemical staining for cleaved caspase 3 (green) and Pdgfra as an OPC marker (magenta, left panels) or Otud7b as a marker for differentiating oligodendrocytes (magenta, right panels) at P7 in the CC and CTX of Sox9 transgenic mice. Scale bar: 1 μm (a), 50 μm (h). Data are presented as mean ± SEM for three biological replicates. Differences to controls were statistically significant for Sox9 transgenic mice as indicated (Student's t test; *p ≤ .05; **p ≤ .01; ***p ≤ .001).
activator (Figure 6d) (Liu et al., 2017;Ming et al., 2022). GO analysis of the upregulated genes furthermore revealed that ensheathment of neurons, regulation of myelination and regulation of gliogenesis were among the top-ranked terms (Figure 6e). A change in these processes is also linked to alterations in the adhesive properties of cells, altered generation of extracellular matrix components and increased production of membrane lipids, which may explain the additional appearance of related terms among the top ranked GO terms. Intriguingly, we also see an enrichment of terms associated with increased cell proliferation. The list of GO terms for the downregulated genes points to changes in neuropeptide signaling and cell adhesion and to a negative regulation of cell proliferation (Figure 6f). Some of these terms mirror GO terms for the upregulated genes. They refer to processes that are likely repressed by high Sox9 levels in OPCs, either directly or indirectly.
Closer inspection of single gene sets validates the stimulatory role of Sox9 transgene expression on cell proliferation, but also on oligodendrocyte differentiation and sphingolipid metabolism (Figure 7a-c).
RNA sequencing data thus confirm our immunohistochemical findings of positive effects on OPCs, pre-myelinating and myelinating oligodendrocytes.
When looking at the expression of single genes, we saw a strong upregulation of Sox9 in the oligodendroglial cells of transgenic mice ( Figure 7d). This was accompanied by increased Sox10 expression, confirming the previously described role of Sox9 as an inducer of Sox10 (Stolt et al., 2003;Wahlbuhl et al., 2011). Cellular Olig2 levels, on the other hand, were not significantly altered. In line with immunohistochemical data and bioinformatics, we also detected an enhanced expression of OPC markers (Cspg4, Pdgfra and Sox6) and markers of pre-myelinating oligodendrocytes (Bcas1, Nkx2.2 and Myrf ) but also of

| Consequences of Sox9 transgene expression in adult OPCs
To extend our analysis on the role of Sox9 to adult OPCs, we first checked for its occurrence in these cells. In contrast to pre-and early postnatal periods, we were unable to detect Sox9 in OPCs of the adult CNS ( Figure S2a To study Sox9 in adult OPCs, we exchanged the Olig2-Cre against a Cspg4-CreERT2 (NG2-CreERT2) transgene and induced Cre activity in 2 month old mice by tamoxifen treatment that then led to tTA induction and tTA-induced expression of the Sox9 transgene ( Figure S3a,b).
In this case, mice served as controls, in which a tdTomato reporter was induced by Cspg4-CreERT2-dependent removal of a stopflox cassette from the Rosa26 locus ( Figure S3a).  (Figures 8a,b and S4c-f). Instead, almost all the tdTomatoexpressing cells co-labeled with Sox10 in control (96 ± 1% in CC, 98 ± 1% in CTX, between 98% and 99% in both white and gray matter of the spinal cord) and Sox9 transgenic (92 ± 2% in CC and 96 ± 1% in CTX, 97 ± 1% in both white and gray matter of the spinal cord) mice at 30 days post tamoxifen induction (dpt) (Figures 8c,d, S6g). This confirmed that expression of the Sox9 transgene was largely restricted to cells of the oligodendroglial lineage. Furthermore, overall numbers of neurons, astrocytes, adult OPCs and oligodendrocytes were comparable in the CC and CTX of both genotypes and did not change during the time of analysis from 15 to 30 dpt .

Both in controls and
Already at the earliest time point when tdTomato expression became visible, we noted a substantial difference in the amount of reporter-expressing cells between both genotypes (Figure 8e,f). Only half as many cells were tdTomato-positive in Sox9 transgenic mice as compared to controls (361 ± 39 cells per mm 2 in CC at 30 dpt for controls vs. 195 ± 29 for Sox9 transgenic mice; 224 ± 14 per mm 2 in CTX for controls vs. 138 ± 24 for Sox9 transgenic mice). Considering the fact that tdTomato is directly activated via Cre in controls and indirectly via Cre and tTA in Sox9 transgenic mice, that transgene expressing cells were already fewer very early after Cre induction and that F I G U R E 7 Altered gene expression in early postnatal oligodendroglial cells expressing the Sox9 transgene. (a-c) Schematic representation of the alteration of select gene sets in oligodendroglial cells expressing the Sox9 transgene according to RNA-seq studies, leading to the determination of the respective normalized enrichment score (NES). (d) Expression levels for select genes associated with the OPC, premyelinating (Pre-OLs) and myelinating stages and with lipid metabolism in oligodendroglial cells expressing the Sox9 transgene relative to controls.
we failed to detect substantial cell death in either genotype (see below), we assume that the difference in labeled oligodendroglial cells is a consequence of the divergent mode of tdTomato induction.
To further specify the identity of the tdTomato-expressing cells in controls and Sox9 transgenic mice, we performed immunohistochemical stainings at 15, 20 and 30 dpt. Most tdTomato-positive cells F I G U R E 8 Characterization of a model of adult onset Sox9 transgene expression in the dorsal forebrain. (a-c) Co-immunohistochemical staining of tdTomato (red) with neuronal marker NeuN (green, a), astroglial marker Gfap (green, b), and oligodendroglial lineage marker Sox10 (green, c) in CC and CTX of control (ctrl) and Sox9 transgenic (tgSox9) mice at 30 dpt. Sox9 transgenic mice combined the Cspg4-CreERT2, Rosa26 stopflox-tTA and TRE::Sox9 alleles, controls the Cspg4-CreERT2 and Rosa26 stopflox-tdTomato alleles (see Figure S3a). Both were tamoxifen-treated at 2 months ( Figure S3b) (d) Quantification of the percentage of transgene-positive cells (tg + ) co-expressing Sox10 in the CC of control (white bars) and Sox9 transgenic (white hatched bars) mice as well as the CTX of control (gray bars) and Sox9 transgenic (gray hatched bars) mice at 15, 20 and 30 dpt. (e, f) Quantification of tdTomato-positive cells per mm 2 in CC and CTX of control (ctrl) and Sox9 transgenic (tgSox9) mice at 15, 20 and 30 dpt (f) based on immunohistochemical staining with tdTomato-specific antibodies (representative pictures in e; scale bar: 50 μm. Data are presented as mean ± SEM for three biological replicates. Differences to control were statistically significant for Sox9 transgenic animals as indicated (Student's t test; *p ≤ .05; **p ≤ .01; ***p ≤ .001).
in control mice were adult OPCs at all times analyzed. At 30 dpt, for instance, 73 ± 3% of the tdTomato-expressing cells in the CC and 84 ± 1% in the CTX of control mice were positive for the OPC marker and proteoglycan NG2 (encoded by the Cspg4 gene) as well as for Pdgfra (Figures 9a,b and S4a,b). In contrast, only 12 ± 1% of the tdTomato-expressing cells in the CC and 12 ± 2% in the CTX of Sox9 F I G U R E 9 Consequences of adult onset Sox9 transgene expression on oligodendroglial cells in the dorsal forebrain. (a-f) Determination of the percentage of transgene-positive cells (tg + ) co-expressing NG2 (a), Myrf (c) and Gstpi (e) in the CC of control (ctrl, white bars) and Sox9 transgenic (tgSox9, white hatched bars) mice as well as the CTX of control (gray bars) and Sox9 transgenic (gray hatched bars) mice at 15, 20 and 30 dpt based on immunohistochemical stainings with marker-specific antibodies (representative pictures in b for NG2, d for Myrf and f for Gstpi). Sox9 transgenic mice combined the Cspg4-CreERT2, Rosa26 stopflox-tTA and TRE::Sox9 alleles, controls the Cspg4-CreERT2 and Rosa26 stopflox-tdTomato alleles. Both underwent the same Cre induction scheme. (g) Staining of Mbp (green) and tdTomato (red) in the CTX of ctrl and Sox9 transgenic mice at 30 dpt recorded on a Zeiss apotome as 12 μm stack, shown for each marker and as merge. Magnifications (right) correspond to boxed areas and are single planes. (h, i) Co-immunohistochemical stainings of tdTomato (red) and proliferation marker Ki67 (green, h) or apoptosis marker cleaved-caspase 3 (cCasp3) (green, i) at 30 dpt in the CC and CTX of ctrl and Sox9 transgenic mice. Scale bar: 50 μm (i) and 10 μm (magnification in g). Data are presented as mean ± SEM for three biological replicates. Differences to control were statistically significant for Sox9-overexpressing animals only as indicated (Student's t test; *p ≤ .05; **p ≤ .01; ***p ≤ .001).
transgenic mice stained positive for NG2 as well as Pdgfra at 30 dpt.
Instead, we observed a reciprocally higher number of Myrf-positive cells among the tdTomato expressing cells in Sox9 transgenic mice as compared to controls (79 ± 1% in CC and 76 ± 1% in CTX of Sox9 transgenic mice compared to 23 ± 1% in CC and 18 ± 2% in CTX of controls at 30 dpt, Figure 9c,d). The same effect was observed for Gstpi as a marker of myelinating oligodendrocytes (Figure 9e for Gstpi in white matter, 73 ± 1% for Myrf and 78 ± 1% for Gstpi in gray matter of Sox9 transgenic mice compared with 27 ± 1% for Myrf and 15 ± 1% for Gstpi in white matter, 19 ± 1% for Myrf and 19 ± 1 for Gstpi in gray matter of controls, Figure S6i,j). Proliferation was correspondingly decreased in tdTomato-expressing cells of Sox9 transgenic mice (1.0 ± 0.2% in white and 0.8 ± 0.1% in gray matter of Sox9 transgenic mice compared with 5 ± 2% in white and 4 ± 1% in gray matter of controls, Figure S6k), whereas cell death did not differ in extent between the two genotypes ( Figure S6l). We conclude from our analysis of both dorsal forebrain and spinal cord that ectopic expression of Sox9 in adult OPCs enhances their capacity to convert into myelinating oligodendrocytes.

| DISCUSSION
In this work, we studied the impact of Sox9 overexpression on proliferation and differentiation of oligodendroglial cells during development and in the adult. Compared to its central role in neural stem cells and for glial specification (Scott et al., 2010;Stolt et al., 2003;Vogel et al., 2020), Sox9 has a lesser influence in cells of the oligodendroglial lineage. This is in part due to the continuous presence of the closely related Sox10 in oligodendroglial cells and its much higher expression relative to Sox9. According to RNA expression data (www.brainrnaseq. org), oligodendroglial Sox10 expression at early postnatal times reaches 103-147 fragments per kilobase of transcript per million mapped fragments (FPKM) depending on the respective stage, whereas Sox9 expression remains low at 5 FPKM or even less. Additionally, Sox9 is not expressed in myelinating oligodendrocytes and furthermore missing from adult OPCs so that any function remains restricted to OPCs during ontogenetic development (Stolt et al., 2003).
Despite its much lower expression relative to Sox10, a clear function has been previously assigned to Sox9 in OPCs of the embryonic spinal cord, where defects in OPC survival and migration become visible upon joint deletion of Sox9 and Sox10, but not upon deletion of either factor alone (Finzsch et al., 2008). This has led to the assumption that Sox9 can compensate for the loss of Sox10 in OPCs during development, further implying at least partial functional redundancy between both Sox proteins. In OPCs, part of the ability to compensate appears to result from the fact that both Sox9 and Sox10 effectively activate Pdgfra expression as the central receptor for survival signals and migratory cues (Baroti et al., 2016;Finzsch et al., 2008). However, there have also been reports on cross-regulation and functional antagonism between both proteins in oligodendroglial cells and other cells outside the CNS, arguing that their relationship is more complicated (John et al., 2011;Reiprich et al., 2017;Shakhova et al., 2015;Wahlbuhl et al., 2011).
To uncover the complexity of Sox9 functions, we used mouse models that allow ectopic expression of a Sox9 transgene in OPCs.
While our mouse models allow a good temporal control of transgene expression, we have little influence on expression levels. Considering that transcription factor functions may be sensitive to amounts, our mouse models may not be able to reveal the full spectrum of Sox9 activities possible over a broad range of concentrations.
We started by determining the consequences of Sox9 transgenic overexpression in OPCs during pre-and early postnatal development.
In these cells, low levels of endogenous Sox9 are expressed. Our RNA-seq data indicate that the transgene increases Sox9 levels several fold. One effect that we see is a substantial increase in OPC numbers. Such an increase has similarly been observed after Sox9 overexpression in neuroepithelial precursor cells and their derivatives (Vogel et al., 2020). However, in the previous study it was not clear whether increased OPC numbers were due to increased OPC specification, OPC proliferation or both. At least in the spinal cord, our current approach restricts Sox9 expression to already specified OPCs. Therefore, we can conclude that higher OPC numbers in the spinal cord are due to increased proliferation. In the forebrain, we likely observe a mix of increased OPC specification and proliferation.
Intriguingly, similar overexpression studies with Sox10 primarily led to premature differentiation of OPCs and lower increases in OPC numbers (Weider et al., 2018) supporting the notion that both proteins are not fully equivalent in their function.
Despite its normally restricted expression to OPCs and its proliferation promoting activity, continued presence of Sox9 did not prevent OPCs from converting in a timely manner into pre-myelinating oligodendrocytes and even into myelinating oligodendrocytes. However, the myelinating oligodendrocytes exhibit increased apoptosis in the presence of the Sox9 transgene. Increased cell death in myelinating oligodendrocytes may be due to Sox9-induced changes in the expression profile. It could for instance be envisaged that the continued presence of Sox9 and the evoked changes in gene expression interferes with transition from an actively myelinating into the maintenance state. Alternatively, the increased apoptosis may simply result from pruning of surplus oligodendrocytes by more general homeostatic mechanisms.
The newly identified role of Sox9 in boosting oligodendrogenesis through all stages may also explain our results from transgene expression in adult OPCs that lack endogenous Sox9 expression and exhibit a limited proliferation capacity. In these cells and under normal conditions of a healthy brain, Sox9 transgene expression does not trigger increased proliferation. Instead, the presence of Sox9 helps OPCs to overcome their intrinsic inefficiency of converting into myelinating oligodendrocytes. Currently, we do not know the mechanism of Sox9 action. Considering the stimulatory effect of Sox9 on Sox10 expression, it may involve Sox10 as a downstream effector. Experiments in which we attempted to delete Sox10 simultaneously with Sox9 induction in adult OPCs using the Cspg4-CreERT2 line turned out to be difficult to interpret as Sox10 deletion remained incomplete in Sox9 overexpressing adult OPCs.
In contrast to oligodendrocytes generated during developmental myelination, the oligodendrocytes generated in the context of the adult brain do not exhibit a greater sensitivity toward apoptosis in the presence of Sox9. They form new myelin sheaths and appear to integrate into existing circuits. Consequences on locomotor functions and behavior may be expected, but have not been analyzed in the current study. Importantly, Sox9 may also be an interesting therapeutical target in demyelinating diseases of the CNS such as Multiple Sclerosis where inefficient differentiation of adult OPCs in the lesion sites is one of the most important obstacles to effective remyelination. Induction of endogenous Sox9 expression in these cells or delivery of Sox9 to these cells may substantially increase the ability of resident OPCs to differentiate and regenerate myelin.