Sox9 overexpression exerts multiple stage‐dependent effects on mouse spinal cord development

The high‐mobility‐group (HMG)‐domain protein Sox9 is one of few transcription factors implicated in gliogenesis in the vertebrate central nervous system. To further study the role of Sox9 in early spinal cord development, we generated a mouse that allows expression of Sox9 in a temporally and spatially controlled manner. Using this mouse, we show that premature Sox9 expression in neural precursor cells disrupted the neuroepithelium of the ventricular zone. Sox9 also compromised development and survival of neuronal precursors and neurons. Additionally, we observed in these mice substantial increases in oligodendroglial and astroglial cells. Reversing the normal order of appearance of essential transcriptional regulators during oligodendrogenesis, Sox10 preceded Olig2. Our study reinforces the notion that Sox9 has a strong gliogenic activity. It also argues that Sox9 expression has to be tightly controlled to prevent negative effects on early spinal cord structure and neuronal development.


| INTRODUCTION
In the central nervous system (CNS) gliogenesis follows neurogenesis.
Whereas the mechanisms that are responsible for commitment to the neuronal lineage are fairly well studied, much less is known about the corresponding mechanisms for glial commitment. One of the few factors that has been implicated in the process is the high-mobility-group (HMG) domain-containing transcription factor Sox9. An initial study had shown that in the absence of Sox9, glial cells were generated in much fewer numbers in the mouse spinal cord (Stolt et al., 2003). This concerned the generation of oligodendrocyte precursor cells (OPCs) from the ventral pMN-domain of the ventricular zone (VZ) as well as the generation of astroglial cells from the adjacent p2 domain. A simultaneous prolonged production and increase in the number of those neurons that are generated at earlier times from these domains (i.e., motoneurons and V2 interneurons) was taken as evidence that Sox9 is part of a switch mechanism from neuro-to gliogenesis. In its gliogenic function, Sox9 appears to interact with the Notch signaling pathway (Taylor, Yeager, & Morrison, 2007). It induces Nfia as a second gliogenic factor whose temporal expression pattern and interaction with other transcription factors decides over the exact glial fate (Glasgow et al., 2014;Kang et al., 2012). A transient expression of Nfia in the presence of Sox10 causes commitment to the oligodendroglial lineage. In contrast, prolonged expression in the absence of Sox10 allows Nfia to interact with its inducer Sox9, synergistically activate an astrocyte-specific regulatory cascade and thereby induce astrogliogenesis (Glasgow et al., 2014).
In addition to a role in glial commitment, Sox9 has also been reported to be already required in neuroepithelial precursor cells (NEPs) to induce glial competence and thereby transform NEPs into multipotent neural stem cells (Scott et al., 2010). For the cerebellum it has furthermore been postulated that Sox9 is more important for the suppression of neurogenesis than for the initiation of gliogenesis (Vong, Leung, Behringer, & Kwan, 2015). Thus, multiple functions have been proposed for Sox9, including functions as a glial competence and multipotentiality factor, a neuron-glia switch molecule and an inhibitor of neurogenesis.
All of these studies primarily relied on loss-of-function approaches. If at all performed, gain-of-function experiments were largely carried out in vitro. Therefore we decided to assess Sox9 function in the early developing spinal cord by overexpressing the protein from a tetracycline-controlled transgene. These studies provide evidence for multiple stage-dependent and cell type-specific functions and show that they are at least in part separable from each other.

| Generation, husbandry, and analysis of transgenic animals
For the generation of Sox9 overexpressing mice, a transgenic construct was generated based on pBI-EGFP, a plasmid with bidirectional tetracycline-controlled promoter (TRE) that drives enhanced green fluorescent protein (EGFP) expression in one direction and contains a multiple cloning site on the other side of the TRE for insertion of an expression cassette. In pBI-EGFP, EGFP sequences were replaced by tdTomato sequences and an aminoterminal FLAG-and HA-tagged Sox9 coding sequence was inserted in the multiple cloning site on the other side of the TRE (Figure 1a). This TRE::Sox9 construct was microinjected after linearization into male pronuclei of fertilized oocytes derived from C57Bl6/J × CBA matings according to standard techniques, and a stable line with transgene insertion on chromosome 12qD1 was established from the obtained founders. Sox9 expression was achieved by combining TRE::Sox9 in the heterozygous state with a Rosa26 stopflox-tTA allele (Wang et al., 2008) and Brn4::Cre (Ahn, Mishina, Hanks, Behringer, & Crenshaw, 2001) or Nestin::Cre (Tronche et al., 1999) transgenes. Genotyping for the Rosa26 stopflox-tTA allele (Wang et al., 2008) and the Cre transgenes was as described in the original reports. The TRE::Sox9 transgene was detected by genomic PCR as a 465 bp fragment using 5 0 -ACCACAACTAGAA -TGCAGTGAAAA-3 0 and 5 0 -GATCCCAGAATGACCACCAC-3 0 as primers. For experiments, a mixed C57Bl6/J × C3HeB/FeJ background was used. Mice were kept under standard housing conditions with 12:12 hr light-dark cycles and continuous access to food and F I G U R E 1 Characterization of TetSox9 Brn4 mice. (a) Schematic representation of the TRE::Sox9 transgene. The transgene carries tdTomato sequences and a Sox9 open reading frame tagged at its aminoterminus with three FLAG and three HA epitopes on either side of the bidirectional TRE promoter. bGI, β-globin intron; pA, polyA signal from the rabbit β-globin gene. (b) Transient transfection of Neuro2a cells with luciferase reporters containing the Col2a1 promoter or the Sox10 U3 enhancer and expression plasmids for Sox9 in wild type and tagged version. Reporter gene expression was determined in extracts 48 hr after transfection and effector-dependent activation rates are presented as fold inductions ± SEM with transfections in the absence of Sox9 arbitrarily set to 1 (n = 3). (c) Schematic representation of the Brn4::Cre transgene and the Rosa26 stopflox-tTA allele that are combined with the TRE::Sox9 transgene in TetSox9 Brn4 mice. (d) Detection of Sox9 expression in control and TetSox9 Brn4 brains at E18.5 by Western blot. Protein amounts were adjusted using Gapdh levels. The size of molecular weight markers in kilodaltons is indicated on the left. (e) From these blots, relative amounts of transgenic and endogenous Sox9 were determined. Endogenous Sox9 levels in control brains were arbitrarily set to 1 (n = 3). Differences were statistically significant as indicated (Student's t test; *p ≤ .05; **p ≤ .01; ***p ≤ .001). (f-w) Immunohistochemical detection of Sox9 (f-q) and tdTomato (r-w) in the spinal cord of control (ctr, f-k) and TetSox9 Brn4 (l-w) mice at E9.5 (f, l, r), E10.5 (g, m, s), E11.5 (h, n, t), E12.5 (i, o, u), E14.5 (j, p, v), and E18.5 (k, q, w). Spinal cords are demarcated by a stippled line and were placed on a black background. Scale bar: 100 μm. TRE, tetracycline-controlled promoter water in accordance with animal welfare laws and under observance of all ethical regulations. For delaying transgenic Sox9 expression, 0.2 mg/ml tetracycline was added to sucrose-containing (0.05 g/ml) drinking water and given to pregnant mice from Day 7.5 until Day 10.5 or Day 12.5 of pregnancy. For proliferation studies, 100 μg BrdU (Sigma, #B5002) per gram body weight were injected intraperitoneally into pregnant females 1 hr before tissue preparation. Embryos and tissues were obtained at embryonic days (E) 9. 5, 10.5, 11.5, 12.5, 14.5, and 18.5 and underwent fixation in 4% paraformaldehyde, cryoprotection in 30% sucrose, embedding and freezing at −80 C.
Cryotome sections were prepared at 10 μm thickness . Both male and female embryos were used for the study. Littermates without Rosa26 stopflox-tTA allele or Brn4::Cre were interchangeably used as controls after confirming that both were comparable to age-matched wild-type embryos.

| Luciferase assays
Mouse Neuro2a neuroblastoma (obtained from American Type Culture Collection, authenticated by polymerase chain reaction) were grown in DMEM supplemented with 10% fetal calf serum and transfected with Col2a1 promoter or Sox10 U3 enhancer containing luciferase reporters and pCMV5-based expression plasmids for Sox9 and the tagged version generated in this study for transgenesis Werner, Hammer, Wahlbuhl, Bösl, & Wegner, 2007). Per 3.5 cm plate, 0.5 μg reporter plasmid and 0.5 μg effector plasmid were used adding empty pCMV5 where necessary. Whole cell extracts were prepared 48 hr after transfection for chemiluminescent measurements of luciferase activities (Hornig et al., 2013).

| Statistical analysis
To determine whether differences in cell numbers or luciferase activities were statistically significant, a two-tailed Student's t test or oneway ANOVA with Bonferroni correction was performed (*p ≤ .05; **p ≤ .01, ***p ≤ .001). Results from independent embryos or transfections were treated as biological replicates (n ≥ 3).

| Data availability
All data generated or analyzed during this study are included in this article and its Supporting Information. Sox9 carried at its aminoterminal end three copies of a FLAG and three copies of a HA epitope. This tag did not interfere with Sox9 expression nor did it impair the ability of Sox9 to activate reporter genes under control of regulatory regions that had previously been shown to be Sox9-responsive such as the Col2a1 promoter and the U3 enhancer of the Sox10 gene ( Figure 1b).
In analogy to a previous study on Sox10 (Weider et al., 2018), we combined the TRE::Sox9 transgene with a Brn4::Cre transgene and a Rosa26 stopflox-tTA allele ( Figure 1c). Following Cre activation in neuroepithelial cells of the early VZ, tTA started to be expressed after removal of the stopflox cassette in the Rosa26 locus and induced Sox9 expression around E9.0, roughly 1-1.5 days before widespread endogenous Sox9 expression commenced in the mouse spinal cord (Figure 1f,g,l,m;Stolt et al., 2003). Western blot analysis of brain extracts at E18.5 furthermore argued that transgenic Sox9 levels were approximately 2.6-fold higher than those of the endogenous Sox9 in 3.2 | TetSox9 Brn4 spinal cords show decreased proliferation and increased apoptosis TetSox9 Brn4 embryos were already recognizable at E14.5 by their smaller head (Figure S1g,h). As evident from tdTomato expression, the transgene was also expressed throughout the brain (Figure S1k,l) where it caused a mild size reduction of the forebrain and a very severe one in mid-and hindbrain regions ( Figure S1i,j). These brain defects may explain why TetSox9 Brn4 mice did not survive birth, thus restricting our analysis to prenatal times. Already at E11.5, TetSox9 Brn4 embryos exhibited a substantially smaller spinal cord (Figure 1h,n,t).
Therefore, we performed DAPI staining on control embryos and Tet-Sox9 Brn4 littermates from E9.5 to E18.5 (Figure 2a When the number of BrdU incorporating cells was set into relation to total cell numbers, both genotypes were comparable at all-time points F I G U R E 2 Total cell numbers in the TetSox9 Brn4 spinal cord. (a-l) DAPI staining of control (ctr, a-f) and TetSox9 Brn4 (g-l) spinal cords at E9.5 (a, g), E10.5 (b, h), E11.5 (c, i), E12.5 (d, j), E14.5 (e, k), and E18.5 (f, l). Spinal cords are demarcated by a stippled line and were placed on a black background. Scale bar: 100 μm. (m) Determination of cell numbers in control (gray dots) and TetSox9 Brn4 (open triangles) spinal cords from DAPI stainings. Mean cell numbers ± SEM per section are presented (n = 3 spinal cords per genotype, counting three separate sections each). Statistical significance was determined by two-tailed Student's t test (**p ≤ .01; ***p ≤ .001). DAPI, 4 0 ,6-diamidin-2-phenylindole except E10.5 where proliferation rates were slightly higher in controls (20 ± 2%) than in TetSox9 Brn4 (14 ± 1%) spinal cords (Figure 3n). This argues that proliferation rates in TetSox9 Brn4 spinal cords are not substantially altered. were obtained when apoptosis was monitored with TUNEL instead of cleaved caspase 3 staining ( Figure S2a-n). We conclude from these findings that apoptosis rates are dramatically increased during most of embryogenesis and that increased apoptosis is the primary cause of reduced cell numbers in the TetSox9 Brn4 spinal cord.
In support of strongly reduced NEP numbers, immunohistochemistry with antibodies directed against Nestin revealed a dramatically decreased number of stained NEP-derived radial fibers ( Figure S3a-j).
Quantification was not feasible. Interestingly, apoptosis was not increased in NEPs (Figure 5n) arguing that these cells may lose their epithelial character and/or develop into a different cell type.

| Neuronal lineage cells are strongly reduced in TetSox9 Brn4 spinal cord
To follow the fate of neuronal cells, we used Sox11 as a marker for cells that are committed to the neuronal lineage but have not yet matured (Sock et al., 2004). These neuronal precursors and immature neurons (from now on referred to as NPs) are found in normal numbers in TetSox9 Brn4 spinal cords until E10.5 (Figure 6a,b,g,h). However, differences between genotypes became apparent from E11.5 onwards. Whereas NP numbers rose steeply in controls at E11.5, reached their peak at E12.5 and then gradually declined as neurons mature (Figure 6c-

m) Determination of neuronal precursors in control (gray dots) and
TetSox9 Brn4 (open triangles) spinal cords as mean ± SEM per section (n = 3 spinal cords per genotype, counting three separate sections each). (n) Quantification of cleaved caspase 3-positive apoptotic neuronal precursors in control and TetSox9 Brn4 spinal cords at E11.5 as mean ± SEM per section (n = 3). Statistical significance was determined by two-tailed Student's t test (**p ≤ .01; ***p ≤ .001) F I G U R E 7 Neurons in the TetSox9 Brn4 spinal cord. (a-l) Immunohistochemical detection of NeuN-positive neurons in control (ctr, a-f) and TetSox9 Brn4 (g-l) spinal cords at E9.5 (a, g), E10.5 (b, h), E11.5 (c, i), E12.5 (d, j), E14.5 (e, k), and E18.5 (f, l). Spinal cords are surrounded by a stippled line and were placed on a black background. Scale bar: 100 μm. (m) Determination of neurons in control (gray dots) and TetSox9 Brn4 (open triangles) spinal cords as mean ± SEM per section (n = 3 spinal cords per genotype, counting three separate sections each). (n) Quantification of cleaved caspase 3-positive apoptotic neurons in control and TetSox9 Brn4 spinal cords at E11.5 as mean ± SEM per section (n = 3). Statistical significance was determined by two-tailed Student's t test (**p ≤ .01; ***p ≤ .001). VZ, ventricular zone 3.5 | Oligodendroglial lineage cells are strongly increased in TetSox9 Brn4 spinal cord We next studied the oligodendroglial lineage. Olig2 is continuously expressed in these cells and can serve as a lineage marker (Lu et al., 2000;Zhou, Wang, & Anderson, 2000). However, Olig2 also labels the NEPs in the pMN domain of the ventral VZ that gives rise to motoneurons and most spinal cord oligodendrocytes. For quantifications of oligodendroglial cells, we therefore excluded the pMN NEPs and only counted the Olig2-positive cells in the mantle zone.
At E9.5, Olig2 stainings were comparable between wild type and TetSox9 Brn4 spinal cords (Figure 8a,g). Almost all Olig2-positive cells F I G U R E 8 Legend on next page. To corroborate our findings, we employed additional markers.
Stainings for Pdgfra as an OPC marker (Pringle & Richardson, 1993) and for Sox10 as an oligodendroglial lineage marker (Kuhlbrodt et al., 1998) confirmed the precocious appearance of OPCs and the drastic increase In line with increased and precocious generation of OPCs in the TetSox9 Brn4 spinal cord, we also observed a premature appearance of Mbp-expressing differentiating oligodendrocytes at E14.5 (Figure 8t, u,w). These cells were predominantly localized in the dorsal spinal cord, whereas Mbp-expressing cells normally appear first in ventral regions. Their number remained increased at E18.5 (Figure 8t,v,x).
However, even at E18.5 only a minor fraction of oligodendroglial cells in the TetSox9 Brn4 spinal cord had started to express Mbp (Figure 8t).

| Astroglial cells are increased in TetSox9 Brn4 spinal cord
Astroglial cells represent the second main type of neuroectodermal, macroglial cells in the CNS. However, few specific markers are available and none of the markers typically used to identify astroglial cells, such as Glast, Fgfr3, Nfia, or Nfib, is able to unambiguously identify this cell type during early spinal cord development. Glast and Fgfr3, for instance, label NEPs in addition to astroglial cells (Stolt et al., 2003). Therefore, they cannot distinguish between an astroglial cell and a displaced former NEP cell in the TetSox9 Brn4 spinal cord. Nfia and its relative Nfib are expressed even more broadly, as they label NEPs, astroglial cells, OPCs, and subsets of neurons ( Figure S7a-l).
In contrast to other astroglial markers, S100b does not stain NEPs.
However, S100b is expressed in a subset of Sox10-positive oligodendroglial cells. We therefore defined astroglial cells as S100b-positive cells that do not express Sox10.
Such cells were undetectable in control and TetSox9 Brn4 spinal cords before E12.5 and increased steadily thereafter (Figure 9a-h). Quantification showed that the increase was much stronger in TetSox9 Brn4 spinal cords than controls (Figure 9i) arguing that Sox9 overexpression also promotes the generation of astrocytes as the second macroglial cell type. However, the increase in astroglial numbers was less pronounced than the one in oligodendrocytes at any given time and we did not find evidence for a precocious generation of astroglial cells.
These cells are normally regarded as astroglial cells. Similarly, Gfap staining was increased and found in mantle and marginal zones of Tet-Sox9 Brn4 spinal cords at E18.5, whereas its occurrence was restricted in age-matched control spinal cords to the marginal zone (Figure 9l,m).
F I G U R E 9 Astroglial cells in the TetSox9 Brn4 spinal cord. (a-h) Immunohistochemical detection of cells that were positive for S100b (green) and negative for Sox10 (magenta) in control (ctr, a-d) and TetSox9 Brn4 (e-h) spinal cords at E10.5 (a, e), E12.5 (b, f), E14.5 (c, g), and E18.5 (d, h). (i) Determination of astroglial cells in control (gray dots) and TetSox9 Brn4 (open triangles) spinal cords. Numbers of S100b-positive and Sox10-negative cells are shown and presented as mean ± SEM per section (n = 3 spinal cords per genotype, counting three separate sections each). (j-m) Immunohistochemical staining of control (j, l) and TetSox9 Brn4 (k, m) spinal cords section at E18.5 with antibodies directed against Aldh1l1 (j, k) and Gfap (l, m). For documentation of immunohistochemical stainings, spinal cords are surrounded by a stippled line and were placed on a black background. Scale bars: 100 μm (h, m). Statistical significance was determined by two-tailed Student's t test (*p ≤ .05; ***p ≤ .001) 3.7 | Effects of Sox9 transgene induction by Nestin:: Cre resemble those by Brn4::Cre

| Delayed Sox9 overexpression leaves NEPs and neuronal cells unaffected
As Sox9 expression in transgenic mice was under control of tTA, we were able to alter its onset by treatment with tetracycline or derivatives such as doxycycline. We chose tetracycline over doxycycline because resumption of tTA activity has been reported to be faster after the end of tetracycline application and thus easier to control (A-Mohammadi, Alvarez-Vallina, Ashworth, & Hawkins, 1997). We employed two different schemes using TetSox9 Brn4 mice.

| Even delayed Sox9 overexpression strongly increases gliogenesis
Considering that delayed Sox9 overexpression had only mild effects on NEPs and neurogenesis, it was interesting to study the consequences of tetracycline treatment on gliogenesis in TetSox9 Brn4 embryos. Both Olig2-and Sox10-positive cells were still increased in number relative to controls, only slightly less than in untreated F I G U R E 1 1 Glial cells in the TetSox9 Brn4 spinal cord after delayed Sox9 induction. (a-o) Staining of spinal cords from tetracycline-treated (a, b, f, g, k, l) and untreated (c, d, h, i, m, n) TetSox9 Brn4 mice at E12.5 (a, c, f, h, k, m) and E14.5 (b, d, g, i, l, n) for Olig2-positive (a-d), Sox10-positive (f-i), as well as S100b-positive, Sox10-negative cells (k-n), and quantification of oligodendroglial (e, j) and astroglial cells (o). Scale bar, 100 μm. Values for untreated TetSox9 Brn4 spinal cords are depicted as open triangles, those for treated TetSox9 Brn4 spinal cords as blue squares and those for controls as gray dots. All numbers are presented as mean ± SEM per section (n = 3 spinal cords per genotype, counting three separate sections each). Statistical significance was determined by oneway ANOVA with Bonferroni's multiple comparison test (*p ≤ .00332; **p ≤ .0021; ***p ≤ .0002) TetSox9 Brn4 embryos (Figure 11a-j). Compared to Olig2, Sox10 was detected in cells closer to the VZ especially at E12.5 arguing that Sox10 was induced earlier in these cells than Olig2. There was also a substantial increase in the number of S100b-positive, Sox10-negative astroglial cells in tetracycline-treated TetSox9 Brn4 embryos relative to controls, although the increase did not reach statistical significance (Figure 11k-o). In summary, this argues that Sox9 is able to promote gliogenesis even if induced late during embryonic development.

| DISCUSSION
Sox9 is one of the few transcriptional regulators that has been associated in the past with commitment to the glial cell lineage in the CNS (Kang et al., 2012;Scott et al., 2010;Stolt et al., 2003). Most of these studies were based on loss-of-function approaches and the few gainof-function studies were performed ex vivo in cultured cells. Here, we have employed an in vivo gain-of-function approach and selectively overexpressed Sox9 in the CNS using an inducible, tetracyclineresponsive tTA-dependent expression system. In the spinal cord as our site of analysis, this strategy allowed Sox9 expression from E9.0 onwards in all NEPs and cells derived from them. However, the system also allowed us to delay Sox9 overexpression until E10.5 or E12.5 by administering tetracycline to pregnant mothers from E7.5 onwards.
In the standard early induction paradigm without a tetracyclinedependent delay, expression of the Sox9 transgene preceded expression of endogenous Sox9 in NEPs. As a consequence of the ectopic expression, NEPs lose their epithelial characteristics and the VZ disappears. This argues that premature Sox9 expression in NEPs is detrimental. As we did not detect increased apoptosis among NEPs, it appears likely that Sox9 interferes with maintenance of the NEP pool either by counteracting the epithelial character or by promoting conversion into different cells. It has previously been reported that Sox9 expression in neural crest precursors promotes epithelial to mesenchymal transition and is essential for their emigration from the roof of the neural tube (Cheung et al., 2005;Cheung & Briscoe, 2003). Inappropriate expression in NEPs may have similar effects and thereby cause resolution of the VZ. However, we have no evidence to assume that Sox9 expression leads to a conversion of NEPs into neural crest cells.
Ectopic Sox9 expression furthermore continues in all neuronal cells that are generated from these NEPs. Both Sox9 overexpressing neuronal precursors and neurons exhibited increased rates of apoptosis. We therefore conclude that Sox9 compromises survival of neuronal cells during their maturation. Interestingly, ventral neurons were largely spared. This is due to the fact that most ventral neurons in our transgenic animals do not express Sox9 as their specification precedes transgene induction.
In contrast, in the late induction paradigm following tetracycline treatment Sox9 transgene expression was induced in NEPs that already expressed endogenous Sox9 and preferentially give rise to glial cells. Surprisingly, neuronal cells were largely Sox9 negative.
One plausible explanation for this unexpected observation may be that the TRE::Sox9 transgene was epigenetically inactivated during neuronal differentiation so that it could no longer be induced once tTA regained activity. As a consequence, transgenic Sox9 is largely expressed in cells that already express Sox9 endogenously, and increases Sox9 levels in these cells. Under these conditions, NEPs and the VZ remain largely unaffected, as do neurons and their precursors.
However, under both conditions, transgenic Sox9 expression causes a dramatic increase in cells with glial identity. This concerns both oligodendroglial and astroglial cells. The changed distribution of cells among NEPs, neuronal, oligodendroglial and astroglial cells is summarized for E12.5 and E14.5 in Figure 12a,b.
The pro-gliogenic activity of Sox9 is in agreement with previous conclusions from loss-of-function studies in the mouse (Stolt et al., 2003). According to our studies, surplus oligodendroglial cells are F I G U R E 1 2 Influence of transgenic Sox9 expression on numbers of neuroectodermal cell types in the spinal cord. (a, b) Pie charts showing the distribution of neuroectodermal cells among neuronal cells (blue), NEPs (light gray), oligodendroglial cells (red), astroglial cells (orange), and others (dark gray) in control (ctr), untreated and tetracyclinetreated TetSox9 Brn4 spinal cords at E12.5 (a) and E14.5 (b). NEP, neuroepithelial precursor cell generated earlier and in greater number than astroglial cells. This may simply reflect the normal situation in the developing spinal cord where oligodendrogenesis precedes astrogliogenesis. However, this result has to be taken with caution because of the dearth of markers that reliably distinguish astroglial cells from NEPs or oligodendroglial cells at early developmental times.
When using Sox10 or Olig2 as oligodendroglial lineage markers, we furthermore observed substantially more Sox10-positive cells than Olig2-positive cells especially at early times. These represent a sizeable population (labeled as "other" in Figure 12a,b). In the late, tetracycline-delayed induction paradigm where the structure of the spinal cord is much better preserved, it is additionally obvious that cells with exclusive Sox10 expression are localized closer to the VZ than double-positive cells. This argues that Sox10 precedes Olig2 expression in a large fraction of the supernumerary oligodendroglial cells induced by Sox9 transgene expression. A likely scenario thus posits that Sox9 induces Sox10 that in turn activates Olig2 expression. This is in line with previous reports that have shown that Sox10 is directly induced by Sox9 at least in the emerging neural crest (Cheung et al., 2005;Cheung & Briscoe, 2003;Wahlbuhl, Reiprich, Vogl, Bosl, & Wegner, 2011), and once present has the ability to stimulate Olig2 expression (Liu et al., 2007;Weider et al., 2015). Currently, we do not know whether all Sox10-positive cells eventually convert into Sox10-and Olig2-positive oligodendroglial cells or not. However, we do know that the Sox10-positive cells without Olig2 do not have a neural crest identity. The eventual convergence of Sox10-and Olig2-positive cell numbers and our failure to detect increased rates of apoptosis among Sox10-positive cells argues that the large majority of Sox10-positive cells become oligodendroglial cells.
The direct induction of Sox10 by Sox9 may also explain why more ectopic oligodendroglial than astroglial cells are generated in our transgenic mice as Sox10 has been reported to promote oligodendrogenesis and inhibit astrogenesis (Glasgow et al., 2014). If this is the case, then effective ways must exist in normal development that suppress Sox10 induction or counteract its activity in those cells that become astrocytes. The transcription factor Nfia has been described as one such factor (Glasgow et al., 2014). It remains to be shown whether Nfia is sufficient or whether other factors contribute.
Finally, it deserves to be noted that most of surplus glia that were generated, remained in the precursor state until birth. Given the premature generation and high number of oligodendroglial cells, a premature maturation may have been expected as well. Our analysis showed that the number of Mbp-expressing cells is indeed increased in the spinal cord shortly before birth. However, they remained fairly low in absolute numbers and represented only a minor fraction of the oligodendroglial cell population. They were furthermore predominantly localized in dorsal regions, whereas differentiating oligodendrocytes under normal conditions appear first in the ventral spinal cord.
This may argue that conditions in the spinal cord are not yet conducive for differentiation, that proliferation promoting and antidifferentiation factors in OPCs such as the Hes5, Sox5, Sox6, and Id4 have not yet been sufficiently downregulated and that conditions in ventral regions are even worse than in dorsal ones (Liu et al., 2006;Marin-Husstege et al., 2006;Stolt et al., 2006). Future experiments will have to distinguish between these options.
In any case, our study provides unequivocal evidence for a progliogenic role of Sox9 in the spinal cord. In addition, we show that the onset of Sox9 expression has to be tightly controlled to preserve the integrity of the VZ and prevent detrimental influences on neurogenesis. This dual function of Sox9 as a pro-gliogenic factor and an impediment to neurogenesis may be one of the reasons why gliogenesis sets in after neurogenesis.