ASCL1 regulates proliferation of NG2‐glia in the embryonic and adult spinal cord

Abstract NG2‐glia are highly proliferative oligodendrocyte precursor cells (OPCs) that are widely distributed throughout the central nervous system (CNS). During development, NG2‐glia predominantly differentiate into oligodendrocytes (OLs) to myelinate axon fibers, but they can also remain as OPCs persisting into the mature CNS. Interestingly, NG2‐glia in the gray matter (GM) are intrinsically different from those in the white matter (WM) in terms of proliferation, differentiation, gene expression, and electrophysiological properties. Here we investigate the role of the transcriptional regulator, ASCL1, in controlling NG2‐glia distribution and development in the GM and WM. In the spinal cord, ASCL1 levels are higher in WM NG2‐glia than those in the GM. This differential level of ASCL1 in WM and GM NG2‐glia is maintained into adult stages. Long‐term clonal lineage analysis reveals that the progeny of single ASCL1+ oligodendrocyte progenitors (OLPs) and NG2‐glia are primarily restricted to the GM or WM, even though they undergo extensive proliferation to give rise to large clusters of OLs in the postnatal spinal cord. Conditional deletion of Ascl1 specifically in NG2‐glia in the embryonic or adult spinal cord resulted in a significant reduction in the proliferation but not differentiation of these cells. These findings illustrate that ASCL1 is an intrinsic regulator of the proliferative property of NG2‐glia in the CNS.

Indeed, local induction of NG2-glia proliferation and differentiation by neuronal activity to re-myelinate projection neurons has been shown to be a critical step for the learning of complex motor skills and behaviors in mice (Gibson et al., 2014;McKenzie et al., 2014;Xiao et al., 2016). Thus, NG2-glia not only serve as a source to supply the necessary OLs to maintain the function and metabolic needs of neurons but also to stabilize the plasticity and dynamics of neuronal circuits.
Initially, NG2-glia was presumed to be a homogenous cell population throughout the various regions of the CNS, mostly because of a shared expression profiles of OL-lineage markers such as OLIG1/2, SOX10, and PDGFRa (Lu et al., 2002;Nishiyama et al., 2009;Stolt, Lommes, Friedrich, & Wegner, 2004;Zhou & Anderson, 2002). However, with the advent of transgenic mice, extensive characterization of NG2-glia in vivo revealed that they are far more heterogeneous than previously appreciated. For instance, electrophysiological recordings of labeled NG2-glia in the cortical GM demonstrate that they exhibit distinct membrane potentials and expression profiles of potassium (K1) and sodium (Na1) channels than their respective counterparts in the subcortical WM or corpus callosum (Chittajallu, Aguirre, & Gallo, 2004).
Similarly, GM NG2-glia in the brain and spinal cord, whether during neonatal development or at adult stages, are generally less proliferative, differentiate at a slower pace, and respond differently to plateletderived-growth-factor (PDGF) in comparison to WM NG2-glia (Dimou, Simon, Kirchhoff, Takebayashi, & Gotz, 2008;Hill, Patel, Medved, Reiss, & Nishiyama, 2013;Kang et al., 2010;Kang et al., 2013;Psachoulia, Jamen, Young, & Richardson, 2009;Rivers et al., 2008;Zhu et al., 2011). Transplantation experiments suggest that GM and WM NG2glia are intrinsically unique (Vigano, Mobius, Gotz, & Dimou, 2013), which may be directly related to their function within these regions in the CNS. However, at present it is unclear how the intrinsic properties of NG2-glia in the GM or WM are regulated, or whether NG2-glia in the GM are derived from the same or a separate OLP lineage than those in the WM.
Previously, we and others reported that the bHLH transcription factor ASCL1, which is well known for its roles during neurogenesis, is broadly expressed in glial progenitor cells throughout the ventricular zone (VZ) at the onset of gliogenesis in the spinal cord and in the cortex (Nakatani et al., 2013;Parras et al., 2007;Sugimori et al., 2007;Sugimori et al., 2008;Vue, Kim, Parras, Guillemot, & Johnson, 2014).
Notably, ASCL1 expression is maintained in NG2-glia as they migrate out of the ventricular zone to populate the surrounding GM and WM, but is downregulated once NG2-glia differentiate to become mature OLs (Nakatani et al., 2013;Vue et al., 2014). Accordingly, Ascl1 mutant and conditional-knock out mice exhibit a significant decrease in the number of NG2-glia and OLs that are generated, particularly in the WM of the spinal cord (Battiste et al., 2007;Nakatani et al., 2013;Parras et al., 2007;Sugimori et al., 2007;Sugimori et al., 2008;Vue et al., 2014), suggesting that ASCL1 may play an important role in regulating the generation of NG2-glia in the CNS. However, the precise function of ASCL1 specifically in NG2-glia during embryonic development or in the adult CNS remains unclear.
In this study, we show that the level of ASCL1 is substantially higher in WM NG2-glia than in GM NG2-glia during development of the spinal cord. Furthermore, clonal analysis using Ascl1 CreERT2 knock-in mice carrying the stochastic multicolor R26R-loxP-stop-loxP-Confetti reporters (R26R LSL-Confetti ; Schepers et al., 2012;Snippert et al., 2010) demonstrate that progeny from single ASCL11 OLPs labeled at embryonic day 14.5 (E14.5) are spatially restricted to the GM or WM, even though they proliferate extensively to form large clusters of SOX101; Confetti1 OLs in the adult spinal cord. This suggests that OLPs are restricted in fate to the GM or WM by the time they express ASCL1.
Finally, conditional deletion of Ascl1 specifically in labeled (tdTomato1 or Confetti1) NG2-glia using Ng2-CreER TM mice at E14.5 or adult postnatal day (P) 30 resulted in a significant decrease in the proliferation, but not differentiation, of NG2-glia. Taken together, these findings illustrate that the level of ASCL1 plays an important role in ensuring the proper generation of the number of NG2-glia in the GM and WM of the spinal cord.

| Tamoxifen and BrdU administration
The appearance of a vaginal plug was considered E0.5 and the day of birth was noted as P0. Tamoxifen (T5648, Sigma-Aldrich Co., St. Louis, MO 63103, United States) was dissolved in 10% ethanol/90% sunflower oil mixed and administered via intraperitoneal injection. For population analysis using Ng2-CreER TM crossed with R26R LSL-tdTOM mice, a single dose of 2.5 mg tamoxifen/40 g body weight was injected to pregnant females at E14.5, or double doses of 2.5 mg tamoxifen/40 g body weight per day were injected to adult mice for two consecutive days starting at P30. For clonal analysis using R26R LSL-Confetti mice, a single dose of 2.5 mg tamoxifen/40 g body weight was injected into pregnant females when crossed with Ascl1 CreERT2 mice, and single dose of 0.625 mg tamoxifen/40 g body weight was injected into pregnant females at E14.5 or adult mice at P30 when crossed with Ng2-CreER TM mice. Due to the effects of tamoxifen on birth complications, cesarean section was performed on the day of birth (E19 or P0) on pregnant females that were injected with tamoxifen at E14.5. Pups were carefully introduced and raised by a foster female until date of analysis.
BrdU (Roche 10280879001) was dissolved at 5 mg/ml in 0.007 M NaOH. Ng2-CreER TM ; R26R LSL-tdTOM mice that received tamoxifen injection at P30 were administered intraperitoneally with 3 mg BrdU/25 g body weight per day for 10 consecutive days starting at 7-16 days post-tamoxifen injection. Statistical analyses were performed as previously described (Vue et al., 2014) to assess the differences in NG2-glia cell density or per-  in spinal cord at E14.5. ASCL1 co-localizes with OLIG2 in the VZ (arrows) and outside (arrowheads) of the VZ. (f-j) Immunofluorescence for ASCL1, NEUN, OLIG2, PDGFRa, Ki67, and CC1 in P2 (f-h) and P30 (i and j) spinal cords. ASCL1 expression at P2 is higher in WM than GM, and co-localizes with OLIG2, PDGFRa, and Ki67 (arrowheads) but not with CC1 (arrow, insets). Differential expression of ASCL1 in GM and WM is maintained at P30. (k-q) Immunofluorescence for GFP in P60 Ascl1 GFP/1 spinal cord. GFP is expressed in PDGFRa1, OLIG21, and NG21 cells (l-o) in GM and WM, but not in CC11 cells (p and q). Dotted lines indicate GM (NEUN1) and WM boundary. Scale bar is 100 lm for a-c, f-g; 25 lm for d, e, h; and 12.5 lm for insets in h, i, j, and l-q. (r) Quantification of ASCL1 immunofluorescent intensity (AU-arbitrary units) in OLIG21;PDGFRa1 NG2-glia at P2 (N 5 3) and P30 (N 5 2). Student's t test, *p < .01. Error bars are SEM. are also Ki671 (arrowheads in middle panels, Figure 1h,s). In contrast, none of the ASCL11 cells are positive for CC1 (arrows in bottom panels, Figure 1h), a marker of immature and mature OLs. Thus, ASCL1 expression in the GM and WM in the spinal cord is limited specifically to OPCs/NG2-glia and not differentiated OLs.

| Tissue preparation and immunofluorescence
Interestingly, a fundamental difference that we observe at this neonatal stage is that ASCL1 is expressed at a higher level in OLIG21 NG2-glia in the WM compared wit those in the GM (top panel, Figure   1h). Indeed, by quantifying the immunofluorescent intensity of ASCL1, we found that the level of ASCL1 is twice as high in WM NG2-glia than in GM NG2-glia at P2 (Figure 1r). To deterhmine if ASCL1 is also differentially expressed in NG2-glia in the GM and WM of the adult spinal cord, we performed immunofluorescence for ASCL1 at P30 (Figure 1i,j). We were able to detect ASCL1 expression in OLIG21 cells in the WM (Figure 1j), whereas ASCL1 expression in the GM was just barely detectable in some OLIG21 cells (Figure 1i). Quantification shows that the immunofluorescence level of ASCL1 at P30 is still twice as high in the WM compared with the GM, but overall it had decreased to about half of P2 in both the GM and WM ( Figure 1r). This indicates that ASCL1 expression in NG2-glia is not only spatially but also temporally dynamic.
To confirm if ASCL1 expression in the mature spinal cord is specific to OPC/NG2-glia, we next performed immunofluorescence for GFP on spinal cord sections of P60 ASCL1 GFP/1 knock-in mice along with OPC and OL markers. Because GFP is a stable protein, immunofluorescence for GFP should mark all ASCL1 expressing cells, including those in the GM that express very low or undetectable levels of ASCL1 at this stage. As expected, we found that GFP is expressed in cells that Collectively, these findings demonstrate that ASCL1 is present in OLPs in the ventricular zone at the onset of oligodendrogenesis, is differentially maintained in NG2-glia in the GM and WM, and is downregulated as development proceeds from neonatal to adult stages.
Furthermore, ASCL1 expression is lost when NG2-glia transition to become CC11 immature or mature OLs at all stages of spinal cord development ( Figure 1t).

| Progeny deriving from single ASCL11 glial progenitor clones are restricted to GM or WM in the spinal cord
Currently, it is unclear when NG2-glia and OLs in the GM are developmentally distinct from those in the WM. Since ASCL1 is expressed in OLPs in the ventricular zone and is maintained at a higher level in WM NG2-glia than in GM NG2-glia, it is possible that this differential expression of ASCL1 may play a role in regulating the intrinsic properties of NG2-glia and their distribution and development in the GM and WM of the spinal cord. We predicted that if this were the case, then clonal analysis of the developmental lineages and distribution of single ASCL11 OLP clones should result in progeny that are spatially restricted to the GM or WM. To test this prediction, we crossed Ascl1-CreERT2/1 knock-in mice (Kim et al., 2011) with the multicolor R26R LSL-Confetti reporter mice (Snippert et al., 2010) In agreement with previous reports that WM NG2-glia are more proliferative than those in the GM (Dimou et al., 2008;Young et al., 2013), across an average thickness of about 300 lm along the AP-axis. In contrast, WM SOX101:Confetti1 clones contain an average of 2.5 cells per 30 lm section but span an average of 900 lm thickness along the AP-axis (Figure 2f,g). Therefore, WM NG2-glia not only generate a larger clone of cells but also spread more widely to occupy a thicker volume.
In addition to SOX101 OL-lineage clones, we identified SOX10clones that are also spatially restricted in the GM or WM in the dorsal spinal cord (arrows, Figure 2h Taken together, these results show that during oligodendrogenesis in the spinal cord, ASCL1 is required to ensure a proper generation of the number of NG2-glia but not their differentiation into mature OL.

| Ascl1-CKO decreases the proliferation of adult NG2-glia in the spinal cord
To determine if ASCL1 is also required for regulating the development of NG2-glia in the adult CNS, we analyzed the spinal cords of Ascl1-CKO and control littermates that were administered with two doses of tamoxifen (2.5 mg/40 g body weight) over two days starting at P30 (red arrows, Figure 5a). The number of OLIG21;tdTOM1 cells/mm 2 in both the GM and WM of thoracic spinal cord hemisections at 1, 7, 28, and 56 dpi was determined (Figure 5b,c).
Within control spinal cords (top panels, Figure 5b), at 1 dpi there were 74 6 11 and 39 6 6 OLIG21;tdTOM1 cells/mm 2 in GM and WM, respectively. This number then doubled to 143 6 28 in GM and 66 6 12 in WM by 7 dpi, and more than tripled to 262 6 47 in the GM and 139 6 10 in the WM by 28 dpi. By 56 dpi, the number of OLIG21; tdTOM1 cells/mm 2 was decreased to 216 6 30 in the GM and 99 6 15 in the WM, suggesting that adult NG2-glia labeled at P30 may also start to turnover between 28 and 56 dpi (green bars, Figure 5c). To investigate if the failure for adult NG2-glia to increase after 7 dpi in the Ascl1-CKO was indeed due to a reduction of cell proliferation, we used bromodeoxyuridine (BrdU), a thymidine analog that is incorporated into newly synthesized DNA, to specifically label replicating cells. Control and Ascl1-CKO littermate mice were administered with two pulses of tamoxifen starting at P30, and BrdU (3 mg/25 g body weight) was then injected once per day for 10 consecutive days starting at 8-17 dpi (black arrows, Figure 5a). We observed that there was far fewer BrdU1;tdTOM1 cells in thoracic spinal cord at 17 dpi in the Ascl1-CKO compared with controls. Indeed, by analyzing the proportion of BrdU1;tdTOM1/total OLIG21;tdTOM1 cells, we found that the percentage of proliferating cells in the control was around 18.5% in the GM and 28.1% in the WM (green bars, Figure 5e), and was significantly reduced in the Ascl1-CKO spinal cords to 11.4% in the GM and 14.4% in the WM (blue bars, Figure 5e). This finding strongly suggests that the decrease in the overall number of OLIG21; tdTOM1/mm 2 cells in Ascl1-CKO spinal cords is likely due to a decrease in the ability of adult NG2-glia to proliferate in the absence of ASCL1.
As in embryonic spinal cord development, we did not observe any significant difference in the percentage of CC11;tdTOM1 cells between the control and Ascl1-CKO spinal cords at 1, 7, 28, or 56 dpi following tamoxifen administration at P30 (Figure 5d). Furthermore, both MAG and CNP were also expressed in OLIG21;tdTOM1 cells of Ascl1-CKO (Figure 6f,h) as in the control (Figure 6e,g). The expression of these myelinating OL markers argues that the ability of adult NG2- | 11 glia to differentiate was not affected by the loss of ASCL1. However, a notable difference between OLIG21;tdTOM1 cells that were labeled by Ng2-CreER TM at P30 from those labeled at E14.5 is that around 30% of the cells were already CC11 at 1 dpi (Figure 5b,c), compared with 0% at E14.5. Furthermore, whereas more than 95% of OLIG21; tdTOM1 cells labeled at E14.5 had differentiated to become CC11 by 28 dpi (Figure 4d), only about 70% of OLIG21;tdTOM1 cells labeled at P30 were CC11, even at 56 dpi (Figure 5d). This indicates that adult NG2-glia proliferate and differentiate at a much slower pace than embryonic NG2-glia in the spinal cord, and about 30% of the adult NG2-glia, as labeled by Ng2-CreER TM , may continue to remain undifferentiated on a long-term basis.
3.6 | The size of NG2-glia clones is decreased in the absence of ASCL1   Figure 7b,c), which are dramatically fewer than the overall number of control clones labeled in Ng2-CreER TM ;Ascl1 F/1 ;R26R LSL-Confetti spinal cords. This decrease was not completely unexpected since the number of NG2-glia lineage cells at the population level, as labeled by tdTOM, was also decreased with Ascl1-CKO (Figure 4b,c).
A notable feature that we frequently observed was that Ascl1-CKO Confetti1 clones contained fewer cells per clone than the control Confetti1 clones (Figure 7a). In particular, when we evaluated the number of Ascl1-CKO Confetti1 clones based on clone size, we found that in the GM there was an increase in clone numbers for those that contained between 2 and 4 cells per clone (pink arrowhead, Figure   7b), along with a concomitant decrease in clone numbers for those that contained more than 5 cells per clone (red arrowheads, Figure   7b). Similarly, Ascl1-CKO Confetti clones in the WM also showed an increased number of clones containing 2-4 cells per clone (pink arrowhead, Figure 7c), and a concomitant decrease in clones containing between 5 and 10 cells per clone (red arrowhead, Figure 7c). Interestingly, there was no decrease in the number of clones in WM Ascl1-CKO Confetti mice that contained more than 10 cells per clone compared with the control (Figure 7c). This suggests that Ascl1-CKO may not negatively affect the proliferation of highly proliferative NG2-glia in the WM, or that these cells may have escaped Ascl1-CKO altogether. Indeed, when we quantified the overall average number of cells per clone, GM but not WM Ascl1-CKO Confetti1 clones exhibit a 46% decrease (5.0 6 0.6 GM and 5.3 6 0.6 WM; blue bars, Figure   7d), in comparison to their respective controls (9.2 6 0.9 GM, and 6.1 6 0.8 WM; green bars, Figure 7d).
To further assess if Ascl1-CKO has a similar effect on the development of NG2 clones in the adult spinal cord, we administered a single dose of tamoxifen (0.625 mg/40 g body weight) at P30 to Ng2-CreER TM ;Ascl1 F/1 ;R26R LSL-Confetti or Ng2-CreER TM ;Ascl1 F/F ;R26R LSL-Confetti mice. Equal number (N 5 2) and volume of spinal cords were serially sectioned and analyzed for Confetti1 clones at 56 dpi for each genotype ( Figure 7e). We identified about 280 control Confetti1 clones in Ng2-CreER TM ;Ascl1 F/1 ;R26R LSL-Confetti spinal cords, which comprised a total of about 800 labeled cells that were similarly distributed into the GM and WM. About 45% of these clones were single cell clones, 50% were clones that contained between 2 and 9 cells, while the remaining 5% were larger clones containing between 10 and 15 cells (green bars, with control clones (green bars, Figure 7h).
Collectively, these findings demonstrate that at the clonal level, the proliferation of NG2-glia is highly dynamic, and on average is reduced by two to three-fold as development proceeds from the embryo into the adult. Furthermore, ASCL1 seems to play a vital role in shaping the survival as well as the proliferation of NG2-glia, as the number of Confetti-labeled NG2-glia clones is reduced with Ascl1-CKO and the size of NG2-glia clones seems to favor the generation of smaller clones at the expense of larger highly proliferative NG2-glia clones.
3.7 | Complete deletion of Ascl1 with Olig1 Cre decreases the number but not differentiation of OLIG21 cells in the spinal cord  (Figure 8b,c). As seen in germline Ascl1 mutants Sugimori et al., 2008), the loss of ASCL1 did not compromise the generation of NG2-glia, as marked by OLIG2 and PDGFRa in the Ascl1-CKO (Figure 8d,k), but there was a noticeable decrease in the The Olig1 Cre/6 ;Ascl1-CKO mice were able to survive postnatally and continued to exhibit a decrease of OLIG21 cells in the WM at P14 ( Figure 8l). Quantification of the number of OLIG21 cells/mm 2 in the GM revealed that it was similar between control (622 6 35) and Ascl1-CKO (594 6 64) at P0 but was slightly decreased in the Ascl1-CKO at P14 (797 6 43 Ascl1-CKO vs. 883 6 42 control), though this decreased was not significant (blue vs. green bars, Figure 8o). In the WM, the number of OLIG21 cells/mm 2 was decreased by about 40% in the Ascl1-CKO compared with control at both P0 and P14 (blue vs. green bars, Figure 8o). This decreased in OLIG21 cells in the WM but not GM was similar to our previous study in which Ascl1 was conditionally knocked out in OLPs at E14.5 using Ascl1 CreERT2 (Nakatani et al., 2013;Vue et al., 2014).
To determine if the differentiation of NG2-glia was affected by the complete loss of ASCL1, we next performed immunofluorescence

| DISCUSSION
The generation, development, and differentiation of NG2-glia in the CNS is a complex process across both space and time, and requires the precise regulation and function of multiple genes, particularly those encoding transcription factors (Olig1/2, Nkx2.2, Ascl1, Sox10, Prox1, Smarca4/Brg1, Myrf, and Sox17;Hornig et al., 2013;Kato et al., 2015;Li, Lu, Smith, & Richardson, 2007;Lu et al., 2002;Stolt et al., 2004;Sugimori et al., 2007;Sugimori et al., 2008;Yu et al., 2013;Zhou, Choi, & Anderson, 2001;Zhou & Anderson, 2002;Zhu et al., 2014). We observed that one of the first fundamental differences between GM and WM NG2-glia in the spinal cord is the level of ASCL1 expression. Indeed, by quantifying the immunofluorescence of ASCL1 specifically in OLIG21;PDGFR1 NG2-glia, we found that it is twice as high in WM NG2-glia than in GM NG2-glia (Figure 1g,r). This difference was apparent from embryonic development up until a month of age, was specific to ASCL1 but not OLIG2 (not shown), and is suggestive that the high and low levels of ASCL1 is a direct reflection of the diversification and heterogeneity of NG2-lineage cells in the WM and GM, respectively ( Figure 9a). Several lines of evidence support this interpretation. First, in loss of function studies, Ascl1-mutant mice, though neonatal lethal, exhibit a dramatic decrease in the number of OLIG21 cells that are generated in the WM Sugimori et al., 2008). Similarly, conditional knock-out of Ascl1 in glial progenitors in the embryonic spinal cord or neonatal brain resulted in a significant and persistent reduction in the number of OL-lineage (OLIG21 and SOX101) cells that are generated in the WM (Figure 8; Nakatani et al., 2013;Vue et al., 2014). In some of these instances, generation of OLIG21 cells in GM was not as adversely affected, which might be expected since WM NG2-glia express a higher level of ASCL1 and therefore may be more dependent on ASCL1 for generation than GM NG2-glia.
The second line of evidence comes from studies which demonstrate that there are intrinsic differences between NG2-glia in the GM and WM in terms of gene expression, electrophysiological properties, proliferation and differentiation rates, and response to PDGF or the surrounding environment (Chittajallu et al., 2004;Dimou et al., 2008;Hill et al., 2013;Kang et al., 2010;Psachoulia et al., 2009;Vigano et al., 2013). Whether these differences are directly determined by or attributable to the differential level of ASCL1 remains to be tested. Nevertheless, they are an indication that GM and WM NG2-glia are distinct and may be derived from separate OLP populations. In agreement with this, our clonal analysis of the spinal cords of Ascl1 CreERT2 ;R26R LSL-Confetti double transgenic mice reveal that when individual ASCL11 OLPs are sparsely labeled at E14.5, they form clones of SOX101;Confetti1 cells that are restricted to the GM or WM at one month of age ( Figure 2).
This restriction was noted for small cell clones, which are probably derived from more mature ASCL11 OLPs or NG2-glia, as well as clones that are comprised of hundreds of cells and are likely to be derived from OLPs in the ventricular zone (Figures 2d,e and 7a). Particularly noteworthy is that the WM clones on average contain more cells per clone than the GM clones. This could be due in part to the higher expression of ASCL1 in WM NG2-glia, given that the proliferation of NG2-glia as measured by their ability to incorporate BrdU is reduced in the absence of Ascl1 ( Figure 5). Interestingly, we also observed that the WM contained more than 70% (21/29) of the clones that were labeled, while the remaining clones (8/29) were found in the GM (Figure 2e).
This uneven labeling distribution could be explained by the fact that ASCL1 marks two distinct OLP-lineages in the spinal cord. For instance, ASCL1 is expressed at a higher level in OLPs, as well as NG2-glia, that are fated to the WM than those that are fated to the GM (Figure 7a), thereby increasing the probability of labeling WM clusters as seen in

| Loss of ASCL1 decreases the proliferation of NG2-glia in the embryonic and adult spinal cord
Previous studies showed that loss of ASCL1 resulted in reduced OPC generation in the spinal cord at the onset of oligodendrogenesis Sugimori et al., 2008), and in the first few weeks following conditional deletion of Ascl1 in OLPs embryonically (Vue et al., 2014). We extend these findings and provide insight into the long-term effect of Ascl1 deletion specifically in NG2-glia during embryonic and adult spinal cord oligodendrogenesis. We observed a significant, persistent reduction in the number of NG2-glia derived tdTOM1 cells in the GM and WM of the spinal cord following conditional deletion of Ascl1 at E14.5 or P30 (Figures 4c and 5c). The reduction of BrdU incorporation in the tdTOM1 cells indicates that this decrease in cell number is due to decreased proliferation, although it is possible that an increase in cell death may also occur for NG2-glia in the absence of ASCL1. This interpretation was supported by long-term clonal lineage analysis of NG2-glia clones, which on average contain not only more cells per clone but also more clones overall in the presence rather than the absence of ASCL1 (Figure 7).
Interestingly, in contrast to the observed effect on the number of OLIG21;tdTOM1 NG2-glia in the embryonic and adult spinal cord following Ascl1 deletion, the proportion of these labeled NG2-glia in the GM or WM that had differentiated into mature CC11 OLs was not significantly altered at any time point analyzed (Figures 4 and 5). These results are in stark contrast with a prior report showing a drastic reduction in the number of myelin-expressing OLs in the WM of the spinal cords of Ascl1-mutant mice at P0, a finding that was attributed to impaired or delayed differentiation of Ascl1-mutant NG2-glia Sugimori et al., 2008). These contrasting findings suggest that while ASCL1 may be required to ensure proper generation and differentiation of NG2-glia, the direct long-term consequences of ASCL1 loss are on proliferation, and potentially survival, but not differentiation of NG2-glia in both the embryonic and adult spinal cord (Figure 9b).
This finding is not surprising considering that key factors such as NKX2.2, OLIG1/2 and SOX10, which have been shown to be crucial for NG2-glia differentiation, are still present and are not affected by the loss of ASCL1. In contrast, Olig1 Cre conditional knock-out of Pdgfra resulted in a drastic decrease in the number of OPC/NG2-glia in the spinal cord (Zhu et al., 2014), a phenotype that is similar but more severe than the Olig1 Cre ;Ascl1-CKO observed in this study (Figure 8).
Because PDGFRa expression is unaffected by ASCL1 loss (Figure 8k), this suggests a possible mechanism by which ASCL1 functions in parallel or downstream of PDGFRa to regulate NG2-glia proliferation.

| Potential mechanism and targets of ASCL1 function in NG2-glia
A fundamental question that needs to be addressed is how ASCL1 differentially regulates NG2-glia proliferation in the GM and WM. Our finding that higher level of ASCL1 is correlated with a higher proliferation rate of NG2-glia in the WM over the GM, and in embryonic and neonatal stages over adult stages implies that the level of ASCL1 is critical for controlling the differential proliferation of NG2 glia. This implication of different levels of ASCL1 controlling its function within progenitor cells has been previously described. For instance, time-lapse imaging of brain slices or neural progenitors in culture revealed that the level of ASCL1, along with OLIG2 and HES1, when fused with reporter proteins oscillates within the cells. Moreover, sustained high levels of ASCL1 promotes cell fate specification and differentiation, whereas lower levels of ASCL1 oscillation promotes cell division (Imayoshi et al., 2013). In agreement with this, ASCL1 expression in neural progenitors in the ventricular zone, whether during neurogenesis or gliogenesis (Figure 1a-e), is relatively high, and may allow ASCL1 to function to ensure proper specification and differentiation of interneurons and OPCs/ NG2-glia (Helms et al., 2005;Parras et al., 2004;Parras et al., 2007;Sugimori et al., 2007). In contrast, the overall level of ASCL1 in NG2-glia in the GM and WM, though can be readily detected in neonatal, and to some extent, adult stages (Figure 1f-j,r), is much lower and may only be at levels sufficient to promote cell division but not differentiation. This would explain how a higher level of ASCL1 in WM NG2-glia would promote these cells to proliferate at a faster rate than GM NG2-glia ( Figure 2), and the lack of ASCL1's role in NG2-glia differentiation.
ASCL1, like other bHLH proteins, functions by forming heterodimers with ubiquitously expressed E-protein (E47) to bind to DNA to regulate gene expression (Castro et al., 2006). Currently, it is unknown whether ASCL1 dimerizes with E47 or different dimerization partners to differentially regulate the proliferation of NG2-glia in the GM and WM.
However, since ASCL1 shows strong preferences in binding to canonical E-box (CANNTG) motifs in the genome of highly proliferative neural progenitors and cancer cells (Castro et al., 2006;Borromeo et al., 2014;Borromeo et al., 2016), it is likely that E47 may also be a dimerization partner in NG2-glia. Recently, a critical role for ASCL1 in regulating neural progenitor cell proliferation has been reported. In particular, Castro et al. (2011) showed that although ASCL1 directly regulates genes critical for the specification and terminal differentiation of neural progenitors in the embryonic ventral telencephalon, a number of direct transcriptional targets also include positive cell cycle genes and oncogenic transcription factors such as E2f1, FoxM1, Cdca7, Tead1/2, and Taz. More recently, Garcez et al. (2015) also demonstrated that ASCL1 directly regulates Cenpj/CPAP, a gene that is critical for centrosome formation and normal mitosis in neural progenitors. Given ASCL1's specific role in regulating NG2-glia proliferation, it is possible that some of these genes may also be directly or differentially regulated by ASCL1 in NG2-glia in the GM and WM of the spinal cord, and in other regions of the CNS. gives rises to OLs capable of re-myelinating projection neurons, a step that was recently shown to be critical for the learning of complex motors tasks in mice (Gibson et al., 2014;McKenzie et al., 2014;Xiao et al., 2016). Lastly, prior studies have suggested that NG2-glia represent a potential cell of origin for some gliomas, such as glioblastoma multiforme, an aggressive, treatment-resistant brain tumor characterized by high ASCL1 expression (Lei et al., 2011;Liu et al., 2011). An understanding of ASCL1's role in regulating the proliferation of NG2glia may therefore provide new insights into the formation and progression of these cancers.