Olig2 defines a subset of neural stem cells that produce specific olfactory bulb interneuron subtypes in the subventricular zone of adult mice

Neural stem cells (NSCs) persist and continuously produce new neurons in a few specialized regions of the adult mammalian brain, including the subventricular zone (SVZ) lining the lateral ventricle (LV) (Nakafuku and Grande, 2020). In these so-called ‘neurogenic niches’, NSCs remain mostly quiescent or divide infrequently (Götz et al., 2016). These quiescent stem cells (qNSCs) share some features with radial glia in embryos and astrocytes in adults, such as the expression of the intermediate filament protein Nestin and glial high-affinity glutamate transporter GLAST (Pastrana et al., 2009). However, a subset of these qNSCs are mobilized to become activated NSCs (aNSCs) and generate rapidly proliferating secondary progenitors called transient amplifying progenitors (TAPs). These TAPs subsequently differentiate into neuroblasts (NBs) and/or glial cells and migrate to their destinations, such as the olfactory bulb (OB), where they integrate into the existing circuitry.

Recent studies have identified many common regulatory mechanisms between these adult NSCs/progenitors and their embryonic counterparts (Götz et al., 2016; Nakafuku and del Águila, 2020). For example, the homeodomain transcription factor (TF) Gsx2 is expressed in a subset of NSCs/progenitors in both the developing ventral telencephalon and the adult SVZ, and plays an important role in the generation of specific neuronal subtypes in the OB (Toresson et al., 2000; Toresson and Campbell, 2001; Yun et al., 2001, 2003; Waclaw et al., 2009; Wang et al., 2013; Qin et al., 2017; López-Juárez et al., 2013). Likewise, the basic helix-loop-helix (bHLH) TF Ascl1 is commonly expressed and plays an essential role in neurogenesis in NSCs/progenitors in the embryonic ventral telencephalon and adult SVZ (Torii et al., 1999; Casarosa et al., 1999; Yun et al., 2002; Wang et al., 2009; Andersen et al., 2014).

Another important common feature between embryos and adults is that NSCs/progenitors in different regions are heterogeneous and produce distinct subtypes of neurons (Götz et al., 2016). For example, NSCs/progenitors in different parts of the ventral telencephalon generate distinct subtypes of OB interneurons during embryogenesis (Stenman et al., 2003; Kohwi et al., 2007; Long et al., 2007; Waclaw et al., 2009; Qin et al., 2017). NSCs that reside in distinct subregions of the adult SVZ also produce different OB interneuron subtypes (Young et al., 2007; Merkle et al., 2007, 2014; López-Juárez et al., 2013; Fuentealba et al., 2015; Delgado and Lim, 2015; Lledo and Valley, 2016; Paul et al., 2017). However, the molecular mechanisms underlying such heterogeneity of adult NSCs remain largely unknown.

It is widely known that the bHLH TF Olig2 is expressed in cells in the oligodendrocyte (OL) lineage and plays an essential role in their development (Emery and Lu, 2015). Importantly, Olig2 is also expressed in other neural cell types, including NSCs/progenitors (Takebayashi et al., 2002; Cai et al., 2007; Qin et al., 2017; Chapman et al., 2018). In this study, we show that Olig2 is expressed in a ventrally enriched subset of NSCs in the early postnatal and adult SVZ and plays an important role in the specification of OB interneuron subtypes. Single cell transcriptomics and genetic studies also reveal important similarities and differences between Olig2-expressing NSCs/progenitors in the embryonic and postnatal brains.

In a search for molecular markers for NSCs, we found that the bHLH TF Olig2 is expressed in the adult SVZ (Fig. 1A-E). Similar to its embryonic expression (Takebayashi et al., 2002; Chapman et al., 2013, 2018), Olig2 expression was found broadly along the dorsoventral (DV) and mediolateral (ML) axes of the LV. However, the density of Olig2⁺ cells among total cells in the dorsolateral (dl) SVZ is significantly lower than in other subdomains (Fig. 1F). Moreover, because both the density and total number of cells are much higher in the ventral (v) SVZ than in the lateral (l) and medial (m) SVZ subdomains (for details, see López-Juárez et al., 2013), the actual number of Olig2⁺ cells is highest in the vSVZ among the four subdomains examined (Fig. 1F; the locations of distinct SVZ subdomains are indicated in the lower panel of Fig. 1C). By contrast, Gsx2⁺ cells were predominantly detected in the dlSVZ. Moreover, although Olig2⁺ and Gsx2⁺ cells were found in all subdomains, few, if any, cells co-expressing both TFs were found (0 out of 63 Olig2⁺ cells in the vSVZ and 0 out of 42 Gsx2⁺ cells in the dlSVZ examined) (Fig. 1B-E,G). A similar regionally biased distribution of Olig2⁺ and Gsx2⁺ cells was also observed in the early postnatal day (P) 16 brain (Fig. 1F). Olig2⁺ cells were also detected in the dorsal roof (callosal) region of the SVZ. However, this SVZ subdomain is immediately adjacent to the overlaying corpus callosum, where Olig2⁺ oligodendrocytes are abundant. Therefore, it is difficult to distinguish Olig2⁺ NSCs from Olig2⁺ oligodendrocytes in this region. Along the anteroposterior (AP) axis of the LV, both Olig2⁺ and Gsx2⁺ cells were broadly detected (Fig. 1A-E), although few Olig2⁺ cells were detected in the rostral migratory stream (RMS) or the most anterior or posterior portions of the SVZ, whereas a significant number of Gsx2⁺ cells occurred in these regions (López-Juárez et al., 2013).

We next examined the cell types that express Olig2 in the NSC lineage. Although only a few Olig2⁺ cells were immunopositive for glial fibrillary acidic protein (Gfap) or Dcx, a large fraction of Olig2⁺ cells co-expressed the TFs Ascl1 and Dlx2, as well as the proliferation marker Ki67, in the vSVZ of the P16 and adult brain (Fig. 1H-N). Ascl1 is expressed in aNSCs and TAPs and plays an essential role in both the proliferation and neuronal differentiation of NSCs, whereas Dlx2 is mostly expressed in late-stage TAPs and NBs and promotes neurogenesis (Doetsch et al., 2002; Kohwi et al., 2007; Brill et al., 2008; López-Juárez et al., 2013; Andersen et al., 2014) (Fig. 1N). These results collectively support the idea that Olig2 is expressed in aNSCs and TAPs in the vSVZ (Fig. 1N), which is reminiscent of the expression pattern of Gsx2 in the dlSVZ (López-Juárez et al., 2013). Although we detected only a few Olig2⁺/Gfap⁺ cells, this might be because of a technical issue given that weak and/or diffuse cytoplasmic signals of Gfap staining in the postnatal SVZ often make it difficult to distinguish individual immunoreactive cells in conventional histology, unlike clear nuclear signals detectable by Ascl1 staining (Pastrana et al., 2009; López-Juárez et al., 2013). Moreover, more than 50% of Ascl1⁺ cells in the vSVZ co-expressed Olig2 [55% (72 Ascl1⁺/Olig2⁺ cells among 130 Ascl1⁺ cells examined) and 62% (120 Ascl1⁺/Olig2⁺ cells among 195 Ascl1⁺ cells examined) in the adult and P16 brains, respectively], demonstrating that Olig2⁺ cells are a predominant population among aNSCs/TAPs in the vSVZ.

The aforementioned results demonstrate that Olig2 is expressed in the NSC lineage. However, it is still possible that some of these Olig2⁺ cells are OLs newly generated from NSCs. In fact, it has been shown that a small, but significant, number of new OLs are generated by adult NSCs (Hack et al., 2005; Menn et al., 2006; Ortega et al., 2013). Recently, single cell RNA-sequencing (scRNA-seq) analysis has begun to be widely used to reveal molecular properties of postnatal NSCs and their progeny (Dulken et al., 2017; Yuzwa et al., 2017; Zywitza et al., 2018; Mizrak et al., 2019; Xie et al., 2020; Coré et al., 2020; Cebrian-Silla et al., 2021). To distinguish the expression of Olig2 in NSCs and other cell types, we also performed scRNA-seq analysis of the early postnatal and adult SVZ. A thin strip of tissue around the LV was dissected out from coronal sections of P14 and adult brains, and dissociated single cells were subjected to scRNA-seq using the 10x Genomics platform. From two independent experiments for both stages, transcriptome data with a median of ∼3000 genes/cell were obtained from a total of 17,941 and 19,505 cells from the adult and P14 SVZs, respectively. Unsupervised cell clustering was performed by the Louvain algorithm together with dimension reduction using the Python implementation of uniform manifold approximation and projection (UMAP) using the top 18 significant principal components determined by JackStraw plot (Becht et al., 2018). Such clustering identified 17 and 14 distinct cell groups in the adult and P14 cells, respectively, based on their transcriptome signatures (Fig. 2A,B; Table S1). As expected, a large number of Olig2-expressing cells belonged to OL clusters (blue outlines in Fig. 2A,B), including oligodendrocyte progenitor cells (OPCs), pre-oligodendrocytes (Pre-OLs), and mature OLs at both stages (Fig. 2C,E,G,I). By contrast, fewer Gsx2-expressing cells were detected in these OL clusters (Fig. 2D,F,G,I). Moreover, numerous Olig2-expressing cells were detected in the NSC lineage clusters (red outlines in Fig. 2A,B), and their numbers were comparable to those of Gsx2-expressing cells (Fig. 2C-I). For example, 13.9% and 8.0% of cells in the adult NSC lineage expressed Olig2 and Gsx2 transcripts, respectively, whereas 17.1% and 15.9% were Olig2⁺ and Gsx2⁺, respectively, in the P14 SVZ (Fig. 2G,I). Furthermore, within the aNSC/TAP cell cluster (clusters 4 and 1 for adult and P14 SVZs, respectively), Olig2⁺ and Gsx2⁺ cells corresponded to 24.1% and 29.9% in adult animals and 37.2% and 34.1% in P14 animals, respectively. These results reiterate the aforementioned histological finding that both Olig2- and Gsx2-expressing cells are major populations among postnatal NSCs. We also identified the occurrence of Olig2-expressing NSCs in previously reported scRNA-seq datasets (Coré et al., 2020; Cebrian-Silla et al., 2021). Interestingly, approximately one-third of Olig2⁺ and Gsx2⁺ aNSCs/TAPs co-expressed Olig2 and Gsx2 transcripts (Fig. 2G,I). Such results appear to be contradictory to the aforementioned results that Olig2 and Gsx2 proteins are mostly expressed in distinct cells. However, further analysis demonstrated that cells expressing Olig2 transcripts at high levels contained Gsx2 transcripts mostly at a low level, whereas cells with many Gsx2 transcripts mainly expressed only a few Olig2 transcripts (Fig. 2H,J). Thus, cells with high copy numbers of Olig2 and Gsx2 transcripts were mostly mutually exclusive, similar to the protein expression results. These results suggest that the expression of Olig2 and Gsx2 proteins in NSCs is regulated by both transcriptional and post-transcriptional mechanisms, as demonstrated in previous studies for embryonic NSCs (Shibata et al., 2011; de Chevigny et al., 2012).

We next examined the fate of Olig2-expressing NSCs in the adult SVZ. We took advantage of the Cre/LoxP lineage-tracing strategy using Olig2-CreER mice, in which CreER cDNA was knocked into the endogenous Olig2 locus (Takebayashi et al., 2002). Adult mice carrying Olig2^CreER/+;Rosa-tdTomato^+/− or ^+/+ were treated with the CreER activator tamoxifen (Tx) once daily for 2 and 10 days and analyzed 3 and 56 days after the first Tx treatment, respectively (hereafter, D3 and D56) (Madisen et al., 2010). To detect adult-born neurons, D56 animals also received intraperitoneal injections of 5-bromo-2′-deoxyuridine (BrdU) twice a day for 5 days 3 weeks prior to histological analysis (from D35 to D39). In these animals, recombined Olig2-expressing NSCs and their progeny were detected as Rosa-tdTomato-expressing (tdTomato⁺) cells.

In Olig2-CreER mice at D3, all tdTomato⁺ cells detected in the SVZ co-expressed Olig2 as expected (62 cells examined in two animals). However, in D56 animals, only approximately a half of tdTomato⁺ cells expressed Olig2 in the SVZ (Fig. 3A,G). Thus, a significant fraction of the fate-mapped cells lost Olig2 expression between D3 and D56. Moreover, approximately 8% and 15% of tdTomato⁺ cells expressed Ascl1 and Ki67, respectively, at D56, indicating that tdTomato-labeled Olig2⁺ NSCs keep proliferating and generating aNSCs and TAPs during this 8-week period (Fig. 3B,C,G). In addition, many tdTomato⁺ cells also co-expressed Dlx2, Dcx and Pax6, which mark mainly late TAPs and NBs (Fig. 3D-G) (Kohwi et al., 2005; Brill et al., 2008; de Chevigny et al., 2012). Thus, tdTomato⁺ cells derived from Olig2⁺ cells include not only NSCs, but also TAPs and NBs, the two major progeny of NSCs in Olig2-CreER mice at D56.

We next examined the OB of Olig2-CreER mice. Only a few tdTomato⁺ cells were detected in D3 mice, and most of these were Olig2⁺ with typical OL morphology (among more than 60 cells examined). However, we found numerous tdTomato⁺ cells, corresponding to ∼4% of all 4′,6-diamidino-2-phenylindole (DAPI)⁺ cells (4.1±0.5%, n=3 animals), in the OB of Olig2-CreER mice at D56 (Fig. 3H-J). Olig2 is expressed in OLs and astrocytes but not in neurons in the adult OB (Wang et al., 2021). Consistent with this, ∼40% of tdTomato⁺ cells in the OB expressed Olig2 and Olig1 (Fig. 3K-K″,L-L″,P). These tdTomato⁺ OLs were either pre-existing resident OLs or those that were newly generated by NSCs in the SVZ and had subsequently migrated to the OB (Hack et al., 2005; Menn et al., 2006; Ortega et al., 2013). Importantly, however, we found that ∼60% of tdTomato⁺ cells in the OB were negative for Olig2 or Olig1 (Fig. 3P). These Olig2-negative tdTomato⁺ cells were most likely to be newborn neurons generated by Olig2⁺ NSCs. In fact, we found that ∼40% of tdTomato⁺ cells across multiple layers co-expressed Dlx2 and NeuN (Fig. 3M-N″,P). In particular, ∼60% of tdTomato⁺ cells in the granule cell layer (GCL) were Dlx2⁺ and NeuN⁺ neurons (Fig. 3P). By contrast, because many neurons in the glomerular layer (GL) and plexiform layer (PL: external and internal plexiform layers combined with the mitral cell layer in the middle) remain immuno-negative for Dlx2 or NeuN even at mature stages (Parrish-Aungst et al., 2007; Long et al., 2007), the percent overlaps between tdTomato and these neuronal markers in these layers were lower than in the GCL (Fig. 3P). Moreover, a small but significant fraction of tdTomato⁺ cells became BrdU⁺ 3 weeks after the animals received BrdU (Fig. 3O-O″,P). Together, these results demonstrate that Olig2-expressing NSCs at the adult stage generate new neurons that migrate to the OB.

We next examined the subtype identity of neurons generated by Olig2⁺ NSCs. After excluding Olig2⁺ OLs from the total tdTomato⁺ cells, ∼60% of Olig2-CreER fate-mapped cells belonged to the GCL, whereas 30% and 10% were detected in the GL and PL, respectively (Fig. 4A). Thus, Olig2⁺ NSCs generate new neurons in multiple layers of the OB. In the GL, neurons expressing calretinin (CR), neurocalcin (NC), calbindin (CB), tyrosine hydroxylase (TH) and parvalbumin (PV) represent discrete interneuron populations, except for a partial overlap of CR⁺ (∼15%) and PV⁺ (∼20%) cells with NC⁺ cells (Briñón et al., 1999; Parrish-Aungst et al., 2007). We found that significant fractions of tdTomato⁺ cells in the GL expressed CR, NC and TH, but no tdTomato⁺ cells were CB⁺ or PV⁺ (Fig. 4B-F,K). A small fraction of tdTomato⁺ cells in the PL and GCL also co-expressed CR (Fig. 4G,J,K). Given that adult-born OB neurons require a long time period (6-8 weeks) after their birth to express mature subtype markers fully (Kohwi et al., 2007), it is likely that more tdTomato⁺ cells would express subtype-specific markers if analyzed at later time points. In fact, ∼40% of tdTomato⁺ cells expressed the homeodomain TF Pax6 in the GL and GCL, and ∼25% express the MADS box TF Mef2c in the GCL (Fig. 4H,I,L) (Lyons et al., 1995; Hack et al., 2005; Kohwi et al., 2005; Brill et al., 2008). These results demonstrate that Olig2⁺ NSCs generate multiple neuronal subtypes across all layers in the adult OB.

The aforementioned results demonstrated that Olig2-expressing NSCs in the adult SVZ generate new OB neurons. Previous studies have also shown that Olig2⁺ NSCs/progenitors continuously produce new neurons and glia from early to late embryonic stages (Miyoshi et al., 2007; Ono et al., 2008). Thus, we next sought to interrogate how the molecular properties of Olig2⁺ NSCs/progenitors change over time by comparing their transcriptome profiles. We obtained scRNA-seq data on the whole ventral telencephalon at embryonic day (E) 12.5 and E18.5 using the 10x Genomics platform and compared these datasets with those from P14 and adult SVZ. First, cells that belonged to NSC/progenitor clusters (qNSC/astrocyte and aNSC/TAP) at each stage were identified based on their transcriptome profiles, and Olig2-expressing cells within these clusters were isolated (Fig. S1 and Table S1). Subsequently, these cells were combined and subjected to reclustering using UMAP plots. In these plots, the physical distance between individual cells reflects the extent of the similarity of their transcriptome profiles, unlike conventional tSNE analysis, so that one can visualize the overall similarities and differences between cells at different stages (Becht et al., 2018). Such an analysis revealed that Olig2⁺ NSCs/progenitors at different stages as a whole exhibited clearly distinct transcriptome profiles (Fig. 5A-C; Table S2) and were divided into partially overlapping, but clearly distinct, subclusters (Fig. 5D,E; Table S3). For example, Olig2⁺ NSCs/progenitors at E12.5 and E18.5 (green and blue dots, respectively, in Fig. 5A,B) were divided into four (0, 1, 3 and 5) and three (3, 4 and 7) distinct subclusters, respectively (compare Fig. 5D with Fig. 5A,B) (Haubensak et al., 2004; Noctor et al., 2004; Fish et al., 2008; Wilsch-Bräuninger et al., 2016). Among these subclusters, subcluster 3 was the only population shared between E12.5 and E18.5 (Fig. 5D). Transcriptome profiling further indicated that cells in subcluster 7 showed enriched expression of many genes related to glial progenitors and immature astrocytes, such as vimentin (Vim), solute carrier family 1 member 3 (Slc1a3 or GLAST), fatty acid-binding protein 7 (Fabp7 or Blbp) and aldehyde dehydrogenase 1 family member L1 (Aldh1l1) (Fig. 5E; Table S3). The occurrence of these unique cells at E18.5 is consistent with the fact that Olig2⁺ progenitors at late embryonic stages generate not only neurons, but also astrocytes and OLs (Miyoshi et al., 2007; Ono et al., 2008). Moreover, the majority of cells remaining at E18.5 formed subcluster 4, which showed enriched expression of genes related to basal progenitors, such as Ascl1, Dlx1/2, and Sox4, but their overall transcriptome profile was clearly distinguishable from the basal progenitor subcluster at E12.5 (subcluster 3) (Fig. 5D,E). Thus, the properties of Olig2⁺ progenitors at E18.5 are molecularly distinct from those at E12.5.

Clusters of postnatal Olig2⁺ cells also partially overlapped with E18.5 cells, but the majority belonged to discrete clusters with very different transcriptome profiles (Fig. 5A-C). The majority of P14 cells belonged to a single cluster (subcluster 2 in Fig. 5D), and their transcriptome signature was reminiscent of the properties of aNSCs and TAPs (Fig. 5E; Table S3). This gene expression signature of P14 cells reflects a higher neurogenic activity of NSCs at early postnatal stages than during the adult stage (Batista-Brito et al., 2008; Sakamoto et al., 2014). By contrast, a significant fraction of cells at the adult stage formed a unique cluster (subcluster 6 in Fig. 5D), characterized by the enriched expression of genes related to qNSCs and astrocytes, such as retinoid-related orphan receptor beta (Rorb), solute carrier family 1 member 2 (Slc1a2 or Eaat2) and gap junction protein alpha 1 (Gja1) (Fig. 5E; Table S3). The occurrence of this adult-specific subcluster is consistent with the fact that, while NSCs mature postnatally, their neurogenic activity gradually declines, and the majority become quiescent or divide only slowly (Götz et al., 2016).

The cell cycle profile also revealed distinct properties of Olig2⁺ NSCs/progenitors at different stages. Based on the relative expression levels of a set of cell cycle phase-specific genes (Table S4), individual cells could be divided into three groups: cells in the G0/G1, S and G2/M phases (Fig. 5F,G). This analysis demonstrates that more than 90% of Olig2⁺ NSCs/progenitors at E12.5 were either in the S or G2/M phase (Fig. 5G), and that the transcriptome profile of cells in the G0/G1 phase (red dots) more closely resembles cells in the S phase (blue dots) than those in the G2/M phase (green dots) (Fig. 5F). By contrast, among Olig2⁺ NSCs/progenitors at E18.5, there were fewer cells in the S phase, and ∼35% of cells were in the G0/G1 phase (Fig. 5G). These G0/G1 phase cells were mostly found in the cluster of glial progenitors/immature astrocytes (subcluster 7), reflecting active gliogenesis at late embryonic stages (compare Fig. 5D with Fig. 5F). Postnatally, cells in the G0/G1 phase further increased and reached 72% at the adult stage (Fig. 5G). Moreover, subcluster 2, which corresponds to aNSCs/TAPs in the P14 and adult SVZ, is a mixture of cells in all phases of the cell cycle, whereas subcluster 6, which contains adult qNSCs and astrocytes, mainly comprised cells in the G0/G1 phase (Fig. 5F). These results clearly demonstrate a stage-dependent stepwise decline in overall proliferative activity during the postnatal development of NSCs.

Comparisons of transcriptome profiles related to various metabolic and signaling pathways by ReactomeGSA also revealed significant stage-dependent changes (https://reactome.org; Griss et al., 2020). A gene set variation analysis (GSVA) of genes associated with specific metabolic and signaling pathways was performed using datasets from all four stages, and the relative activity of each pathway was scored and compared as described in a previous study (Hänzelmann et al., 2013) (for details, see Materials and Methods and Table S5). We found that the activity of many pathways increased over time from E12.5 to adult stages, but the timing of their increase varied significantly. For example, the activity of the signaling pathways for the neurotrophin receptors TrkA and TrkB was low in cells at E12.5 but showed gradual increases from E18.5 to the adult stage (blue font in Fig. 5H). By contrast, the activity of various lipid metabolism and neurotransmission/ion transport-related pathways (red and green fonts, respectively, in Fig. 5H) remained very low during embryogenesis, but those pathways became significantly active at P14 and reached higher levels at the adult stage. These changes probably reflect that NSCs/progenitors at postnatal stages gradually acquire responsiveness to various local and systemic environmental cues (Nakafuku and Grande, 2020). Altogether, these results demonstrate that the properties of Olig2⁺ NSCs/progenitors markedly change over time from early embryonic to adult stages.

In the developing telencephalon, both Olig2 and Gsx2 are expressed in broad progenitor domains, including the lateral ganglionic eminence (LGE), medial ganglionic eminence (MGE), preoptic area (POA) and septum (Toresson et al., 2000; Toresson and Campbell, 2001; Yun et al., 2001, 2003; Takebayashi et al., 2002; Petryniak et al., 2007; Wang et al., 2013; Chapman et al., 2013, 2018). In particular, Gsx2 expression is initially broad and uniform at early stages (E9.5-E11.5), but subsequently becomes refined to a high dorsal-low ventral gradient within the LGE from E12.5 onward (Stenman et al., 2003; Waclaw et al., 2009). Thus, the Gsx2-high and Olig2-high progenitor domains become mostly complementary to each other at late embryonic stages (Chapman et al., 2018). Such complementary patterns are reminiscent of their expression in the postnatal SVZ described above.

Importantly, our recent study demonstrated that the dorsal limit of Olig2 expression in the developing LGE expands in Gsx2 mutant embryos (Chapman et al., 2018). Thus, cross-repression between Olig2 and Gsx2 may play an important role in establishing and/or maintaining their mutual expression patterns. To address this possibility, we compared the expression of Olig2 and Gsx2 in both Gsx2 and Olig2 germline mutant embryos at E15.5. In control embryos, the ventral limit of Gsx2-high expression was confined to the dLGE (Fig. 6A, arrow), whereas the dorsal limit of Olig2 expression remained in the ventral half of the vLGE (Fig. 6A, arrowhead). By contrast, Gsx2-high expression expanded far ventrally and reached the MGE in Olig2 knockout (KO) embryos (Fig. 6B, arrow). Conversely, in Gsx2 KO embryos, the dorsal limit of Olig2 expression expanded dorsally and reached the ventral limit of the dLGE (Fig. 6C, arrowhead), where both the dLGE and vLGE were markedly reduced in size as a result of Gsx2 loss (compare dashed lines in Fig. 6A with those in Fig. 6C) (Toresson et al., 2000; Yun et al., 2001). These results demonstrate that Gsx2 limits the expression of Olig2 dorsally, whereas Olig2 limits the high expression of Gsx2 ventrally in embryonic NSCs/progenitors. Whether such mutual cross-repression is attributable to direct transcriptional repression remains to be investigated.

Given that these germline mutant animals die at birth (Toresson et al., 2000; Dessaud et al., 2007), we next used the CreER-loxP system for the conditional inactivation of these genes in adult NSCs. We generated mice homozygous for a floxed allele of Olig2 (Olig2^flox) (Yue et al., 2006; Cai et al., 2007) and heterozygous for a GLAST-CreER knock-in allele (Mori et al., 2006). The latter contained a CreER transgene expressed from the locus of the glutamate transporter gene GLAST so that we were able to evaluate the impact of Olig2 inactivation in the most primitive/undifferentiated GLAST⁺ NSCs. Olig2 conditional knockout (cKO) (Olig2^flox/flox;GLAST-CreER^+/−;Rosa-tdTomato^+/− or^+/+) and their littermate control (Olig2^flox/+ or ^+/+;GLAST-CreER^+/−;Rosa-tdTomato^+/− or^+/+) mice were treated with Tx at the age of 10-12 weeks, and subsequently analyzed at D28. As expected, the percentage of Olig2⁺ cells among tdTomato⁺ recombined cells decreased by more than 90% in the vSVZ of Olig2 cKO mice compared with control animals (Fig. 6D,D′,G). However, in these mice, the dorsally restricted expression pattern of Gsx2 did not significantly change, and the percentages of Gsx2⁺ cells among tdTomato⁺ cells in the vSVZ remained very low (Fig. 6G). Two other homeodomain TFs, Nkx2.1 and Gsx1, are also expressed in a subset of NSCs in the vSVZ (López-Juárez et al., 2013; Merkle et al., 2014; Paul et al., 2017). The ventrally enriched expression of these TFs among tdTomato⁺ cells did not significantly change either (Fig. 6E-G). Thus, the inactivation of Olig2 does not alter the region-specific expression patterns of either Gsx2 or other NSC identity genes. We also previously reported that Olig2 expression in the dlSVZ is not significantly affected by conditional inactivation of Gsx2 in adult NSCs (López-Juárez et al., 2013). Thus, unlike in embryos, cross-repression between Olig2 and Gsx2 does not play a major role in maintaining their region-selective expression patterns in NSCs in the adult stage.

The aforementioned immunohistochemical and scRNA-seq analyses demonstrated that Olig2⁺ cells are a predominant (>50%) NSC population in the vSVZ of the adult brain (Fig. 1M and Fig. 2G). Thus, GLAST-CreER-based Olig2 cKO mice allowed us to examine the role of Olig2 in neuronal differentiation of NSCs in the vSVZ. In fact, only a small fraction of tdTomato⁺ cells expressed Olig2⁺, whereas many tdTomato⁺ cells became Dcx⁺ and Pax6⁺ NBs in these Olig2 cKO mice at D28 (Fig. 6G and Fig. 7D-F). Interestingly, the percentages of Ascl1⁺ aNSCs/TAPs and Ki67⁺ proliferative cells among tdTomato⁺ cells were modestly but significantly increased at D28 compared with control animals (1.4- and 1.2-fold for Ascl1 and Ki67, respectively), suggesting that activation and/or proliferation of NSCs/TAPs are moderately stimulated by the loss of Olig2 (Fig. 7B,B′,C,C′,F). Nevertheless, Dcx⁺/tdTomato⁺ and Pax6⁺/tdTomato⁺ NBs were detected at comparable levels in Olig2 cKO and control mice, demonstrating that Olig2 inactivation does not significantly affect neuronal output from adult NSCs (Fig. 7D-F). Thus, certain feedback mechanisms may operate to maintain a steady level of NB production despite a slight increase in the activation and/or proliferation of NSCs/TAPs. Moreover, Gfap⁺/tdTomato⁺ cells also remained unchanged between Olig2 cKO and control mice, suggesting that Gfap⁺ stem cells are maintained in Olig2 cKO mice (Fig. 7A,A′,F). Together, these results demonstrate that Olig2 is dispensable for neurogenesis in adult NSCs.

In these GLAST-CreER-based Olig2 cKO mice, gene recombination and subsequent lineage tracing occurred not only in Olig2⁺ NSCs, but also in other NSCs, such that tdTomato⁺ fate-mapped cells derive from heterogeneous NSC populations. Therefore, they are not suitable for examining the impact of Olig2 inactivation on the specificity of neuronal subtypes. For this purpose, we next used Olig2^CreER/flox;Rosa-tdTomato^+/− or ^+/+ (Olig2-CreER cKO) mice, in which Olig2 is inactivated selectively in Olig2⁺ NSCs. As expected, the percentage of Olig2⁺ cells among tdTomato⁺ cells was markedly reduced (by more than 95%) in the GL of the OB in Olig2-CreER cKO mice compared with control at D56 (Fig. 7G). In parallel, Olig1⁺ OLs were markedly reduced among tdTomato⁺ cells and, conversely, Dlx2⁺ neurons were dramatically increased and comprised more than 70% of all tdTomato⁺ cells in the mutant OB (Fig. 7G). Thus, inactivation of Olig2 resulted in a loss of OLs among tdTomato⁺ cells, and the majority of tdTomato⁺ cells remaining became neurons in the mutant GL (97.8±2.2% in Olig2-CreER cKO mice compared with 51.3±18.2% in control mice, n=3). This loss of OLs could be because Olig2 is required for the survival of existing OLs and/or production of new OLs from NSCs.

Importantly, among these remaining tdTomato⁺ cells in Olig2-CreER cKO mice, the percentage of CR⁺ neurons was reduced by half compared with control animals (Fig. 7H). Furthermore, given that the proportion of new neurons among tdTomato⁺ cells was doubled in Olig2 mutants compared with control animals, the percentage of CR⁺ neurons among new neurons needs to be normalized based on the percentage of Olig2-negative cells among all tdTomato⁺ cells. Comparison of these normalized values demonstrated that the actual extent of the reduction in CR⁺ neurons among Olig2⁺ NSC-derived new neurons (Olig2-negative tdTomato⁺ cells) was even more drastic (approximately a quarter of the control level) (Fig. 7I). By contrast, although the percentages of NC⁺ and TH⁺ cells among all tdTomato⁺ cells were increased in the mutants (Fig. 7H), their normalized fractions among new neurons were comparable between control and mutant mice (Fig. 7I). Given that Olig2 is not expressed in either Dcx⁺ immature neurons in the SVZ or mature neurons in the OB, it is unlikely that the selective loss of CR⁺ neurons in the mutant OB results from the defective survival of CR⁺ neurons. Rather, these results support the idea that Olig2 expression in NSCs is selectively required for the specification of CR⁺ neurons in the GL.

Our data showed that Olig2-expressing NSCs reside widely along the DV and ML axes of the postnatal SVZ, but that they are most enriched in the ventral subdomain of the SVZ (Fig. 1). Importantly, this ventrally biased distribution pattern of Olig2⁺ cells is largely complementary to that of dorsally enriched Gsx2-expressing cells in the adult SVZ (López-Juárez et al., 2013). Moreover, Olig2⁺ and Gsx2⁺ NSCs were mostly mutually exclusive (Fig. 1). scRNA-seq data also demonstrated that, when combined, Olig2⁺ and Gsx2⁺ cells comprised 46% of all aNSCs/TAPs in the adult SVZ and 59% at P14, demonstrating that they are major NSC subtypes in the postnatal brain (Fig. 2). The homeodomain TFs Gsx1 and Nkx2.1, and the zinc finger TF Zic1, are also expressed in a subset of NSCs in the vSVZ and mSVZ, respectively (López-Juárez et al., 2013; Merkle et al., 2014; Paul et al., 2017). However, our data showed that inactivation of Olig2 does not significantly change the expression of these TFs in the vSVZ, Thus, the relationships of these NSCs with Olig2⁺ and Gsx2⁺ NSCs are currently unknown and remain to be further investigated.

The region-selective expression of Olig2 and Gsx2 defines a similar topological organization of NSC/progenitor domains in embryos and adults. Nevertheless, we found a clear difference in the relationship between Olig2 and Gsx2 in embryonic and adult NSCs. In embryos, genetic inactivation of Olig2 led to a marked ventral expansion of the Gsx2-high region, whereas inactivation of Gsx2 resulted in expansion of the dorsal limit of Olig2 expression (Fig. 6). Thus, the border between Gsx2-high and Olig2⁺ domains is established and/or maintained by cross-repression during embryogenesis. By contrast, conditional inactivation of Olig2 and Gsx2 in adult NSCs did not significantly alter the expression of the other gene (this study and López-Juárez et al., 2013). It could be that, once the mutually exclusive expression patterns of Olig2 and Gsx2 are established during embryogenesis, their expression is subsequently maintained independently by certain cell-autonomous mechanisms in the postnatal brain.

Our results demonstrated that Olig2⁺ adult NSCs contribute to the generation of both CR⁺ and TH⁺ neurons in the GL, which are mutually exclusive and the predominant neuronal subtypes among adult-born GL neurons (Lemasson et al., 2005; Kohwi et al., 2007; Batista-Brito et al., 2008; Sakamoto et al., 2014). Our data showed that Olig2⁺ NSCs were present in all SVZ subdomains, although they were most enriched in the vSVZ. Thus, it is possible that Olig2⁺ NSCs are heterogeneous and that those in distinct subdomains may be responsible for the generation of distinct neuronal subtypes. Interestingly, it has been proposed that CR⁺, TH⁺ and CB⁺ GL neurons derive predominantly from anterior/medial, carousal/dorsal, and ventral subpopulations of NSCs, respectively (Merkle et al., 2007, 2014). However, experimental data reported in these studies indicate that the correlation between these neuronal subtypes and the topological locations of NSCs is not a strict match but reflects an overall trend only. In fact, the current study, together with our previous report (López-Juárez et al., 2013), demonstrates that both Olig2⁺ and Gsx2⁺ NSCs generate CR⁺ neurons in the GL despite the fact that they are mutually exclusive stem cell pools in the postnatal SVZ. Moreover, although Olig2⁺ NSCs are abundant in the vSVZ, our fate-mapping study did not detect the generation of CB⁺ GL neurons by these ventral Olig2⁺ NSCs. Thus, not all NSCs in the vSVZ are responsible for the generation of CB⁺ neurons. As such, the notion that particular subtypes of OB neurons marked by existing markers are generated solely by a specific group of NSCs that exist in particular SVZ subdomains only is not supported by available data and certainly an oversimplification. Rather, our study and existing data support the idea that neurons expressing the same known molecular markers are likely to be produced by heterogeneous NSC populations that may exist in multiple SVZ subdomains.

Similarities and differences between NSCs/progenitors in the developing and adult brain have been extensively studied (Fuentealba et al., 2015; Götz et al., 2016). However, a difficulty in such comparisons is that NSCs/progenitors are highly heterogeneous in different regions of the embryonic and postnatal brains, and such heterogeneity changes over time. scRNA-seq analysis of Olig2⁺ NSCs/progenitors at different stages provides us with a unique opportunity to compare equivalent NSC populations at distinct stages at the molecular level. We found that, although Olig2⁺ NSCs/progenitors at each developmental stage intrinsically comprise heterogeneous cell populations, cells belonging to each stage as a whole form a unique cell cluster compared with cells at other developmental stages (Fig. 5). Our analysis further revealed that changes in cell cycle and metabolic/signaling pathways contribute significantly to such stepwise developmental transitions. For example, the fraction of cells in the G0/G1 phase markedly increased from 7% to 35% between E12.5 and E18.5 and was further elevated at the P14 (41%) and adult (72%) stages. Such stepwise increases in cells in the G0/G1 phase were accompanied by gradual decreases of cells in both the S and G2/M phases. Molecular pathway analysis also revealed that the activity of the signaling pathways for NTF receptors was initially low at E12.5 but gradually increased from E18.5 to the adult stage, whereas a group of lipid metabolism and neurotransmission/ion transport-related pathways did not become active until NSCs reached the adult stage. Further investigations of these scRNA-seq datasets will help us to better understand the molecular mechanisms underlying the developmental transitions of NSCs/progenitors from early embryonic to adult stages.

All animal procedures were performed according to the guidelines of the National Institute of Health and the Institutional Animal Care and Use Committee. Embryos were collected at E12.5, E15.5 and E18.5 (the day on which the copulatory plug was found was considered to be E0.5). Brains of P14 and P16 pups and adult animals from both females and males at the age of 10-12 weeks were collected after cardiac perfusion with 4% (v/v) paraformaldehyde (PFA) in PBS. Details of the mouse lines and their handling are described in the supplementary Materials and Methods.

Cryosections (12 μm) of embryos and vibratome sections (30 μm) of postnatal and adult brains were used for immunostaining. The method used to identify specific subdomains of the SVZ and the antibodies used for immunostaining are described in the supplementary Materials and Methods. Fluorescent images were captured using a Nikon A1R si laser scanning confocal microscope. Quantitative results are expressed as the mean±standard deviation (s.d.) of data obtained from three to four animals for each genotype. All statistical analyses were performed using R software (version R 3.5.0: https://www.r-project.org/) (for details, see the supplementary Materials and Methods).

scRNA-seq data on E12.5 ventral telencephalon have been described previously (GSE42768: Roychoudhury et al., 2020), and the other three datasets on E18.5, P14 and adult cells generated in this study are available from GEO under the accession number GSE174392. Details of the generation and analysis of scRNA-seq data are described in the supplementary Materials and Methods.

We are grateful to Magdalena Götz, Jane Johnson, Hirohide Takebayashi and Kazuaki Yoshikawa for reagents, and to Matthew Kofron, Evan Meyer, Sarah McLeod, Diane Girard, Jordan Burge and Amber Niederhelman for technical support.

The peer review history is available online at https://journals.biologists.com/dev/article-lookup/doi/10.1242/dev.200028.