The telencephalon is patterned into two major subdivisions, the dorsal and ventral regions, called the pallium and subpallium, respectively (Puelles et al., 2000). NSPCs in the pallium produce excitatory neurons, whereas those in the subpallium produce interneurons (Anderson et al., 1997; Gorski et al., 2002). For proper patterning along the dorsoventral (DV) axis, it is essential that several transcription factors work in concert. For example, the loss of Pax6, a transcription factor expressed in the dorsal region, shows severe ventralization with a dorsal expansion of the expression of Gsx2, Ascl1, and Dlx1, which are involved in the production of interneurons from the ventral region; thus, Pax6 is crucial for patterning the dorsal and ventral telencephalon (Stoykova et al., 2000; Toresson et al., 2000). In particular, the positioning of the DV boundary region is defined by Pax6-Gsx2 mutual antagonism (Corbin et al., 2000; Toresson et al., 2000; Yun et al., 2001).

DmrtAs are also crucial for the determination of the DV axis. Recent Dmrt3 and Dmrta2 knock-out (KO) studies have clarified how they specify DV regional identity in progenitors as follows. Dmrt3;Dmrta2 double-KO (dKO) mice showed the most severe phenotype of DV regionalization of the telencephalon than Dmrta2 single-KO (sKO) mice, and Dmrt3 sKO mice showed the mildest phenotype (Desmaris et al., 2018; Konno et al., 2019). It is noteworthy that the expression of subpallium-specific markers (e.g., Gsx2, Dlx2, Ascl1) is shifted dorsally to the lateral part of the dorsal telencephalon in Dmrt3;Dmrta2 dKO mice (Desmaris et al., 2018). Furthermore, although Dmrta1 sKO mice did not have a phenotype of ventralization, the overexpression of Dmrta1 in the ventral telencephalon in rats induced the expression of the dorsal marker Neurog2 and repressed that of the ventral marker Ascl1 (Kikkawa et al., 2013). These results indicate that DmrtAs function as an essential determinant of progenitor cell DV identity by repressing ventralization.

Therefore, a question arises: what are the main target genes of DmrtA transcription factors that determine telencephalic identity? The enhancer activity of the Gsx2 locus, to which Dmrta2 binds, is found in the ventral telencephalon (Desmaris et al., 2018). A more comprehensive method, whole-genome chromatin immunoprecipitation-sequencing (ChIP-seq), revealed that Dmrt3 and Dmrta2 bind to the Gsx2 and Pax6 loci within ±100 kb from the transcription start site (Konno et al., 2019). The authors also generated transgenic mice to visualize the enhancer activity of Dmrt3/Dmrta2-binding sites at Gsx2 and Pax6. They found that the enhancer activities of the loci at Gsx2 and Pax6 bound by Dmrts are detected in the dorsal lateral ganglionic eminence (dLGE) and the dorsal telencephalon, where Gsx2 and Pax6 are expressed, respectively. Interestingly, a suppressive histone moiety, H3K27me3, is found in the Gsx2 enhancer bound by Dmrt3/Dmrta2 in their study, suggesting that DmrtAs may suppress the expression of Gsx2. Since the dLGE neighboring the DV boundary expresses a high level of Gsx2, its direct regulation may modulate the fate specification of the progenitors in the dorsal and ventral telencephalon.

Pax6 regulates the expression of various genes in the embryonic telencephalon (Sansom et al., 2009; Kikkawa et al., 2013, 2019; Walcher et al., 2013; Xie et al., 2013; Sun et al., 2015). We initially searched for downstream genes of Pax6 by comparing the transcriptomic profiles of telencephalic samples from wild-type and Pax6 null mutant rat embryos and found Dmrta1 as a novel target (Kikkawa et al., 2013) in addition to the various downstream molecules in the embryonic CNS (Shimoda et al., 2002; Arai et al., 2005; Numayama-Tsuruta et al., 2010; Shinohara et al., 2013; Inada et al., 2018). Pax6 ChIP-seq has later proved the direct regulation of the Dmrta1 gene by the Pax6 transcription factor in the E12.5 mouse forebrain (Sun et al., 2015). However, this Pax6-Dmrta1 pathway seems to work only in specific dorsal-lateral telencephalic regions. Dmrta1 expression was markedly decreased in the neocortex of Pax6 homozygous mutant rat embryos, although it remained in the CH (Kikkawa et al., 2013). Focusing on the functions in the formation of the CH, Dmrta1 supports the generation of the CH (Kikkawa et al., 2020), whereas Pax6 suppresses its fate specification of the CH (Godbole et al., 2017). Therefore, the Pax6-Dmrta1 pathway may work differently in the neocortex and extra-neocortical areas, such as the CH.

The dorsal telencephalon is divided into the neocortex, hippocampus, and dorsal midline, giving rise to the CH and choroid plexus. The dorsal midline of the telencephalon is known to express BMPs and Wnts (Furuta et al., 1997; Grove et al., 1998). BMP and Wnt signaling are required for the formation of the medial telencephalic structure that develops into the hippocampus and the dorsal midline region (Lee et al., 2000; Hebert et al., 2002), and both signaling pathways also regulate the expression of transcription factors such as Emx2 and Lhx2, which specify and expand the medial and dorsal parts of the telencephalon (Monuki et al., 2001; Theil et al., 2002). These molecular pathways are essential for patterning the dorsal telencephalon along the mediolateral axis.

In addition to these critical molecules that regulate the development of the medial telencephalon, it has become clear that DmrtAs contribute to the following events. Previous studies have demonstrated malformation of caudomedial telencephalic structures, that is, the CH that expresses Wnt3a and Bmp4, and the hippocampus, in Dmrta2 sKO mice (Konno et al., 2012; Saulnier et al., 2013; De Clercq et al., 2018). Dmrt3 sKO embryos also have shown defects in the CH formation (De Clercq et al., 2018; Kikkawa et al., 2020). These phenotypes are reasonable because Dmrt3 and Dmrta2 are expressed with medial-high/lateral-low and rostral-low/caudal-high gradients. Although the expression level of Dmrta1 was weak in the medial telencephalon, Dmrt3;Dmrta1 dKO mice exhibited more severe defects in medial structures compared to Dmrt3 sKO mice (Kikkawa et al., 2020). Therefore, Dmrta1 may have an additional support function for patterning medial telencephalic structures in cooperation with Dmrt3.

When do the defects in the medial telencephalon of Dmrt mutants begin? It has been reported that DmrtAs are expressed in the forebrain at early developmental stages, that is, E10.5–E12.5 (Konno et al., 2012; Kikkawa et al., 2013). Because Wnt3a^Cre-driven Dmrta2 conditional KO (cKO) mice showed normal morphology of the medial telencephalon (De Clercq et al., 2018), the apparent defects in the medial telencephalon by the ablation of DmrtAs may be due to the loss of their expression during earlier developmental stages before the initiation of recombination by Wnt3a-Cre at E10 in mice (Yoshida et al., 2006). These findings suggest that DmrtAs establish the dorsal midline structure, a signaling center that expresses Wnt and BMP in patterning the cerebral cortex.

Then, what molecules function upstream of DmrtAs to determine the dorsomedial patterning of the telencephalon? Dmrt gene expression was downregulated in Gli mutant mice, in which Wnt expression in the forebrain was severely affected (Hasenpusch-Theil et al., 2012). This study also clarifies that Dmrt3 is a direct Wnt target gene in the dorsomedial telencephalon by DNA binding and reporter gene assays. Furthermore, the expression of Dmrt3 and Dmrta2 were decreased by the overexpression of a dominant-negative form of Tcf3 and increased by the overexpression of a constitutively active form of β-catenin (Konno et al., 2012). Moreover, Dmrta2 expression was induced in organotypic slice cultures of mouse embryonic telencephalons treated with Chir that selectively inhibits GSK3β and activates Wnt signaling (Saulnier et al., 2013). This means that the secretory factor Wnts can directly regulate Dmrt expression and could contribute to adjusting the amount of Dmrts.

The neocortex has “primary” areas: the primary motor (M1, controls the voluntary movement of body parts), somatosensory (S1, processes the information received from the body), and visual (V1, processes the information received from eyes) areas. One of the hypotheses to determine the cortical arealization is the “protomap model.” The feature of progenitors in early development is predetermined by the combination of molecules, and then the progenitors differentiate into specific neurons, leading to the formation of different cortical regions (Rakic, 1988). Some transcription factors that contribute to the formation of the “protomap.” Rostral^high-Pax6 and caudal^high-Emx2 in progenitors preferentially impart the identities of the rostrolateral and caudomedial areas, respectively (Bishop et al., 2002). A study on the loss of Pax6/Emx2 indicates that they suppress each other’s expression (Muzio et al., 2002). Lhx2 is expressed in a graded manner (caudal^high/medial^high) in the telencephalon (Nakagawa et al., 1999; Monuki et al., 2001), acts as an essential determinant of cortical identity (Mangale et al., 2008), and is further required for the neocortical-paleocortical subdivision (Chou et al., 2009). Pax6 expression is decreased in the Lhx2 mutant in the dorsal telencephalon due to direct regulation by Lhx2 (Hou et al., 2013; Shetty et al., 2013). The combinatorial expression patterns of these transcription factors regulate the regionalization of the cerebral cortex.

Since the loss of CH affects cortical size and patterning (Caronia-Brown et al., 2014), the reduction in cortical size in Dmrt mutants may be due to the developmental impairment of CH formation. However, deleting Dmrta2 after CH formation still decreases the cortical size and changes the area map; the V1 area, which is the caudomedial neocortical region, was reduced in Emx1^Cre- and Nestin^Cre-driven Dmrta2 cKO mice on postnatal day 7 (De Clercq et al., 2018). Conversely, Dmrta2^Tg/+; Emx1^Cre mice with excess Dmrta2 in the cortical primordium showed enlarged V1 and reduced size of the S1 and M1 areas. They also revealed a decrease in Lhx2 and Emx2 and the expansion of Pax6 to the caudal cortex in Dmrta2 cKO embryos. From these results, the level of Dmrta2 seems to regulate the expression of these transcription factors that determine the rostral-caudal patterning during cortical development and construct the proper neocortical area map.

NSPCs maintain their population by self-renewal and produce neurons by differentiation during embryogenesis. At an early stage of development, that is, E9–E11 in the mouse telencephalon, NSPCs undergo symmetric division in the VZ, producing daughter cells with similar fates to give rise to more progenitors. As development proceeds, NSPCs start to divide asymmetrically, producing one apical progenitor (AP) positive for Pax6 with self-renewing capability and one differentiated neuron, or one intermediate progenitor (IP) positive for Tbr2, which divides symmetrically in the subventricular zone (SVZ) and generates a pair of IPs or neurons (Gotz and Huttner, 2005; Huttner and Kosodo, 2005). Subsequently, newly born neurons migrate radially from the VZ/SVZ to the upper area of the telencephalon. In this section, we introduce the multiple roles of DmrtAs expressed in NSPCs in the telencephalon in cell proliferation and differentiation regulation.

Dmrta2 is expressed in NSPCs derived from mouse embryonic stem cells (ESCs) (Young et al., 2017). It maintains their proliferation by positively regulating Hes1 expression via Dmrta2 bound to the locus of Hes1 (Young et al., 2017), a significant Notch target gene that inhibits neuronal differentiation through negative regulation of proneural genes (Kageyama et al., 2020). Consistently, the neurogenic gene Neurog2 was upregulated in the medial region of the dorsal telencephalon of Dmrta2 sKO mice (Saulnier et al., 2013). Furthermore, in Dmrt3, transgenic mice overexpressing Dmrt3 in NSPCs under the control of the Nestin enhancer showed a planar expansion of the ventricular surface, indicating an increased NSPC pool and decreased Tbr2-positive IP cells (Konno et al., 2019). These results suggest that Dmrt3 and Dmrta2 are involved in the maintenance of NSPCs in the dorsal telencephalon.

In contrast, DmrtAs seem to contribute to neuronal differentiation. A zebrafish mutant of the ha2 locus, encoding Dmrta2, reduced the expression of neurog1 and impaired telencephalic neurogenesis (Yoshizawa et al., 2011). In addition to the brain, Xenopus Dmrta1 and Dmrta2 promoted neurogenesis in the olfactory placode (Huang et al., 2005; Parlier et al., 2013). Although Dmrt3 and Dmrta2 are involved in the maintenance of NSPCs, as mentioned above, the ratio of Tbr2⁺ to the total number of progenitors was lower, while that of Pax6⁺ cells was higher, in Dmrta2 sKO mice than in wild type mice in the lateral cortex (Ratie et al., 2020). This suggests that the APs could generate fewer IPs and/or that the timing of their differentiation could be delayed. Our data showed that the knockdown of Dmrta1 or Dmrt3 in the rat dorsal telencephalon reduces Neurog2, while the overexpression of Dmrta1 in the ventral telencephalon induces ectopic Neurog2 expression in the restricted lateral cortex and near the DV boundary region (Kikkawa et al., 2013). Based on the above reports, the function of DmrtAs in neurogenesis is not simple, and its function may be altered in a region-dependent manner within the telencephalon.

The sequential production of neurons and glia from NSPCs is a critical event during CNS development. In the mammalian neocortex, NSPCs at the early developmental stage generate neuronal cells, whereas those at the late stage generate mainly glial cells, including astrocytes and oligodendrocytes (Qian et al., 2000). The appropriate neurogenic-to-gliogenic switch in NSPCs is essential for the production of proper numbers of neurons and glia. Various molecular mechanisms have been clarified; for example, BMP, Fgf, and Notch signaling induce the cell fate switch from neuronal to glial cells (Miller and Gauthier, 2007). The transcription factor Lhx2 suppresses astrogliogenesis and promotes neurogenesis in the developing hippocampus, but not in the neocortex, indicating the spatial-specific regulation of NSPCs to neuron/glia fate specification (Subramanian et al., 2011).

There is an exciting finding that Dmrta2 is involved in the cell fate switch from neurons to astrocytes. The loss of Dmrta2 increased the population of GFAP-positive cultured astrocytes taken from the embryonic hippocampus (Muralidharan et al., 2017). They also showed that the loss of Lhx2 also induces astrogliogenesis, and the phenotype is rescued by the overexpression of Dmrta2. These results suggest that Dmrta2 and Lhx2 reciprocally regulate each other and that this pathway is involved in the neuron-glia cell-fate switch, emphasizing the novel role of Dmrta2 as a neurogenic factor. It is reasonable to assume that this phenotype is limited to the hippocampus, but not the cortex, because both Lhx2 and Dmrta2 are strongly expressed in the medial region and have already been reported to be involved in hippocampal development (Bulchand et al., 2001; Konno et al., 2012; Saulnier et al., 2013; De Clercq et al., 2018).

In the Lhx2-Dmrta2 pathway, ChIP-quantitative PCR using the embryonic hippocampus showed that Lhx2 binds to the Dmrta2 locus, indicating that Lhx2 could directly regulate the expression of Dmrta2 (Muralidharan et al., 2017). Interestingly, the locus bound by Dmrta2 is evolutionarily highly conserved among species. If this mechanism of the neuron-glia cell fate switch by the Lhx-Dmrta2 pathway could be conserved among different organisms, it may provide a deeper insight into the poorly understood neuron-glia switch in other species.

It is unclear which neuronal subtypes are regulated by DmrtAs. The expression of DmrtAs is abundant in developmental stages when early born neurons called Cajal-Retzius (CR) cells are produced (Konno et al., 2012; Kikkawa et al., 2013). CR cells are among the first neurons to be generated (between E9.5 and E13.5) in mice (Hevner et al., 2003; Takiguchi-Hayashi et al., 2004). CR cells are generated from specific extra-neocortical regions, that is, (1) the pallial septum, which is located adjacent to a rostral signaling center in the rostromedial telencephalon; (2) the boundary between the pallium and subpallium (pallial-subpallial boundary; PSB); and (3) the CH in the caudomedial telencephalon, and later migrate tangentially over long distances from their original production sites, gradually covering the surface of the cortex (Meyer et al., 2002; Takiguchi-Hayashi et al., 2004; Bielle et al., 2005; Yoshida et al., 2006; Garcia-Moreno et al., 2007; Imayoshi et al., 2008; Gu et al., 2009; Tissir et al., 2009). CR cells appear to play critical roles in the radial migration of subsequently born cortical neurons and eventually in the laminar organization of the cortex (D’Arcangelo et al., 1995; Ogawa et al., 1995; Super et al., 2000).

We revealed the reduced production of CH-derived CR cells in Dmrt3 sKO mice, especially in Dmrta1 and Dmrt3 dKO mice (Kikkawa et al., 2020). The reduction in CR cell production is consistent with the developmental impairment of the CH from which CR cells are produced. Dmrta2 sKO mice also have defects in the CR cells (Saulnier et al., 2013). These results suggest that DmrtAs cooperatively maintain the appropriate number of CR cells derived from specific source regions by defining the patterning of the medial telencephalon. For CR cell production, one of the candidate molecules working upstream of Dmrt3 and Dmrta1 may be forkhead box G1 (Foxg1). Foxg1 cKO mice (Foxg1^tetOFoxg1 line, repressing Foxg1 transgene expression in the presence of doxycycline) showed an upregulated expression of Dmrt3 and Dmrta1 in the dorsal telencephalon (Kumamoto et al., 2013). ChIP-seq for Foxg1 revealed direct binding to the intronic sequences of Dmrt3 upstream and Dmrta1 downstream. Thus, the pathway by which Foxg1 represses Dmrt3 and Dmrta1 expression may critically affect telencephalic regionalization. This hypothesis is consistent with the converse phenotypes in the formation of CH in Dmrt mutants versus Foxg1 KO mice, showing region expansion (Dou et al., 1999; Martynoga et al., 2005; Hanashima et al., 2007).

Another early born neurons are subplate (SP) neurons generated before the excitatory neurons in the cortex; that is, between E10.5 and E12.5, in the mouse (Price et al., 1997). Glutamatergic SP neurons are derived from the cortex and rostral medial telencephalic wall (Yoshida et al., 1997; Shinozaki et al., 2002; Garcia-Moreno et al., 2008; Pedraza et al., 2014). SP neurons contribute to the establishment of the initial neural circuits between the cortex and thalamus during neocortical development (McConnell et al., 1989; Ghosh et al., 1990). SP neurons are also critical for inducing the multipolar-to-bipolar transition, which leads to a faster migration mode (Ohtaka-Maruyama et al., 2018).

A recent study has demonstrated that the production of SP neurons is significantly decreased by the loss of Dmrt3 and Dmrta2 (Ratie et al., 2020). To clarify the time window of the effect of Dmrta2 in SP neuron formation, they disrupted Dmrta2 in cortical progenitors from E10.5 in Dmrta2^Lox/Lox;Emx1^Cre mice and from E11.5 in Dmrta2^Lox/Lox;Nestin^Cre mice. The SP neurogenesis was only developmentally delayed in Dmrta2^Lox/Lox;Emx1^Cre mice and no phenotypic change was observed in Dmrta2^Lox/Lox;Nestin^Cre mice. Thus, the continuous expression of Dmrta2 from an earlier stage is likely necessary for the production and specification of SP neurons. Interestingly, the Dmrta2 protein is continuously detected in SP neurons, where Dmrta2 may regulate cortical neuron migration (Ratie et al., 2020).