The telencephalic subpallium is synonymous with embryonic basal ganglia (also termed ganglionic eminences [GEs]). These are divided into five regions with distinct dorso-ventral and rostro-caudal positions: medial ganglionic eminence (MGE), caudal ganglionic eminence (CGE), lateral ganglionic eminence (LGE), septum, and the pre-optic area (POA). The GEs generate distinct nuclei and cell types (Puelles et al., 2016; Rubenstein & Campbell, 2020).

Progenitor cells in these domains give rise to projection neurons and INs, which are largely GABAergic, although smaller numbers of cholinergic and dopaminergic neurons are also generated. Over development these progenitor domains become progressively gliogenic; by birth in rodents they become largely non-neurogenic ependymal zones. However, neurogenesis within these domains may continue into neonatal stages in primate and human (Paredes et al., 2016). In this review, we will mainly focus on the findings from rodent studies.

There is evidence that the MGE, CGE, LGE, POA, and septal progenitor domains are further subdivided into spatial regions that generate distinct cell types. These domains have distinct gene regulatory element (enhancer) activation patterns and express distinct TF combinations and signaling molecules (e.g., sonic hedgehog [SHH]) (Flames et al., 2007; Silberberg et al., 2016; Hu et al., 2017a; Su-Feher et al., 2021).

We expect that further studies will reinforce this concept and provide further granularity into the spatio-temporal mechanisms that contribute to cell fate specification from the subpallial progenitor subdomains.

The ventriculo-pial (mediolateral) axis of the central nervous system (CNS) is the axis of differentiation. Adjacent to the lateral ventricle lies the highly proliferative neuroepithelium, which consists of a pseudostratified layer of subpallial stem cells generally called the ventricular zone (VZ). Around embryonic day (E)9.5, these transition from neuroepithelial cells to radial glia stem cells (apical progenitors); the latter generate intermediate progenitors (basal progenitors), neurons, and glia. Superficial to the VZ are the intermediate progenitors that are in the subventricular zone (SVZ). There is evidence for two SVZs: SVZ1 lies adjacent to the VZ, and SVZ2 is adjacent to SVZ1 (Petryniak et al., 2007). SVZ1 cells appear to be more immature and SVZ2 cells are highly intermixed with immature postmitotic cells. Superficial to the SVZs is the mantle zone (MZ) which contains migrating neurons, glia, and post-migratory cells forming the nuclei of the basal ganglia, such as the striatum (St), globus pallidus (GP), and the central and medial amygdala.

Below, we will briefly introduce each subdivision of the mouse embryonic basal ganglia and their main functions (Fig. 47–1).

Figure 47–1.

Illustration of the subpallial anatomy, representative transcription factors (TFs), and neuronal subtype derivatives. Anatomical mouse developing basal ganglia subregions, for the LGE, MGE, Septum, CGE, and POA, are illustrated and coded by colors. LGE (more...)

Medial Ganglionic Eminence

The MGE generates pallidal projection neurons, glia, and INs; the latter tangentially migrate throughout the telencephalon, particularly into the striatum, cortex, hippocampus, and amygdala (Marín et al., 2000; Kessaris et al., 2006; Xu et al., 2008; Fragkouli et al., 2009; Flandin et al., 2010; Nóbrega-Pereira et al., 2010; Bupesh et al., 2011; Silberberg et al., 2016).

The MGE is the major source of CINs (~70%) and generates roughly equal groups that are somatostatin+ (SST+) or parvalbumin+ (PV+) CINs (Wonders and Anderson, 2006; Miyoshi et al., 2010; Batista-Brito et al., 2020). MGE-derived CINs migrate tangentially to the neocortex where they integrate into cortical lamina (in an inside-out manner based on when they are generated) (Guo and Anton, 2014). Immature CINs form synaptic connections with excitatory projection neurons and other CINs.

TF expression and enhancer activity provide evidence that the MGE VZ has six subdomains (MGE0-5) along the dorsoventral axis (MGE0 most dorsal and MGE5 most ventral) that are biased in producing different neuronal subtypes (Flames et al., 2007; Silberberg et al., 2016; Hu et al., 2017a). MGE0 and MGE1 express the TFs Nr2f2 and Nkx6.2, but not Nkx2.1; its cellular output is uncertain but can produce progenitors that are displaced ventrally where they intercalate with other progenitors (Hu and Rubenstein, unpublished). MGE1 and MGE2 express Nkx2.1, Nr2f2, and Nkx6.2. Nkx6.2 loss of function and fate-mapping show that these subregions generate mostly SST CINs (Fogarty et al., 2007; Sousa et al., 2009). MGE3–5 encompass the ventral MGE and border the POA. These regions express Otx2 and Nkx2-1, but not Nr2f2 and Nkx6.2. Fate mapping using an enhancer expressed in this domain shows a bias for PV CINs (Silberberg et al., 2016; Hu et al., 2017a). Shh-Cre (active in MGE5) fate mapping suggests this ventral MGE subregion generates striatal cholinergic and PV+ INs, few PV+ CINs, and GP PV+ projection neurons (Flandin et al., 2010).

Biased production of PV and SST CINs is also seen across the MGE’s rostral-caudal axis. WNT signaling and Nr2f2 expression provide evidence for two rostral-caudal subregions. Transplantation of rostral MGE (low WNT signaling and absent of Nr2f2) resulted in mostly PV CINs, whereas the caudal domain (high WNT signaling and Nr2f2+) generates mostly SST CINs. Findings were supported by loss and gain of Nr2f2 function and gain of RYK function, a noncanonical WNT receptor (Hu et al., 2017b; McKenzie et al., 2018).

Caudal Ganglionic Eminence

The CGE progenitor zone, caudal to the MGE and LGE, lacks Nkx2.1 and has molecular properties similar to the LGE. The CGE produces CINs with distinct laminar positions, electrical properties, and molecular/cellular features (Nery et al., 2002; Miyoshi et al., 2010). There is no anatomical structure that clearly defines where the CGE begins and ends. A small zone of Nkx2.1+ MGE extends caudally adjacent to the CGE (Hu et al., 2017a). CINs derived from the CGE are delayed in their generation by about 3 days compared to the MGE (Wonders and Anderson, 2006). However, unlike MGE-derived CINs that laminate in an inside-out fashion, most CGE-derived CINs occupy upper cortical lamina regardless of birthdate (Guo and Anton, 2014).

CGE-lineage CINs comprise two broad groups based on mostly nonoverlapping molecular markers: (1) vasoactive intestinal peptide (VIP+) and (2) NDNF+; REELIN+; SST- (Miyoshi et al., 2010; Schuman et al., 2019). The latter group consists of neurogliaform CINs (Taniguchi et al., 2011; Silberberg et al., 2016; Abs et al., 2018). Notably, each of these groups expresses the CGE-enriched TFs, Sp8 and Prox1. The Serotonin-receptor 3a (5HT3aR) is also a CGE-lineage marker (Miyoshi et al., 2010; Lee et al., 2010; Rudy et al., 2011; Ma et al., 2012; Rubin and Kessaris, 2013; Miyoshi et al., 2015; Touzot et al., 2016; Wei et al., 2019).

There is no known TF that specifies the CGE, as Nkx2.1 does for the MGE. Perhaps the LGE/CGE regions are patterned by a default mechanism, while Nkx2.1 is required to drive MGE identity. This is supported by data showing that in the absence of Nkx2.1, LGE molecular identity is present in the space where the putative MGE would be (Sussel et al., 1999). However, some notable TFs have a role in CGE cell fate and IN development, including Gsx1, Gsx2, Nr2f1, Nr2f2, and Prox1. We will discuss these TFs below in the regional fate specification section (Wang et al., 2009, 2013; Pei et al., 2011; Chapman et al., 2018).

Lateral Ganglionic Eminence

The lateral ganglionic eminence (LGE) is dorsal to the MGE and rostral to the CGE. Unlike the sulcus that separates the MGE and LGE, the LGE and CGE do not have a clear border, and they share many molecular properties (Flames et al., 2007; Long et al., 2009a,b). The LGE produces at least three major groups of neurons: striatal projection neurons, olfactory bulb (OB) INs (Deacon et al., 1994; Olsson et al., 1997; Olsson, Björklund, and Campbell, 1998; Wichterle et al., 2001; Ma et al., 2012), and perhaps a small subset of CINs (Torigoe et al., 2016).

Evidence suggests the existence of different LGE progenitors in the SVZ is responsible for LGE-lineage neuronal diversity. For example, the Dlx+; Isl1+ progenitors in the SVZ predominantly generate striatal projection neurons, while the Dlx+; Isl1–; ER81+ progenitors that mainly reside in the dorsal LGE give rise to OB INs. As OB INs are born, they take the rostral migratory stream (RMS) to populate the granular/periglomerular layers in the OB (Stenman et al., 2003). The most dorsal LGE has strong expression of Sp8, which with Sp9, regulates the development of OB and CGE INs (Li et al., 2018; Wei et al., 2019).

Pre-Optic Area

The embryonic POA is interposed between the MGE and the hypothalamus (Puelles et al., 2000; Gelman et al., 2009). While both the MGE and the POA express Nkx2.1, they have molecular differences (Flames et al., 2007). The MGE expresses Lhx6, Lhx8, and Olig2, whereas the POA expresses Dbx1 and Nkx5.1.

The POA generates ~10% of CINs, most of which are NPY+/SST–, similar to CGE INs (Gelman et al., 2009, 2011). The POA also generates neurogliaform INs, neurons that migrate to GP and amygdala, and oligodendrocytes (Hirata et al., 2009; Flandin et al., 2010; Torigoe et al., 2016; Niquille et al., 2018). Nkx5.1 (Hmx3) demarcates a POA progenitor subpopulation that gives rise to neurogliaform INs (Niquille et al., 2018).

Septum

The septum is a rostral extension of the LGE and the MGE, but with distinct properties governed in part by FGF signaling and expression of the Zic TFs (Inoue et al., 2007; Hoch et al., 2015b). Patterning of the lateral septum is regulated by the Dlx and Gsx genes (Long et al., 2007, 2009a; Wang et al., 2009; Qin et al., 2017), whereas patterning of the medial septum is controlled by Lhx6, Lhx8, and Nkx2-1 (Zhao et al., 2003, 2008; Flandin et al., 2011; Magno et al., 2017).

The septum and adjacent regions of the MGE generate GABAergic and cholinergic neurons that populate the septum diagonal band, striatum, GP, and some pre-optic nuclei based on Fgf8-Cre and Fgf17-Cre fate mapping (Hoch et al., 2015b). Shh-Cre fate mapping showed that specific regions of the MGE generate neurons of the medial septum and diagonal band (Flandin et al., 2010). Combining Nkx2.1 and Zic4 in vivo genetic labeling, the lineages of septal progenitors can be traced temporally (Magno et al., 2017). Many of these neurons form projections to the hippocampus and are crucial for modulating synaptic plasticity, rhythm generation, and learning/memory (Colom, 2006; Drever, Riedel, and Platt, 2011; Magno et al., 2017).

MGE Specification and LGE/CGE Identity Repression

Nkx2.1 and Otx2 play crucial roles in MGE specification through their VZ expression. Nkx2.1 expression starts at ~E9.5 and is responsible for the induction of Lhx6 and Lhx8 in the MGE SVZ to specify IN and GP progenitors (Sussel et al., 1999; Flandin et al., 2011; Sandberg et al., 2016). As CINs are specified, they lose the expression of Nkx2.1 and Lhx8 but maintain Lhx6, which drives CIN differentiation, migration, and the subsequent expression of SST and PV (Liodis et al., 2007; Zhao et al., 2008; Vogt et al., 2014). In addition to promoting IN production, Nkx2.1 represses CGE and LGE fate. Constitutive loss of Nkx2.1 leads to loss of MGE CIN production and ectopic generation of striatal medium spiny neurons (Sussel et al., 1999); later removal (E12.5) of Nkx2.1 leads to CIN subclass switch from MGE- to CGE- lineage properties (Butt et al., 2005, 2008).

Although broadly expressed in the MGE VZ, the Otx2 homeodomain TF preferentially guides the fate specification of the ventral MGE (vMGE), a subregion that mainly produces projection neurons that migrate to the globus pallidus (GP) (Flandin et al., 2010). Conditional deletion of Otx2 in the Nkx2.1-cre defining MGE leads to diminished GP neuron production. Otx2 also interacts with the Fgf-signaling pathway and promotes the expression of Sall3, Tal2, and Tll2 that delineate vMGE from the POA and other parts of MGE (Hoch et al., 2015b).

Roles of Gsx1 and Gsx2 in Establishing LGE and CGE

Gsx1 and Gsx2 are homeodomain TFs that specify LGE and CGE identity. Gsx2 is expressed in the VZ and SVZ, whereas Gsx1 is expressed in the SVZ. Loss of Gsx2 and combined deletion of Gsx1/Gsx2 leads to progressively severe phenotypes in the LGE and CGE (Toresson and Campbell, 2001; Yun et al., 2003). No clear phenotypes have yet been seen in the Gsx1 mutant. In addition, Gsx1/Gsx2 may not be critical for MGE development (Xu et al., 2010).

Gsx2 is particularly important for dorsal LGE specification, a region that produces many OB INs and parts of the striatum. Gsx2 also represses expression of cortical TFs Ngn2, Tbr2, and Pax6 in the LGE (Fode et al., 2000; Toresson, Potter, and Campbell, 2000; Toresson and Campbell, 2001; Yun et al., 2003). Gsx1 functions with Gsx2 for LGE specification, the production of striatal neurons, and guidance of corticofugal/thalamocortical axons (Toresson and Campbell, 2001; Yun et al., 2003). Loss of Gsx1/Gsx2 leads to reductions in the expression of Dlx genes in the LGE SVZ, TFs required in maintaining the SVZ progenitor activity and the differentiation of striatal neurons (Anderson et al., 1997a; Toresson and Campbell, 2001; Yun, Fischman et al., 2002; Yun et al., 2003; Pei et al., 2011).

Gsx2 also facilitates CGE patterning and neurogenesis (Corbin et al., 2000, 2003; Xu et al., 2010; Wang et al., 2013). Constitutive Gsx2 mutants have a hypoplastic CGE (Xu et al., 2010). Gsx2 gain-of-function in the MGE induces a CGE-like identity to produce bipolar Calretinin+/SST– INs (Butt et al., 2008; Xu et al., 2010; Wang et al., 2013).

Ventral and Rostral Patterning Centers Together Induce Nkx2.1 and the MGE through SHH and FGF8 Signaling

Sonic hedgehog (SHH) is a secreted morphogen that promotes ventral regional identity throughout the developing CNS. Through its expression in the rostral hypothalamus and preoptic area, it promotes ventral telencephalic fate by inducing Nkx2.1, Dlx2, and Gsx2 expression (Xu et al., 2005, 2010). Reduced SHH signaling leads to a hypoplastic ventral telencephalon and ectopic expression of dorsal telencephalic markers, such as Pax6 and Emx2 (Ohkubo et al., 2002; Fuccillo et al., 2004). SHH signaling functions in part by transcriptionally activating Gli1 and Gli2. On the other hand, Gli3 can serve as a transcriptional repressor and antagonize Shh function to promote telencephalic dorsal identity (Litingtung and Chiang, 2000; Motoyama and Aoto, 2007; Zaki, Martynoga, and Price, 2008; Ruiz i Altaba, 1998; Jiang and Hui, 2008; Hui and Angers, 2011). During early MGE development, Lhx6 and Lhx8 have compensatory roles to induce the expression of Shh in the MGE mantle zone via regulation of a Shh enhancer (Flandin et al., 2011). In turn, the induction of Shh provides a feed-forward loop to maintain the expression of Lhx6, Lhx8, and Nkx2.1, which facilitate MGE-derived pallial IN and subpallial neuron generation (Xu et al., 2005; Flandin et al., 2011). Depletion of Shh in the Dlx1/2-Cre lineage leads to increased apoptosis, which results in a similar reduction of SST and PV CINs in adult cortices (Flandin et al., 2011).

Despite the antagonism between Shh and Gli3, Shh/Gli3 double deletion renders no phenotype in dorsoventral telencephalic patterning. This suggested that other morphogens may be involved in telencephalic patterning (Rallu et al., 2002; Fuccillo et al., 2004). Fibroblastic growth factors (FGFs) are expressed in the rostral midline of the telencephalon, also known as the rostral patterning center (RPC). One such FGF is FGF8, which is implicated in the induction of telencephalic identify by promoting Foxg1 (BF1) expression (Shimamura and Rubenstein, 1997; Storm et al., 2006). Fgf8 and Otx2 reciprocally repress each other in the RPC and the midbrain/ hindbrain boundary that instructs proper telencephalic morphogenesis (Crossley et al., 2001; Hoch et al., 2015a). Furthermore, Fgf8 is required for the induction of Shh in the ventral telencephalon (Storm et al., 2006). It is proposed that positive regulation between FGF and SHH signaling, and the convergence of the two pathways, is required for induction of Nkx2.1 (Gutin et al., 2006; Storm et al., 2006; Hoch, Rubenstein, and Pleasure, 2009). FGF function, downstream of Shh and Gli3, is required for ventral progenitor generation and differentiation (Fuccillo et al., 2004; Gutin et al., 2006; Storm et al., 2006). Although the rostral patterning center expresses several Fgf genes, FGF signaling in the ventral telencephalon appears to be dominated by Fgf8 expression and its interaction with FGFR1/FGFR2 (Meyers et al., 1998; Gutin et al., 2006; Storm et al., 2006). In addition to promoting ventral progenitor fate, FGF signaling is also required to maintain the ventral progenitor pool by promoting proliferation and preventing apoptosis (Storm et al., 2006).

Transcriptomic analyses (Porteus et al., 1992; Long et al., 2009a,b) and mutational analysis identified TFs and other genes that are crucial for subpallial subdomain fate specification. However, the neurogenesis in the GE is far more complicated, as each of the germinal domains can at different developmental stages produce distinct types of cells that populate unique brain regions.

To investigate this complexity, various molecular and genetic tools are used. Nucleotide labeling during DNA replication (EdU/BrdU) is used to determine when neuron progenitors had last divided (birthdate). In vivo transplantation assays can assess intrinsic cell fates of progenitors in temporal and spatial manners. Finally, the use of Cre or Flip recombinases and recombinase-dependent reporters is a powerful means to follow the descendants of progenitors that have regionally and temporally restricted expression of genes. This can be coupled with further temporal regulation by fusing part of the estrogen protein (ER) to the recombinase to make it only functional in the presence of an estrogen-like ligand (e.g., tamoxifen) (Taniguchi et al., 2011; Silberberg et al., 2016). The application of recombinases (sometimes multiple in combination), and recombinase-ER fusions, allows cell-type and temporal specificity to permanently label developing lineages to enable the anatomical, molecular, and physiological characterization of their descendants (Nagy et al., 2000; Kessaris et al., 2006; Xu et al., 2009). Fate mapping using transgenic Cre lines can have complications (Luo et al., 2020). Genetic background and germline recombination can strongly influence the results from many loci. For a given locus, Cre recombinase expression can vary across development, as can the degree of tamoxifen-induced recombination. In this section, we will discuss several approaches.

Identification of MGE-Derived Descendants

As noted above, Nkx2.1 expression marks MGE progenitors and the GP, but not CINs. A transgenic line carrying Cre-recombinase driven by Nkx2.1 regulatory elements was generated to determine the descendants of Nkx2.1-expressing cells (Xu et al., 2008). Fate mapping revealed that Nkx2-1-Cre activity labeled 70%–80% of CINs and HINs, with PV- and SST-expressing IN subgroups being evenly targeted (Xu et al., 2008). Nkx2.1-Cre marks other neurons produced by the MGE, such as striatal GABAergic (SST+ and PV+), cholinergic INs, and GP projection neurons (Xu et al., 2008; Flandin et al., 2010). Unlike pallial INs, many of these neurons express Nkx2.1 and Lhx8 (Marín et al., 2001; Zhao et al., 2008; Flandin et al., 2011).

Nkx2.1 facilitates MGE-derived CIN fate determination through promoting the expression of Lhx6 (Du et al., 2008; Sandberg et al., 2016). A similar Cre-recombinase transgenic line was generated for Lhx6. Lhx6-Cre is also an excellent tool for studying MGE-lineage CINs/HINs (Fogarty et al., 2007). Almost 100% of the PV and SST CINs/HINS are labeled by the Lhx6-Cre, consistent with the Nkx2.1-Cre;Nkx6.2-Cre combined fate-mapping result (Fogarty et al., 2007; Batista-Brito et al., 2009). In addition, Lhx6-Cre demonstrated high (~100%) efficiency in labeling SST/CR-expressing Martinotti CINs and NPY- expressing CA1 HINs (Fogarty et al., 2007).

Olig2 expression is restricted primarily to GE progenitors and later oligodendrocytes. A tamoxifen-regulated Cre (Cre-ER) was inserted in the Olig2 locus and allowed a temporal specificity to fate-mapping early-born and late-born INs, followed by morphological and electrophysiological analyses (Kimmel et al., 2000; Miyoshi et al., 2007, Takebayashi et al., 2000; Nery et al., 2001; Lu et al., 2002). Consistent with the Nkx2.1-Cre finding, SST-expressing IN production was enriched at earlier ages (E9.5–E12.5) (Xu et al., 2008). Finer IN subtyping was feasible when combining molecular marker, morphology, laminar location, and intrinsic firing patterns. The diversity of IN subtypes in the Olig2-iCre lineage correlated with their birthdates. Like the nucleotide-birthdating analyses, they found that MGE-derived CINs incorporated into the circuit in an inside-out manner (Valcanis and Tan, 2003; Pla et al., 2006; Rymar and Sadikot, 2007; Xu et al., 2008; Lim et al., 2018).

Identification of CGE-Derived Descendants

To determine birthdates, subtypes, and physiological properties of the CGE lineage, several Cre lines have been used. Mash1BAC-CreER, although widely active in the GEs, can be used to follow CGE-derived INs by taking advantage of temporal differences in IN production (Battiste et al., 2007; Miyoshi et al., 2010). Tamoxifen-induced recombination suggested that CGE neurogenesis begins ~E12.5, ~3 days after the MGE. ~75% of CGE-derived INs occupied upper cortical lamina (layer I–III) (Miyoshi et al., 2010), consistent with the proposal of Rymar and Sadikot (2007) that some INs do not undergo inside-out lamination. Taking into account the morphology and physiology diversity, at least nine subtypes of CGE-derived CINs were documented (Miyoshi et al., 2007; G. Miyoshi et al., 2010), including VIP+ and REELIN+ (SST–) CGE-derived INs. At least half of the REELIN CINs were late-spiking cells, while the majority of the VIP CINs were fast-adapting cells.

Silberberg et al. (2016) generated stable transgenic mouse lines expressing CreERT2 and GFP driven by two different enhancer elements (enhancer-CreERT2-IRES-GFP) (Silberberg et al., 2016). These enhancers, 953 and 1060, map close to Sp8 and Nr2f1, respectively. The 953 CreERT2 and 1060 CreERT2 enhancer lines exhibited specific activity in the CGE MZ. Tamoxifen induction at E10.5, E12.5, or E14.5 revealed CGE-derived (SP8+) amygdala neurons labeling in both lines (Silberberg et al., 2016). CGE-derived CINs were labeled by both lines when tamoxifen was administered post E14.5. ~80% of the 1060 CreERT2 fate-mapped CINs were REELIN+ and/ or SP8+ but not VIP+, likely due to preferential labeling of neurogliaform INs (Miyoshi et al., 2010; Silberberg et al., 2016). Approximately 30% of the 953 CreERT2 fate-mapped CINs were SP8+, suitable for experiments that require sparse CGE CIN targeting.

POA-Derived INs

Although the majority of CINs and HINs are derived from the MGE and CGE, the diversity and complexity of IN subtypes and properties are constantly being discovered, suggesting that there might be more regions in the brain that contribute to diverse IN populations. POA and MGE share high levels of Nkx2.1 expression (Sussel et al., 1999; Nery et al., 2002; Flames et al., 2007; Xu et al., 2008). When a Cre-dependent GFP reporter was injected in utero into the POA of Nkx2.1-cre animals, some of the POA-originated GFP+ cells tangentially migrated into the neocortex, yet lacked Lhx6 expression (Gelman et al., 2009). Further analyses, using Nkx5.1-Cre showed that the POA generated INs for the neocortex, piriform cortex, and the hippocampus (Fogarty et al., 2005; Gelman et al., 2009). These POA-originated INs are GABAergic and express NPY but not traditional IN markers, such as PV, Calretinin, SST, VIP, and nNOS; their morphology and electrophysiological properties are also distinct from MGE/CGE INs (Fogarty et al., 2005; Fogarty et al., 2007; Gelman et al., 2009).

Pallial Progenitors Contribute to Olfactory Bulb IN Diversity

The dorsal LGE is the major prenatal source of OB INs (Wichterle et al., 1999; Toresson and Campbell, 2001; Stenman et al., 2003; Yun et al., 2003). However, the pallium also generates OB INs, and this continues postnatally (Wichterle et al., 2001; Kohwi et al., 2007; Merkle, Mirzadeh, and Alvarez-Buylla, 2007; Kriegstein and Alvarez-Buylla, 2009; Obernier and Alvarez-Buylla, 2019; Zhang et al., 2020). Emx1-Cre lineage tracing at P0 and adulthood identified a unique subset of SVZ progenitors that are LGE-like (Dlx2/Gsx2-expressing) and can produce calretinin+ OB INs (Gorski et al., 2002; Kohwi et al., 2007; Zhang et al., 2020). It is proposed that Dlx2/Gsx2/Sp8 expression in the SVZ intermediate progenitors (mostly Emx1+) promotes OB IN fate specification and represses the default oligogenic potential (Gorski et al., 2002; Chapman et al., 2012, 2018; Li et al., 2018; Zhang et al., 2020).

Shh Fate Mapping

Shh expression in the MGE is enriched in the ventral MGE and dorsal POA (Flames et al., 2007). Shh-Cre fate mapping showed that ventral MGE generated (1) very few CINs; these were PV+; (2) neurons that primarily express PV and migrate to GP, septal/preoptic complexes, and a subset of amygdaloid nuclei; and (3) INs that were predominantly PV+ that migrate to the striatum (Flandin et al., 2010). In the GP, Shh-Cre preferentially captures the NKX2-1+/NPAS1+ subpopulation (~70%–80%), rather than the NKX2-1–/NPAS1+ cells (~35%) (Flandin et al., 2010). When Nkx2.1 is depleted in the Shh domain using Shh-Cre, expression of Er81, Npas1, and Lhx8 were greatly reduced in the GP, further supporting the roles of Shh and Nkx2.1 in specifying GP neuron progenitors (Sussel et al., 1999; Flandin et al., 2010).

Fgf8 and Fgf17 Fate Mapping

FGFs are produced by the rostral patterning center (RPC) to guide regional patterning and progenitor fate commitment. RPC lineages were assessed using tamoxifen-inducible Cre under the control of either Fgf8 or Fgf17 (Hoch et al., 2015b). Induction of Cre recombination at E7.5/E8.5/E9.5 and assessment at E18.5 to P40 fate mapping revealed that Fgf8-CreER and Fgf17-CreER lineages give rise to distinct groups of neurons that populate different parts of the brain. In the OB, sparse Fgf17+ neurons are observed, while more neurons belong to the Fgf8 lineage. The Fgf8+ OB neurons can be GABAergic INs (periglomerular and granule cells) and glutamatergic projection neurons (mitral cells). The Fgf8-lineage contribution to OB neurons tapers around E10–E11.5. Projection neurons in the rostro-dorsal cortex also preferentially come from the Fgf8 lineage. The basal ganglia and CINs are also the destination for cells in the Fgf8- and Fgf17 lineages. The RPC largely generates cells of the septum. A surprising result is that both Fgf8-CreER and Fgf17-CreER generate subpallial cholinergic projection neurons and INs. In the striatum, Fgf8 lineages label medium spiny neurons (CTIP2+) and INs (PV+/SST+; ChAT); in the GP, the Fgf8 lineage co-localizes with different subtype markers (Nkx2.1, Ctip2, and Npas1).

The Dlx TFs were identified in the early 1990s, and some of the initial Dlx studies focused on the expression of the Dlx genes during craniofacial and forebrain development (Porteus et al., 1991, 1992; Price et al., 1991; Robinson et al., 1991; Depew et al., 2002). In the brachial arches, differential expression of Dlx1/2/5/6 controls regional fate specification (Depew et al., 2002; Panganiban and Rubenstein, 2002; Jeong et al., 2008, 2012). Dlx1/2 mutants had skeletal defects specifically in the upper jaw (Qiu et al., 1997; Depew et al., 2005), whereas deletion of Dlx5 resulted in lower jaw defects (Depew et al., 1999); in Dlx5/Dlx6 mutants the lower jaw transformed into an upper jaw (Depew et al., 2002).

Dlx genes were the first described TFs whose expression is largely restricted to forebrain GABAergic neurons (Panganiban and Rubenstein, 2002; Stühmer et al., 2002a; Stühmer et al., 2002)b. The Dlx TFs control GE cell fate specification and differentiation through their regulation of enhancer activity and epigenetic state (Anderson et al., 1997a, 1997b; Long et al., 2007, 2009a, 2009b; Lindtner et al., 2019). In the ganglionic eminences, Dlx TFs are expressed in a sequential manner (Dlx2 → Dlx1 → Dlx5 → Dlx6) starting from E9.5. Their enriched and overlapping expression in the SVZ roughly separates GE progenitor zones into VZ, early progenitors, SVZ1 and SVZ2, intermediate progenitors, and MZ, immature neurons (Eisenstat et al., 1999; Yun et al., 2002; Petryniak et al., 2007). In this section, we will focus on the roles of Dlx and other key TFs on GE progenitor cell biology.

GE Ventricular Zone

The GE ventricular zone (VZ) consists of proliferating stem cells that have the potential to generate glia and several neuronal subtypes. The stemness and the proliferative capacity of the progenitors are promoted by Notch signaling. Mash1 (Ascl1) enhances Notch activity by driving the expression of Dll1, Dll3, and Hes5 (Casarosa et al.,1999). The Hes TF genes are critical in specifying and maintaining the stem cell state (Kobayashi and Kageyama, 2014). Loss of Mash1 results in reduced Notch signaling and premature expression of Dlx2, Dlx5, and Gad67 in the VZ (Yun et al., 2002).

As VZ progenitors mature and migrate into SVZ, Dlx1 and Dlx2 expression is initiated, which represses Hes1/5, Gsx1/2 (Gsh), and Mash1 to facilitate differentiation. In line with this, Dlx1/2 constitutive mutants exhibit increased expression of Dll1, Gsx1/2, Hes5, Mash1, and other VZ markers in the SVZ, and show prolonged proliferation and impeded differentiation (Anderson et al., 1997a, 1997b; Yun, Fischman et al., 2002; Long et al., 2009a, 2009b). Note that Dlx1/2 constitutive mutants fail to induce expression of Dlx5&6; thus, these mutants lack expression of any Dlx gene (Anderson et al., 1997a). In addition to the role of the Mash1/Dlx1/Dlx2 transcriptional cascade in promoting neuronal differentiation, they also repress gliogenic differentiation (Yun et al., 2002; Petryniak et al., 2007). In the Dlx1/Dlx2 mutants, while GE neurogenesis was nearly abolished, Olig2 expression was increased (Petryniak et al., 2007). Dlx1/2 have a modest role in GE regional patterning; in Dlx1/2 mutants, the dorsal LGE is partially transformed toward a cortical fate (Long et al., 2009a).

What is the mechanism that balances gliogenesis versus neurogenesis in the VZ? Petryniak et al. and Silbereis et al. demonstrated that Olig2 and Olig1 are key TFs that control the neuron-glial fate switch in the GE VZ (Petryniak et al., 2007; Silbereis et al., 2014). Olig1 expression preserves progenitors’ gliogenic potential, and it represses Dlx1/2 through the DlxI12b enhancers (Ghanem et al., 2007a; Silbereis et al., 2014). Constitutive deletion of Olig1 results in (1) excessive expression of Lhx6/Dlx1/Dlx2 in the ventral MGE, (2) a decrease in oligodendrocyte progenitors in the GE, and (3) ~30% excess production of INs that are primarily PV+ and CR+. Of note, PV+ and CR+ INs are produced by late progenitors in the GE that also have the potential to generate oligodendrocytes. Olig1 null mutants had no changes in NPY+ and SST+ INs, perhaps because they are produced before oligodendrogliosis (Petros et al., 2015). It is likely that Olig1 controls the neuronal-glial fate switch only within a restricted time window. Olig2 shares overlapping expression pattern and redundant functions with Olig1.

There are several VZ-expressing TFs that play critical roles in guiding GE regional specification. For instance, Nkx2.1 and Otx2 guide MGE specification through their VZ expression to repress LGE/CGE fate (Sussel et al., 1999; Hoch et al., 2015a). Nkx2.1 loss-of-function impedes MGE neurogenesis and MGE-lineage reduction, and expansion of LGE/CGE-derived neuronal population (Sussel et al., 1999; Butt et al., 2005, 2008). Gsx1 and Gsx2, on the other hand, drive the identity of the LGE and CGE. Gsx1/Gsx2 double deletion resulted in hypoplastic CGE, loss of Dlx expression in the LGE SVZ, and a subsequent loss of LGE-derived striatal and OB neurons (Toresson and Campbell, 2001; Corbin et al., 2003; Yun et al., 2003).

GE Subventricular Zones

Mechanisms Proposed for the Generation of SST and PV CIN

The contribution of VZ and SVZ progenitors to IN fate specification has been best studied in the MGE, with progress in understanding the mechanisms of PV and SST CIN specification. Several TFs have been shown to differentially mediate IN subtype specification along the MGE VZ-SVZ1-SVZ2 axis. For instance, Mafb and c-Maf in the SVZ repress SST IN neurogenesis and promote PV IN generation (Pai et al., 2019). In contrast, Nr2f2 in the VZ and SVZ promotes SST and represses PV IN production (Hu et al., 2017a).

Four MGE cell fate specification models have been proposed (Hu et al., 2017b). In model 1, different dorso-ventral and rostro-caudal regions of the MGE VZ/SVZ are biased but may not be specific for generating distinct MGE-derived INs and projection neurons (Flames et al., 2007; Flandin et al., 2010). The subsequent models proposed mechanisms within a given regional subdomain. In model 2, the “mosaic model,” distinct progenitors in the VZ are committed (or strongly biased) to generate either SST+ or PV+ CINs. Currently there is no strong evidence for this. In model 3, the “homogenous VZ model,” VZ progenitors can generate both subgroups. Retroviral labeling and lineage tracing of VZ progenitors has shown that individual MGE progenitors can generate both SST+ and PV+ CINs (Brown et al., 2011; Harwell et al., 2015; Mayer et al., 2015). In model 4, the “direct versus indirect neurogenesis model,” VZ progenitors generate SST+ CINs (direct neurogenesis), whereas SVZ progenitors generate PV+ CINs (indirect neurogenesis). The support for model 4 comes from evidence that the timing of progenitor cell cycle exit can independently dictate neuronal fates (Glickstein et al., 2007; Lodato et al., 2011; Petros et al., 2015). Petros et al. showed that SST INs are mainly produced by MGE VZ progenitors that exit the cell cycle and initiate differentiation early, while PV INs are mainly produced by MGE progenitors that proliferate and cycle longer (from VZ to SVZ) (Petros et al., 2015). By repressing Notch signaling and pushing progenitors to exit the cell cycle in the SVZ, more SST IN production was observed; when the cell cycle was extended in the SVZ, more PV IN production was noticed (Glickstein et al., 2009; Petros et al., 2015). Finally, the observation of ventrally displaced progenitors from MGE0/1 adds another mechanism that may contribute to the specification of neuronal diversity (Hu and Rubenstein, unpublished). Thus, multiple mechanisms could contribute to cell fate specification in the GEs. Future studies using single cell in vivo lineage analysis, spatial transcriptomics, single-cell RNA-seq coupled with RNA velocity, and cell cycle status analyses are needed to better understand these mechanisms (Kowalczyk et al., 2015; La Manno et al., 2018; Hsiao et al., 2020).

After their roles in progenitors are complete, many TFs also mediate CIN migration, differentiation, maturation, and function. Here, we will summarize some of the TF functions in the order of when these TFs start to be expressed in the GEs (VZ/SVZ: Dlx, Nr2f2; SVZ: Arx, Npas1/Npas3, Mafb/c-Maf; MZ and immature INs: Satb1, Mef2c).

Postmitotic Roles of Dlx TFs in IN Survival, Morphogenesis, and Synapse Formation and Function

Dlx expression is restricted to forebrain GABAergic progenitors (GEs) and several subtypes of GABAergic neurons, including CINs/HINs/OB INs. In the stem cell biology section, we addressed that Dlx1/2 initiates the expression of Dlx5/6 in the GE progenitors to drive neurogenesis. Dlx1/2/5/6 expression and their downstream transcriptional cascades also provide guidance for INs to properly migrate and mature (Anderson et al., 1997a, 1997b; Eisenstat et al., 1999; Stühmer et al., 2002a, 2002b; Cobos et al., 2005b; Petryniak et al., 2007; Long et al., 2009a, 2009b; McKinsey et al., 2013). For instance, Dlx1/2 promote the expression of cytokine receptors (Cxcr4 and Cxcr7), which allow INs to migrate along pathways secreting the ligand Cxcl12 (Wang et al., 2011; Wu et al., 2017; Cobos et al., 2007). To facilitate tangential migration, Dlx1/2 also delay the initiation of IN axon and dendritic growth (Cobos et al., 2007).

To further understand the postnatal function of Dlx1/2, postmitotic and postnatal Dlx1/2 conditional deletion mouse models were generated using Calretinin (CR)-Cre, Sst-Cre, PV-Cre, and DlxI12b-Cre (Hippenmeyer et al., 2005; Potter et al., 2009; Taniguchi et al., 2011; Silbereis et al., 2014; Pla et al., 2018; Sohal and Rubenstein, 2019). Deletion of Dlx1/2 in postmitotic neurons does not disrupt the expression of Dlx5/6. As a result, IN generation and migration are not affected in these mutants. However, several changes were observed after Dlx1/2 loss, including IN survival, dendritic morphogenesis, and synaptic number/function. These phenotypes derive likely in part from the decreased expression of Gad1, Gad2, Vgat, Cxcr4, and Grin2b (Cobos et al., 2007; Long et al., 2009; Le et al., 2017; Pla et al., 2018). Specifically, the conditional Dlx1/2 mutants had extensive CIN cell death affecting all classes of CINs by P7; at later stages, CINs failed to receive proper excitatory inputs due to hypoplastic dendritic arbors and formed fewer inhibitory synapses onto excitatory neurons, resulting in reduced mIPSC frequency (Pla et al., 2018). HINs with constitutive loss of Dlx1 also exhibit similar hypoexcitability and reduced glutamatergic excitation, supporting a role for Dlx1 in IN maturation (Jones et al., 2011; Howard et al., 2014). Of note, while Dlx genes are generally imporant for postnatal IN survival, PV INs maybe be less affected as Dlx1 expression is downregulated earlier in these progenitors. In addition. Dlx1 constitutive mutants exhibited ~30% pan CIN/HIN loss except for the PV subpopulation (Cobos et al., 2005a).

Dlx5/6 homozygous mutants have exencephaly, which makes it difficult to study Dlx5/6 in CIN development (Schüle et al., 2007). Using an MGE transplantation assay, Dlx5/6 loss of function was found to preferentially result in reduced PV CINs (Wang et al., 2010). This impact on PV CINs was corroborated in Dlx5/6 heterozygous mutants, where electrophysiologcal firing defects were observed in PV INs, leading to alterations in behavior and epilepsy (Cho et al., 2015). The transcriptional cascades downstream of Dlx5/6 are currently an active pursuit in the field.

Nr2f1/Nr2f2 Promotes CGE Fate Specification and MGE Generation of SST CIN

Nr2f1 and Nr2f2 (CoupTFI and CoupTFII) nuclear receptor TFs have regional expression within the MGE and pan-CGE expression (Kanatani et al., 2008; Lodato et al., 2011; Cai et al., 2013; Hu et al., 2017a). Conditional deletion of Nr2f1/Nr2f2 in GE intermediate progenitors, using Dlx5/6-Cre, reduced CGE-derived INs (VIP+; CR+) and increased MGE proliferation, including an excessive production of PV-expressing INs (Lodato et al., 2011). In addition, Nr2f1/Nr2f2 have instructive roles in guiding CGE-derived CINs to integrate into the neocortical circuit through the caudal migratory stream (Tripodi et al., 2004; Kanatani et al., 2008; Touzot et al., 2016). Forced expression of Nr2f2 is sufficient to convert the MGE-derived INs to migrate in a manner similar to the CGE-derived INs (Kanatani et al., 2008), suggesting the regulation of potential guidance cues by these genes during development.

In the MGE, Nr2f1/Nr2f2 expression is in the dorsal domain with a caudo-rostral gradient and is complementary to the Otx2-expressing ventral rosto-caudal gradient (Hoch et al., 2015a; Hu et al., 2017a). Nr2f1/Nr2f2 represses Ccnd2 and promotes Sox6 to drive SST+ IN fate specification (Glickstein et al., 2007; Azim et al., 2009; Batista-Brito et al., 2009; Hu et al., 2017a). Nr2f1/Nr2f2 conditional nulls also exhibited lengthened cell cycle and a shifted SST/PV ratio in the MGE that led to an increase in PV+ CINs (Glickstein et al., 2007; Hu et al., 2017a). Finally, forced expression of Nr2f2 in MGE cells increased Sox6 expression, suppressed PV fate, and promoted SST fate; expression of Sox6 in MGE cells also resulted in increased SST+ fate at the expense of PV (Vogt et al., 2014; Hu et al., 2017a).

Role of Arx in MGE-Derived IN Migration and PV IN Fate Specification

Arx encodes an X-linked homeodomain protein; when mutated, humans have severe epilepsy and cognitive impairment (Kitamura et al., 2002). In the developing telencephalon, Arx is expressed in pallial VZ progenitors, subpallial SVZ progenitors, and immature neurons, including INs tangentially migrating to the pallium (Friocourt et al., 2006; Colasante et al., 2008, 2009; Friocourt and Parnavelas, 2010). Constitutive mouse Arx mutants, deletions, or triplet repeat expansions modeling human mutations have molecular defects in the MGE and decreased numbers of CINs, as well as epilepsy (Kitamura et al., 2002; Colasante et al., 2009; Price et al., 2009). Deletion of Arx in newborm subpallial neurons also results in CIN defects that appear to result in part from migration defects (Marsh et al., 2009, 2016); some of the abnormal migration is linked to reduced expression of Cxcr4 and Cxcr7 (Fulp et al., 2008).

Arx expression is promoted and maintained by the Dlx and Lhx6 genes (Cobos et al., 2005; Colombo et al., 2007; Colasante et al., 2008; Zhao et al., 2008; Vogt et al., 2014). Furthermore, Arx re-expression in the MGE of telencephalic slices of E14.5 Dlx1/2 mutants enables immature INs to tangentially migrate to the cortex (Colasante et al., 2008). In the Lhx6 null mutant, MGE-derived CINs fail to express PV and SST. Arx complementation of Lhx6 loss-of-function INs, using an MGE transplantation assay, was sufficient to induce PV expression, suggesting Arx has a role in fate promotion in postmitotic MGE-derived INs (Vogt et al., 2014). Arx’s role in postmitotic CGE-derived INs is uncertain.

Opposing Roles of Npas1 and Npas3 in the MGE/CGE-Derived IN Generation and Differentiation

Npas1 and Npas3 bHLH-PAS TFs are expressed in MGE/CGE progenitor zones and in immature and mature CINs (Zhou et al., 1997; Erbel-Sieler et al., 2004; Stanco et al., 2014). While coexpression of Npas1 and Npas3 had been identified in GAD67-expressing CINs, these two TFs exhibit distinct functions to guide IN progenitor proliferation and differentiation (Erbel-Sieler et al., 2004; Stanco et al., 2014). In the MGE and CGE, Npas1^–/– had increased proliferation, ERK signaling, and Arx expression. They also generated an excess of SST and VIP CINs. In contrast, Npas3^–/– showed decreased progenitor proliferation and ERK signaling and generated fewer SST+ and VIP+ INs. The deletion of Npas1 and Npas3, however, did not affect the generation of PV CINs.

In addition, Npas1 and Npas3, or Npas3 alone may be crucial in guiding Reelin+ CIN production and differentiation. A drastic reduction of Reelin+ CINs is observed both in the Npas3 null and the Npas1/Npas3 double mutants (Erbel-Sieler et al., 2004; Stanco et al., 2014). Of note, Npas1 also mediates IN spontaneous activity. In the Npas1 null, upper-layer pyramidal neurons receive higher frequency inhibitory inputs from CINs, which implies a role of Npas1 in regulating IN excitability (Stanco et al., 2014).

Roles of the Mafb and c-Maf TFs in PV/SST IN Fate Specification, Migration, and Maturation

Mafb and c-Maf are BZIP TFs that are expressed in the MGE SVZ and in immature/mature INs generated from the MGE. The roles of Mafb and c-Maf were explored in single and double mutants (Pai et al., 2019, 2020). Nkx2.1-cre-depleted single Mafb or c-Maf conditional mutants did not show severe phenotypes, whereas double deletion mice (Maf cDKO) exhibited strong phenotypes, suggesting compensatory roles. First, there is ~70% reduction of total MGE-derived INs in the adult neocortex, indicating that the Mafb and c-Maf work together to produce the proper amount of mature INs. This cell loss did not start until P0. Second, of the INs that remained, the proportion expressing PV was decreased by about half, while there was no change in the proportion of SST INs. Looking at earlier developmental stages, ~50% increase in SST INs was observed at neonatal ages, just before IN programmed apoptosis would occur (Southwell et al., 2012). While the increase of the SST CIN phenotype is transient, and is diminished by ~P7, it highlights a potential mechanism whereby an early increase in one subtype of IN could potentially outcompete another during early stages of development resulting in skewed ratios by the time the brain reaches adulthood. Third, cells were becoming postmitotic/neuron fated at a higher rate in the mutants in the MGE progenitor domain, potentially in line with one hypothesis of how SST CINs could be preferentially generated (Petros et al., 2015). Finally, INs prematurely migrated into the cortical plate layer of the developing neocortex, suggesting further migratory regulation by these TFs. This could be partially explained by the decrease in Cxcr4 expression, as Cxcr4 constitutive null mice exhibit similar migration defects (Stumm et al., 2003; Wang et al., 2011; Pai et al., 2019, 2020).

In a later study, single-cell RNA sequencing showed that Mef2c was downregulated in the Maf cDKO neonatal pallial INs. Mef2c conditional ablation in postmitotic INs showed preferential loss of PV CINs, similar to what was observed in the Maf cDKO (Mayer et al., 2018; Pai et al., 2020). By restoring the expression of Mef2c in Maf cDKOs, PV expression was rescued (Pai et al., 2020). A transcriptional pathway was proposed in which Mafb and c-Maf promote the expression of Mef2c to drive PV IN specification (Mayer et al., 2018; Pai et al., 2020).

When Maf single and double mutant CINs were evaluated through whole-cell patch-clamp recordings, Mafb cKO and c-Maf cKO showed opposite spontaneous firing patterns, while Maf cDKO remained relatively normal. This evidence suggests that Mafb and c-Maf have distinct roles in CIN differentiation and physiology (Pai et al., 2019). A separate study focused on the function of Mafb in migratory SST+ Martinotti CINs. Their result suggested that Mafb can serve as an intrinsic factor that preprograms the migratory route for Martinotti cells and their layer I dendritic arborization (Lim et al., 2018). After Mafb depletion (using SST-Cre), Martinotti CINs failed to migrate to the neocortex through the marginal zone. These mutant CINs also formed poor layer I dendritic trees and failed to inhibit layer II/III pyramidal cells properly (Lim et al., 2018).

Activity-Dependent Expression of Satb1 Drives SST CIN Maturation

The expression of Satb1 homeobox TF initiates several days after the induction of Lhx6; it becomes more prominent when MGE-derived INs terminate their tangential migration and undergo radial migration into the cortical plate (Denaxa et al., 2012). Loss of Satb1 expression in postmitotic MGE-derived INs leads to reduced survival of postnatal PV and SST CINs and more severe deficit in the SST subtype (Close et al., 2012).

IN morphology and connectivity diversification occurs as immature INs integrate into the cortical circuitry (Moody and Bosma, 2005; Cossart, 2011; Denaxa et al., 2012). Satb1 may translate environmental cues during SST CIN differentiation (Denaxa et al., 2012). At perinatal stages, Satb1 expression is downstream of Lhx6 and is also activity-dependent; potassium chloride-induced depolarization in the INs is sufficient to promote Satb1 expression (Denaxa et al., 2012; Neves et al., 2013). Satb1 expression then promotes (1) Sst and other SST IN subtype marker (Npy, Calretinin) coexpression and (2) the spontaneous excitability of SST INs (Close et al., 2012; Wonders and Anderson, 2006; Denaxa et al., 2012).

Mef2c Controls PV CIN Differentiation

The Mef2c MAD box TF is expressed in both excitatory and inhibitory cortical neurons. It is also a risk gene for autism spectrum disorder and schizophrenia. Haploinsufficiency of MEF2C in humans can lead to a range of symptoms, including but not limited to autistic behaviors and intellectual disability (Li et al., 2008; Le Meur et al., 2010; Novara et al., 2010; Paciorkowski et al., 2013; Harrington et al., 2016; Tu et al., 2017). In excitatory neurons, Mef2c is a transcriptional repressor that regulates expression of synaptosomal genes. Mef2c depletion in the Emx1-Cre lineage (pyramidal neurons) results in an inverted excitation/inhibition circuit balance in the somatosensory cortex. Mutant excitatory neurons exhibited increased inhibitory inputs and a decrease in the formation of excitatory spine density (Harrington et al., 2016).

The role of Mef2c was surveyed in the mouse GE and CINs across different developmental time points (E13.5, E14.5, E18.5, P0, and P56) using single-cell transcriptomic and computational techniques (Mayer et al., 2018). Based on the presumption that each GE progenitor and their progenies share a group of developmentally conserved markers, transcriptomic features of mature PV CINs were traced back to MGE progenitors. In this way they identified that Mef2c is a gene that is enriched in immature MGE cells that will become mature PV CINs (Mayer et al., 2018). To test the function of Mef2c in PV IN development, Mef2c expression was deleted in GEs using the Dlx6a-Cre (Mayer et al., 2018). These mutants showed PV CIN reductions, similar to what was observed in the Nkx2.1-Cre generated Mafb/c-Maf double mutants (Mayer et al., 2018; Pai et al., 2019). A MGE transplantation rescue assay, together with MAFB ChIP-seq, suggested that Mafb/c-Maf could be the upstream TFs driving Mef2c expression that promotes PV IN differentiation (Pai et al., 2020).

Genomic-level understanding about the transcriptional control of CINs has come from single-cell RNA sequencing (scRNA-seq), epigenomic assessment on chromatin state, and the identification of noncoding DNA regulatory elements (REs; also named enhancers) (Ghanem et al., 2007; Visel et al., 2013; Sandberg et al., 2016; Silberberg et al., 2016; Lim et al., 2018; Mayer et al., 2018; Mi et al., 2018; Lindtner et al., 2019; Rubin et al., 2020). In this last section, we will review several recent scientific discoveries that refine GE progenitor regulation and CIN subtyping using these novel technologies.

Forebrain Enhancer Identification and Functional Characterization

The sequential and overlapping expression of Dlx1/2/5/6 drives basal ganglia development. Several intergenic regulatory elements (enhancers) of Dlx genes have been identified, including Dlx1/2 I12b and URE2, and Dlx5/6 I56i and I56ii. Their regional activities and function have been documented (Zerucha et al., 2000; Ghanem et al., 2007a; Fazel Darbandi et al., 2016). URE2 was the only Dlx RE active in the VZ between E11.5 and E15.5. The URE2 active progenitors formed radial clusters, suggesting that these VZ stem cells might be clonally related. I12b/I56i exhibited strong and overlapping expression in the SVZ, complementary to the weak URE2 activity in the same domain. This differential RE expression pattern could be one mechanism that delineates the border between VZ/SVZ/MZ. In addition, the majority of migrating immature INs that insert into the neocortex in multiple zones were equally labeled by I12b and I56i, although a small subset of INs migrating along the marginal zone showed single labeling for URE2 but not I12b/I56i. The activity of these three REs continues to be observed in mature CINs, and I12b/I56i broadly labels derivatives of the LGE, MGE, and CGE lineages; on the other hand, URE2 may be preferentially active in PV+, CR+, and NPY+ (but not SST+) INs (Ghanem et al., 2007a). Thus, distinct REs, within either the Dlx1/2 or Dlx5/6 loci, drive expression in distinct distributions of cells at different developmental stages.

Higher throughput approaches have greatly expanded our understanding of the complexity and specificity of gene regulatory elements that are active in the developing forebrain. In 2013, Visel et al. identified nearly 5000 candidate embryonic forebrain enhancers using DNA conservation, proximity to genes that regulate forebrain development, and p300 ChIP-seq followed by an in vivo transgenic assay of transcriptional activity (Visel et al., 2013). Through RE-driven transient transgene genetic tools, the activity and expression pattern of 145 evolutionarily conserved forebrain enhancers were studied in vivo, which further lead to a digital atlas that remains one of the most comprehensive resources for RE studies (Visel et al., 2013; https://enhancer.lbl.gov/). Some interesting discoveries are listed here: (1) a subset of enhancers whose expression pattern mirrors their associated genes, which makes RE-based transgenic tool a potential surrogate for studying gene functions (Frankel et al., 2010; Visel et al., 2013). (2) Some enhancers show spatial and temporal redundancy. (3) Approximately two-thirds of the candidate enhancers identified from human fetal brain tissues have orthologous sites in rodents, some of which may be implicated in neurodevelopmental or neuropsychiatric disorders (Bejerano et al., 2004; Sebat et al., 2007; Walsh et al., 2008; Cooper et al., 2011; Visel et al., 2013).

Hundreds of ultraconserved enhancers have been conserved over millions of years of evolution (Bejerano et al., 2004). While initial mutagenesis analyses found no overt phenotypes of deletion of some enhancers (Ahituv et al., 2007), it was likely that functional redundancy was the cause, particularly for the Arx locus because it has two pairs of enhancers that are active in the GEs or in the cortex (Ahituv et al., 2007; Colasante et al., 2008; Chiang et al., 2008; Visel et al., 2013). Dickel et al. (2018) explored this further by generating single or multiple enhancer deletion mouse lines, specifically focusing on the region containing the Arx locus and eight nearby enhancers (Bejerano et al., 2004; Dickel et al., 2018). CRISPR/Cas9 was utilized to delete individual and pairs of enhancers. While Arx null mice are lethal (Kitamura et al., 2002; Colombo et al., 2007), these enhancer double mutants were viable and fertile (Dickel et al., 2018). However, there was significant decreased Arx expression in the corresponding embryonic brain regions. In line with Arx’s function in GE-derived IN and striatal neuron migration and differentiation (Kitamura et al., 2002; Colombo et al., 2007; Colasante et al., 2009; Vogt et al., 2014), one double mutant (hs119 & hs121 null) had a near total loss of cholinergic striatal neurons, and a ~20%–30% decrease in PV and VIP CINs (Dickel et al., 2018).

Epigenetic Functions of Dlx1/2/5, Gsx2, Lhx6, Nkx2.1, and Otx2 during GE Development

Recent genomic studies have identified noncoding REs that possess dynamic regional and temporal-specific activity across forebrain development (Visel et al., 2009, 2013; Nord et al., 2013, 2015; Silberberg et al., 2016). How TFs interact with REs that drive cell fate and differentiation is a burgeoning field. Two recent studies revealed sets of Dlx and Nkx2.1-bound REs, and how Dlx and Nkx2.1 transcriptionally interact with REs to drive GE subregion fate specification by combining bulk RNAseq, TF chromatin-immunoprecipitation followed by sequencing (ChIP-seq), and histone ChIP-seq (Sandberg et al., 2019).

To understand how Dlx TFs coordinately regulate the transcriptional programs that drive GE development, Lindtner et al. (2019) utilized a functional genomic approach by conducting integrated analyses on (1) DLX1/2/5 ChIP-seq and histone ChIP-seq from E11.5, E13.5, and E16.5 Dlx1/2 wild-type and null embryonic GEs, and (2) RNA seq from E13.5 Dlx1/2 wild-type and null GEs (Lindtner et al., 2019). Sequential expression of Dlx2/1/5 in GE VZ and SVZ represses TFs Hes5 and Otx2 to enable GE stem cells to undergo neurogenesis (Eisenstat et al., 1999; Yun et al., 2002; Petryniak et al., 2007; Long et al., 2009; Hoch et al., 2015a; Lindtner et al., 2019). Their data suggested that Dlx TFs bind at enhancers and transcriptional starting sites (TSS, promoters) to repress the genes that drive proliferative cell states (Lindtner et al., 2019). Dlx TFs also bind distal regulator elements, which promote gene expression that drives GABAeric neuron fate specification and differentiation. For instance, Dlx promotes Sp8 expression, which drives LGE fate. Through DLX and H3K27ac ChIP-seq, they identified several Sp8 distal regulatory sequences bound by DLX that correspond to characterized GE enhancers. Transcription assays validated that DLX TFs promote RE-dependent transcription, with some enhancers having stronger effects. When the RE binding affinity was interfered (using CRISPRi), E13.5 cultured LGE neurons exhibited down-regulated Sp8 RNA expression. Furthermore, TF binding motif analyses revealed multiple active REs (a.RE) and repressive REs (r.RE) that modulate gene expression to facilitate Dlx function in GE biology. These include (1) a.REs for Gad1/2, Arx, Cxcr4/ Cxcr7, and Nrxn3 genes that promotes IN fate specification and maturation, and (2) r.REs for Pax7, Lhx2, and Otp, whose repression prevents GE progenitors from adopting hypothalamus and midbrain neuronal fates (Long et al., 2009; Lindtner et al., 2019). Together these results are elucidating the transcriptional circuitry that drives the development of forebrain GABAergic neurons.

A similar functional genomic approach was adopted to explore the genome-wide binding activity of NKX2-1 in the developing basal ganglia (Sandberg et al., 2016). Their data suggested that NKX2-1 has both activating and repressive roles on RE regulation to mediate MGE development. NKX proteins recruit Gro/TLE proteins that facilitate chromatin condensation, which leads to reduced expression of TFs and regulators in the SHH/WNT/BMP signaling pathways to pattern the ventral CNS (Muhr et al., 2001; Patel et al., 2012).

NKX2-1 acts as a repressor in the MGE VZ, where it binds to r.REs to quell the expression of LGE genes (Sandberg et al., 2016). However, NKX2-1 adopts an activator role in the MGE SVZ/MZ and binds to a.REs to promote the expression of Lhx6/Lhx8 and other genes that promote MGE-neuronal differentiation. In the MGE-derived CIN lineage, Nkx2.1 expression becomes repressed, in part via Zfhx1b (McKinsey et al., 2013), while Lhx6 is necessary and sufficient for the MGE-derived CIN maturation (Du et al., 2008; Vogt et al., 2014). Indeed, several genes expressed in CINs, such as Arx, Cxcr7, Maf, and Nxph1, are activated through LHX6-bound REs that lack NKX2-1 binding. On the other hand, two-thirds of the LHX6-bound REs have the NKX2-1 DNA-binding motif. These dual-bound distal REs are hypothesized to drive the expression of genes that are either specific to MGE-derived pallidal projection neurons, or with broad expression in cells within the MGE lineage.

These DLX1/2/5, NKX2-1, and LHX6 studies are unveiling the epigenetic complexity that coordinates gene regulation in GE development. They also support the hypothesis that REs serve as competitive binding hubs for cell-type-specific TFs, and it is the RE-TF interaction-driven combinatorial gene expression that facilitates neuronal subtype fate specification and development. Indeed, computational analyses of 10 different TFs binding to MGE-specific and LGE-specific enhancers showed that NKX2-1 and LHX6 were among the most highly correlated TFs binding to enhancers with MGE-specific activities (Silberberg et al., 2016).

MGE regional specification is driven by parallel functions of Nkx2.1 and Otx2 (Hoch et al., 2015a). OTX2 genomic binding provided evidence for its direct roles in regulating enhancers that control MGE regional specification and in promoting neurogenesis and oligodendrogenesis (Hoch et al., 2015a). As described earlier, LGE specification is driven by Gsx2. This TF has been shown to bind to as monomer to a monomer motif, and dimer to a motif; the monomeric motif is associated with repression and the dimer motif with activation (Salomone et al., 2021). The site where GSX2 and DLX co-bind is associated with gene activation in the LGE.

Molecular and Genetic Tools Utilizing Enhancers

Transgenic approaches to study cell lineages and genetic modifications in specific cell types have been widely adopted for rodent research, with a plethora of developed models (e.g., Taniguchi et al., 2011). It is difficult to translate these approaches to primate, human, or other organisms that are not readily accessible to gene manipulations. However, the use of viral vectors, such as lentiviruses and adeno-associated viruses (AAVs), has opened the door to such approaches. Recent progress with viral vectors that utilize conserved DNA enhancers to bias expression in GABAergic cells has drawn attention for cell type-targeted labeling. This has been recently highlighted and easily applied to different model organisms, with the potential for developing gene therapies (Nathanson et al., 2009; Betley and Sternson, 2011; Lee et al., 2014; Dimidschstein et al., 2016; Sun et al., 2016; Rubin et al., 2020; Vormstein-Schneider et al., 2020). While each virus type has advantages, some limitations exist: (1) there are multiple AAV serotypes with different tropisms, some with poor neuron transduction; (2) AAVs have a more limited genome packaging capacity (Wu et al., 2010; Choi et al., 2014); (3) lentiviruses have greater genomic packaging load but lack the variety of serotypes that have been worked out for AAVs; (4) unlike AAVs, lentiviruses integrate randomly into the host genome, potentially disrupting gene function in a subset of cells.

Regulatory elements (REs, also named enhancers) are noncoding DNA fragments that regulate gene transcription through DNA interactions with TFs (Long et al., 2016; Visel et al., 2013). Transgenic in vivo studies have identified evolutionarily conserved REs that have activity patterns similar to nearby genes (Visel et al., 2013; Silberberg et al., 2016). Coupled with a fluorescent reporter and/or Cre expression, REs have become a useful tool for lineage tracing. For instance, I12b, a Dlx1/2 RE was utilized to generate a Cre line. This Dlx I12b-Cre is active during the development of most MGE and CGE-derived INs in the neocortex and hippocampus, recapitulating the expression pattern of Dlx genes (Potter et al., 2009). In addition, the 1056- CreERT2 and 1538-CreERT2 transgenic alleles are active in ventral and rostro-dorsal MGE progenitors, respectively (Silberberg et al., 2016). 799 is another MGE-specific RE that is proximal to the neurexophilin1 gene. The 799-CreERT2 line targets both PV and SST CINs (Silberberg et al., 2016). Finally, the line 1060-CreERT2 has restricted expression in the CGE MZ where postmitotic neurons accumulate and targets ~80% of the CGE-derived CINs as well as a subset of CGE-originated amygdala neurons (Silberberg et al., 2016).

Building from the successes of the enhancer-driven transgenic mouse lines, subsequent studies asked whether these REs could drive biased expression in GABAergic neurons using viral transduction. One of the first set of studies used the mouse DlxI1/2b enhancer to specifically drive GFP in GABAergic CINs using lentiviral delivery (Arguello et al., 2013; Vogt et al., 2014). Using a number of conserved enhancers, including DlxI12b, Chen et al. showed that specific progenitor and neuronal cells types could be labeled using enhancer-driven lentiviruses in mouse embryonic stem cells programmed to be ventral telencephalic cells (Chen et al., 2013). Moreover, the DlxI12b enhancer has been used in human stem cells differentiated into mixed cultures including GABAergic neurons to label these cells in vitro for manipulation (Sun et al., 2016). Due to their capacity to carry large DNA cargoes, lentiviruses have also been used to combine biased expression of a reporter using an enhancer such as DlxI12b with the delivery of other genes. These approaches were able to drive Cre and/or a fluorescent reporter and specific genes under the control of an enhancer. These studies included assays to complement/rescue genetic loss of function in CINs, as well as screen how human genomic variants impact CIN development at the molecular, cellular, and electrophysiological levels (Vogt et al., 2014, 2015, 2018; Elbert et al., 2019; Malik et al., 2019; Wundrach et al., 2020). Overall, lentiviruses have shown promise, but in part due to their biosafety concerns, other viral types have been explored.

AAVs are safer than lentiviruses but carry less DNA. To this end, many AAV studies have focused on solely driving reporters/single genes to manipulate specific cell types using distinct enhancers to drive expression in GABAergic and other cell types. Several studies have taken advantage of the increased knowledge about RE activity patterns, TF interactions with REs, and the small size of REs. These studies utilized REs to give cell type and temporal specificity for targeting expression to cortical pyramidal and INs (Visel et al., 2013; Pattabiraman et al., 2014; Dimidschstein et al., 2016; Silberberg et al., 2016; Graybuck et al., 2021). For instance, REs around Dlx1/2 (I12b) and Dlx5/6 (I56i and I56ii) have been successfully utilized to generate AAVs, which reliably bias expression in a pan-GABAergic fashion across different species, and can be applied in vitro and in vivo (Lee et al., 2014; Cho et al., 2015; Dimidschstein et al., 2016; Mehta et al., 2019; Rubin et al., 2020).

When these enhancers are utilized to drive chemogenetic receptor DREADD and opsin expression in conjunction with AAV shuttling, it allows in vivo manipulation and activity to be monitored in INs (Dimidschstein et al., 2016; Vormstein-Schneider et al., 2020). Furthermore, REs close to genes Arl4d, Dlgap1, and Scn1a were recently identified that demonstrate robust expression in AAV vectors and a strong tropism toward fast-spiking PV CINs (Rubin et al., 2020; Vormstein-Schneider et al., 2020). When coupling Arld4d/Dlgap1 RE-driven AAV vector with channelrhodopsin expression, behavioral deficits derived from the lack of PV IN activity can be optogenetically rescued in the Dlx5/6 constitutive mutant, further extending the RE-AAV application (Cho et al., 2015; Rubin et al., 2020). Intriguingly, aside from IN enhancers, Graybuck et al. validated several enhancers—mscRE4 and mscRE16 provide specific labeling toward deep layer (layer V/VI) pyramidal neurons through single-cell chromatin accessibility analysis (Graybuck et al., 2021). They further showcased that multiple RE copies can be cloned into AAV vectors to enhance gene expression, and that the compatibility of the RE vector with the PHP.eB serotype allows better blood–brain barrier penetration and broader application when used with different transgenic lines (Graybuck et al., 2021).

All the above examples showcase the broad utility of enhancer elements (Fig. 47–2). Continuing efforts strive to better understand RE-TF temporal-spatial dynamic interactions and to polish RE-based molecular approaches that can facilitate cell-type-specific manipulation and therapeutics.

Figure 47–2.

Enhancer applications. Enhancers, also known as DNA regulatory elements (REs), regulate gene expression through their interactions with transcription factors (TFs). Complex TF-RE networks regulate temporal and spatial aspects of lineage specification. (more...)

Interneuron Classification and New Marker Discovery through Single-Cell RNA Sequencing

Historically, cell morphology, histochemical properties, intrinsic firing properties, and inter-neuronal connectivity have been utilized to classify neuronal subtypes (Rudy et al., 2011; DeFelipe et al., 2013; Greig et al., 2013). More recently, RNA expression has been employed to characterize cell subtypes. Tissue-based RNA sequencing facilitated identification of genes that specify pallial and subpallial subdomains and lineages across development. The temporal and spatial resolution from this method comes from researcher-driven microdissection of different parts of the brain and/or fluorescence- activated sorting (FACS). However, this can be labor-intensive and is subject to the accuracy of the dissection. The throughput and depth of the analyses are also limited.

Droplet and microfluidic-based high-throughput single-cell transcriptomic platforms that enable the profiling of neuronal and non-neuronal subtypes at higher temporal and spatial resolution have emerged in the past decade (Tasic et al., 2016, 2018; Paul et al., 2017; Mayer et al., 2018; Mi et al., 2018; Saunders et al., 2018). Taking advantage of the laminar-specific and IN-specific transgenic lines for targeted cell population enrichment, Tasic et al. used scRNA-seq to identify 49 clusters of cells from adult mouse visual cortex. Eighteen clusters belong to three broad GABAergic IN families (PV, SST, and VIP), and several “unique” markers were validated that can better subclassify INs (Tasic et al., 2016). For instance, cerebellum 4 (Cbln4) coexpression with Sst and Calretinin is predominantly present in upper-layer INs, suggesting the specific role of Cbln4 in labeling Martinotti cells. Similarly, Cpne5 was identified to preferentially label upper-layer PV+ chandelier INs, which was corroborated in a chandelier cell-enriched RNA-seq dataset (Tasic et al., 2016, 2018). Ndnf coexpresses with Reelin in the superficial layer in the absence of MGE-lineage markers, suggesting that it marks neurogliaform cells. Indeed, Ndnf-cre-lineage INs are mostly present in layer I and have morphological and physiological features of neurogliaform CINs (Tasic et al., 2016, 2018).

Several studies aimed to uncover heterogeneity and diversity in GEs and along IN developmental trajectories using scRNA-seq (Chen et al., 2017; Mayer et al., 2018; Mi et al., 2018). One study profiled the transcriptomes from dMGE, vMGE, and CGE at E12.5 and E14.5 (Mi et al., 2018). Through semi-supervised analysis (clustering using researcher annotated gene list), temporal factor appeared to be the main driving force that segregates E12.5 and E14.5 VZ/SVZ progenitor cells. This finding implies that GE progenitors, especially in the SVZ, may have high dynamic turnovers that allow generation of different types of neurons. When cross-referencing conserved variable genes identified from adult IN subtypes (Tasic et al., 2016), at least 11 (out of 23) IN subclasses were identified at E12.5 and E14.5. Based upon the hypothesis that progenitors and newborn neurons share cardinal genes that determine lineages, potential PV and SST IN progenitors were identified and compared. Several markers were found enriched in the potential SST IN progenitors, including Epha5, Cdk14, and c-Maf (Mi et al., 2018). Nkx2.1-Cre generated c-Maf conditional deletion and c-Maf overexpression through MGE transplantation were adopted for validation. Loss of c-Maf in the Nkx2.1-Cre lineage resulted in SST+ CIN reduction in adulthood, while overt c-Maf expression led to a slight increase in the ratio of SST/PV INs, supporting the hypothesis that c-Maf may have roles in promoting SST fate (Mi et al., 2018). Based on a similar assumption that certain genes are conserved from precursor to mature INs, Mayer et al. conducted an integrated study of MGE/CGE cells (progenitors and immature neurons) from E13.5, E14.5, and Dlx6a-Cre enriched pan-GABAergic INs harvested at E18.5 and P10 (Mayer et al., 2018). Several candidate early markers were identified for SST progenitors (Sst, Tspan7, and Satb1) and PV progenitors (Mef2c, Erbb4, and Plcxd3). Encouragingly, Mef2c stands as a marker enriched in PV IN subtype across evolution, in rodents and in human (Habib et al., 2017; Mayer et al., 2018; Pai et al., 2020). Preferential loss of PV CINs was also observed in Dlx6a-Cre-generated conditional Mef2c mutant neocortices, reinforcing the essential role of Mef2c in PV IN production (Mayer et al., 2018).

The advancement of scRNA-seq technology has enhanced our knowledge of the temporal and spatial specification and classification of INs. However, many factors can confound the interpretation of these studies, which include but are not limited to sample handling, sequencing method, sequencing depth, and computational analysis algorithms. Therefore, in-depth molecular and cell biology studies should be pursued in parallel to validate these sequencing results.

The Mef2c finding of Mayer et al. (2018) was corroborated and extended by a study that surveyed the roles of Mafb and c-Maf in CIN development and maturation (Pai et al., 2020). Double deletion of Mafb and c-Maf led to reduced Mef2c expression in Nkx2.1-Cre- lineage CINs. Furthermore, both the Mafb/c-Maf double and Mef2c conditional mutants show preferential loss of PV CINs. When Mef2c was transduced into the Mafb/c-Maf double mutant MGE cells, PV expression was rescued, suggesting that Mafb/c-Maf and Mef2c may be in the same core transcriptional pathway that promotes PV IN fate (Mayer et al., 2018; Pai et al., 2019, 2020).

Contrary to the conclusion that c-Maf alone promotes SST fate (Mi et al., 2018), Pai et al. (2019) came to a different conclusion. c-Maf mutants did not show a reduction of SST CINs at E12.5, E15.5, and P0. Second, both c-Maf single and double deletion in the Nkx2.1-Cre lineage resulted in loss of MGE INs (both PV and SST) starting at P7. Thus, c-Maf function may be implicated in CIN survival and/or maturation, and it may not necessarily be required to promote specification of SST CINs (Mi et al., 2018; Pai et al., 2019, 2020).

Finally, studies using other species are required to study the transcriptomic features of CINs during vertebrate evolution (Ma et al., 2013; Boldog et al., 2018; Liu et al., 2021). This information will likely be of import in humans to help elucidate genetic contributions to neuropsychiatric disorders.

J. L. R. is the cofounder, stockholder, and a current scientific board member of Neurona Therapeutics, a company that studies the potential translational application of interneuron transplantation.

This work was supported by the research grants to: JLR from NIMH R01-MH081880 and NIMH R01-MH049428; DV from Spectrum Health-Michigan State Alliance Corporation. Both figures were created with BioRender.com; Figure 47–2 was adapted from the “Regulation of Transcription in Eukaryotic Cells” template (2021). Author contribution: ELLP, DV, and JLR conceptualized, drafted, and finalized the article; JSH had partial contribution to the second and fifth sections of this chapter.