Loss of Gsx1 and Gsx2 Function Rescues Distinct Phenotypes in Dlx1/2 Mutants

Mice lacking the Dlx1 and Dlx2 homeobox genes (Dlx1/2 mutants) have severe deficits in subpallial differentiation, including overexpression of the Gsx1 and Gsx2 homeobox genes. To investigate whether Gsx overexpression contributes to the Dlx1/2 mutant phenotypes, we made compound loss-of-function mutants. Eliminating Gsx2 function from the Dlx1/2 mutants rescued the increased expression of Ascl1 and Hes5 (Notch signaling mediators) and Olig2 (oligodendrogenesis mediator). In addition, Dlx1/2;Gsx2 mutants, like Dlx1/2;Ascl1 mutants, exacerbated the Gsx2 and Dlx1/2 patterning and differentiation phenotypes, particularly in the lateral ganglionic eminence (LGE) caudal ganglionic eminence (CGE), and septum, including loss of GAD1 expression. On the other hand, eliminating Gsx1 function from the Dlx1/2 mutants (Dlx1/2;Gsx1 mutants) did not severely exacerbate their phenotype; on the contrary, it resulted in a partial rescue of medial ganglionic eminence (MGE) properties, including interneuron migration to the cortex. Thus, despite their redundant properties, Gsx1 and -2 have distinct interactions with Dlx1 and -2. Gsx2 interaction is strongest in the LGE, CGE, and septum, whereas the Gsx1 interaction is strongest in the MGE. From these studies, and earlier studies, we present a model of the transcriptional network that regulates early steps of subcortical development. J. Comp. Neurol. 521:1561–1584, 2013. © 2012 Wiley Periodicals, Inc.

The Gsx2 and Dlx1 and -2 genes mediate their subcortical transcriptional programs in combination with the Ascl1 (Mash1) bHLH gene. A feature of the Dlx1/2 mutants is overexpression of Ascl1, Gsx1, and Gsx2 (Yun et al., 2002;Long et al., 2009a,b). We hypothesized that some of the Dlx1/2 mutant phenotype is caused by the increased levels of Ascl1, Gsx1, and Gsx2 and therefore set out to make compound mutants that reduce Ascl1, Gsx1, or Gsx2 dosage in Dlx1/2 mutants. Dlx1/2;Ascl1 compound mutants do not exhibit a rescue of Dlx1/2 mutant properties; rather, their phenotype is much more severe than that of the individual mutants, because Dlx1/ 2 and Ascl1 regulate parallel pathways of subcortical differentiation (Long et al., 2009a,b).
Here we explored the effect of reducing Gsx1 or Gsx2 expression in the Dlx1/2 mutants by making compound mutants. We performed our phenotypic analysis on the CGE, LGE, MGE, and septum and focused on the expression of transcription factors that are abnormally expressed in the Dlx1/2 mutants (Long et al., 2009a,b).
We also focused on GAD1 expression, given its fundamental importance in defining the GABAergic phenotype.
Eliminating Gsx2 function from the Dlx1/2 mutants rescued the increased expression of Hes5 (Notch signaling indicator), Olig2 (oligodendrogenesis indicator), and Gbx1 (unknown function). In addition, Dlx1/2;Gsx2 mutants, like Dlx1/2;Ascl1 mutants, showed an exacerbation of the Gsx2 and Dlx1/2 mutant phenotypes, including GAD1 expression, particularly in the LGE, CGE, and septum. On the other hand, eliminating Gsx1 function from the Dlx1/2 mutants (Dlx1/2;Gsx1 mutants) did not severely exacerbate their phenotype; rather, the mutants exhibited a partial rescue of MGE properties and MGE interneuron migration to the cortex. Thus, despite their partially redundant properties, Gsx1 and Gsx2 have distinct interactions with the Dlx1/2 mutants. We present a model of the transcriptional network that regulates early steps of subcortical development.

MATERIALS AND METHODS Animals
Mice were maintained under standard conditions with food and water ad libitum. All experimental procedures were approved by the Committee on Animal Health and Care at the University of California, San Francisco (UCSF). Mouse colonies were maintained at UCSF, in accordance with National Institutes of Health and UCSF guidelines. Mouse strains with a null allele of Dlx1, Dlx2, Gsx1, and Gsx2 were used in this study (Anderson et al., 1997b;Casarosa et al., 1999). These strains were maintained on a CD-1 background. For staging of embryos, midday of the vaginal plug was calculated as embryonic day 0.5 (E0.5). PCR genotyping was performed as described elsewhere (Anderson et al., 1997b;Casarosa et al., 1999). Gsx1 and -2 genotyping was performed as described by Yun et al. (2003) and Wang et al. (2009). Tissue preparation, in situ hybridization, and immunofluoresence Preparation of sectioned embryos, immunofluoresence, and in situ hybridization were performed using digoxigenin riboprobes on 20-lm frozen sections cut on a cryostat using methods described by Long et al. (2007Long et al. ( , 2009a. We used a rabbit polyclonal anti-GSX2 antibody (Toresson et al., 2000), a guinea pig polyclonal anti-DLX2 (Kuwajima et al., 2006), and a mouse monoclonal anti-MASH1 (BD Pharmingen, San Jose, CA) Riboprobes have been described by Long et al. (2009a,b), except for Ngn2 (1.5-kb mouse full coding sequence from Francois Guillemot).

Antibody specificity characterization
DLX2 immunoreactivity closely matches endogenous Dlx2 RNA expression and disappears in the brain of Dlx1/2 À/À mutants (Long et al., 2007). In Dlx1/2 mutants, a deletion removes most of the coding exons for both Dlx1 and Dlx2 but does not remove exon 1, so an Nterminal truncated protein could be produced. Because the guinea pig anti-DLX2 antibody was made to the N-terminal amino acids (1-154; Kuwajima et al., 2006), we conclude that very little of this protein is present in the Dlx1/2 À/À mutant brain (Table 1).
GSX2 immunoreactivity closely matches endogenous Gsx2 RNA expression and disappears in the brain of Gsx2 À/À mutants (Toresson et al., 2000). Moreover, it does not recognize GSX1 even when Gsx1 is overexpressed (Pei et al., 2011). ASCL1 antibody specificity was tested using lysate from rat embryonic brain by Western blot; the antibody specifically recognized a 34-kDa protein (information provided by the manufacturer; BD Pharmingen product 556004). Furthermore, its immunoreactivity in the sections from the embryonic mouse brain closely resembles the expression of ASCL1 RNA, as detected here by in situ hybridization.

Microscopy
Images of in situ hybridization results were captured with a Zeiss AxioCam MR (Zeiss, Thornwood, NY) and saved as TIFF files. Images of immunofluoresence were captured with a Zeiss LSM 510 confocal microscope. The images were then processed in Adobe Photoshop CS3 to optimize the contrast and brightness to illustrate best the gene expression patterns.
Analysis of the number of DLX2-, GSX2-, and ASCL1-expressing cells  were visualized on 20-lm coronal forebrain sections from E10.5 and E12.5 wild-type mice by immunofluorescence confocal microscopy. The images were imported into Adobe Photoshop CS3, and a rectangle encompassing the VZ and SVZ domains was placed orthogonal to the ventricular surface (see Fig. 1). The labeled cells in the VZ, SVZ1, and SVZ2 within each rectangle were manually counted from one section by using the Photoshop counting tool. We calculated the mean size of the stained part of the cell (nucleus) separately for each population of neurons we counted, and at each age, and then corrected the profile counts separately for each population using the Abercrombie equation with the mean nuclear diameter for that population (Guillery, 2002). The average size of the staining was as follows at E10.5: Dlx2, 4.3 6 0.2 lm (SEM); Gsx2,. 4.2 6 0.2 lm; Ascl1, 4.0 6 0.2 lm; and, at E12.5: Dlx2, 4.3 6 0.2 lm (SEM); Gsx2, 4.5 6 0.2 lm; Ascl1, 4.0 6 0.2 lm. For Table 2, we used the following scoring system to describe the number of positive nuclei/section: 1þ ¼ 1-9, 2þ ¼ 10-29, 3þ ¼ 30-59, 4þ ¼ !60.

Qualitative analysis of gene expression changes
In describing the gene expression changes between control and mutant brains, in the text and in Figures 16  and 17, we made our best judgment assessments, independently by two or three people during at least two separate rating sessions. Similar approaches and figures were used by Long et al. (2009a,b) in describing gene expression changes in the Dlx1/2 À/À , Ascl1 À/À , and Dlx1/2;Ascl1 À/À mutants, so we have used a related system in this paper to allow comparison between these mutants. In Figures 16 and 17, we used a color-coded scoring system in which black indicates that expression was not analyzed (if no squares are listed, this also means that this analysis was not performed); gray indicates that expression was not clearly changed in the mutant; white indicates no detectable expression; red indicates severe reduction in expression; orange indicates moderate/mild reduction in expression; green indicates ectopic expression; and blue indicates increased expression. If the box is subdivided diagonally, the top part corresponds to the dorsal region, the bottom to the ventral region.
Expression at E12.5 and E15.5 showed similar results ( Fig. 1D-F, and not shown). The main difference was the formation of the SVZ. Previously, we presented evidence that the SVZ consists of two layers, SVZ1 (adjacent to the VZ) and SVZ2 (Yun et al., 2002;Petryniak et al., 2007). DLX2 was robustly expressed in nearly all cells in SVZ1 and SVZ2. The intensity of GSX2 expression decreased as cells moved from the VZ to SVZ1, although most cells continued to express detectable GSX2 and coexpressed DLX2 and ASCL1 (Table 1). However, in SVZ2, GSX2 and LGE expression of ASCL1, DLX2, and GSX2 proteins using two-color immunofluoresence, at E10.5 (A-C) and E12.5 (D-F). Cells (nuclei) expressing both proteins are yellow; green and red correspond to specific transcription factors defined at the top of each panel. The boxed areas in each panel are shown at higher magnification in A 0 -F 0 . E12.5 data showing previously published information about the relationship between Gsx and Dlx expression in Gsx2 and Dlx1/2 mutants. Dlx1/2 mutants overexpress Gsx1 (G,H) and Gsx2 (I,J). Gsx2 mutants express less Dlx1, especially in the dorsal LGE (K,L). LGE, lateral ganglionic eminence; MZ, mantle zone; SE, septum; SVZ1 and SVZ2, subventricular zones 1 and 2. Scale bars ¼ 20 lm in F (applies to A-F); 20 lm in F 0 (applies to A 0 -F 0 ); 500 lm in L (applies to G-L). See data in Figure 1. A dash indicates that Gsx2 þ cell density was too high to count with accuracy but clearly was greater than Ascl1 density. The scoring system for positive nuclei/section is 1þ ¼ 1-9, 2þ ¼ 10-29, 3þ ¼ 30-59, 4þ ¼ !60.
ASCL1 expression was at background levels, except in occasional cells. Analysis of GSX1 protein expression was hampered by the lack of a specific antibody. However, previous analyses showed that expression of Gsx1 RNA, and EGFP from a BAC transgenic, is largely complementary to that of Gsx2 at E12.5 and later stages. Whereas Gsx2 was highly expressed in the VZ, Gsx1 expression began in SVZ1. Furthermore, although their expression overlapped along the dorsoventral axis of the subpallium, Gsx2 was most strongly expressed in the LGE, CGE, and septum, and Gsx1 was most strongly expressed in the MGE (Toresson et al., 2000;Yun et al., 2001;Pei et al., 2011). Overall, GSX2 expression in the VZ was temporally upstream of DLX2 expression; as progenitors mature to the SVZ1 state, there generally was coexpression of GSX2, ASCL1, and DLX2. We used analysis of Gsx2 À/À ;Dlx1/2 À/À (Gsx2;Dlx1/2) compound mutants to assess the effects of losing expression of these transcription factors in the same progenitor cells.
Here we further investigated Gsx2 function, with the goal of understanding the ramifications of its upregulation in the Dlx1/2 mutant SVZ (Long et al., 2009a,b). We studied the phenotype of the Gsx2;Dlx1/2 compound mutants to determine 1) the combined functions of Dlx1/ 2 and Gsx2 and 2) whether some of the Dlx1/2 mutant phenotype was reversed by removing expression of Gsx2.
At E12.5, Gsx2 mutants lose expression of VZ progenitor regulators (Ascl1, Dlx1, and Vax1) in the dLGE and likewise have reduced vLGE expression of these transcription factors ( Fig. 2B,F,J). Loss of Dlx1/2 has modest effects on the LGE VZ properties at E12.5 (Fig. 2C,G,K,O), but combined loss of Dlx1/2 and Gsx2 function (Gsx2;Dlx1/2 compound mutants) greatly reduced LGE VZ properties, judged from the accentuated reduction of Ascl1, Dlx1, and Vax1 expression as well as the more ventral expansion of Ngn2 (cortical) expression ( Fig. 2D,H,L,P). SVZ properties of the LGE are also greatly reduced in the triple mutant, judged from expression of Arx, Dlx1, Gad1, and Vax1 (Figs. 2, 5). The LGE generates the striatum, so we assessed expression of LGE MZ markers in the Gsx2;Dlx1/2 compound mutants. Expression of Gad1 was not detected, but expression of Ebf1, FoxP4, and Islet1 were preserved, albeit reduced (see Fig. 5H,L,P,T).
On the other hand, by E15.5, the LGE of Gsx2;Dlx1/2 compound mutant largely recovered its morphology, progenitor cell properties, and expression of some striatal markers (e.g., FoxP4; Figs. 5, 12), presumably because of the Gsx1-mediated rescue (see Toresson and Campbell, 2001;Yun et al., 2003). However, despite the recovery of some LGE/striatal properties, expressions of Arx and Gad1 were less than in the single mutants (Figs. 5D,P, 12D,L); thus, together, Gsx2 and Dlx1/2 have a central role in promoting expression of Arx and Gad1 in striatal progenitors and neurons.
Overall, at E12.5 the Gsx2;Dlx1/2 compound mutants had a severe defect in regional specification of the LGE and CGE that was more severe than that in either the Dlx1/2 or the Gsx2 mutants. By E15.5, although there was partial recovery of the LGE phenotype, the CGE phenotype was not rescued. Furthermore, together, Gsx2 and Dlx1/2 were essential for expression of Gad1 in LGEand CGE-derived neurons.
Combined functions of Gsx2 and Dlx1/2 are required to maintain Arx and Gad1 expression in the MGE Specification of MGE progenitors (based on expression of Nkx2.1 and Lhx6), unlike the LGE and CGE, was not strongly affected by loss of either Gsx2 or Dlx1/2 (Yun et al., 2003;Long et al., 2009b). To test whether this could be due to compensation by Gsx2 and Dlx1/2, we studied the Gsx2;Dlx1/2 compound mutants. However, as in the individual mutants, the triple mutant continued to show relative normal indices of MGE regional identify, including VZ expression of Nkx2.1, Nkx6.2, and Olig2 (Figs. 4L, 9P,S).
Loss of Gsx1 and Gsx2 function in dlx1/2 mutants The Journal of Comparative Neurology | Research in Systems Neuroscience MGE MZ properties were also defective in Dlx1/2 mutants, including a small globus pallidus (Gbx1 þ , Lhx6 þ , Nkx2.1 þ ; Long et al., 2009b;Fig. 10O,R,V). This defect was similar in the Gsx2;Dlx1/2 compound mutant, although there may be increased numbers of Lhx6 þ and Nkx2.1 þ cells in the MZ at E15.5 (Fig. 10S,W). Tangential migration of Lhx6 þ cells to the cortex remained strongly blocked (Fig. 10S); this should be compared with a partial rescue of this phenotype in the Gsx1;Dlx1/2 compound mutant (discussed below; see Fig. 14).

Loss of Gsx1 and Gsx2 function in dlx1/2 mutants
The Journal of Comparative Neurology | Research in Systems Neuroscience from ectopic Dbx1 expression (Toresson and Campbell, 2001;Yun et al., 2003). Here we investigated Gsx1 function because its expression is increased in the Dlx1/2 mutants (Long et al., 2009a,b). We studied the phenotype of the Gsx1;Dlx1/2 compound mutants to determine whether some of the Dlx1/2 phenotype was caused by overexpression of Gsx1. Analysis was performed at E15.5.
First, we identified some subtle phenotypes in the Gsx1 mutant. The septum had the most obvious phenotypes, with reduced expression of Dlx1 and Dlx2 in the VZ; reduced Hes5 in the SVZ; and reduced Dlx2, Gbx1, Lhx6, and Nkx2.1 in the MZ (particularly medial septum); Gbx1 expression was increased in the SVZ (Fig. 13R). The LGE showed reduced expression of Dlx1 and Dlx2 in the VZ (ventral more severe than dorsal; Fig. 13F,J) and reduced Gad1 expression in the SVZ (Fig. 13N); Gbx1 expression was increased in the VZ (Fig. 13R).
GSX2. As the VZ cells mature, scattered cells express ASCL1 and DLX2, most of which coexpress GSX2. By E12.5, many LGE progenitors (VZ þ SVZ) coexpress GSX2, ASCL1, and DLX2 (Table 1). Coexpression is strongest in SVZ1, the part of the SVZ adjacent to the VZ. The Dlx1/2;Ascl1 mutant phenotype of subpallial progenitors and neurons showed much more severe defects than either the Dlx1/2 or the Ascl1 mutants (Long et al., 2009a,b). Likewise, Gsx2 and Ascl1 double mutants showed more severe defects than the single mutants ). Here we demonstrated a functional interaction between Gsx1 and Gsx2 with Dlx1/2 and provided evidence for the functional hierarchy of Gsx2, Gsx1, Ascl1, and Dlx1/2. We suggest that these phenotypes are due in large part to cell autonomous defects, particularly in the SVZ1, where GSX2, ASCL1, and DLX2 are coexpressed.

Gsx2 homeodomain: top of the hierarchy of dLGE/dCGE identity
We propose that Gsx2 promotes the identity of primary progenitors in the VZ of the dLGE and dCGE. Gsx2 null mutants fail to specify dorsal parts of the LGE and CGE, showing reduced expression of other transcription factors that mark the VZ of these regions (Ascl1, Dlx2, Olig2). Our loss-of-function analysis is consistent with ectopic expression experiments in which cortical misexpression of Gsx2 induces Ascl1 and Dlx1/2 . Therefore, we hypothesize that Gsx2 promotes the expression of Ascl1, Dlx2, and Olig2, from which emanate three major pathways (Fig. 18): 1) neural differentiation driven by Dlx1 and -2; 2) lateral inhibition to promote the maintenance of multipotent progenitors driven by Ascl1 promoting Delta expression, which in turn increases Notch signaling and Hes5 expression; and 3) progenitor cell maintenance through Hes5 and competence to produce oligodendrocytes through Olig2.
Gsx1 homedomain is upregulated in the absence of Gsx2 and Dlx1/2 and contributes to MGE phenotypes in the Dlx1/ 2 mutants Previous studies showed that Gsx1 mutants have a very mild telencephalic phenotype. They have increased Gsx2 expression (Pei et al., 2011) and ectopic expression of Dbx1, a marker of the ventral cortex and preoptic area; the ramifications of this are not known. Gsx2 mutants are partially rescued by Gsx1, providing evidence that Gsx1 can compensate for Gsx2. Combined removal of Gsx1 and -2 leads to misspecification of the dorsal and ventral LGE (Toresson and Campbell, 2001;Yun et al., 2003). Overexpression of Gsx1 throughout the telencephalon Figure 16. Expression of transcription factors in the ventricular zone (VZ), subventricular zone (SVZ), and mantle zone (MZ) of the LGE, MGE, and CGE in the Gsx2 À/À (Gsx2), Dlx1/2 À/À (Dlx), and Gsx2 À/À ;Dlx1/2 À/À (Gsx2/Dlx) mutants at E12.5 and E15.5. This figures depicts, as discrete boxes, the VZ, SVZ, and MZ of the CGE, LGE, MGE, and septum. The genes are listed alphabetically. The effect of each mutation on transcription factor expression in each box is indicated using a color code. Black indicates that expression was not analyzed (if no squares are listed, this also means that this analysis was not performed). Gray indicates that expression was not clearly changed in the mutant. White indicates no detectable expression. Red indicates severe reduction in expression. Orange indicates moderate/mild reduction in expression. Green indicates ectopic expression. Blue indicates increased expression. If the box is subdivided diagonally, the top part correspond to the dorsal region, the bottom to the ventral region.
Loss of Gsx1 and Gsx2 function in dlx1/2 mutants The Journal of Comparative Neurology | Research in Systems Neuroscience (tetO-Gsx1-IRES-EGFP;Foxg1tTA/þ mice) induced Ascl1 and Dlx1/2. This result is similar to the phenotype of Gsx2 overexpression, confirming that Gsx1 and Gsx2 share properties (Pei et al., 2011). On the other hand, misexpressions of Gsx1 and Gsx2 show opposite effects on the switch between proliferation and differentiation: Gsx1 lengthens the cell cycle and promotes neurogenesis (and represses Gsx2 expression), whereas Gsx2 maintains the progenitor state Pei et al., 2011). Thus, Gsx1 and Gsx2 share functions in promoting subpallial identity and appear to have opposite functions in regulating the switch between progenitor and neuronal fates.
Gsx2 and Dlx1 and -2 are negative regulators of Gsx1 (Fig. 1G-L ;Toresson et al., 2000;Yun et al., 2001;Long et al., 2009a,b). Dlx1 and -2 repression of Gsx1 was explored here by making Gsx1;Dlx1/2 mutants. We found that loss of Gsx1 partially rescued MGE phenotypes in the Dlx1/2 mutant, including interneuron migration to the cortex (see below). As noted above, overexpression of Gsx1 promoted neurogenesis and repressed proliferation (Pei et al., 2011); therefore, we had anticipated that removing Gsx1 in the Dlx1/2 mutants might further block their differentiation. On the contrary, the MGE of the Gsx1;Dlx1/2 mutants had a subtle reduction of Hes5 expression (indicator of Notch signaling) compared with Dlx1/2 mutants and had increased GAD1 and Lhx6 expression (indicators of differentiation; Fig. 14). Thus, in the Dlx1/2 mutant MGE, removing Gsx1 reduced Hes5 (Notch signaling), which is similar to the effect of removing Gsx2 in the Dlx1/2 mutant (Fig. 4).
Two functions of Ascl1 (Mash1) bHLH: promoting the subcortical progenitor state through Notch signaling and promoting subcortical differentiation with Gsx2 and Dlx1/2 Previous studies demonstrated that Ascl1 promotes the subcortical progenitor state through cell autonomously increasing Delta1 expression and cell nonautonomously (through lateral inhibition) increasing Notch Figure 17. Expression of transcription factors in the ventricular zone (VZ), subventricular zone (SVZ), and mantle zone (MZ) of the LGE, MGE, and CGE in the Gsx1 À/À (Gsx1), Dlx1/2 À/À (Dlx), and Gsx1 À/À ;Dlx1/2 À/À (Gsx1/Dlx) mutants at E15.5. This figure depicts, as discrete boxes, the VZ, SVZ, and MZ of the CGE, LGE, MGE, and septum. The genes are listed alphabetically. The effect of each mutation on transcription factor expression in each box is indicated using a color code. Gray indicates that expression was not clearly changed in the mutant. White indicates no detectable expression. Red indicates severe reduction in expression. Orange indicates moderate/mild reduction in expression. Green indicates ectopic expression. Blue indicates increased expression. If no squares are listed, this means that this analysis was not performed. If the box is subdivided diagonally, the top part correspond to the dorsal region, the bottom to the ventral region. Asterisk indicates that diagonal band GAD1 expression was absent. signaling and repressing Dlx expression (Casarosa et al., 1999;Horton et al., 1999;Yun et al., 2002;Castro et al., 2006); this has the effect of repressing neurogenesis and promoting gliogenesis, including oligodendrogenesis (Parras et al., 2007;Petryniak et al. 2007). Ascl1 mutants continue to express Gsx2 at roughly normal levels at E12.5 (Wang et al., 2009) and E15.5 (Long and Rubenstein, unpublished).
Ascl1;Gsx2 compound mutants have a severe reduction in LGE differentiation , despite continued expression of Gsx1. Thus, Gsx2 and Ascl1 together contribute to specifying the LGE developmental program.
Analysis of Ascl1;Dlx1/2 individual and compound mutants provided evidence for distinct Ascl1-and Dlx1/ 2dependent pathways of LGE/dCGE development; we have proposed that the Ascl1 pathway operates through promoting the expression of Hes5, Olig2, and Sp9; the Dlx pathway components are described below (Long et al., 2007(Long et al., , 2009a. Gsx2 also is a positive regulator of Ascl1, Hes5, and Sp9 (which could occur through Ascl1;Figs. 2, 6, 9;Wang et al., 2009). Thus, Gsx2 and Ascl1 share common regulatory functions for promoting Notch signaling (based on Hes5 expression) and Sp9 expression that distinguish them from Dlx1/2 function. Ascl1;Dlx1/2 compound mutants have greatly reduced subcortical differentiation but continue to express limited aspects of subcortical identity, based on expression of GAD1 and truncated Ascl1 and Dlx1 RNAs; we have postulated that subcortical identity is maintained in these mutants through the function of a few key transcription factors, including Gsx1 and -2 and Islet1 (Long et al., 2009a,b). Thus, to evaluate the core functions of Gsx1 and -2, we generated the Gsx2;Dlx1/2 mutants and Gsx1;Dlx1/2 mutants.

Elimination of Gsx2 partially rescues subpallial Notch signaling and oligodendrogenesis deficits in Dlx1/2 mutants
The Dlx genes promote LGE/dCGE development through controlling the expression of multiple transcription factors (Figs. 2, 4-6, 11;Long et al., 2009a,b). Generally, they repress the expression of transcription factors that promote the progenitor and/or glia cell state, including Ascl1, Gsx1 and -2, Hes5, and Olig2. The block in subcortical neural differentiation in Dlx1 and -2 mutants may be due, in part, to persistent expression of transcription factors that promote progenitor cell properties. For instance, in the Dlx1 and -2 mutants, there is overexpression of Olig2 that is linked to their overproduction of oligodendrocytes (Petryniak et al., 2007). This phenotype is reversed in Ascl1;Dlx1/2 compound mutants (Petryniak et al., 2007). Compound Gsx2;Dlx1/2 mutants also show a recovery of specific aspects of the Dlx1/2 mutant phenotype. In the Gsx2;Dlx1/2 mutants, there was rescue of progenitor zone overexpression of several transcription factors, including Ascl1, Hes5, Olig2, and Gbx1 (Figs. 2-7, 11). The reduction in Ascl1 and Hes5 expression at E12.5 and E15.5 provides evidence that the increase in Notch signaling present in the Dlx1/2 mutants is mediated by Gsx2 expression. The reduction in Olig2 expression provides evidence that the Ascl1-mediated increase in oligodendrogensis present in the Dlx1/2 mutants (Petryniak et al., 2007) is mediated via Gsx2 promoting Ascl1 expression.
Elimination of Gsx2 exacerbates LGE/dCGE specification and differentiation defects in Dlx1/2 mutants Although certain features of the subpallial progenitors are partially rescued in the Gsx2;Dlx1/2 compound Figure 18. Model of transcription factor network interactions in the developing LGE based largely on loss-of-function analyses (see Discussion). Green arrows indicate activation; magenta squares indicate inhibition. Genes activated by Dlx1 and -2 that have asterisks correspond to genes whose expression is most strongly reduced in the Dlx1/2 À/À mutant. At the top of the hierarchy is Gsx2; Gsx2(1) indicates that Gsx1 can compensate for loss of Gsx2. For comparison see Long et al. (2009a,b) for summaries of gene expression changes in the Dlx1/2, Ascl1, and Dlx1/2;Ascl1 mutants, using the same schemata. mutants, other features are worsened compared with the Gsx2 and Dlx1/2 mutants, including regional specification/patterning and neural differentiation. The LGE, CGE, and septum show regional patterning defects, including ectopic expression of pallial markers (Ngn2), loss of subpallial markers (Ascl1, Dlx1, Vax1), and frank hypoplasia, especially of the CGE (Figs. 2, 4, 6, 11). We postulate that reducing Ascl1 levels in the Gsx2;Dlx1/2 compound mutants is an important mechanism that contributes to their these phenotypes.
Elimination of Gsx1 partially rescues Dlx1/2 mutant MGE properties, including interneuron migration Whereas Gsx, Mash, and Dlx participate in MGE differentiation, regional and cellular fate specification in this region also operates through a parallel/overlapping program mediated by the Nkx2.1 and the Lhx6/7(8) genes (Sussel et al., 1999;Liodis et al., 2007;Zhao et al., 2008;Flandin et al., 2011). This may explain why MGE properties are relatively better preserved than LGE/CGE/septal properties in the Gsx2;Dlx1/2 mutants. For instance, at E12.5 Arx, GAD1, and Sp9 expression are maintained only in the MGE (Fig. 9). We suggest that the preserved expression of Ascl1 and Nkx2.1 plays a major role in maintaining the MGE developmental program in the Gsx2;Dlx1/2 mutants. Gsx1 appears to be more important than Gsx2 in MGE development. Remarkably, the Gsx1;Dlx1/2 mutants, but not the Gsx2;Dlx1/2 mutants, show a partial rescue of the migration of Lhx6 þ cells to the cortex (Figs. 13-15). Loss of Gsx1 normalizes molecular properties of the Dlx1/2 mutant MGE, including increasing Lhx6 and GAD1 expression and reducing Hes5 expression (Notch signaling marker; Fig. 14). Thus, Gsx1 overexpression in the MGE may repress its differentiation, including its ability to produce interneurons that can migrate to the cortex. Alternatively, Gsx1 overexpression may alter the expression of factors that directly promote interneuron migration. In either case, overexpression of Gsx1 contributes to the block of interneuron migration of the Dlx1/2 mutants. It should be noted that, despite the partial rescue, most MGE-derived interneurons in the Gsx1;Dlx1/2 mutants remain in the subpallium, as in the Dlx1/2 mutant. Many of these cells appear to coalesce as an ectopia in the caudoventral subpallium, which expresses Gad1, Lhx6, Nkx2.1, and Sp9 (Fig. 15).
Independent and combined functions of Gsx1 and Gsx2, with Dlx1/2, in septal development Septal and LGE development share many similarities, but a key difference in the transcription programs of the LGE and the septum is that septal expression of Dlx5/6 is preserved in the Dlx1/2 mutant (Long et al., 2009a). Furthermore, the septum and the ventral LGE are particularly sensitive to loss of Ascl1 function (Long et al., 2009a).