Dlx6 Regulates Molecular Properties of the Striatum and Central Nucleus of the Amygdala

Animals

Mice were maintained in 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.

We used the following strains in this study: Dlx6^LacZ/+ (Jeong et al., 2008), BAC-Dopamine receptor 1-GFP (D1R); BAC-Dopamine-receptor 2-GFP (D2R); and BAC muscarinic receptor M4-GFP (M4R) (http://www.gensat.org/index.html; Lobo et al., 2006). Through the information available to us, the expression patterns of these BAC-GFP strains appeared indistinguishable from the authentic expression patterns of these genes in the striatum, based on our in situ hybridization analyses (Long et al., 2009a and data not shown), the analyses of Lobo et al. (2006), and gene expression results shown at the Allen Institute for Brain Science (http://www.brainmap.org/). However, we did not examine the BAC-GFP transgene expression outside of the striatum and its projections, nor did we perform double labeling with antibody or in situ probes for the authentic genes; therefore we do not wish to imply that their expression is definitely eutopic.

For staging of embryos, midday of the vaginal plug was calculated as embryonic day 0.5 (E0.5). Polymerase chain reaction (PCR) genotyping was performed as described (Anderson et al., 1997b; Parras et al., 2004). Because no obvious differences in the phenotypes of Dlx6^+/+ and Dlx6^+/− brains have been detected, they were both used as controls. In the figures, we state explicitly which genotype was used as the control.

To generate the Dlx6-lacZ (Dlx6^LacZ) allele, a 4.3-kb SpeI-SnaBI genomic fragment spanning Dlx6 exon 3 (which encodes the homeodomain) was subcloned into the SpeI-SmaI sites of pBS KS+. Site-directed mutagenesis was performed to engineer a unique NruI site immediately following the 219th amino acid from the N-terminus of the Dlx6 protein (F of the sequence VKIWFQNKRS). Flanking genomic sequences were added between the unique 5′ XhoI site (5.8-kb 5′ homology arm) and the 3′ NotI site (3.5-kb 3′ homology arm). The reporter cassette IRES-lacZ-PGKneo (Robledo et al., 2002) was cloned into the unique NruI site to generate the final targeting construct. The Dlx6^LacZ allele therefore interrupts the Dlx6 protein immediately following the amino acid F, within the homeodomain sequence. This insertion prevents translation of one-third of the homeodomain and the entire C-terminal domain, including the nuclear localization signal (amino acids 220–228); thus, this allele is likely to be null, although, because no effective anti-DLX6 antibody is available, a precise analysis of the protein products from this allele has not been performed.

RNA preparation and gene expression array analysis

We dissected the rostral striatum from Vibratome sections of E18.5 wild type and Dlx6^LacZ/LacZ brains. RNA was purified and shipped to the NINDS/NIMH Microarray Consortium (TGEN; http://arrayconsortium.tgen.org/) where biotin-labeled cRNA hybridization probes were generated by using Affymetrix’s GeneChip IVT Labeling Kit (Santa Clara, CA), which simultaneously performs in vitro transcription (a linear ~20–60-fold amplification) and biotin labeling. The samples were hybridized to the Affymetrix Mouse Genome 430 2.0 array. TGEN uses GeneChip Operating Software (GCOS) to scan the arrays and to perform a statistical algorithm that determines the signal intensity of each gene (see Cobos et al., 2007; Long et al., 2009a, b for details; the hybridization results of these arrays are available at http://arrayconsortium.tgen.org).

Tissue preparation, in situ hybridization, and immunohistochemistry

We compared gene expression by either in situ hybridization or immunohistochemistry in wild type and Dlx6^LacZ/LacZ mutants. Preparation of sectioned embryos, immunohistochemistry, and in situ hybridization were performed with antibodies and digoxigenin riboprobes on 20-μm frozen sections cut on a cryostat by using methods described in Wang et al. (2009) and Long et al. (2009a, b). The antibodies and plasmids used in this analysis are described in Tables 1 and 2. The specificity of the antibodies is as follows:

1. Calbindin (CB). The calbindin D28k antibody was characterized by Western blot. The antibody reacts specifically to a single band at 28 kDa on Western blot of rat hippocampal cells (Kim et al., 2006). Immunostaining of mouse embryo brain sections produces a specific staining pattern of calbindin-positive neurons identical to that published previously (in the mouse brain atlas of Jacobowitz and Abbott, 1998).

2. Calretinin (CR). The calretinin antibody was characterized as described (Schwaller et al., 1993), and is also described on the antibody database of this journal (Fischer et al., 2007). Briefly, this antibody detects a single band corresponding to the size of calretinin on Western blot of brain homogenates from human, monkey, rat, and mouse. It labels a subpopulation of neurons in the wild-type mouse brain sections, but does not stain the brain of calretinin knock-out mice (Schiffmann et al., 1999).

3. CTIP2. The anti-CTIP2 monoclonal antibody 25B6 was characterized as described (Senawong et al., 2003). On immunoblot prepared from nuclear extract from Jurkat cells immunoprecipitated with anti-Sir2 antibody, the antibody detects two bands corresponding to two isoforms of CTIP2 at about 120 kDa. Immunohistochemistry in mouse brain sections shows expression patterns that are indistinguishable from the RNA expression patterns (Zhao et al., 2008).

4. DARPP-32. This antibody was characterized by the manufacturer with Western blot on Jurkat lysate, which shows that the antibody detects a single band at 23 kDa corresponding to DARPP-32. Imunohistochemistry shows patterns of cellular morphology and distribution in the cortex and striatum consistent with published reports (Pardo et al., 2006).

5. FoxP1. The FoxP1antibody was characterized by Western blot on Jurkat nuclear extract. According to the manufacturer, this antibody detects the largest three of four isoforms that have been shown to migrate at approximately 97, 68, and 50 kDa. Immunostaining shows a pattern of cellular morphology and distribution in the mouse brain consistent with published reports (Regad et al., 2009).

6. GFP. According to the manufacturer, this anti-green fluorescent protein (GFP) antibody detects a band on Western blot corresponding to the size of GFP. No staining was detected on brain sections from wild-type mice.

7. PH3. The antibody to phospho-histone H3 was characterized by the manufacturer with Western blot on acid extracts of colcemid-treated HeLa cells. It shows that the antibody detects a single band at 17 kDa corresponding to phospho-histone H3. Immunohistochemistry shows that the antibody labels mitotic cells.

X-gal histochemistry

Brains used for X-gal staining were fixed in 4% paraformaldehyde for 1 hour (for E12.5) or 2 hours (for E15.5 and E18.5) at 4°C, and then sectioned with a cryostat at 20 μm. Sections were rehydrated in 1X phosphate-buffered saline (PBS), incubated in a solution of 10 mM Tris-HCl, pH 7.3, 0.005% Na deoxycholate, 0.01% Nonidet P40, 5 mM K₄FeCN₆, 5 mM K₃FeCN₆, 2 mM MgCl₂, and 0.8 mg/ml X-gal for 24 hours to reveal cells expressing LacZ. Samples were washed twice with PBS, dehydrated in an ethanol series (50%, 70%, 90%, 100%) and xylene for 30 seconds each, and then mounted with Aqua poly-mount and a coverslip (Polysciences, Warrington, PA, #18606).

Microscopy

Images of in situ hybridization and immunohistochemistry results were captured by using a Zeiss AxioCam MR (Thornwood, NY) and saved as TIFF files. The images were then processed in Adobe Photoshop CS3 (San Jose, CA). Image backgrounds of the mutants and controls were adjusted to the same level.

Quantitative analysis of LacZ expression in the striatum and gene expression changes in central nucleus of the amygdala (CeA)

The density of LacZ expression in the striatum of Dlx6^LacZ/+ and Dlx6^LacZ/LacZ mice, and gene expression signals in the CeA of wild-type and Dlx6^LacZ/LacZ mice were quantified by using ImageJ. The images were imported into ImageJ, and regions of interest (ROIs) were drawn with the polygon selection tool. For each image, a ROI was drawn in the CeA region and another ROI in an area devoid of specific signals to measure background signal. The average brightness value in each ROI was then measured by the Histogram function. LacZ intensity in the striatum (Str) and in situ signal intensity in CeA were calculated by the following equations: I_Str equals; B_background − B_Str and I_CEA equals; B_background − B_CEA, respectively. I signifies signal intensity, and B signifies brightness. Because stronger (LacZ or in situ) signals result in smaller brightness values, B_background − B_Str will yield a positive value. In each pair of Dlx6^LacZ/+ and Dlx6^LacZ/LacZ or wild-type and Dlx6^LacZ/LacZ images, the ROIs have the same shape and size. The graph of gene expression signal intensity was then generated in Microsoft Excel.

Dlx6 and Dlx6^LacZ basal ganglia expression at E12.5, E15.5, and E18.5 in Dlx6^LacZ/+ and Dlx6^LacZ/LacZ

Previous studies have briefly described Dlx6 expression in the developing basal ganglia (Anderson et al., 1997b; Liu et al., 1997; Long et al., 2009a, b). Here, we examined in detail the expression of Dlx6 RNA by using in situ hybridization and β-galactosidase expression from the Dlx6^LacZ allele. (This allele appears to be a null, or at least a severe hypomorph; see Materials and Methods, and Jeong et al., 2008).

The level of Dlx6 RNA expression was low compared with other transcription factors in the E15.5 ganglionic eminences (Long et al., 2009a, b) or E18.5 striatum (Supplementary Table 1). For instance, at E15.5 Dlx6 expression in the MGE was roughly 20-fold less than Dlx1 and roughly 4-fold less than Dlx5 (see Table 1 in Long et al., 2009a). In the E18.5 striatum Dlx6 expression was roughly 10-fold less than Dlx1 and roughly 5-fold less than Dlx5. The relatively low level of Dlx6 RNA limited the sensitivity of the in situ RNA hybridization array analysis. Nevertheless, we did not detect an obvious difference between the Dlx6 RNA and βgalactosidase expression (Figs. 1, 2). Therefore, we concentrated our description on the β-galactosidase expression from the Dlx6^LacZ allele because it was easier to detect and it gave better cellular resolution.

Dlx6 expression at E12.5, E15.5, and E18.5, like other Dlx genes, encompasses most subpallial regions (Eisenstat et al., 1999; Stühmer et al., 2002b). However, unlike Dlx1 and Dlx2, Dlx6 expression was restricted to the SVZ of the septum, LGE, MGE, POA, and CGE, and mantle of the septum, nucleus accumbens (NAc), striatum (Str), and central nucleus of the amygdala (CeA).

Dlx6 expression was conspicuously low in the olfactory bulb (and the SVZ of the dLGE), olfactory tubercle (OT), and preoptic area, and was not detectable in the ventral pallidum, globus pallidus, medial amygdale, and cortical/hippocampal interneurons. The low expression in these regions contrasts with the expression of its closest paralogue, Dlx1 (Fig. 8C; Cobos et al., 2005; Long et al., 2007). This suggests that Dlx1 function is more important than Dlx6 in the olfactory bulb, olfactory tubercle, pre-optic area, ventral pallidum, globus pallidus, medial amygdale, and cortical/hippocampal interneurons, whereas Dlx6 may be more important than Dlx1 in the development of the CeA. Dlx5 and Dlx6 share similar patterns of expression, although Dlx5 expression was higher in the SVZ of the LGE and higher in the OT and ventral striatum/NAc (Figs. 3A, B, 4A, B). Dlx6, and its antisense transcript (Liu et al., 1997; Faedo et al., 2004; Feng et al., 2006; Bond et al., 2009), showed inverse expression gradients in the striatum and OT, and complementary pallial/subpallial expression (Fig. 3B, C). The function of the Dlx6 antisense transcript is uncertain, but it could regulate expression of Dlx6.

Next we assessed the effect of the Dlx6^LacZ/LacZ mutation on the expression of Dlx6. Generation of the Dlx6 mutation involved an insertion of the LacZ gene into exon 3 (see Materials and Methods and Robledo et al., 2002; Jeong et al., 2008). Thus, we could study the effect of the mutation on the expression of both the 5′ end of the Dlx6 gene (exons 1 and 2) and the LacZ gene. Expression of both the truncated Dlx6 RNA (exons 1 and 2) and β-galactosidase was increased in most regions of the subpallium in Dlx6^LacZ/LacZ compared with Dlx6^LacZ/+ (Figs. 1, 2). Within the E18.5 striatum (Str), the distribution of LacZ expression was non-uniform; it appeared patchy (lower expression in the patch-like domains) and showed a gradient (high lateral-low medial); this distribution of LacZ expression was disrupted in the Dlx6^LacZ/LacZ mutant. Furthermore, the mutant showed much higher, and perhaps ectopic, expression in the OT and core of nucleus accumbens (NAcC). The increase in β-galactosidase in the Dlx6^LacZ/LacZ mutant in the striatum and central nucleus of the amygdala (CeA) was roughly twofold more than in the Dlx6^LacZ/+ heterozygotes (Supplementary Fig. 5).

Only three structures in the mutant showed reduced Dlx6 RNA expression: the SVZ of the LGE and septum, and the mantle zone of the CeA (Figs. 1D, D′, H, H′, L, L′ P, P′, 3B, B′, 7B, B′, 8B, B′). Despite the reduced RNA expression, LacZ expression remained strong in the CEA (Fig. 2). This suggests that the CeA is produced, but may have a molecular defect. Below we describe molecular defects in the NAcC, Str, OT, and CeA in the Dlx6^LacZ/LacZ mutant.

Gene expression array analysis of the E18.5 Dlx6^LacZ/LacZ striatum

To obtain an unbiased identification of gene dysregulation in the Dlx6^LacZ/LacZ mutants, we performed gene expression array analysis in the E18.5 striatum. We dissected the rostral striatum from Vibratome sections of wild-type and Dlx6^LacZ/LacZ brains. RNA was prepared and used for hybridization to Affymetrix Mouse Genome 430 2.0 arrays. From these data we generated a list of transcription factors that are expressed in the wild-type striatum, and arranged the list as a function of their relative expression levels (Supplementary Table 1). We found that Dlx6 was expressed at much lower levels than most other transcription factors. For instance FoxP1, FoxG1, and Isl1 were expressed at 32-, 24-, and 15-fold higher levels, respectively. Unfortunately, comparison of wild type and Dlx6^LacZ/LacZ mutants did not reveal any robust changes in gene expression, including genes studied by in situ hybridization (Supplementary Table 2); all of the data are at http://arrayconsortium.tgen.org, even for genes showing expression changes found by the histochemical analyses described below. We believe that this is because the phenotype is subtle, and that many of the gene expression changes are alterations in RNA distributions rather than frank loss or gain of expression.

Changes in expression of LGE and striatal markers at E15.5 and E18.5 in Dlx6^LacZ/LacZ

β-Galactosidase expression in Dlx6^LacZ/LacZ telencephalon was disrupted in the striatum (changes in its gradient and patchy distribution), and showed much higher (perhaps ectopic) expression in the OT and NAcC (Fig. 2). We compared the expression of ~40 genes by either in situ hybridization or immunohistochemistry in wild type and Dlx6^LacZ/LacZ mutants; most of these markers were chosen because their expression was altered in the basal ganglia of the Dlx1/2^−/− mutants (Long et al., 2009a; Long et al., 2009b). Below we describe abnormalities in the expression for several of these genes (Figs. 1–4; Supplementary Figs. 1, 2).

At E15.5, Dlx5 expression was reduced in the mutant OT (Fig. 3A, A′). Furthermore, the relative levels of Dlx5 expression in the mutant LGE SVZ and striatum were similar (also seen in the septum), unlike in the wild type. At E18.5, these phenotypes were not detected (Fig. 4A, A′). Antisense Dlx6 expression at E15.5 showed a subtle increase throughout its expression domain (Fig. 3C, C′); at E18.5 its expression appeared increased throughout the striatum and septum (not shown). Reduced striatal expression was noted for Golf, RXRγ, and Tiam2 at E15.5 and E18.5 (Figs. 3D, D′, G, G′, H, H′, 4D, D′, E, E′, H. H′). Increased striatal expression was noted for Ikaros at E15.5 (Fig. 3E, E′); at E18.5 the increased expression became even more prominent in neurons adjacent to the SVZ (arrow; Fig. 4C, C′). Preprotachykinin (PPT) expression was lost in scattered cells in the region of the OT at E15.5 (arrow; Fig. 3F, F′); at E18.5, the patchy striatal PPT expression was greatly reduced, especially along the outer margin of the striatum (arrow; Fig. 4G, G′); Tiam2 also lost expression in this region (arrow; Fig. 4E,’E). Likewise, DARPP-32 striatal expression in patch-like domains was less distinct (Fig. 4I, I′). The gradient of Pbx3 striatal expression was altered, and its expression appeared increased in the NAc (arrow; Fig. 4F, F′).

Next, to study the effect of the Dlx6 mutation on axon projections from the striatum, we introduced three different BAC-GFP transgenes into the Dlx6^+/− background: BAC-Dopamine receptor 1-GFP (D1R); BAC-Dopamine-receptor 2-GFP (D2R); and BAC muscarinic receptor M4-GFP (M4R) (http://www.gensat.org/index.html; Lobo et al., 2006). D1R and D2R are preferentially expressed in the striatal neurons that project in the direct and indirect pathways, respectively, whereas M4R is robustly expressed in both (Lobo et al., 2006). For each transgene, we generated mice that were Dlx6^−/−; BAC-GFP^+/−, and then studied GFP expression at E18.5 (Figs. 5, 6; Supplementary Figs. 1, 2). For each of these transgenes, we did not observe major changes in expression in the striatum, nor an obvious defect in striatal axonal projects. However, in the Dlx6 mutant the M4R-GFP line showed increased GFP⁺ staining in the NAc (Fig. 5A, A′, B, B′; similar to Pbx3 in Fig. 4F, F′), and may have an abnormal accumulation of axons leaving the striatum (Fig. 5D, D′).

Finally, we followed the trajectory of the M4R-GFP axons to the midbrain and their target zone in the substantia nigra (Fig. 6). We compared expression of GFP, with glutamic acid decarboyxylase 1 (GAD1; marker of pars reticulata) and tyrosine hydroxylase (TH; marker of pars compacta). We did not observe a clear defect in the ability of the GFP⁺ striatal axons to project to the substantia nigra (Fig. 6).

Changes in expression of central nucleus of the amygdala (CeA) markers at E15.5 and E18.5 in wild type and Dlx6^LacZ/LacZ

In wild-type mice Dlx5 and Dlx6 were expressed in the CeA (Figs. 1H, P, 7A, B, 8A, B), whereas Dlx1 was expressed in the medial amygdala (MeA), bed nucleus stria terminalis intraamygdaloid (STIA), basomedial (BM), and corticoid (Co) nuclei, as well as in scattered cells of the intercalated nuclei (ITC) and within the pallial amygdala (lateral and basolateral) (Fig. 8C). In the Dlx6^LacZ/LacZ mutant, Dlx1 expression may have decreased in the STIA, BM, and Co (Fig. 8C′). Dlx2 expression in the amygdala was difficult to detect (data not shown).

Dlx6 RNA expression in Dlx6^LacZ/LacZ mutant showed a conspicuous loss of expression in the central nucleus of the amygdala (CeA) at E15.5 and E18.5 (Figs. 1H, H′, P, P′, 7B, B′, 8B, B′; Supplementary Fig. 4). Thus, we examined the CeA in greater detail by using markers we found that label the CeA, or nearby subpallial components of the amygdala. Like Dlx6, Dlx5 RNA was strongly decreased in the CeA at E15.5 and E18.5 (Figs. 7A, A′, 8A, A′; Supplementary Fig. 4). Reduced expression in the region of the CeA was also detected for Golf (arrow; Figs. 7C, C′, 8D, D′; Supplementary Fig. 4) and Pbx3 (Fig. 7D, D′). PPT expression in scattered cells in the CeA was reduced (arrow; Figs. 7E, E′, 8F, F′), as was its expression in the caudoventral striatum (Figs. 7E, E′, 8F, F′).

Dlx6 expression: comparison with Dlx1, Dlx2, and Dlx5: overlapping and unique features

The Dlx6^LacZ/LacZ mutant dies the day of birth with conspicuous craniofacial defects (Jeong et al., 2008). Although its forebrain shows no obvious morphological phenotypes, it does exhibit gene expression abnormalities in striatal and amygdaloid structures. The lack of severe forebrain defects probably reflects compensatory functions of Dlx1, Dlx2, and Dlx5 (Anderson et al., 1997a, b; Depew et al., 2002; Jeong et al., 2008). For instance, although the forebrain of Dlx1 mutants appears normal, subsets of its cortical and hippocampal neurons have molecular and functional defects that only appear postnatally (Cobos et al., 2005).

Dlx5/6^−/− mutants have been generated (Beverdam et al., 2002; Robledo et al., 2002), but their forebrains are difficult to study because of exencephaly (Depew et al., 2002). Nevertheless, we have obtained evidence that Dlx5/6^−/− mutants have defects in the migration and differentiation (based on transplantation of mutant cells into a wild-type cortex) of MGE-derived interneurons (Wang et al., 2010). We did not observe evidence of LGE molecular defects at E15.5; analysis of their striatal development was not possible due to the abnormal morphology of this region. Generation of a Dlx5&6 conditional mutant will be important in establishing the function of Dlx5&6 in the developing and postnatal brain.

The low level of Dlx6 RNA expression (Supplementary Table 1) may also contribute to lack of robust forebrain phenotypes in the Dlx6^LacZ/LacZ mutants. At E15.5 Dlx6 expression in the MGE was nearly 20-fold less than Dlx1 and roughly 4-fold less than Dlx5 (see Table 1 in Long et al., 2009a). In the E18.5 striatum Dlx6 expression was about 10-fold less than Dlx1 and roughly 5-fold less than Dlx5 (Supplementary Table 1).

Dlx6 expression was restricted to the SVZ of the septum, LGE, MGE, POA, and CGE, and mantle of the septum, nucleus accumbens (NAc), striatum (Str), and central nucleus of the amygdala (CeA) (Figs. 1, 3, 4, 7, 8). Future studies are needed to carefully describe Dlx6 expression in the diencephalon, where it is prominent in Dlx⁺ zones of the prethalamus (reticular nucleus, zona incerta), anterior hypothalamus, and dorsomedial, suprachiasmatic, and arcuate nuclei (Puelles and Rubenstein, 2003; Yee et al., 2009).

Dlx6 expression was conspicuously low in the olfactory bulb (and the SVZ of the dLGE), olfactory tubercle (OT), and preoptic area, and was not detectable in the ventral pallidum, globus pallidus, medial amygdale, and cortical/hippocampal interneurons. The low expression in these regions contrasts with the moderate/high expression of its closest paralogue, Dlx1, in these regions (Fig. 8C; Cobos et al., 2005). Furthermore, the low Dlx6 expression in interneurons of the olfactory bulb, cortex, and hippocampus contrasts with expression and function of Dlx1, Dlx2, and Dlx5 in these cells (Figs. 1, 3, 4, 7, 8; Long et al., 2003, 2007; Cobos et al., 2005; Wang et al., 2010). Thus, we suggest that Dlx6^−/− mutants may primarily have molecular and functional defects in forebrain neurons/regions where its expression has minimal overlap with the other Dlx genes. However, functional analysis of Dlx6^−/− neurons will probably require study of postnatal mutants, and thus necessitate the generation of a Dlx6 condition allele.

Dlx6^LacZ/LacZ mutants show molecular defects in the ventral striatum, dorsal striatum, and central nucleus of the amygdala

Gene expression defects were detected in Dlx6^LacZ/LacZ mutants in the SE, NAcC, Str, OT, and CeA. The OT and NAcC normally express low levels of Dlx6 (compared with the dorsal striatum; Figs. 1, 3, 4); Dlx6^LacZ/LacZ mutants showed increased Dlx6 and LacZ transcripts in the OT and NAcC (Figs. 1, 2). In addition, these regions had increased expression of Pbx3 (Fig. 4F, F′). These molecular changes in the ventral striatum (OT and NAcC) in the Dlx6^LacZ/LacZ mutant provide evidence that Dlx6 regulates the development of this region.

Molecular defects in the Str and SE appear in their SVZ, which have reduced Dlx6 expression (Fig. 3B, B′), suggesting that secondary progenitors may be abnormal. However, we did not detect obvious changes in proliferation based on phosphohistone-3 staining (Supplementary Fig. 3). Striatal mantle defects in the Dlx6^LacZ/LacZ mutants included reduced expression of several genes that are downregulated in Dlx1/2^−/− mutants (Long et al., 2009a, b): Golf, PPT, RXRγ, and Tiam2 (Figs. 3, 4). These striatal defects included reduced expression of DARPP-32 and PPT in patchy domains (Fig. 4), and alterations in the gradients of expression of Ikaros and Pbx3 (Fig. 4). Therefore, we hypothesize that loss of Dlx6 expression in the Dlx1/2^−/− mutants contributes to their LGE/striatal differentiation defects. Further work is needed to definitively establish whether these molecular defects are indicative of alterations in striatal patch (striosomal) development in the Dlx6^LacZ/LacZ mutant..

Analysis of striatal projections to the globus pallidus and substantia nigra failed to identify a clear phenotype in the Dlx6^LacZ/LacZ mutant (Figs. 5, 6; Supplementary Figs. 1, 2). However, some of these axons appeared abnormal where they exit the telencephalon into the diencephalon (Fig. 5D, D′). Even so, striatal axons expressing M4R reach the substantia nigra (Fig. 6). In sum, striatal molecular architecture was disrupted in the absence of Dlx6.

Perhaps the most surprising result of this study was the differential expression of Dlx1 versus Dlx5&6 in the amygdala. Whereas Dlx5 and Dlx6 were expressed in the CeA, Dlx1 was expressed in the medial amygdala (MeA), in other subpallial amygdala nuclei, and in pallial amygdala interneurons (Fig. 8A–C). Dlx2 expression was difficult to detect in these regions (data not shown).

The Dlx6^LacZ/LacZ mutant showed greatly reduced expression of Dlx5 and Dlx6 RNA in the CeA at E15.5 and E18.5 (Figs. 1H, H′, 7A, A′, B, B′, 8A, A′, B, B′; Supplementary Fig. 4). LacZ expression from the Dlx6 locus was maintained in the region of the CeA (Fig. 2), supporting the hypothesis that the CeA was present. Thus, we suggest that the CeA is formed in the Dlx6^LacZ/LacZ mutant, but it has molecular defects. This model is further supported by subtle reduction in the expression of Golf, Pbx3, PPT, and Six3 (Figs. 7C, C′, D, D′, E, E′, 8D, D′, E, E′, F, F′). These results raise the possibility that Dlx5 and Dlx6 are important regulators of CeA development and function, and therefore alterations in Dlx5 and Dlx6 function will affect behaviors linked to the CeA (e.g., anxiety and fear). In this regard, it is interesting that mice with reduced Dlx1 dosage (Dlx1^−/− and Dlx1^+/− mutants) have reduced fear conditioning (Mao et al., 2009). Therefore, distinct members of the Dlx family may have important roles in regulating amygdala development and function: Dlx5 and Dlx6 in the CeA, and Dlx1 in the MeA, in other subpallial amygdala nuclei, and in pallial amygdala interneurons.