Ectopic expression of the Dlx genes induces glutamic acid decarboxylase and Dlx expression

Recent evidence suggests that neurotransmitter subtype specification is linked to dorsoventral patterning (Marín et al., 2000; Wilson and Rubenstein, 2000). For example, in the ventral spinal cord, longitudinal progenitor domains, which are arrayed as distinct dorsoventral tiers, give rise to distinct types of neurons: cholinergic motoneurons arise from more ventral positions than most GABAergic interneurons (Briscoe et al., 2000; Sander et al., 2000). A similar organization appears to be found in the embryonic telencephalon (Marín et al., 2000). In both of these tissues, Nkx genes regulate specification of cholinergic neurons (Sussel et al., 1999; Sander et al., 2000; Briscoe et al., 2000). Transcription factors that are expressed in progenitor domains for GABAergic neuronal zones are candidates for regulating GABAergic specification. In the telencephalon, members of the Dlx homeobox gene family are expressed in, and regulate the development of, the primordia of the basal ganglia (Anderson et al., 1997a; Marín et al., 2000). These large nuclei consist of GABAergic projection neurons. In addition, the Dlx genes are expressed in, and regulate the development of, neurons that tangentially migrate from the basal telencephalon into the cerebral cortex (Anderson et al., 1997b; Anderson et al., 1999; Anderson et al., 2001; Pleasure et al., 2000). These neurons give rise to GABAergic interneurons of the cerebral cortex, hippocampus and olfactory bulb (Anderson et al., 1997a; Anderson et al., 1997b; Bulfone et al., 1998; Lavdas et al., 1999; Pleasure et al., 2000; Sussel et al., 1999; Wichterle et al., 1999). These data support a model that cortical GABAergic and glutamatergic neurons arise from distinct telencephalic progenitor zones and are under distinct genetic controls (Anderson et al., 1997b; Anderson et al., 1999; Anderson et al., 2001; Bulfone et al., 1998; Parnavelas, 2000; Hevner et al., 2001).

The molecular mechanisms through which the Dlx genes regulate the development of forebrain GABAergic neurons remain to be elucidated. To complement Dlx loss-of-function analyses, we have implemented a novel electroporation gain-of-function assay in slice cultures of the embryonic mouse forebrain. Using this approach, we demonstrate that Dlx2 and Dlx5, but not Dlx1, can robustly induce cortical cells to express the glutamic acid decarboxylase (Gad) genes. Furthermore, this assay provides evidence for a cross-regulatory cascade of the Dlx genes.

Coding regions of mouse Dlx1, Dlx2 and Dlx5 genes (Eisenstat et al., 1999; Zerucha et al., 2000) were cloned into pCAGGS, a chicken β-actin-promoter driven expression vector (Niwa et al., 1991). The original vector was modified to include a greater number of restriction sites 3′ of the promoter (plasmids pCAGGS/ES and pCAGGS/SE). The genes for Dlx2 and Dlx5 consisted of their complete and unmodified coding sequences, whereas that for Dlx1 lacked coding sequence for the two C-terminal amino acids and, potentially, for 5 amino acids (MTMTT) at the predicted N terminus, which contains three closely spaced methionines. Untranslated gene sequences were kept as short as feasible, ranging from 0-7 bases at the 5′ end to maximally 8 bases at the 3′ end.

Preparation and maintenance of slice tissue cultures followed the methods of Anderson et al. (Anderson et al., 1997a).

A tissue vibratome slice (in 4% low melting point agarose in 1× Krebs), with its supporting membrane (Nucleopore Track-Etch membrane, Whatman), was placed onto a 1% agarose block (in 1× Krebs buffer) within a setup of two horizontally oriented platinum electrodes (System CUY-700-1 and CUY-700-2; Protech International, San Antonio, TX) (see Fig. 1 for a drawing of the apparatus). A tiny agarose column, punched with a clipped hypodermic needle (gauge 19 or 21) from a 1% agarose gel in 1× Krebs, was attached to the mobile upper electrode. The agarose blocks and columns were made form a gel that was cast between two glass plates (the layer of agarose should not be thinner than 1.5 mm). A small amount of plasmid solution was applied to the lower end of the agarose column, and then the electrode was lowered to let the solution contact the tissue. The system was powered by a T820 Electro Square Porator (BTX). In empirical trials, we determined that plasmid concentrations of about 1 mg/ml at charging voltages of >100 V (with two pulses of 5 ms each) yielded acceptable numbers of electroporated cells.

After incubation for up to 48 hours, slices were fixed in 4% paraformaldehyde/phosphate-buffered saline (PBS), washed in PBS and either embedded in Tissue-Tek or 5% low gelling temperature agarose/PBS for further sectioning on a cryostat (10 μm) or a vibrating blade microtome (50 μm), respectively. This material was subjected to standard in situ hybridization and/or immunohistochemistry procedures.

In situ hybridization was performed along the lines detailed in Porteus et al. (Porteus et al., 1992).

Immunohistochemistry was performed on 50 μm free-floating sections. Following pre-incubation for 2 hours in a solution of 2% normal goat serum in PBST (PBS with 0.1% Triton X-100) and 0.1% NaN₃, the liquid was changed for a fresh aliquot containing the primary antibody, and the tissue incubated for up to 48 hours at 4°C. Sections were subsequently washed 3× in PBST and the secondary antibody (1:200 goat anti-rabbit IgG, conjugated to either the red fluorescent dye Alexa594 (Molecular Probes), or biotin (Vector)) applied, and incubated for 4 hours at room temperature. The biotinylated antibody was visualized with the ABC kit (Vector), according to the manufacturers instructions. We used the following primary antibodies: anti-distal-less (rabbit polyclonal, 1:400; a kind gift from Dr G. Panganiban, University of Wisconsin-Madison); anti-GAD65 (rabbit polyclonal, 1:2000; Chemicon AB5082); and anti-β-galactosidase (rabbit polyclonal, 1:2000; 5prime-3prime).

In the embryonic forebrain, the expression pattern of the DLX family of transcription factors is nearly identical to that of the glutamic acid decarboxylases (GAD65 and GAD67), enzymes that synthesize the neurotransmitter γ-aminobutyric acid (GABA). For example, at early stages of embryonic telencephalon development, the subpallium expresses Gad65 (Gad2 – Mouse Genome Informatics) and Gad67 (Gad1 – Mouse Genome Informatics), and Dlx1, Dlx2, Dlx5 and Dlx6 (Fig. 2) (Eisenstat et al., 1999; Fode et al., 2000; Liu et al., 1997; Zerucha et al., 2000; Katarova et al., 2000), whereas the pallium does not. Once the subpallial-to-pallial tangential migration begins, bringing GABAergic cells into cortical regions, GAD65 and DLX expression show similar distributions within the cerebral cortex (Fig. 2). Coincident expression of GAD65 and DLX is also notable in subdivisions of the rostral diencephalon (Fig. 2c,d,h,i). Expression of GAD posterior to the zona limitans, is not associated with DLX expression, as central nervous system expression of the Dlx genes appears to be restricted to the anterior forebrain (Porteus et al., 1991; Price et al., 1991; Robinson et al., 1991; Eisenstat et al., 1999). The spatial and temporal coincidence between DLX and GAD expression suggests that the Dlx genes may have an important role in regulating GAD expression. To test this, we used a gain-of-function assay.

We adapted an electroporation method (Funahashi et al., 1999) to transfect Dlx expression vectors into coronal slices of the embryonic brain grown in vitro (Anderson et al., 1997a; Anderson et al., 1997b) (Fig. 1). To test whether the DLX proteins could induce GADs, we electroporated the expression vectors into the cerebral cortex from E12.5 mouse embryos. At this age, the cortex is essentially free of GAD- or DLX-positive cells, and selecting the caudal-most sections of the telencephalon for the experiments (Fig. 2d,i) assured that they were severed from the subpallial sources of tangentially migrating GAD⁺/DLX⁺ neurons (Anderson et al., 1997a; Anderson et al., 2001). The caudal-most sections were also consistently more efficiently electroporated.

An expression vector encoding green fluorescent protein (GFP) was co-electroporated with the Dlx expression vectors to identify the regions that expressed the transfected plasmids (Fig. 3a,c,e,g), and to assess the efficiency of each electroporation. Electroporation of Dlx1, Dlx2 and Dlx5 expression vectors led to the appearance of ectopic DLX immunoreactivity within 3 hours (not shown); by 7 hours, strong expression (comparable with endogenous levels) was detectable (Fig. 5b). Expression from the Dlx1, Dlx2 and Dlx5 vectors produced roughly equivalent levels of DLX expression, as judged by immunofluorescence with an anti-distal-less antibody (Fig. 3b,d,f). The patterns of GFP and DLX expression were virtually identical (Fig. 3a-f), suggesting extensive co-transfection of the electroporated plasmids. Counting of GFP- and DLX-positive cells showed that >95% of electroporated cells expressed both plasmids.

Electroporation of either the Dlx2 or Dlx5 expression plasmid readily induced GAD65 immunoreactivity (27/30 experiments for Dlx2; 12/20 for Dlx5) (Fig. 4b,f; Fig. 5). In the most effective Dlx2 electroporations, more than 85% of the GFP-positive cells expressed GAD65 (Fig. 4a,b). Dlx5 was less efficient at inducing GAD65 (Fig. 4e,f), with the most effective electroporations inducing GAD65 in ∼50% of the GFP-positive cells. Dlx1 was inefficient at inducing GAD65 (data not shown); only in four out of 12 electroporations were any GAD65-positive cells detected, and these constituted less than 1% of the GFP-positive cells.

Electroporations of binary combinations of the Dlx2 and Dlx5 expression vectors did not appear to increase the amount of GAD65 expression compared with levels seen with the Dlx2 plasmid alone (n=3). Electroporation of the GFP expression plasmid alone never induced expression of GAD65 (n>50) (Fig. 4c,d,g,h).

The kinetics of Dlx2-mediated GAD65 induction was assessed by analyzing electroporated slices after 7, 19 and 35 hours in culture (Fig. 5). Whereas strong expression of DLX2 could be detected after 7 hours, ectopic expression of GAD65 was not (n=3). GAD65 expression was detectable by 19 hours, and was increased by 35 hours (Fig. 5).

Electroporation of the Dlx2 vector also induced the expression of Gad67. As no antibody specific to GAD67 was available, we studied this induction using in situ RNA hybridization (Fig. 4i-k). Although only 5/28 experiments showed a clear induction of Gad67 by Dlx2, we suggest that this is an underestimate, as most of the negative experiments were performed during early stages of the project, when we were establishing the protocol and often encountered low levels of electroporation. The last three experiments show Gad67 induction by Dlx2. Attempts to compare the efficiencies of GAD65 and GAD67 induction have been hampered by our inability to use the same method to study both of their expression (in situ hybridization for Gad65 has not worked, and we have no antibody specific for GAD67).

Thus, these experiments indicate that Dlx2 and Dlx5 are efficient inducers of a fundamental element of the GABAergic phenotype in cortical cells. However, Dlx2 and Dlx5 failed to clearly induce other markers that are characteristic of basal telencephalic neurons or cortical interneurons such as NPY, nNOS, substance P, enkephalin, OCT6, calretinin or calbindin, although none of these shows as close a temporal or spatial correlation during development to Dlx expression as do the Gad genes. Furthermore, ectopic expression of the DLX proteins did not seem to affect expression of TBR1, a transcription factor that marks most early cortical plate neurons (Hevner et al., 2001) (data not shown).

Previous loss of function studies provided evidence that Dlx1 and Dlx2 together regulate Dlx5 and Dlx6 in most of the forebrain (Anderson et al., 1997a; Zerucha et al., 2000). In addition, an intergenic enhancer for mouse Dlx5 and Dlx6 and zebrafish dlx4 and dlx6 has been shown to be regulated by Dlx1 and Dlx2 in transgenic mice and in tissue culture cells (Zerucha et al., 2000). Here, we used the gain-of-function assay to test whether ectopic expression of Dlx1, Dlx2 and/or Dlx5 is sufficient to induce expression from endogenous Dlx genes and from a co-electroporated Dlx5/6 enhancer/reporter plasmid (mI5/6i-lacZ) (Zerucha et al., 2000).

Electroporation of Dlx2 in the cortex induced Dlx5 RNA, as judged by in situ hybridization (Fig. 4i,j,l). Although the efficiency of this induction was not robust (occurring in only 4/23 cases), we suggest this is an underestimate for the same reasons discussed above regarding the induction of Gad67. Electroporation of Dlx1 did not induce expression of the endogenous Dlx5 gene (data not shown). Induction of Dlx6 expression was not detected following electroporation with either Dlx1, Dlx2 or Dlx5 (data not shown), nor with binary combinations of these expression vectors (n=28).

Next we tested whether the Dlx expression vectors could induce expression from a co-electroporated mI5/6i-lacZ enhancer/reporter plasmid. We performed the electroporations at various rostrocaudal and dorsoventral positions of the E12.5 brain, and assessed the pattern of β-galactosidase through immunohistochemistry. Control experiments showed that electroporation of the mI5/6i-lacZ plasmid alone resulted in β-galactosidase expression that was restricted to areas that endogenously express the Dlx genes, such as the ganglionic eminences, septum and ventral thalamus (Fig. 6a-f). Thus, Dlx5/6 enhancer-mediated expression appears to critically depend on the transcription factors that are present in the Dlx⁺ regions.

To test whether Dlx2 is sufficient to induce mI5/6i-lacZ expression, we co-electroporated the Dlx2 expression vector with the mI5/6i-lacZ and the GFP expression vectors into the cerebral cortex, midbrain and hindbrain. The Dlx2 expression vector induced robust β-galactosidase expression in the cerebral cortex (Fig. 6h,j), the dorsal thalamus (Fig. 6j), the tectum (Fig. 6l) and the hindbrain (n=15). Co-expression of GFP and β-galactosidase regularly exceeded 90% in these experiments (Fig. 6), implying that Dlx2 induction of expression from mI5/6i-lacZ was very efficient. This result also suggests that mixtures of three different plasmids are efficiently electroporated into the same cells.

While the Dlx5 expression vector was also a potent inducer of β-galactosidase expression throughout the brain (Fig. 6n,r; n=15), the Dlx1 expression vector produced a very different result from the Dlx2 and Dlx5 vectors. Dlx1 induced very little β-galactosidase expression in the cerebral cortex (not shown) and the tectum (Fig. 6p). However, the Dlx1 vector was capable of inducing β-galactosidase in the dorsal thalamus, the tegmentum (ventral midbrain) and in the hindbrain (Fig. 6t) (n=8).

We demonstrate that Dlx and Gad expression in the embryonic forebrain is nearly identical. We also show that ectopic expression of a subset of Dlx genes in the developing cerebral cortex is sufficient to induce the Gad genes. This study establishes that the Dlx genes are sufficient to induce the fundamental phenotype of GABAergic neurons. The results support the hypothesis that the Dlx genes play a significant role in the specification of telencephalic GABAergic neurons (e.g. the projection neurons of the basal ganglia and interneurons of the cortex). Furthermore, these findings may also prove useful for engineering GABAergic neurons from appropriate progenitor cells.

This gain-of-function study complements the analysis of the Dlx1 and Dlx2 loss-of-function double mutants, which show differentiation and migration defects in the development of most telencephalic GABAergic neurons (Anderson et al., 1997a; Anderson et al., 1997b; Anderson et al., 1999; Anderson et al., 2001; Bulfone et al., 1998; Pleasure et al., 2000). Despite the severe block in GABAergic neuron development in the Dlx1 and Dlx2 mutants, which includes decreased expression of Dlx5 and Dlx6, GAD expression persists, showing that other transcription factors are also capable of regulating GAD expression in the forebrain. Candidates for these genes include Mash1 (Ascl1 – Mouse Genome Informatics), and Gsh1 and Gsh2. These transcription factors continue to be expressed in the Dlx1 and Dlx2 mutants (Yun and J. L. R. R., unpublished observations), and gain-of-function evidence suggests that Mash1 is able to induce GABAergic neurons (Fode et al., 2000).

This analysis has begun to dissect the specific roles for the Dlx genes in the development of GABAergic neurons. While they can induce both Gad65 and Gad67 within 20 hours, we did not find induction of other markers of GABAergic projection neurons (e.g. enkephalin and substance P) or of GABAergic interneurons (e.g. calbindin and nNOS) (data not shown). Perhaps additional time is needed to see expression of these genes. Alternatively, cortical cells may not be fully competent to express all genes found in GABAergic neurons, as other transcription factors may be needed in parallel, or in conjunction with, the DLX proteins.

While Dlx2 and Dlx5 were robust inducers of GADs, Dlx1 was not. Based on their amino acid sequences, the Dlx genes fall into two major homology groups: Type A (Dlx2, Dlx3 and Dlx5) and Type B (Dlx1, Dlx6 and Dlx7) (Stock et al., 1996; Liu et al., 1997). These are the first results that suggest biochemical differences in the functions of A and B subtypes. Furthermore, while Dlx2 and Dlx5 efficiently transactivated expression from a co-electroporated Dlx5/6 enhancer/reporter plasmid in every CNS region tested (Fig. 6h,j,l,n,r), Dlx1 appeared more restricted in its ability to activate the Dlx5/6 enhancer. For example, while Dlx1 activated Dlx5/6 enhancer expression in the ventral midbrain (tegmentum) and in the ventral hindbrain (Fig. 6t), we did not observe any effect of Dlx1 on the Dlx5/6 enhancer in the dorsal midbrain (superior colliculus) (Fig. 6p). We also noted that whereas Dlx2 could induce both endogenous Dlx5 and the Dlx5/6 enhancer, it did not effectively induce endogenous Dlx6 expression. This may be explained by the observation that expression from the Dlx5/6 enhancer more closely resembles endogenous Dlx5 than Dlx6 expression. These results suggest that there is a separate Dlx6 enhancer, which may be less sensitive to activation by Dlx2.

Therefore, the results of this and our previous studies suggest both redundant and distinct functions for different members of the Dlx gene family. Dlx1 and Dlx2 are redundant for the control of late-born neurons of the basal telencephalon to efficiently migrate away from the subventricular zone and to express markers of more differentiated neurons (such as DARPP32) (Anderson et al., 1997; Marín et al., 2000) (Yun and J. L. R. R., unpublished). Neither single mutant shows this phenotype. However, we provide evidence that Dlx1 and Dlx2 show different abilities to ectopically induce GAD expression, and to regulate the Dlx5/6 enhancer. Thus, perhaps both Type A (Dlx2 and Dlx5) and Type B (Dlx1 and Dlx6) Dlx genes have redundant functions in regulating aspects of differentiation related to migration, but Type A and Type B Dlx genes may have distinct functions with respect to cross-regulation of Dlx genes and GAD expression. The hypothesis that Type A and B Dlx genes have different functions in vivo, is consistent with the observation that Dlx1 and Dlx2 mutants have distinct maxillary dysmorphologies, despite their similar expression patterns in the first branchial arch (Qiu et al., 1997).

The electroporation method used in these experiments is a novel adaptation of the approach devised for in ovo electroporation of chicken embryos (Funahashi et al., 1999). Slice culture electroporation is spatially more precise and it can be readily applied to diverse species. Given the large number of mouse mutants now available, this opens the possibility of rescuing mutant phenotypes. Furthermore, many developmental and physiological processes are largely unperturbed in slice cultures (e.g. neuronal migrations) (Anderson et al., 1997b; Anderson et al., 2001); thus electroporation that does not restrict the size of the transfected plasmid (unlike most viral vectors) will be an effective method with which to study rapidly the effects of genes on specific processes. Finally, because one can efficiently co-electroporate two or more plasmids into single cells, this approach promises to be an important method to study multi-component processes, such as transcriptional regulation in physiologically relevant cells. This proved to be the case for establishing that Dlx2 and Dlx5 are effective inducers of the Dlx5/6 enhancer.

This work was supported by a research grant to T. S. from the Deutsche Forschungsgemeinschaft (DFG), to S. A. from NIMH K08 MH 01620-1, and to J. L. R. R. from Nina Ireland, NARSAD, NIMH RO1 MH49428 and NIMH K02 MH01046. We thank Susan Yu for her excellent assistance, Prof. Jun-ichi Miyazaki for the pCAGGS expression vector, Dr Grace Panganiban for the anti-distal-less antibody and Oscar Marín for drawing the schematic in Fig. 1.