In this paper we analyse the expression pattern of a zebrafish dlx4/6 enhancer/reporter construct in embryonic transgenic mice. We show that the pattern of LacZ/β-galactosidase in cells that tangentially migrate from the ganglionic eminences to the cerebral cortex is identical to that of various subpallial markers, namely Dlx and GAD genes, that are known to label this population. Because β-galactosidase activity persists long after expression of the Dlx genes and the transgene becomes undetectable, we were able to analyse the β-galactosidase-positive cell population of the mature cortex through X-gal staining and immunohistochemistry. We show that this population is largely identical with the adult cortical and hippocampal interneuron population, providing further evidence for their subpallial origin.
Gene expression data and vital dye labeling experiments suggest that during forebrain development there is a tangential migration of cells from the basal telencephalon to the cerebral cortex (Porteus et al., 1994; De Carlos et al., 1996; Anderson et al., 1997a; Tamamaki et al., 1997). Additional evidence shows that this migration is responsible for bringing GABAergic neurons into the cerebral cortex, where they develop into interneurons of the neocortex, olfactory cortex, olfactory bulb and hippocampus of neonatal mice (DeDiego et al., 1994; Anderson et al., 1997a,b,1999,2001; Casarosa et al., 1999; Lavdas et al., 1999; Sussel et al., 1999; Wichterle et al., 1999; Pleasure et al., 2000). Here, we provide evidence that interneurons of the mature cortex are similarly derived from these prenatally tangentially migrating cells.
Most of the tangentially migrating cells in the telencephalon appear to require the function of the Dlx homeobox genes. There are four known Dlx genes expressed in the vertebrate forebrain: Dlx1, 2, 5 and 6 (Liu et al., 1997; Zerucha et al., 2000). Generally, these genes are expressed in the following temporal sequence: Dlx2, Dlx1, Dlx5 and Dlx6 (Anderson et al., 1997b; Liu et al., 1997; Eisenstat et al., 1999). In the mouse Dlx1 & 2 double mutant, expression of Dlx5 and 6 is greatly reduced and differentiation of late-born neurons in the basal telencephalon is blocked (Anderson et al., 1997b; Marín et al., 2000). This leads to an accumulation of tangentially migrating cells in periventricular ectopia (Marín et al., 2000). As a result, the cerebral cortex is depleted in GABAergic interneurons (Anderson et al., 1997a) and the olfactory bulb and hippocampus have almost no GABAergic interneurons (Anderson et al., 1997b; Bulfone et al., 1998; Pleasure et al., 2000).
Dlx expression marks the tangentially migrating cells (Porteus et al., 1994; Anderson et al., 1997a,1999). Enhancer elements from the mouse and from the zebrafish Dlx5/6 locus (known as dlx4/6 in zebrafish) have been identified that can drive expression of LacZ in a pattern that closely resembles that of Dlx5 in the embryonic basal ganglia (Zerucha et al., 2000). Here we demonstrate that transgenic mice harboring the zebrafish dlx4/6-LacZ enhancer/reporter (zfdlx4/6-LacZ) express β-galactosidase prenatally in the tangentially migrating neurons. Although RNA expression from zfdlx4/6-LacZ is not detectable in the adult cerebral cortex, β-galactosidase activity persists. We found that the vast majority of mature GABAergic cortical interneurons expressed zfdlx4/6-LacZ, thereby further supporting their origin in the basal telencephalon.
Materials and Methods
The construction and initial characterization of the zfdlx4/6-LacZ transgenic animals is described in Zerucha et al. (Zerucha et al., 2000). The transgene was maintained in C57 Bl/6 mice. No changes in either the pattern or strength of expression of the reporter protein were observed over time.
Preparation of Tissue
To obtain embryonic nervous tissue, dams were killed by cervical dislocation, embryos removed and the brains dissected and fixed for up to 60 min in 4% PFA/PBS at 4°C. Older animals were anaesthetized and perfused with cold 4% PFA/PBS, the brains removed and post-fixed for up to 3 h. Tissue intended for in situ hybridization was equilibrated in 30% sucrose/1 × PBS, embedded in Tissue-Tek and sectioned at 10 μm with a cryostat. Tissue of older animals intended for immunohistochemistry was equilibrated in 30% sucrose/PBS and sectioned at 30 μm with a freezing microtome. Embryonic and perinatal tissue intended for immunohistochemistry was embedded in 5% low gelling temperature agarose/ PBS and cut at 50 μm with a vibrating blade microtome. For long-term storage the tissue was kept at −20°C in a solution of 0.4 × PBS, 30% glycerol, 30% ethylene glycol.
Sections were immersed in a solution of 10 mM Tris–HCl pH 7.3, 0.005% Na-desoxycholate, 0.01% Nonidet P40, 5 mM K4Fe(CN)6, 5 mM K3Fe(CN)6, 2 mM MgCl2 and 0.8 mg/ml X-gal and incubated at 37°C until the X-gal precipitate was clearly visible. X-gal staining was always performed prior to immunohistochemistry.
In situ Hybridization
Cryosectioned tissue was post-fixed for 20 min in 4% PFA/PBS, washed 1 × 4 min in 3 × PBS and 2 × 4 min in 1 ′ PBS, rinsed in H2O and treated for 20 min with 1 μg/ml proteinase K in 0.1 M Tris–HCl, 0.05 M EDTA, pH 8.0. Slides were rinsed in 1 × PBS, the fixation (for 5 min) and washing steps repeated, and acetylated for 10 min in 0.1 M triethanolamine/0.4% (v/v) acetic anhydride. Slides were rinsed twice in H2O, dehydrated though an ethanol series and air-dried. RNA probes labeled to specific activities of ~5 × 109 c.p.m./mg were applied to the slides in a buffer containing 50% formamide, 10% dextran sulfate, 2 × Denhardt's solution, 5 × SSC, 10 mM β-mercaptoethanol (β-ME), 250 μg/ml yeast t-RNA and 500 μg/ml salmon sperm DNA, and hybridized overnight at 55°C. Slides were washed for 30 min at 37°C in 2 × SSC, 10 mM β-ME, for 30 min at 55°C in 2 × SSC, 50% formamide, 1 mM EDTA, 10 mM β-ME, and again for 30 min in 2 × SSC, 10 mM β-ME. They were RNase-treated for 1 h at room temperature (2 × SSC, 1 mM EDTA, 20 μg/ml RNase A, 1U/ml RNase T1), washed twice for 30 min at 55°C in 50% formamide, 2 × SSC, 1 mM EDTA, 10 mM β-ME, rinsed for 5 min in 0.2 × SSC at room temperature, dehydrated through an ascending ethanol series and dipped in NTB-2 photographic emulsion (Kodak) to localize the signal.
Immunohistochemistry was performed on 30 μm (adult tissue) or 50 μm (embryonic tissue) free-floating sections. The tissue was preincubated for 2 h in a solution of 2% normal goat serum and 0.1% NaN3 in PBST (PBS with 0.1% Triton X-100), the solution exchanged for a fresh aliquot containing the primary antibody and incubated for up to 48 h at 4°C. Sections were subsequently washed in PBST, treated with 1% H2O2 in PBST for 30 min at room temperature, thoroughly rinsed with four washes in PBST and reacted with the secondary antibody (1:200 goat anti-rabbit IgG conjugated to biotin, in PBST) for 4 h at room temperature. The ABC-kit (Vector) in combination with diaminobenzidine (DAB) staining was used to visualize the signal. For immunofluorescence, dye-coupled anti-rabbit IgG was employed as secondary antibody (Molecular Probes). The following primary antibodies were used (all of them rabbit polyclonal): anti-β-galactosidase (1:2000; 5′–3′), anticalbindin (1:5000; SWANT CB-38), anti-calretinin (1:5000; Chemicon AB149), anti-distal-less (1:400; a kind gift from Dr G. Panganiban, University of Wisconsin-Madison), anti-GABA (1:5000; Sigma A-2052), anti-GAD65 (1:2000; Chemicon AB5082), anti-nNOS (1:3000; ZYMED Z-RNN), anti-NPY (1:3000; Incstar) and anti-parvalbumin (1:5000; SWANT PV-28).
Expression from the Zebrafish dlx4/6 Enhancer Labels most Dlx-expressing Cells in the Central Nervous System
Previously we showed that the zfdlx4/6 enhancer expresses LacZ in a pattern that is nearly identical to the endogenous expression of mouse Dlx5 (Zerucha et al., 2000). Here, we closely examined its expression in the neurons that tangentially migrate from the basal telencephalon to the cerebral cortex. Following the expression of zfdlx4/6-LacZ provides evidence that these cells give rise to the interneurons of the mature cerebral cortex.
Analysis of β-galactosidase expression from the zfdlx4/6-LacZ locus in a transgenic mouse, by the X-gal histochemical stain (Fig. 1a′,b,c), in situ hybridization to LacZ transcripts (Fig. 1a) or immunofluorescence (Fig. 1d,i,j), shows that it marks the same regions where Dlx5 (Fig. 1g) and glutamic acid decarboxylase 65 (GAD65; Fig. 1f) are expressed at E13.5 and E15.5. Immunofluorescent labeling with an antibody that cross-reacts with DLX proteins also shows the remarkable fidelity of the zfdlx4/6 enhancer's expression (Fig. 1e).
Like Dlx5, GAD65 and GAD67, zfdlx4/6-LacZ is expressed in the basal telencephalon, ventral thalamus and parts of the hypothalamus (Liu et al., 1997; Katarova et al., 2000; Zerucha et al., 2000). In addition, these genes and the transgene are robustly expressed in the tangentially migrating cells that transit from the basal telencephalon and contribute interneurons to the entire cerebral cortex. At E13.5, there are two major streams of tangentially migrating cells that pass through the cortical intermediate and marginal zones; these cells largely avoid the cortical plate and cortical proliferative zone (Fig. 1a′–f). By E15.5, the transgene now labels a third migratory pathway, that runs through the cortical subventricular zone (Figs 1i,j and 2a,b). The subventricular zone expression is particularly prominent at P0 (Fig. 2h,j). By this age, labeled cells have penetrated into the cortical plate (Fig. 2i,k).
RNA expression from the zfdlx4/6-LacZ transgene and the Dlx genes decreases in the early postnatal period (Figs 2d,e,f and 3a–c,e,g). In adult mice LacZ and Dlx2, 5 and 6 RNA was not detectable by in situ hybridization in the cerebral cortex and basal ganglia (Fig. 3e,g and not shown). Of all Dlx genes tested (Dlx1, 2, 3, 5, 6, 7), only Dlx1 RNA expression was slightly above background (not shown). An exception to the downregulation of zfdlx4/6-LacZ and Dlx5 RNA expression occurs in the granule and periglomerular layers of the adult olfactory bulb (Fig. 3i,k).
Thus, whereas expression from GAD65 and GAD67 is strongly correlated with zfdlx4/6-LacZ and Dlx5 expression during developmental stages, this relationship does not continue in the mature brain, where high levels of GAD expression are maintained in the cortex and basal ganglia (Fig. 3h). On the other hand, regional co-expression between the LacZ, Dlx5 and GAD67 genes persists in the olfactory bulb (Fig. 3i,k,l).
Strong β-Galactosidase Expression from zfdlx4/6-LacZ Persists after LacZ RNA Expression Wanes
Although LacZ RNA expression is undetectable by in situ hybridization in most of the adolescent and adult mouse brain, β-galactosidase activity remains initially strong throughout the forebrain in a distribution consistent with the earlier expression of LacZ and Dlx5. Although in adult animals (P70 and older) the signal gradually wanes in the basal ganglia, β-galactosidase activity is detected in the periventricular parts of the striatum, septum, the cerebral cortex and in the interneuron layers of the olfactory bulb (Fig. 3f,j). The distribution closely resembles the pattern of GAD67 expression (Fig. 3h,l). The blue X-gal reaction product generally appears as one or more dots within the cytoplasm of labeled cells.
zfdlx4/6-LacZ is Expressed in GABAergic Neurons
The presumed perdurability of β-galactosidase activity (prolonged expression of the β-galactosidase protein after transcription of the LacZ gene has diminished) enabled us to determine the histochemical identity of the X-gal-labeled cells. We began by determining whether the X-gal cells entering the cerebral cortex expressed TBR1, a marker of cortical glutamatergic neurons (Hevner et al., 2001). Double labeling at E15.5 showed that β-galactosidase and TBR1 expression in the cerebral cortex and olfactory bulb were in separate cellular populations (Fig. 2a–c). This analysis allowed us to detect a thin stream of X-gal+ cells entering the olfactory bulb from the region of the septum; the blue column of cells followed around the ventricle and then was continuous with the marginal zone of the cortex (Fig. 2a,c), similar to the labeling pattern seen for DLX2 (Porteus et al., 1994) and for DLX1 (Anderson et al., 1999).
Since the X-gal+ cells in the cortex did not appear to be glutamatergic neurons, we tested whether they were GABAergic interneurons (Fig. 4). We performed double-labeling studies for GABA and X-gal at multiple ages (P0, P4, P20, P150) and show the results in the P150 adults (Fig. 4). At P150, >90% of X-gal+ cells also expressed GABA. Similarly, >90% of GABAergic cells were X-gal positive. This high concordance of X-gal and GABA co-expression was also found in the hippocampus (Fig. 4g–i), olfactory bulb (not shown) and all parts of the amygdala (not shown). Even though we were unable to detect in situ hybridization signals for the Dlx RNAs (except perhaps for Dlx1), most of the cells that express the zfdlx4/6-LacZ transgene also expressed some DLX proteins, as detected by immunoreactivity with the anti-distal-less and the anti-DLX1 antibodies (Fig. 4e,f).
zfdlx4/6-LacZ is Expressed in Subtypes of GABAergic Neurons
Based on immunolabeling, there are several major types of cortical GABAergic neurons (Parnavelas et al., 1977; Hendry et al., 1987; Meinecke and Peters, 1987; Van Eden et al., 1989; Del Rio et al., 1992; DeFelipe, 1993; Freund and Buzsdáki, 1996; Gonchar and Burkhalter, 1997; Meyer et al., 1998; Sloviter et al., 2001). We found expression of β-galactosidase in all of the subtypes tested in the P20 and P150 cortex (Figs 5–7). There were no apparent differences in either the number or the distribution of any of the labeled cells between these ages. The calcium binding proteins calretinin and calbindin are expressed in both GABAergic and glutamatergic cortical neurons (Celio, 1990; DeFelipe, 1993). In the adult mouse cortex, calretinin is primarily expressed in local circuit neurons; the vast majority of these cells also have β-galactosidase activity (Fig. 5a,b). Calbindin is expressed in both projection and local circuit neurons (Celio, 1990). In deep cortical layers, calbindin expression appeared in scattered small cells and β-galactosidase labeled most of these (Fig. 5c,e). In superficial cortical layers, calbindin weakly labeled radial columns of cells that are probably pyramidal cells and intensely labeled smaller scattered cells (Sánchez et al., 1992); only the scattered cells were co-labeled with β-galactosidase activity (Fig. 5c,d). In the hippocampus, calbindin labeled the pyramidal cells of the CA fields, and scattered cells in stratum oriens and stratum radiatum; only the cells in stratum oriens and stratum radiatum were co-labeled with the X-gal reaction (Fig. 5f,g).
Parvalbumin is expressed in ~50% of cortical GABAergic neurons (Gonchar and Burkhalter, 1997) and is not detectable in glutamatergic neurons. All parvalbumin+ neurons in our material expressed β-galactosidase activity (Fig. 6).
Expression of neuronal nitric oxide synthase (nNOS), neuropeptide Y (NPY) and somatostatin defines a specific type of cortical GABAergic interneuron (Freund and Buzsáki, 1996; Gonchar and Burkhalter, 1997). These neurons do not appear to develop in mice lacking Nkx2.1 homeobox gene function (Anderson et al., 2001). Nkx2.1 is required for specification of the medial ganglionic eminence (Sussel et al., 1999), a major source of the tangentially migrating neurons (Sussel et al., 1999; Wichterle et al., 1999; Lavdas et al., 1999; Marín et al., 2000; Anderson et al., 2001). Nkx2.1 is co-expressed in many Dlx-positive MGE cells (T. Stühmer and J.L.R. Rubenstein, unpublished data). One hundred per cent of the nNOS+ neurons are also positive for β-galactosidase (Fig. 7a–d). These large cells are present mainly in the deep and superficial layers of the neocortex, and in the pyramidal cell layer of the hippocampus (Fig. 7c,d). Likewise, there is a population of large NPY+ cells that are all β-galactosidase+ (Fig. 7e–g) and that show the same spatial distribution as the nNOS+ cells. In addition, there is a larger number of smaller NPY+ cells (or at least they look smaller because their neurites are not labeled) scattered over all of the neocortical layers (Fig. 7e,f). The majority of these cells are also clearly β-galactosidase+.
Here we have provided evidence that Dlx homeobox gene expression marks the vast majority of GABAergic cortical neurons, from the beginning of their development in the basal telencephalon, through their migration to the cortex, and in the adult mouse. Using an enhancer element from the zebrafish dlx4/6 locus, which previously was shown to express LacZ in a pattern nearly identical to endogenous Dlx5 in transgenic mice (Zerucha et al., 2000), we obtained evidence that virtually all cortical GABAergic neurons express the Dlx genes, at least at one point in their development.
Perdurability of β-Galactosidase Activity after the Reduction of Expression from Dlx Genes and the zfdlx4/6-LacZ Transgene
As reported for early stages of telencephalon development (Zerucha et al., 2000), the zfdlx4/6-LacZ transgenic mouse expresses β-galactosidase in a pattern extremely similar to mouse Dlx5 throughout gestation. We found that both RNA expression from the zfdlx4/6-LacZ transgene and from the endogenous Dlx genes wanes in the neonatal period, such that its expression, as judged by in situ hybridization, was undetectable in the cerebral cortex of the adult mouse (Fig. 3e). Nonetheless, activity of β-galactosidase continued into adulthood (Figs. 3f and 4–7).
Whereas β-galactosidase in developing/migrating neurons is present throughout the cell (see Fig. 1d), it becomes almost exclusively localized to one or more cytoplasmic deposits in differentiated neurons (Figs 4–7). These deposits are apparent from very early stages in development and appear to constitute a metabolically stable accumulation of the enzyme that marks the cell even after RNA levels have decreased below our means of detection. We suggest that β-galactosidase expression, in the form of these intracellular particles, may be a lineage marker for cells that expressed the zfdlx4/6-LacZ transgene during development.
On the other hand, some or all of the β-galactosidase activity in the postnatal cortex could be due to very low levels of transcription from the zfdlx4/6-LacZ locus that we were not able to detect by in situ hybridization.
Exceptions to these observations are some neurons in middle layers of the neocortex, that were strongly positive for β-galactosidase and for GABA, but that did not show co-labeling with the interneuronal markers tested (Fig. 6c). Perhaps this cell type maintains transcription from the zfdlx4/6-LacZ transgene into adulthood.
Thus, it is conceivable that we are following the lineage of Dlx expressing cells using this method. Of course, this hypothesis will need to be verified using lineage analysis with tissue-specific recombination methods (Dymecki and Tomasiewicz, 1998).
Although RNA expression from the zfdlx4/6-LacZ transgene parallels the down-regulation of Dlx5 in the postnatal cortex (Fig. 3a–c,e,g), we did observe some DLX protein immuno reactivity in GABAergic neurons with an anti-distal-less antiserum (Fig. 4f). A similar result was found with anti-DLX1 (Fig. 4e). DLX2 immunoreactivity was difficult to judge due to the low signal (not shown).
Co-expression of the Dlx and GAD genes
The results shown in this paper highlight the fact that in the cerebral cortex the Dlx genes are co-expressed with the GAD65 and GAD67 genes, both of which are hallmarks of an interneuronal phenotype. From the outset of their tangential migration from the ganglionic eminences and through their migrations in the marginal, intermediate and subventricular zone, to their settling in the cortical plate, the distribution of Dlx and GAD gene expression shows a remarkable coincidence. Furthermore, Dlx expression appears to be excluded from TBR1+ glutamatergic neurons (Fig. 2a–c), consistent with genetic evidence that cortical (neocortex, olfactory bulb and hippocampus) GABAergic and glutamatergic neurons have distinct genetic controls (Anderson et al., 1997a,1999; Bulfone et al., 1998; Pleasure et al., 2000;Hevner et al., 2001) and descend from distinct lineages (Tan et al., 1998; Parnavelas, 2000).
The close link of the Dlx and GAD genes suggests a functional relationship. Although GAD65 and GAD67 expression is not lost from the basal ganglia in the Dlx1 & 2 mutants (Anderson et al., 1997b) (Yun and J.L.R. Rubenstein, unpublished data), there are several lines of evidence indicating that the Dlx genes are essential for the development of GABAergic cortical neurons. In the Dlx1 & 2 double mutants there is an ~80% reduction in neocortical GABAergic neurons (Anderson et al., 1997a) and a >95% reduction in hippocampal and olfactory bulb interneurons (Anderson et al., 1997b; Bulfone et al., 1998; Pleasure et al., 2000). These reductions are probably primarily due to a block in the tangential migration of these cells from the ganglionic eminences. However, there is gain of function evidence that the Dlx genes may have additional roles in governing the phenotype of GABAergic cells. Introduction in vitro of a retroviral vector encoding Dlx2 into dispersed embryonic cortical cells from Dlx1 & 2 mutants induced the expression of GABA and GAD67 (Anderson et al., 1999). In addition, electroporation of Dlx expression vectors into slices of embryonic cerebral cortex ectopically induced GAD65 and GAD67 (T. Stühmer et al., in preparation). Furthermore, Dlx expression vectors can induce expression from an enhancer element isolated from GAD65 in co-transfection assays (B. Condie, personal communication). Thus, it is our contention that the Dlx genes may be important regulators not only of the differentiation and migration of GABAergic neurons, but also in regulating the expression of the GAD genes. The fact that GAD expression is not lost in the Dlx1 & 2 mutants suggests that there are additional transcription factors that can compensate for the absence of these genes. Since Dlx5 and 6 expression are greatly reduced in the Dlx1 & 2 mutants, it would have to be other genes. Candidates include the Mash1 and Gsh1 and 2 genes, whose expression is not diminished in the Dlx1 & 2 mutants (Yun and J.L.R. Rubenstein, unpublished data). Furthermore, there is evidence that ectopic expression of Mash1 in the cerebral cortex induces ectopic expression of Dlx1 and GAD67 (Fode et al., 2000).
If the Dlx genes regulate GAD expression, then they could play a role in regulating the function of GABAergic neurons of the forebrain. The Dlx mutant mice die either at birth [Dlx2, Dlx5 (Qiu et al., 1997; Depew et al., 1999)], or soon thereafter [Dlx1 (Qiu et al., 1997)]; accordingly, we have not been in a position to study the function of the GABAergic neurons in these mutants. However, the generation of conditional alleles of the Dlx genes offers the possibility to investigate Dlx function in maturing and mature GABAergic neurons.
This work was supported by research grants to T.S. from the Deutscher Akademischer Austauschdienst and the Deutsche Forschungsgemeinschaft in Bonn, to L.P. from CICYT (grant No. PB98-0397) 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, Dr Grace Panganiban for the anti-distal-less antibody and Stewart Anderson and Oscar Marín for their expert advice.