The Arx transcription factor is expressed in the developing ventral telencephalon and subsets of its derivatives. Mutation of human ARX ortholog causes neurological disorders including epilepsy, lissencephaly, and mental retardation. We have isolated the mouse Arx endogenous enhancer modules that control its tightly compartmentalized forebrain expression. Interestingly, they are scattered downstream of its coding region and partially included within the introns of the downstream PolA1 gene. These enhancers are ultraconserved noncoding sequences that are highly conserved throughout the vertebrate phylum. Functional characterization of the Arx GABAergic enhancer element revealed its strict dependence on the activity of Dlx transcription factors. Dlx overexpression induces ectopic expression of endogenous Arx and its isolated enhancer, whereas loss of Dlx expression results in reduced Arx expression, suggesting that Arx is a key mediator of Dlx function. To further elucidate the mechanisms involved, a combination of gain-of-function studies in mutant Arx or Dlx tissues was pursued. This analysis provided evidence that, although Arx is necessary for the Dlx-dependent promotion of interneuron migration, it is not required for the GABAergic cell fate commitment mediated by Dlx factors. Although Arx has additional functions independent of the Dlx pathway, we have established a direct genetic relationship that controls critical steps in the development of telencephalic GABAergic neurons. These findings contribute elucidating the genetic hierarchy that likely underlies the etiology of a variety of human neurodevelopmental disorders.
During telencephalon development, tissue patterning and cell type specification are tightly connected processes that allow the emergence of neural structures coupled with specific cell-type composition. In fact, early on in development, glutamatergic and GABAergic cell fate specification is spatially confined into two nonoverlapping areas, namely, the dorsal pallium and the ventral subpallium, respectively. Despite the identification and characterization of numerous molecular players (mainly using a loss-of-function approach), how their interactions are finely regulated and sequentially determined still remains unclear (Guillemot, 2005, 2007). To address this crucial issue, the identification of the endogenous regulatory sequences controlling the temporal–spatial activities of these factors is a prerequisite to identify the mechanisms underlying gene regulation and specification of the different neural tissues.
Dlx1 and Dlx2 are homeodomain-containing transcription factors highly similar and redundant that act as critical molecular determinants of forebrain development. In the telencephalon, Dlx1/2 expression is restricted to the subventricular zone (SVZ) and mantle regions of the subcortical structures (Panganiban and Rubenstein, 2002). Dlx1/2 mouse mutants exhibit a block in the differentiation of progenitors located in the basal ganglia that was associated with a failure in neuronal tangential migration to the cerebral cortex, olfactory bulb, and hippocampus (Anderson et al., 1997a,b; Bulfone et al., 1998; Pleasure et al., 2000; Long et al., 2007). Furthermore, Dlx1/2 mutant cells exhibit exuberant neuritic growth and premature activation p21-activated serine/threonine kinase PAK3 (Cobos et al., 2007). These data suggest that the Dlx genes are, therefore, required for coordinating the timing of GABAergic cell migration and process formation.
How Dlx1/2 controls these key processes in telencephalon development and which molecular signals lie downstream to these factors is poorly understood. The Dlx1 and Dlx2 genes are located in a tail-to-tail oriented cluster with an intergenic region that contains two conserved domains acting as enhancers for both genes (Ghanem et al., 2003). Specifically, within the I12b enhancer, controlling the expression of Dlx1/2 in the forebrain, a conserved E-box sequence was proved to specify the expression of Dlx1/2 in the subpallium by associating with the basic helix-loop-helix protein Mash1 (Ghanem et al., 2007; Poitras et al., 2007). Along the same line, Fode et al. (2000) demonstrated that Mash1 ectopic expression in the cortical primordium induces the expression of Dlx1/2. These data provided compelling evidence that Dlx1 and Dlx2 genes represent direct targets of Mash1 in vivo. However, the detection of additional transcription factor-binding sites within the I12b enhancer (Poitras et al., 2007), such as Meis1/2-binding sites, indicates that additional transcriptional factors also control Dlx1/2 expression. In addition, downstream factors that may mediate the regulation of subpallium development and GABAergic cell fate commitment and migration by Dlx1/2 are just beginning to be identified.
Arx encodes for a homeodomain-containing transcription factor belonging to a small family of mammalian homologs of the aristaless (al) Drosophila gene (Campbell et al., 1993; Miura et al., 1997; Meijlink et al., 1999). Mutations in the human ARX gene have been identified in a large variety of neuropathological conditions, including West and Partington syndromes, myoclonic epilepsy, lissencephaly, and nonsyndromic mental retardation (Kato et al., 2004; Gécz et al., 2006; Nawara et al., 2006). Both neuropathological studies in autoptic human tissues and analysis in Arx mutant animals provided evidence that Arx is necessary for the proper migration of subpallial neuronal progenitors to the cerebral cortex, and thus for the supply of the majority of GABAergic cortical interneurons (Bonneau et al., 2002; Kitamura et al., 2002; Colombo et al., 2007). Therefore, it has been suggested that, although these ARX-dependent pathological manifestations were thought to emanate from different molecular origins, they may all be included in a unique large class of neuropathological conditions known as “interneuronopathies” (Kato and Dobyns, 2005; Kato, 2006).
The Arx murine ortholog is expressed in the subpallium of the developing telencephalon both in the medial ganglionic eminence (MGE) and lateral ganglionic eminence (LGE), anterior entopeduncular region, anterior preoptic area, and the preoptic-hypothalamic region. Furthermore, Arx expression is maintained in the subpallial neurons migrating tangentially toward the cerebral cortical primordium (Kitamura et al., 2002; Colombo et al., 2004; Cobos et al., 2005a; Yoshihara et al., 2005). We previously demonstrated that Dlx factors control Arx activity (Cobos et al., 2005). However, it is unclear whether Arx represents a direct target of the Dlx proteins; furthermore, it is unknown what aspects of Dlx function are controlled by Arx. To investigate these issues, we identified the regulatory sequences that promote Arx's expression in forebrain GABAergic neurons. We then found that Dlx protein activity, both in vitro and in vivo, positively regulated this enhancer. Furthermore, using a combination of loss- and gain-of-function approaches, we provide evidence that Arx is not implicated in mediating the Dlx-dependent specification of GABAergic neuronal cells, whereas Arx does participate in regulating the Dlx-dependent GABAergic migration.
Materials and Methods
Arx mutant mice (Collombat et al., 2003) and Dlx1/2 mutant mice (Qiu et al., 1997) were maintained by backcrossing with C57BL/6 animals. Genotyping was performed by morphological assessment [when possible, after embryonic day 13.5 (E13.5)] and confirmed by genomic PCR. Mice were maintained at San Raffaele Scientific Institute Institutional mouse facility, and experiments were performed in accordance with experimental protocols approved by local Institutional Animal Care and Use Committees.
Generation of transgenic mice.
Genomic fragments were subcloned in the p1230 vector containing a β-actin minimal promoter followed by a LacZ reporter cassette. Transgenic mice were produced by pronuclear injection in FVB fertilized oocytes using standard procedures. Presence of the transgene was determined by PCR on DNA prepared from extraembryonic tissues of embryos or clipped tails of postnatal animals. For each construct, at least three founders or primary transgenic embryos were analyzed. Founder transgenic mice were maintained by breeding with FVB inbred animals and/or with CD1 outbred mice once it was assessed that transgene expression was no altered.
X-gal staining and immunohistochemistry.
For immunoperoxidase staining, 10-μm-thick frozen sections were treated with 3% hydrogen peroxidase in methanol for 30 min at room temperature. Sections were rehydrated and blocked in 10% fetal calf serum, 0.5% bovine serum albumin (BSA) in PBS for 1 h. Primary antibodies were diluted in the same medium, applied to sections and incubated overnight at 4°C. Then, the slices were washed and incubated with appropriate biotinylated secondary antibody (DakoCytomation) for 90 min. After washing, the sections were processed with a Vectastain ABC kit (Vector Laboratories) for 1 h at room temperature and revealed with DAB peroxidase substratum (SK-4100; Vector Laboratories). Finally, sections were dehydrated, dried, and coverslipped with Eukitt (Electron Microscopy Science).
The primary antibodies used were as follows: mouse anti-MAP2 (1:400; Millipore Bioscience Research Reagents), rabbit anti-NPY (1:1000; Incstar), rabbit anti-calretinin (1:1000; Swant), rabbit anti-DARPP32 (1:50; Immunological Sciences), rabbit anti-ChAT (1:500; Millipore Bioscience Research Reagents). For immunofluorescence, secondary antibodies of the Alexa family were used (Invitrogen) in combination with the fluorescent dye Hoechst 33342 (Invitrogen). Then, slices were washed and mounted in Fluorescent Mounting Medium (DakoCytomation).
Hippocampal primary neuronal cultures.
Primary neuronal cultures were prepared from the hippocampi of Sprague Dawley E18 rat embryos (Charles River) as described previously (Banker and Cowan, 1977). Briefly, hippocampi were dissected in ice-cold HBSS supplemented with antibiotics and incubated in 0.25% trypsin at 37°C for 15 min. Cell dissociation was performed using polished glass pipettes until obtaining a single-cell suspension. Neurons were plated on poly-lysine-treated coverslips. After 4 h, coverslips were moved to Petri dishes with cultured glial cells and cultured in MEM supplemented with 1% N2 supplement (Invitrogen), 2 mm glutamine (BioWhittaker), 0.1% ovalbumin, 1 mm sodium pyruvate (Sigma-Aldrich), and 4 mm glucose. Neurons were cultured up to 15 d at 37°C in a 5% CO2 atmosphere, changing half of the medium every other day.
Transient cotransfection experiments.
Transient cotransfection studies were performed on P19 murine embryonic carcinoma cells cultured in α-MEM, 10% FBS, penicillin/streptomycin, and essential amino acids. Cells (300,000) were seeded in six-well plates and transfected with the Lipofectamine reagent (Invitrogen) using 10 μg of luciferase reporter plasmid, 5 μg of expression construct, and 2 μg of pRSV-β-galactosidase as an internal control of transfection efficiency. Cells were harvested 36–48 h after transfection, lysed incubating in ice for 10 min in lysis buffer (25 mm Gly-Gly, pH 7.8, 15 mm MgSO4, 4 mm EGTA, 1 mm DTT, and 1% Triton X-100) and centrifuged at 4°C for 15 min. Cell extracts were subjected to luciferase and β-galactosidase assays as described previously (Zappavigna et al., 1994).
Electrophoretic mobility shift assays.
Oligonucleotides corresponding to the Dlx putative binding sites were labeled by T4 kinase reactions in the presence of radiolabeled γ32P-ATP (GE Healthcare). Electrophoretic mobility shift assays (EMSAs) were performed with in vitro-translated proteins as previously described using 2 μl of reticulocyte lysate containing the desired combination of proteins mixed with 18 μl of PPH binding buffer [10 mm Tris-Cl, pH 7.5, 75 mm NaCl, 1 mm EDTA, 6% glycerol, 3 mm spermidine, 1 mm dithiothreitol, 0.5 mm phenylmethylsulfonyl fluoride (PMSF), 1 μg of poly(dI-dC), 30,000 cpm 32P-labeled oligonucleotide] to a total volume of 20 μl. After 30 min of incubation on ice, the reaction mixtures were separated by 5% PAGE in 0.5× Tris-buffered EDTA. For the competition assays, a 50- or 100-fold molar excess of unlabeled competitor oligonucleotide was added to the binding reaction mixture 10 min before the labeled probe.
E13.5 mouse embryonic forebrains were isolated and single-cell suspension derived by enzymatic treatment. Cells were cross-linked with 1% formaldehyde for 10 min and chromatin prepared essentially as described previously (Ferrai et al., 2007) by using 10 sonication cycles [35 s at 60–70 W, in an Ultrasonic Processor XL Sonicator (Misonix), followed by a 2 min rest on ice]. Cross-linked chromatin-containing fractions were pooled and stored at −80°C. Each aliquot of cross-linked chromatin (55 μg) was precleared with 17.5 μl of Protein A-Sepharose beads (GE Healthcare), previously coated with 10 μg/ml each of poly-(dI-dC), poly-(dG-dC), and poly-(dA-dT) and with 100 μg/ml BSA in RIPA buffer (1 mm EDTA, 0.5 mm EGTA, 10 mm Tris, pH 8, 1% Triton X-100, 0.1% Na deoxycholate, 0.1% SDS, 140 mm NaCl, and 1 mm PMSF). The aliquots were then incubated overnight with 1 μg of the appropriate antibodies in a total volume of 1 ml of RIPA buffer and immunoprecipitated as described previously (Ferrai et al., 2007). After immunoprecipitation, the material was treated with RNase A (50 μg/ml) for 30 min at 37°C and by proteinase K (500 μg/ml) in 0.5% SDS at the same temperature overnight. Formaldehyde cross-links were reverted by heating the samples at 65°C for 6 h, and the DNA was purified with phenol extraction and resuspended in 200 μl of distilled water. Resuspended material (4 μl) was used as a template in PCRs. PCR primers used are the following: mArx-F2, 5′-GTCTATAAGTACA-ATGGTGACAC-3′; mArx-R2, 5′-CTCCATC-AAGATCCTTCTC-3′ (amplification product, 240 bp); interleukin-1β (IL-1β)-F, 5′-ACCTATCTTCTTCGACACATGGG-3′; IL-1β-R, 5′-GGGCTTATCATCTTTCAACACGC-3′ (amplification product 200 bp). PCR primers for Neuropilin-2 (Npn2) regulatory sequences were used as described by Le et al. (2007). PCR products were analyzed on 2% agarose gels in 0.5× TBE buffer. The specific antibodies used for immunoprecipitations were against acetylated histone H3 (#06-599; Millipore) and Dlx proteins (Kuwajima et al., 2006). For mock controls, chromatin was immunoprecipitated either without antibody or with an unrelated polyclonal antibody against uPAR (urokinase-type plasminogen activator receptor) generated in the laboratory of Prof. Blasi (San Raffaele Scientific Institute, Milan, Italy) (Ferrai et al., 2007).
Organotypic culture, electroporations, and grafting experiments.
Slice cultures of embryonic mouse forebrain were prepared as described previously (Anderson et al., 1997b). Briefly, mouse brains were isolated and embedded in 4% low-melting agarose (Sigma-Aldrich), and 250-μm-thick coronal sections were cut on a vibratome. The sections were then transferred to polycarbonate culture membranes (diameter, 13 mm; pore size, 8 μm; Costar) in organ tissue dishes containing 1.5 ml of serum-containing medium (Invitrogen α-MEM with 10% fetal calf serum, glutamine, penicillin, and streptomycin). Slices were maintained for 1 h at 37°C in 5% CO2 in a standard sterile incubator. Before changing to the Neurobasal/B27 (Invitrogen) medium, sections were electroporated with a square electroporator (ECM830; BTX) using planar electrodes (BTX) as described by Stühmer et al. (2002). Two 5 ms electric pulses of 100 V were applied to targeting one side of the section. In those cases in which cell migration from MGE needed to be assessed, the day after electroporation the targeted MGE was carefully isolated and transplanted into the homologous region in the contralateral side of the same sections. After grafting, slices were cultured up to 48 h.
Results were expressed as mean value ± SD and were tested for statistical significance by the one-tailed Student's t test for paired differences with GraphPad Prism software.
Identification of genomic regions controlling Arx activity
The Arx gene exhibits a fairly elaborated expression pattern: during embryonic development, it is detected in different organs and tissues such as brain, floor plate, somites, pancreas, and gonads. In particular, in the CNS, its expression has been described in different regions, including the ventral thalamus, eminentia thalami, medial and lateral ganglionic eminences, and cerebral cortex (Miura et al., 1997; Bienvenu et al., 2002; Kitamura et al., 2002; Strømme et al., 2002; Colombo et al., 2004; Poirier et al., 2004; Yoshihara et al., 2005). Furthermore, within these neural compartments, Arx was found to act at different stages of differentiation. In fact, whereas in the cerebral cortex Arx is selectively expressed in the ventricular zone (VZ) proliferating cells, in the basal ganglia, most Arx-positive cells correspond either to proliferating SVZ neuroblasts or to postmitotic young neurons (Colombo et al., 2004; Cobos et al., 2005a).
Therefore, to understand how this complex temporal and spatial expression profile was achieved, we sought to search the cis-regulatory elements controlling Arx expression in the brain. We first performed a multiple phylogenetic alignment of genomic DNA including the Arx gene locus and its flanking regions using the VISTA browser tool (http://genome.lbl.gov/vista). This analysis compared orthologous sequences from human, mouse, and zebrafish genomes. Beside a reasonable nucleotide conservation of exonic sequences, this search highlighted three small domains localized downstream of the Arx gene and exhibiting a nucleotide conservation >50% (Fig. 1A). These three domains were termed ultraconserved Arx sequence 1 (UAS1), UAS2, and UAS3. In particular, UAS1 is 651 bp long and located 3.7 kb downstream to Arx poly-A signal (ChrX: 90,546,954–90,547,604; UCSC Genome Browser), UAS2 is a 328 sequence placed 5.1 kb downstream (ChrX: 90,548,333–90,548,660) and, finally, UAS3 is 811 bp long and located 12.6 kb distally (ChrX: 90,555,532–90,556,342).
Remarkably, UAS3 was found included within the last intron of the PolA1 gene, coding for the DNA polymerase α1. Both genes display a tail-to-tail orientation with the last exon of PolA1 being located 6.8 kb downstream of the Arx coding region. Conversely, we did not detect sufficient nucleotide conservation among the different species in the region spanning the 10 kb upstream to the Arx initiation codon (Fig. 1A).
To uncover putative cis-regulatory elements, the 13.5 kb genomic DNA located downstream of the Arx gene was split into two fragments of 3.5 and 10 kb, the first encompassing UAS1 and UAS2, and the second containing UAS3. Both sequences were cloned upstream of a minimal β-actin promoter followed by the LacZ reporter gene. Transgenic animals were generated and analyzed for β-galactosidase activity. In five different transgenic embryos, the 3.5 kb sequence was found able to specifically direct LacZ expression in the developing cerebral cortex, eminentia thalami, and ventral thalamus (Fig. 1B). Conversely, the 10 kb fragment was able to drive reporter expression in the developing basal ganglia, ventral thalamus, floor plate, and pancreas anlage, as tested in four different transgenic embryos (Fig. 1C). These results demonstrated that the two genomic fragments were able to target reporter expression in domains normally expressing the Arx gene. Interestingly, these two sequences were also found capable of independently targeting LacZ expression and recapitulating two complementary Arx-expressing domains. These findings indicate that Arx expression relies on at least two different regulatory elements that may operate independently from each other.
Characterization of the UAS3-mediated spatial–temporal expression profile
To determine in vivo whether the UAS3 sequence, present within the last Pola1 intron, could act as cis-acting regulatory element controlling Arx expression, we designed reporter constructs encompassing a 0.8 kb UAS3-containing DNA fragment and generated transgenic mice. Two stable mouse lines were characterized and found to exhibit identical expressions of the reporter gene. β-Galactosidase activity was first detected at E9 in the diencephalon and, starting from E10, in the ganglionic eminences (ge), where the expression of the reporter was found maintained during further development (Fig. 2A,B).
In E13.5 coronal brain sections, a strong labeling was observed in the olfactory tubercle (ot), LGE, MGE, caudal ganglionic eminence, ventral thalamus (vt), and hypothalamus (Fig. 2D–G). Transgene expression appeared undetectable in the ventricular regions of these structures, in agreement with the Arx expression pattern previously reported in the SVZ and mantle zone, but not in the proliferative periventricular areas (Miura et al., 1997; Colombo et al., 2004; Cobos et al., 2005a). Interestingly, two main streams of tangentially migrating cells starting from the ventral basal ganglia were highlighted (Fig. 2D–F). This observation is consistent with the dynamics of migration of GABAergic cortical interneuron precursors originating from the ventral pallium (Corbin et al., 2001; Marín and Rubenstein, 2001; Métin et al., 2006; Wonders and Anderson, 2006). Similar to interneurons, by E15.5, numerous LacZ-positive cells seemed to have reached the dorsal pallium as three large streams invading the cortical subventricular region and intermediate and marginal zones (Fig. 2H–K, arrows in I). To ascertain the identity of LacZ-labeled cells, an immunofluorescent analysis was performed using antibodies raised against markers of glutamatergic or GABAergic neurons, such as Tbr1 and GAD65, respectively. Although β-galactosidase-labeled cells were found negative for Tbr1, virtually all expressed GAD65, identifying these unambiguously as GABAergic cortical interneurons (Fig. 2L–O). Therefore, the UAS3 genomic element appears to induce both transgene expression dynamics and cell type-specific localization, in a manner closely recapitulating Arx expression pattern in the forebrain.
Arx expression is maintained in postnatal stages and in the adult brain in the rostral migratory stream, neurons of the olfactory bulbs, and GABAergic cortical interneurons (Colombo et al., 2004). Hence, we wondered whether the UAS3 genomic element could also control Arx expression postnatally. Reporter gene expression was detected in adult brains of both transgenic lines. In particular, β-galactosidase activity was found localized in cells of the olfactory bulbs (Fig. 3A), rostral migratory stream (Fig. 3B), cerebral cortex (Fig. 3C), striatum, septum (Fig. 3D), ventral thalamus (Fig. 3E), reticular thalamic nucleus (Fig. 3E), dorsal-medial hypothalamic nucleus (Fig. 3F), and telencephalic subventricular region (Fig. 3G). Thus, this spatial distribution of the enhancer activity closely matched the described Arx endogenous expression profile (Colombo et al., 2004; Yoshihara et al., 2005). β-Galactosidase labeling provides an informative resolution mapping of the Arx expression pattern at a single-cell level. Therefore, transgene-positive cortical cells were assayed for the expression of several markers for GABAergic neurons to determine which subclasses of interneurons UAS3 targets. Interestingly, we found that most, if not all, neurons belonging to the three main subsets of GABAergic interneurons identified by the nonoverlapping expression of parvalbumin, calretinin, and NPY and accounting for ∼80% of all the cortical interneurons were coexpressing the reporter gene (Fig. 3H–O). This finding was in agreement with previous observations regarding Arx expression in the adult cerebral cortex (Colombo et al., 2004). A thorough analysis was next performed using primary cultures of E18.5 wild-type and transgenic hippocampal neurons (Bonanomi et al., 2005). Arx was found present only in the neuronal fraction, but not in glial cells (Fig. 4A), and, in particular, appeared expressed selectively in most, if not all, isolated GABAergic neurons as demonstrated using the GABAergic markers GAD65 and vesicular GABA transporter (VGAT) (Fig. 4B–D, arrows). β-Galactosidase activity in primary neuronal cultures was detected as single dots in the cytoplasm of Arx-positive neurons (Fig. 4F, arrowheads). Furthermore, transgene activity was detected in a large fraction of GAD65+ neurons, but not Vglut1+ neurons, identifying GABAergic and glutamatergic neuronal populations, respectively (Fig. 4G,H, arrows). These findings imply that Arx and the reporter controlled by the UAS3 genomic element share a common expression profile in most of the mature cortical and hippocampal GABAergic interneurons.
Subsequently, the transgene expression in adult olfactory bulb and striatum was analyzed. A strong β-galactosidase staining was detected in the granular cell layer (gcl) and glomerular layer (gl) of the adult olfactory bulb (Fig. 5A,B). The glomerular layer represents the domain in which olfactory axons establish synaptic contacts with the primary dendrites of mitral and tufted projection neurons. Local interneurons surround the glomeruli and are thereby termed periglomerular cells. UAS3 element was found to target transgene expression both in GABAergic and dopaminergic periglomerular interneurons as shown by colabeling with GAD65, calbindin, and TH (Fig. 5C–H). Furthermore, β-galactosidase labeled a significant fraction of GABA+ granular cells in the gcl (Fig. 5C,E). Conversely, transgene expression was not detected in olfactory escheating glia (NPY+) nor in glutamatergic mitral neurons (data not shown). Yet again, this expression pattern closely matched Arx protein distribution previously described in the adult olfactory bulbs (Yoshihara et al., 2005). Finally, UAS3-dependent expression was analyzed in the adult striatum. LacZ staining revealed only a few scattered positive cells in this structure (Fig. 5I). Colocalization experiments with the principal striatal spiny neuron marker DARPP32 or the striatal interneuron markers NPY or ChAT revealed a β-galactosidase coexpression only with the latter cell population (Fig. 5J–L). These findings indicate that the striatal UAS3 transgene expression became confined to GABAergic and cholinergic interneurons in adulthood.
Molecular analysis of the UAS3 enhancer element
To identify the minimal sequence acting as enhancer in the UAS3 domain, we performed a homology search (phylogenetic footprinting) in four different vertebrate species: human, mouse, zebrafish, and fugu (Fig. 6A). Only a stretch of 205 bp in this entire domain revealed a high nucleotide conservation among the four species (from 100% to 91% when compared human to mouse and zebrafish, respectively). In contrast, identity in UAS3 sequences outside this core was generally much lower (42% human to zebrafish). We named the core element mUAS3, for minimal UAS3 enhancer sequence. The mUAS3 sequence was compared with a library of matrix descriptions for transcription-binding sites using the MatInspector software (Genomatix). By DNA-binding module selection and relevance to forebrain development, we chose to further analyze two 5′-TAATT-3′ sites that we termed HD-1 and HD-2, which are highly conserved in all different sequences and are bound by homeodomain transcription factors. Although several different homeodomain proteins are normally expressed in the forebrain, only a few are localized in the ventral telencephalon. In particular, we considered as possible activators those expressed in the ganglionic eminences: Mash1, Nkx2.1, Gsh1/2, Pax6, Isl1, and Dlx1/2/5. To test these putative activators, expression vectors for each of these genes were cotransfected with a luciferase reporter gene controlled by the UAS3 element in P19 embryonal carcinoma cells. Importantly, we detected a dramatic increase in luciferase activity only in the presence of Dlx1/2/5, whereas a negligible expression was observed after cotransfecting all the other genes (Fig. 6B, data not shown). Therefore, despite the fact that mUAS3 element displayed a rather consensual homeodomain binding sequence, only Dlx proteins were able to consistently induce luciferase expression (Fig. 6B, data not shown). This transactivation ability was strongly reduced when the mUAS3 sequence was devoid of the two HD-1/2 sequences (Fig. 6B). Hence, despite the presence of few additional AT palindromic sequences in the UAS3 element, none appeared capable of sustaining Dlx2 activity in the absence of HD-1/2 elements. Next, to unambiguously confirm that Dlx2 interacts with the mUAS3 sequence, we first used EMSAs. Hence, DLX2 protein was translated in vitro and was indeed found able to directly bind to both sequences (Fig. 6C). Subsequently, to determine whether this interaction also occurred in vivo, we performed chromatin immunoprecipitation assays. As a positive control, we used a previously reported Dlx2-binding sequence on the Npn2 promoter, whereas an IL-1β promoter element was chosen as unspecific sequence (Le et al., 2007). The chromatin was isolated from E14.5 mouse forebrains and precipitated with a highly specific Dlx2 antibody (Kuwajima et al., 2006). In all cases (3 of 3), only the Arx and Npn2, but not the IL-1β, sequences were precipitated by the anti-Dlx2 antibody, clearly indicating that Arx indeed is a reliable Dlx2 target gene during embryonic development. Finally, to verify whether the mUAS3 element itself can induce reporter activity in forebrain GABAergic tissue, we generated transgenic mice with multiple oligomerized mUAS3 sequences upstream of the β-galactosidase reporter. All founder embryos (5 of 5) displayed a tissue-specific β-galactosidase activity that overlapped with Arx endogenous expression, although with a much weaker intensity than the UAS3 reporter activity (data not shown).
In vivo requirement of Dlx binding for UAS3 proper activity
To reveal whether Dlx activity was sufficient to promote the UAS3 enhancer activity in vivo, we ectopically expressed Dlx2 either in the forebrain or the midbrain of E13.5 mouse brain slices. At first, we confirmed our experimental design demonstrating that Dlx2 misexpression was indeed able to activate Gad2 expression in the forebrain tissue as previously reported by Stühmer et al. (2002) (Fig. 7A–D). In addition to the forebrain, the exogenous Dlx2 also triggered GAD65 protein activation in the mesencephalon, a neural tissue in which GABAergic neuronal fate is normally controlled in a Dlx2-independent manner (Miyoshi et al., 2004; Guimera et al., 2006; Nakatani et al., 2007). We next assayed whether β-galactosidase expression was activated after Dlx2 misexpression in brain slices of UAS3 transgenic mice. Both in the cerebral cortex and the ventral mesencephalon, Dlx2 was found able to elicit the activation of the reporter gene with a virtually perfect correspondence between Dlx2-misexpressing cells and LacZ ectopic expression (Fig. 7I–P). These results strongly indicate that Dlx2 mediates the activation of the UAS3 enhancer element controlling its spatiotemporal expression. If the UAS3 element represents the faithful Arx GABAergic enhancer, Dlx2 misexpression should trigger Arx endogenous expression in a similar manner. Indeed, using adjacent sections to those analyzed for UAS3-dependent β-galactosidase activity, Arx immunoreactivity was detected in all Dlx2-misexpressing cortical and mesencephalic neurons. Altogether, these data suggest that Arx expression is regulated in vivo by Dlx2 activity most probably through the activation of the UAS3 enhancer.
To further characterize the UAS3-dependent Dlx activity, we crossed UAS3 transgenic mice with Dlx1/2 mutant animals and analyzed reporter activity in a Dlx1/2 null background. Interestingly, β-galactosidase activity was still detectable in these mice in MGE, LGE, and ventral thalamus, albeit to a much lower intensity than in controls. Conversely, LacZ gene expression was not strongly reduced in the caudal ganglionic eminences (Fig. 8H,J). Furthermore, reporter signal was excluded from the cerebral cortex because of the failure of GABAergic neurons to tangentially migrate in Dlx1/2 mutants (Fig. 8B,D,F, H,J,L, arrows). These findings are in agreement with previous observations that revealed a still detectable, although strongly reduced, Arx mRNA expression in Dlx1/2 mutant mice (Cobos et al., 2005a).
Together, these results suggest that Dlx1/2 transcription factors are sufficient to control UAS3 activity expression profile in misexpression studies and are necessary for correct UAS3 activity. Moreover, Dlx1/2-independent molecular mechanisms seem to act to maintain a detectable activity of the UAS3 element at least in some tissues expressing Arx.
Arx modulates the Dlx-dependent migratory activity but not the ability of Dlx to specify the GABAergic cell fate
Dlx1/2 transcription factors are critical throughout forebrain development. They play pleiotropic functions in regulating ventral forebrain structures and GABAergic neuronal differentiation and migration. In the light of the strong downregulation of Arx expression in Dlx1/2 mutants, we wondered whether some of the Dlx-dependent functions might be mediated by Arx activity. For instance, because both Dlx1/2 and Arx loss of function lead to a severe reduction of GABAergic tangential migration (Anderson et al., 1997b, 2001; Kitamura et al., 2002; Colombo et al., 2007), we tested whether Dlx1/2-induced migration is relying on a sustained Arx expression. Therefore, we assessed whether Arx reexpression in a Dlx1/2 mutant background was sufficient to rescue neuronal migration activity. As an experimental system, we misexpressed either EGFP or Arx-IRES-EGFP in wild-type or Dlx1/2 mutant ventral forebrain. On the following day, we transplanted the MGE electroporated tissue in the homologous region of the contralateral side of cultured brain slices. Using this approach, GFP-expressing neurons could be clearly followed during their migration from the transplanted MGE outward in the striatum and cortex (Fig. 9A–C). As expected, 48 h after transplantation, GFP+ Dlx1/2 mutant cells exhibited an almost negligible migration ability from the transplanted MGE, in all cases analyzed (8 of 8) (Fig. 9D–F). On the contrary, a robust number of Dlx1/2 mutant-Arx-overexpressing cells showed a sustained migration from the transplanted MGE, reaching in some cases also the basolateral cortex (Fig. 9G–I). To quantify this rescue in migration ability, we subdivided the forebrain into four consecutive domains centered on the transplanted MGE and counted the number of GFP+ cells present in each domain. Arx overexpression in Dlx1/2 mutant cells induced an approximately fivefold increase in cell migration in the first three domains (Fig. 9J). This represented an ∼65% rescue in first and second proximal domains, and an ∼32% rescue in the distal areas, compared with wild type (Fig. 9J).
Hence, Arx overexpression in Dlx1/2 mutant can efficiently rescue migration activity defects in a Dlx1/2 mutant context, in particular in MGE proximal domains. Conversely, Arx overexpression in Dlx1/2-deficient cells failed to promote long-term migration in the fourth most distal domain, at least in a 48 h time window.
Dlx1/2/5 ectopic expression is sufficient to activate some aspects of GABAergic cell fate specification, such as GAD65/67 expression in the neural progenitors of the cerebral cortex normally committed to a glutamatergic phenotype (Stühmer et al., 2002). Therefore, Dlx genes appear instrumental in promoting initial aspects of GABAergic differentiation at least in a heterologous forebrain domain. We wondered whether this aspect of the Dlx function might be affected by Arx activity, because Dlx ectopic activity triggered Arx expression as previously described. To answer this question, E13.5 Arx-deficient brain slices were electroporated in the lateral cortex with a Dlx2-IRES-GFP expression vector and maintained in culture for 30 h before final analysis. As in wild-type brain slices (data not shown), Dlx2 efficiently activated GAD65 protein in Arx mutant cortical progenitors as detected with a specific antibody (monoclonal GAD65; Sigma) (Fig. 10A–C). Arx inactivation was achieved by replacing the Arx coding sequence with a LacZ reporter gene. This approach also allowed the detection of the activation of the Arx endogenous locus by following β-galactosidase activity (Collombat et al., 2003). Hence, electroporated Dlx2 cortical cells were expected to coexpress GAD65 together with β-galactosidase. Indeed, we detected a virtual complete overlapping expression pattern between GFP and β-galactosidase activation (Fig. 10D–F). These findings suggest that the acquisition of GABAergic cell features promoted by Dlx activation does not rely on a functional Arx allele.
Together, these results provide evidence that Arx is necessary to promote Dlx-dependent GABAergic cell migration, but is dispensable for Dlx ability to induce GABAergic cell fate commitment.
In this study, we report the identification and functional characterization of the Arx GABAergic enhancer and disclose the biological significance of the Dlx–Arx genetic hierarchy. Arx-specific forebrain expression is achieved by the spatial combinatorial expression promoted by at least two independent enhancers located downstream of the Arx coding region and 8 kb apart. The molecular mechanisms regulating Arx expression through these two domains appear completely divergent, because Dlx is not involved in controlling the activity of the cortical Arx enhancer, which presents a specific set of transcription factor-binding elements (G. Colasante and V. Broccoli, unpublished results). Hence, Arx expression is accomplished by independently regulated enhancer modules that appear separated and distributed along the genome. It is intriguing that both Arx enhancer elements are highly conserved from fugu to human, suggesting a relevant function of Arx in neuroanatomical features of the ancestral vertebrate cerebrum. The GABAergic enhancer element is located 13 kb downstream to the last Arx exon and enclosed in the last intron of the PolA1 gene. This is a very peculiar site for an enhancer element, being contained in a different gene from the one it is controlling. This structural organization is conserved in different vertebrates, raising the question of how this could be achieved and positively selected during evolution. A possibility would lie in the fact that the last three introns of the PolA1 gene are extremely large (110 kb altogether), and therefore evolutive pressure may have been applied in these sequences without affecting the PolA1 gene transcription.
Recently, highly evolutionarily conserved noncoding regions were identified in systematic studies analyzing the whole human genome (Bejerano et al., 2004; Woolfe et al., 2005; Pennacchio et al., 2006). In particular, Bejerano et al. (2004) isolated 481 nucleotide segments, coined as ultraconserved elements, which are absolutely conserved between orthologous regions of the rat and mouse genomes (100% identity), and display a significant homology in chicken and fish. Interestingly, five of these conserved sequence modules were detected in the last introns of the POLA1 gene, and one of them (uc.467) was included into the Arx GABAergic enhancer sequence herein analyzed. Because in our functional analysis the enhancers controlling Arx expression in testis and muscles were not identified, it is conceivable that these sequences may lie on the other conserved sequences within the PolA1 gene, but located at least 20 kb away downstream to the Arx coding region. Ultraconserved elements in the human genome have been shown to exhibit almost no natural variation in the overall population, with a rate of nucleotide change that is ∼20 times slower than the average genome (Bejerano et al., 2004). Therefore, it is conceivable that these highly conserved noncoding sequences may be a favorable substrate for the development of disease-causing mutations. The Arx GABAergic enhancer represents a good candidate as a noncoding sequence that, when mutated, may lead to Arx-specific silencing restricted to forebrain interneurons, and therefore causes the development of related neuropathological conditions (e.g., epilepsy, mental retardation, and motor deficits). Mutations in the enhancer elements of developmental genes like Shh, Ret, Pitx2, Gli3, and Pax6 have been already described to cause human congenital diseases (Emison et al., 2005; Kleinjan and van Heyningen, 2005; Gurnett et al., 2007). Thus, it would be valuable to perform a mutational screening of this Arx enhancer sequence in those highly related neurological disorders in which mutations in the Arx coding region failed to be identified.
The analysis of the expression of the GABAergic enhancer confirmed previous results on Arx endogenous expression in the great majority of cortical interneurons belonging to all the different subclasses. Notably, β-galactosidase activity was strongly detected in neurons of the adult brain. β-Galactosidase is a highly stable enzyme, which remains active for long time in the tissue, even longer, in some cases, than the endogenous protein. However, it is plausible that Arx expression is maintained in adult brain as confirmed by immunohistochemistry on mature hippocampal neurons and RNA expression profile in adult brain. Therefore, Arx may exert additional roles on mature and functional cortical interneurons. The significance of these findings remains to be assessed.
The highly conserved core of the Arx GABAergic enhancer (mUAS3) contains two recognizable binding modules for homeodomain transcription factors. Although these binding motives do not fully match the Dlx consensus binding requirement (G-A/C-TAATT-A/G-G/C), we detected a very robust and specific dependence by these proteins (Dlx1/2/5). Other homeodomain-containing proteins, such as Nkx2.1, Gsh1/2, and Isl1, are normally expressed in the ventral telencephalon and could represent putative candidates activating the Arx promoter. However, both in vitro and in vivo transient expression assays revealed that all these other proteins were unable to elicit a reliable activation of the Arx enhancer. This apparent contradiction can be explained by considering the presence of molecular partners of the Dlx proteins that can cooperate with them to promote high enhancer activation. Indeed, other sequences in the Arx enhancer flanking the putative Dlx-binding sites are extremely conserved and may act as binding elements for other transcription factors that we were unable to identify so far. This hypothesis is also supported by the finding that Arx expression is still detectable, although much weaker, in Dlx1/2 mutants, further emphasizing a role for other unidentified proteins.
Dlx proteins are key molecular players of ventral pallium development with multiple functions in GABAergic neurons, regulating various aspects of their differentiation and migration. In particular, Dlx1/2 proteins are redundant in specifying a later subset of neuronal subpallial progenitors and promoting their terminal differentiation by repressing Notch signaling (Anderson et al., 1997a; Yun et al., 2002). In addition, Dlx1/2 can promote GABAergic neuronal cell fate (Stühmer, 2002) while repressing oligodendrocyte precursor cell formation by negatively regulating Olig2 expression (Petryniak et al., 2007). During interneuron migration, Dlx1/2 proteins sustain cell motility by repressing neurite development through the inhibition of the Pak3 kinase, and lead the cellular path by controlling the expression of guidance molecules such as Neuropilin-2 (Le et al., 2007). How Dlx proteins tightly control these different processes at the molecular level is unknown. Few Dlx molecular binding proteins have been recognized so far, albeit their relevance in these processes has not been addressed yet (Sasaki et al., 2002). Therefore, it is conceivable that Dlx proteins could directly act on different molecular targets that mediate each of these different processes. Arx appears to be an important intermediate target of Dlx proteins, and here we show that Dlx proteins directly control its expression. However, our findings support the idea that Arx could be acting to accomplish some, but not other, Dlx-dependent genetic programs. In fact, the Dlx–Arx pathway is not required for ectopic GABAergic cell gene expression, as shown by the retained ability of Dlx2 protein to induce ectopic GAD+ cells in the absence of Arx. This finding is corroborated by both the Arx failure to induce any GABAergic molecular marker when overexpressed in ectopic forebrain regions (V. Broccoli, I. Cobos, and J. Rubenstein, unpublished results) and the maintenance of GAD expression in Arx mutant brains (Colombo et al., 2007). However, these data do not entirely exclude any Arx involvement in GABAergic neuronal specification inside the ganglionic eminences, where still-undetermined molecular mechanisms are controlling GABAergic fate. In contrast, Arx has a remarkable role in promoting GABAergic interneuron migration, similarly to what was described for Dlx proteins (Anderson et al., 1997b; Kitamura et al., 2002; Colombo et al., 2007). We show here that their comparable function in this system is partially redundant, because Arx expression in Dlx1/2 mutant tissue is able to rescue interneuron migration to a certain extent (between 14 and 30%). It has been recently described that ectopic activation of the Pak3 kinase may account in part for the inhibition of neuronal migration in Dlx mutant brains (Cobos et al., 2007). Interestingly, we found that Pak3 is also upregulated in Arx mutants as tested by qPCRs on forebrain lysates (our unpublished results). However, the rescue obtained with Arx is quantitatively more relevant with respect to that achieved by Pak3 overexpression, indicating that Arx should also act on other still-unknown molecular players. Together, Arx may be considered a molecular switch acting downstream to Dlx proteins and contributing in the selection of the different Dlx-dependent processes.
It is worthwhile to note that both Dlx and Arx genes have a strong relevance in human neurodevelopmental disorders. In fact, Dlx genes were shown to be linked to epilepsy and Rett syndrome (Cobos et al., 2005b; Horike et al., 2005). Also, members of the Dlx homeobox family are found in two autism-susceptibility loci, chromosome 2q and 7q (Hamilton et al., 2005). All these neuropathological conditions are closely overlapping with those depending on Arx mutations. These findings provide evidence of a conserved functional relationship between Dlx and Arx in human brain development, which controls an overlapping transcriptional cascade with common molecular targets. Thus, this indicates that discovering other members of this molecular pathway will offer significant novel candidates possibly responsible for other related human diseases. The term interneuronopathies has been recently introduced to describe those human diseases that share all or a set of neurological symptoms, which may range from epilepsy to nonsyndromic mental retardation, that arise from the impairment of tangential GABAergic cortical interneuron migration (Kato and Dobyns, 2005). The analysis of the downstream molecular pathways controlled by Dlx and Arx transcription factors will enhance our knowledge of interneuron development and provide insights into such important neurological disorders.
This work was supported by Telethon (GGP07181), Italian Ministry of Research (Investment Fund for Basic Research project), and “Ricerca Finalizzata-ex Art.56” Italian Ministry of Health (V.B.); Nina Ireland, Human Frontiers Science Program, and National Institute of Mental Health Grants R01 MH49428-01 and K05 MH065670 (J.L.R.R.); Max Planck Society and National Institutes of Health Beta Cell Biology Consortium (U19 DK072495-01) (A.M., P.C.); and Dr. H. Storz and Alte Leipziger Foundation (A.M.). We thank Drs. V. Pachnis, J. Briscoe, J. Gécz, and A. Fairen for sharing reagents and materials. We are grateful to Dr. V. Zappavigna for stimulating discussion. Elisabetta Feretti and Massimiliano Golino are acknowledged for advice on EMSA and initial cloning work, respectively.
- Correspondence should be addressed to Vania Broccoli, Stem Cell Research Institute (SCRI), San Raffaele Scientific Institute, Via Olgettina 58, 20132 Milan, Italy.