Dlx Homeobox Genes Promote Cortical Interneuron Migration from the Basal Forebrain by Direct Repression of the Semaphorin Receptor Neuropilin-2*

Dlx homeobox genes play an important role in vertebrate forebrain development. Dlx1/Dlx2 null mice die at birth with an abnormal cortical phenotype, including impaired differentiation and migration of GABAergic interneurons to the neocortex. However, the molecular basis for these defects downstream of loss of Dlx1/Dlx2 function is unknown. Neuropilin-2 (NRP-2) is a receptor for Class III semaphorins, which inhibit neuronal migration. Herein, we show that Neuropilin-2 is a specific DLX1 and DLX2 transcriptional target by applying chromatin immunoprecipitation to embryonic forebrain tissues. Both homeobox proteins repress Nrp-2 expression in vitro, confirming the functional significance of DLX binding. Furthermore, the homeodomain of DLX1 and DLX2 is necessary for DNA binding and this binding is essential for Dlx repression of Nrp-2 expression. Of importance, there is up-regulated and aberrant expression of NRP-2 in the forebrains of Dlx1/Dlx2 null mice. This is the first report that DLX1 or DLX2 can function as transcriptional repressors. Our data show that DLX proteins specifically mediate the repression of Neuropilin-2 in the developing forebrain. As well, our results support the hypothesis that down-regulation of Neuropilin-2 expression may facilitate tangential interneuron migration from the basal forebrain.

Members of the Dlx homeobox gene family, orthologs of Distal-less in Drosophila melanogaster, are expressed in the developing brain (1,2), retina (3), craniofacial structures, and limbs (4). Four Dlx genes, Dlx1, Dlx2, Dlx5, and Dlx6, are expressed in overlapping domains in the subcortical telencephalon and diencephalon, including the ventral thalamus and the ganglionic eminences. There are distinct boundaries of Dlx1 and Dlx2 expression at the pallial-subpallial boundary (1). Insights into the functional role of Dlx genes in development have been primarily gained from analysis of the phenotypes of mice with targeted deletions of Dlx1/Dlx2 (5)(6)(7)(8), Dlx5 (9), and Dlx5/Dlx6 (10). The single Dlx1 and Dlx2 knockouts have relatively normal forebrain development at birth, which is consistent with functional redundancy between Dlx1 and Dlx2 in this anatomic region (1,4). However, postnatal Dlx1 mutants show specific differentiation defects in interneuron subclasses (11,12). In the absence of both Dlx1 and Dlx2 function, there is abnormal development of the subventricular zone (SVZ) 2 of the ganglionic eminences. There is an almost complete loss of tangential migration of GABAergic interneurons from the medial ganglionic eminence (MGE) to the neocortex and from the lateral ganglionic eminence (LGE) to the olfactory bulb (5,13) 3 . These migrations comprise the major source of cortical inhibitory interneurons in the murine telencephalon (5,14,15). Furthermore, it is evident that there are both Dlx-dependent and Dlxindependent pathways affecting the differentiation and subsequent migration of interneurons to the striatum (6), olfactory bulb (13), and hippocampus (16).
Our understanding of Dlx gene function is limited by the current paucity of established transcriptional targets of Dlx genes. We have commenced isolating and characterizing candidate Dlx gene targets in the developing mouse, particularly in the forebrain, using chromatin immunoaffinity purification (ChIP) of embryonic tissues that regionally express Dlx genes. We have optimized the ChIP protocol using embryonic ganglionic eminence tissues and specific affinity-purified rabbit polyclonal antibodies to the proteins encoded by the Dlx1 and Dlx2 genes (1) to isolate the Dlx5/Dlx6 intergenic enhancer, a DLX target gene promoter sequence (17). In this study, we exploited this approach to identify a novel DLX target in the embryonic forebrain.
Several factors are important for the regulation of differentiation and/or migration of interneurons from the forebrain. These include transcription factors such as Pax6 (18,19), Emx1/Emx2 (20), and Tlx (21), Neuregulin-1/Erb-B4 signaling (22,23), and other subcortical chemorepulsive or cortical che-motactic guidance cues (24 -26). The semaphorins (Sema) 3A and 3F, by binding to their receptors Neuropilin-1 and Neuropilin-2, respectively, provide strong repulsive guidance cues and mediate sorting of tangentially migrating interneurons from the ganglionic eminences to the cortex and striatum (15). Different models to assess the function of members of the neuropilin family have been generated. A dominant negative form of Nrp-1 increases interneuron migration from the MGE to the striatum and reduces migration to the neocortex in vitro (15). In Nrp-2 knock-out mice, sympathetic and hippocampal neurons have reduced repulsive responses to Sema3F but not to Sema3A. There are also axonal pathfinding defects of specific cranial and sensory nerves and hippocampal mossy fiber projections (27)(28)(29). However, similar to the Nrp-1 dominant negative experiments, there are increased numbers of interneurons invading the developing and postnatal striatum in these mice (15). Evidence of ectopic expression of Neuropilin-2 (Nrp-2) in the Dlx1/Dlx2 mutant forebrain (15) provided a rationale to test whether direct DLX1 and/or DLX2 directly regulate Neuropilin-1/2 expression.
Herein, using ChIP of mid-gestation ganglionic eminences, we demonstrate that DLX1 and DLX2 directly bind to a specific region of the Nrp-2 but not to the Nrp-1 promoter in vivo. Furthermore, both DLX1 and DLX2 inhibit transcription of Nrp-2 in vitro. Moreover, loss of Dlx1 and Dlx2 function leads to the increased and ectopic expression of Neuropilin-2 as subventricular ectopias in the medial and lateral ganglionic eminences. These results support our hypothesis that a subpopulation of late-born Dlx-expressing interneurons (Neuropilin-2 low/non-expressing) successfully bypass semaphorin repulsive cues to migrate to the neocortex. Conversely, loss of Dlx1 and Dlx2 function results in increased and aberrant Neuropilin-2 expression in this population, and their resultant responsiveness to semaphorin signaling may contribute to block tangential interneuron migration from the basal telencephalon. These data delineate part of the molecular mechanism by which migrating interneurons may bypass inhibitory cues to reach their correct destinations in the forebrain.

MATERIALS AND METHODS
Animals-Mice with null mutations of Dlx1 and Dlx2 (a kind gift from Dr. J. Rubenstein, University of California, San Francisco) were maintained in a CD1 background at Manitoba Institute of Cell Biology following protocols approved by the University of Manitoba under the auspices of the Canadian Council on Animal Care. Dlx1/Dlx2 mutants were genotyped using published protocols (7,27). Embryonic age was determined by the day of appearance of the vaginal plug (E0.5), and tissues were processed as previously described (3). For comparative studies all mutant embryos were paired with wild-type littermate controls.
Chromatin Immunoprecipitation Assays-ChIP was performed under the following conditions as described (17). 1-2 ϫ 10 7 E13.5 ganglionic eminence cells were fixed with 1% paraformaldehyde for 90 min at room temperature. Sonication of cells (Sonifier cell disruptor 350) in SDS lysis buffer (1% SDS, 50 mM Tris-HCl, pH 8.1, 10 mM EDTA) on ice generated soluble chromatin complexes with DNA fragment lengths ranging between 100 and 300 bp. Specific polyclonal high affinity DLX antibodies were used to immunoprecipitate genomic DNA targets cross-linked to DLX1 or DLX2 homeoproteins. Genomic DNA from an E13.5 mouse embryo was used as a positive control. DNA derived from E13.5 hindbrain was used as a negative control because this tissue does not express any Dlx family members.
Reporter Gene Assays-We used the Lipofectamine 2000 reagent (Invitrogen) to transiently co-transfect luciferase gene reporters (1 g) into 10 6 human embryonic kidney 293 cells with pRSV-␤ galactosidase (Promega) (0.4 g) as an internal control and harvested cell lysates 36 h later. Subsequently, luciferase activities were measured with the Luciferase Reporter Assay System (Promega), normalizing luciferase activity with ␤-galactosidase activity. The Nrp-2 luciferase reporter contains the Nrp-2 promoter (444 -562 nt) (GenBank TM AF022854) directing expression of the luciferase gene. We expressed DLX1 and DLX2 (Dlx1 and Dlx2 cDNAs were a kind gift from Dr. J. Rubenstein, UCSF) and their derivatives from the pcDNA3 expression plasmid (Invitrogen). We generated Q50E variants using the QuikChange site-directed mutagenesis kit (Stratagene, La Jolla, CA). DLX-VP16 and DLX-Engrailed fusion constructs were generated as follows: we excised the N-terminal domain of either Dlx1 or Dlx2 by HindIII/PpuMI or EcoRI/ BsmBI (New England Biolabs) restriction enzyme digestions, respectively, and maintained the nuclear localization signal, homeodomain, and C-terminal domains. Engrailed (888-bp) and VP16 (255-bp) domains (30) were ligated to the 5Ј-end of these N-terminal-modified Dlx constructs.
Tissue Preparations and in Situ Hybridization-Tissues were prepared as previously described (1,17). E13.5 tissues (whole embryos) were fixed using 4% paraformaldehyde and cryopreserved using sucrose gradients followed by embedding in OCT medium (Tissue Tek), whereas central nervous system tissues from E16.5 and E18.5 embryos were prepared following dissection. Tissue samples were also sectioned coronally at a thickness of 15 m using a cryostat (Cryotome, ThermoShandon, Cheshire, UK). Digoxigenin in situ hybridization was carried out as described (1,3,17) using cRNA riboprobes for Nrp-1 and Nrp-2 (a kind gift from Dr. M. Tessier-Lavigne, Genentech, South San Francisco, CA).
Immunohistochemistry, Immunofluorescence, and Immunoblotting-Immunohistochemistry and single and double immunofluorescence experiments were performed as described (1,3,17). For immunohistochemistry and immunofluorescence we used the following primary antibodies: rabbit anti-DLX1 (1:100, N114 affinity-purified), rabbit anti-DLX2 (1:300, C199 affinity-purified) (DLX1 and DLX2 polyclonal antisera were kind gifts from Dr. J. Rubenstein, UCSF, and were affinity-purified (1) Marin et al. (15) established that the Class III semaphorins, semaphorins 3A and 3F, and their receptors, Neuropilin-1 and Neuropilin-2, play important roles in the sorting of interneurons that are migrating to the neocortex and the striatum. The promoters of both of these genes, Neuropilin-1 (Nrp-1) and Neuropilin-2 (Nrp-2), contain putative TAAT/ATTA homeodomain DNA binding sites. Hence, we considered Nrp-1 and Nrp-2 as candidate transcriptional targets of several homeobox genes regionally expressed in the subcortical telencephalon, including members of the Dlx gene family. To test this hypothesis, we assayed for the direct binding to both Nrp promoters by DLX1 and/or DLX2 in vivo. E13.5 ganglionic eminences were treated with formaldehyde to cross-link protein-DNA complexes. Using a modified ChIP procedure, soluble nucleoprotein complexes of ϳ100 -300 base pair fragments were immunoprecipitated using anti-DLX1 and/or DLX2 antibodies. We then used PCR to amplify candidate homeodomain binding regions in the Nrp-1 and Nrp-2 loci. These regions were chosen based on the presence of consensus homeodomain binding motifs and were designated Nrp-1i (nucleotides 104 -283), Nrp-1ii (nucleotides 711-900), Nrp-2i (nucleotides 138 -255), and Nrp-2ii (nucleotides 444 -562) (Fig. 1A). Notably, ChIP assays revealed that both DLX1 and DLX2 bound only to Nrp-2 promoter region 2 (Nrp-2ii) in embryonic striatum but there was no evidence for binding with the other promoter regions in vivo (Fig. 1B). The resulting amplicons were subcloned and sequenced to verify their identity and for subsequent biochemical analyses. As expected, control ChIP assays performed without antibody or with anti-DLX antibodies and chromatin derived from embryonic hindbrain tissues where Dlx genes are not expressed were negative.

DLX1 and DLX2 Homeobox Proteins Bind to a Neuropilin-2 Promoter Region in Embryonic Forebrain in Vivo-
DLX1 and DLX2 Bind to the Neuropilin-2 Promoter in Vitro-To determine whether DLX1 and DLX2 specifically bind to the Nrp-2 promoter region 2 in vitro, we used recombinant DLX1 and DLX2 proteins and radiolabeled Nrp-2ii promoter regions isolated from the ChIP assay. An electrophoretic gel mobility shift assay (EMSA) showed binding of both DLX1 and DLX2 to Nrp-2ii ( Fig. 2A, lanes 2, 6) that was competitively inhibited by unlabeled Nrp-2ii probe (lanes 3, 7). Moreover, the addition of specific anti-DLX1 or anti-DLX2 antibodies to the protein-DNA complex resulted in significant band mobility shifts (lanes 4, 8), whereas a nonspecific polyclonal antibody failed to produce such a "supershift" (lanes 5,9). Neither DLX1 nor DLX2 bind to the first TAAT motif of the Nrp-2ii in vitro. However, both recombinant Dlx proteins bind to the second motif ( Fig.  1A, box, supplemental Fig. S2, and data not shown). These experiments demonstrate that DLX1 and DLX2 specifically bind to the Nrp-2 promoter region 2 in vitro. As well, binding of DLX1 and DLX2 to this region may be mutually exclusive.
EMSA with nuclear extracts derived from embryonic ganglionic eminences showed that endogenous DLX1 and DLX2 proteins can bind to the Nrp-2 promoter (Fig. 2B, lane 2). Unlabeled probe can compete with radiolabeled oligonucleotides for both proteins (lane 3). Experiments using specific DLX1 or DLX2 antibodies and a control antibody confirmed the identity of the Dlx complexes (lanes 4 -6). The recombinant and endogenous DLX proteins do not have identical molecular weights, perhaps reflecting different binding partners or post-translational modifications such as phosphorylation in vivo (32). 4 These experiments demonstrate that DLX1 and DLX2 specifically bind to the Nrp-2 promoter region 2 in vitro.
DLX1 and DLX2 Repress Neuropilin-2 Promoter Expression in Vitro-The functional significance of DLX1 and DLX2 binding to the Nrp-2 promoter was assessed using luciferase reporter gene experiments. We co-transfected human embryonic kidney 293 cells or P19 embryonal carcinoma cells (data not shown) with an expression vector encoding Dlx1 or Dlx2 and a vector in which region 2 of the Nrp-2 promoter (444 -562 nt) drives luciferase expression. Co-transfection with either wild-type Dlx1 or Dlx2 expression constructs resulted in significant reductions of luciferase activity compared with controls (ϳ1.7-fold for Dlx1 and ϳ6-fold for Dlx2, p Ͻ 0.001; Fig. 3C), indicating that both DLX1 and DLX2 proteins can act as transcriptional repressors of Nrp-2 promoter expression.
We generated several different mutant constructs of DLX1 and DLX2 homeodomain proteins to explore the consequences of modifying specific DLX protein domains on activity of the reporter gene (Fig. 3A). The first set of constructs included a mutation that converts glutamine (Q) to glutamate (E) at amino acid position 50 (Q50E) of the DLX homeodomain. This Q50E mutant homeodomain is predicted to abrogate DNA binding, and the mutant protein may act in a dominant negative fashion (33). EMSA assays confirmed that the Q50E mutants do not bind to the Nrp-2 promoter (Fig. 3B). In co-transfection assays, mutant Dlx1 and Dlx2 (Q50E) were unable to repress luciferase gene expression when compared with the wild-type Dlx expression constructs (Fig. 3C). The DNA binding mutant affects a similar level of transcriptional activity for both Dlx1 and Dlx2; however, because the repression of Dlx1 on the Nrp-2 promoter is less than by Dlx2 in the first place, the difference in Dlx1 rescue (from repression) is not as significant as we found in Dlx2 rescue by the Q50E DNA binding mutants. The second group of constructs included an N-terminal domain exchange of DLX1 and DLX2 with either a VP16 activator domain or an Engrailed repressor domain (30). N-terminal DLX1 and DLX2-Engrailed chimeric fusion proteins further repressed transcription of the Nrp-2 promoter region 2 compared with the wild-type Dlx constructs. Conversely, N-terminal DLX1 and DLX2-VP16 chimeric fusion proteins activated transcription of the Nrp-2 promoter region, overcoming all or some of the transcriptional repression resulting from the wildtype Dlx constructs, by ϳ3.7and 1.7-fold, respectively (p Ͻ 0.001, Fig. 3D). In this assay, Dlx2 is a stronger repressor; hence this function is not completely alleviated by replacement by VP16 at its N terminus. Hence, from these data we can conclude that the direct interaction of DLX1 or DLX2 proteins with the Nrp-2 promoter results in the repression of Neuropilin-2 transcription.
DLX1 or DLX2 and Neuropilin-2 Expression Patterns in the Developing Forebrain-Dlx homeobox genes are expressed in interneurons that express ␥-aminobutyric acid (GABA) in the embryonic rostral forebrain (5,6,34,35). During the early stages of embryonic telencephalon development, the subpallium expresses Dlx1 and Dlx2 primarily in the ventricular and subventricular zones (1,2,36,37) with a clear limitation of expression at the LGE/neocortex (subpallialpallial) boundary. Dlx1 and Dlx2 expression are both well established by E10.5 (1). Expression of NRP-2 is reported to be primarily restricted to the mantle zone of the MGE as well as the olfactory cortex (15). We found that endogenous expression of Nrp-2 becomes well established from E13.5 in the mantle zone of the subcortical telencephalon using in situ hybridization (supplemental Fig. S5). Furthermore, we also detected NRP-2 protein expression as early as E13.5 using immunofluorescence (Figs. 4 and 5A). NRP-2 expression becomes more restricted to the mantle zone of the MGE and paleocortex at age E16.5 and E18.5 or postnatal day 0 (Figs. 5, B and C, supplemental Figs. S3 and S4). Co-expression studies of DLX1 or DLX2 and NRP-2 at E13.5 (for E16.5 and E18.5, see supplemental Figs. S3 and S4) show only a minimal overlapping pattern with majority populations of DLX1/DLX2 and NRP-2 single positive cells in the basal telencephalon (Fig.   FIGURE 1. Neuropilin-2, but not Neuropilin-1, is a DLX homeoprotein target in vivo. A, the sequences of candidate regulatory elements within the mouse Neuropilin-1 (GenBank TM AF482432) and Neuropilin-2 (GenBank TM AF022854) promoters, designated Nrp-1i, Nrp-1ii, Nrp-2i, and Nrp-2ii, respectively, contain putative homeodomain DNA binding sites (TAAT/ATTA) in italics. Oligonucleotide primers used for PCR are underlined, and the TAAT motif located in Nrp-2ii required for binding to DLX1 and DLX2 is boxed. B, ChIP assays were performed on E13.5 mouse forebrain tissues using affinity-purified polyclonal DLX1 and DLX2 antibodies following crosslinking of protein-DNA complexes with 1% paraformaldehyde. Specific bands were evident for Nrp-2ii but not Nrp-1i, Nrp-1ii, or Nrp-2i (left panel). Negative controls included performing ChIP without the addition of either primary antibody or the use of E13.5 hindbrain, tissue that does not express Dlx genes (right panel). Positive controls were mouse genomic DNA (gDNA) and primer pairs for the Dlx5/Dlx6 intergenic enhancer (data not shown) (17). PCR bands were subcloned and confirmed by DNA sequencing. 4). Combined in situ RNA hybridization (Nrp-2) and immunohistochemistry (DLX1 or DLX2) experiments at E13.5 and E16.5 confirmed that only subsets of cells co-expressed Nrp-2 and DLX1/DLX2 within this neuroanatomic region (supplemental Fig. S5). Because most cells do not co-express DLX and Nrp-2, these results are consistent with the potential role of DLX proteins as repressors of Neuropilin-2 expression in the developing forebrain.
Neuropilin-2 Is Ectopically Expressed in the Absence of Dlx1 and Dlx2 Gene Function-The above assays show that DLX proteins bind Nrp-2 in vivo and in vitro, that they repress the Nrp-2 promoter in vitro, and that DLX1/DLX2 and Nrp-2 are primarily expressed in separate cell populations in vivo. However, these experiments do not reveal whether DLX proteins are required to repress Nrp-2 in vivo. To address this key issue, we determined the expression patterns of Neuropilin-1 and Neuropilin-2 in vivo by analyzing Nrp-1 RNA, NRP-2 protein, and Nrp-2 RNA expression in the forebrains of Dlx1/Dlx2 double mutant mice. No differences in Neuropilin-2 expression were observed between wild-type and Dlx1/Dlx2 mutant telencephalon at E13.5 (Fig. 5A). Similarly, the expression pattern of Nrp-1 is unaffected throughout development of the embryonic forebrain to P0 (supplemental Fig. S7B and data not shown). Significantly, immunofluorescence and in situ RNA hybridization experiments demonstrate that there is progressive accumulation of Neuropilin-2-expressing cells in the SVZ of the LGE and MGE after E13.5 through to birth (15) when the mutant mice die (Figs. 5, B and C, and 6). These results are consistent with the hypothesis that with the absence of Dlx1 and Dlx2 function there is loss of transcriptional repression by DLX1 or DLX2 leading to the aberrant expression of NRP-2, resulting in the formation of SVZ ectopias in subcortical cells of the mutant ganglionic eminences (Fig. 6, arrows, panels b, d).

DISCUSSION
This study is the first to report that DLX homeoproteins may function as transcriptional repressors in vivo. DLX proteins have been previously characterized as transcriptional activators (17,38,39), although it was shown that DLX proteins could repress transcription of several reporter plasmids in vitro (40). In addition, BP-1, an isoform of DLX7, represses the ␤-globin gene in vitro (41). We have demonstrated that both DLX1 and DLX2 bind to specific homeodomain binding motifs in a cisregulatory domain of the Neuropilin-2 (Nrp-2), but not the Nrp-1 promoter in vivo, and repress transcription of a reporter vector containing these sequences in vitro. We have also confirmed that the homeodomain of DLX1 and DLX2 is necessary for DNA binding and that this binding is essential for Dlx repression of Nrp-2 expression. A single base pair mutation of glutamine (Q) to glutamate (E) at amino acid position 50 (Q50E) of the 60-amino acid homeodomain is sufficient to eliminate the DNA binding ability of the DLX1/DLX2 homeodomain and subsequent transcriptional repression of a reporter gene in vitro (Fig. 3, B and C). Residual DNA binding in the Q50E mutant proteins due to other possible DNA binding sites localized within the homeodomain (not detected by EMSA) or by binding of DLX1/2 to other proteins bound to the Nrp-2 promoter could account for the ability of Dlx1/2Q50E to still marginally repress (although this repression was not statistically significant). Consistent with the evidence that Dlx1 and Dlx2 function as repressors when bound to the Nrp-2 promoter is the observation that Nrp-2 expression is significantly  (lanes 3, 7), with affinity-purifed DLX antibodies (lanes 4, 8), and with nonspecific antibodies (lanes 5, 9). B, EMSA using embryonic forebrain demonstrates that endogenous DLX1 and DLX2 proteins bind In contrast to our finding that DLX1 and DLX2 are transcriptional repressors of the Nrp-2 promoter, DLX1 and DLX2 act as transcriptional activators of a specific target, the Dlx5/Dlx6 intergenic enhancer in vitro and in vivo (17,39) and gonadotropin-releasing hormone regulatory elements in vitro (42). Furthermore, the Dlx family member DLX3 also acts as a transactivator of a model target gene construct, Dlx3-CAT (43) and the osteocalcin gene promoter (44). In support of a repressor function of the Dlx genes, DLX1, DLX2, and DLX5 interact with a homeodomain binding site within the Wnt-1 enhancer in vitro (2,45). Mutation of this site results in extension of the rostral boundary of Wnt-1/lacZ expression in transgenic animals. The authors suggest that this site may mediate repression of Wnt-1 expression in the forebrain, although neither DLX1 nor DLX2 has been implicated in transcriptional repression of Wnt-1 in vivo (45). The amino termini of both DLX1 and DLX2 may mediate, in part, the transcriptional repression activity of these homeodomain proteins, as demonstrated by the results of substituting their N termini with the Engrailed repressor or VP16 activation domains in the reporter gene assays. We have found DLX2 to be a better repressor than DLX1; therefore, the replacement of the DLX1 N-terminal domain with the VP16 activation domain is able to overcome the "weak" repressive function of DLX1. In contrast, DLX2 is a stronger repressor;  (30)). In addition, the glutamine residue at amino acid position 50 of the homeodomain, critical for binding of the homeodomain to DNA (33), was mutated to glutamate (Q50E) in the full-length wild-type and fusion protein constructs. DB, DNA binding mutant. B, both DLX1 and DLX2 Q50E recombinant proteins (left panel) fail to bind Nrp-2ii oligonucleotides (right panel), confirming these proteins as DNA binding mutants. Arrows indicate specific DLX1 or DLX2/Nrp-2ii complexes. C, reporter gene assays following transient co-transfections with Nrp-2ii DNA sequences fused to a pGL3-Luciferase reporter, in the absence or presence of DLX1 or DLX2 expression in human embryonic kidney 293 cells. Q50E mutations of the DLX homeodomain resulted in an almost complete reversal of transcriptional repression of the Nrp-2 promoter evident with wild-type Dlx constructs. D, DLX homeodomains fused with N-terminal Engrailed repressor or VP16 activator domains modify transcriptional activity of Neuropilin-2 promoter. Compared with wild-type DLX proteins, transcriptional activity of Nrp-2ii was further repressed by Eng-DLX fusion proteins but was reversed by using VP16-DLX fusion proteins (more so for DLX1 than DLX2; refer to "Discussion"). For panels C and D, data shown are the mean Ϯ S.D. of at least three trials. ␤-galactosidase activity was used as an internal control. Because each construct has a different basal level of reporter gene expression, luciferase activities were normalized. * denotes a p value Ͻ 0.001. hence, this function is not completely alleviated by replacement by VP16 at its N terminus. Certainly, the VP16-DLX2 mutant represses less than the wild-type protein. Perhaps the fusion protein did not sufficiently eliminate all of the repressor domains of DLX2. For example, there may be another repressor domain at the C terminus of DLX2, given that the chimeric VP16-DLX2 construct was unable to transactivate reporter gene expression to levels seen with the VP16-DLX1 construct. Subsequently, fusion of the DLX2 DNA binding domain alone with VP16 may have been able to activate better than the VP16-DLX2 mutant used in this study. Both N and C termini of DLX3 are required for mediating transcriptional activation in vivo using Xenopus embryo expression assays (43). Further analysis will better delineate functional domains of DLX proteins other than the homeodomain that are important for the modulation of transcription.
DLX2 is more robust than DLX1 as a transcriptional activator (17) or transcriptional repressor (Fig. 3, C and D). Furthermore, co-transfection of Dlx1 and Dlx2 wild-type or chimeric constructs is neither additive nor synergistic, yielding results that are similar to that of transfection of Dlx2 constructs alone (17); data not shown). This lack of potentiation suggests that there may be greater affinity of DLX2 than DLX1 for binding to their specified homeodomain DNA binding sites and/or slower rates of dissociation. In the embryonic telencephalon, it is also possible that DLX2 interacts with one or more co-repressors or co-activators with which DLX1 does not interact. The Evf-2 noncoding RNA is a recently identified DLX2 transcriptional coactivator transcribed from the Dlx5/Dlx6 intergenic region (46). Although DLX2 and DLX5 form homodimers and also heterodimerize with Msx repressor proteins in vitro and in vivo (38), heterodimerization of Dlx family members has not been established.
The DNA binding specificity of MSX1 is determined by its association with cofactors, such as PIAS1, that direct this homeoprotein to subnuclear compartments where its target genes are located (47). It is likely that the specificity of transcriptional regulation by Dlx genes may involve other factors in unique protein-protein complexes with different transcriptional activities. Whether DLX proteins function as activators or repressors of target gene expression may depend on cooperation with other transcription factors as demonstrated for other homeobox proteins with paired-type homeoproteins (48) or the TATA-binding protein (49). DLX transcriptional activity may also depend on post-translational modifications such as phosphorylation as shown for DLX3 (32) or interaction with proteins such as the PDZ protein GRIP1 (40) and the MAGE protein Dlxin (4,50). It has been shown that DLX1 interacts through its homeodomain with the co-Smad Smad4 during hematopoietic differentiation (51). It will also be interesting to determine whether DLX1 and DLX2 have non-overlapping sets of downstream gene targets at various developmental time points or in specified tissues where Dlx genes are expressed. Although a consensus DNA binding sequence has been established for paired-type homeoproteins such as PAX6, 5Ј-TAATN 3 ATTA-3Ј (48), and HOX proteins, 5Ј-(C/G) TAATTG-3Ј (52), one specific for Dlx family members remains to be established (nucleotides conserved among confirmed DLX targets are underlined). Feledy et al. identified an 8-base consensus 5Ј-(A/C/G)TAATT(G/A)(C/G)-3Ј for DLX3 DNA binding sites (43). This sequence was conserved at specific sites of the Dlx5/Dlx6 intergenic enhancer bound by DLX1 or DLX2 in vitro or in situ (17,39), although there was discordance at the most 3Ј-nucleotide. As well, only seven nucleotides match for the region of the Neuropilin-2 promoter, ATAATTAT, bound by DLX1 or DLX2 in vitro (Fig. 1A). These results suggest that although there are sequence similarities between DLX1, DLX2, and DLX3 binding sites, there are currently insufficient verified DLX targets to establish a consensus DNA binding site for the Dlx gene family.
In the developing forebrain, although cell migration along radial glia is predominant, tangential migration is also an important means for other sets of differentiating cells to reach their destinations. For instance, immature GABAergic interneurons produced in the ganglionic eminences (subpallium) migrate tangentially to the neocortex and hippocampus (pal-lium) (5,14,15,(53)(54)(55). Interneurons migrating to the cortex arise primarily from the MGE and the entopeduncular area and follow two major pathways as they traverse the LGE/striatum: a superficial subpial route and a deep route adjacent to the SVZ (25,34,37). The molecular mechanisms that guide and sort these migrating cells are becoming elucidated (25). Recently, ErbB/neuregulin signaling has been implicated in supporting migration through the LGE toward the cerebral cortex (22,23,56). Glial cell line-derived neurotrophic factor signaling via GFR␣1 has a role in the differentiation and migration of cortical GABAergic cells from the MGE (57). Likewise, chemorepellant tissues and factors have been identified that participate in directing these migrations (24,26). There is evidence that striatal expression of semaphorin 3A and 3F repels tangentially migrating interneurons (15). Semaphorin 3A can also act as a chemoattractant in vitro (58,59), whereas another secreted class III semaphorin, semaphorin 3B, mediates both attraction and repulsion in vivo (60). Semaphorins interact with receptor complexes consisting of neuropilins (ligand binding subunits) and class A plexins (signal transduction subunits) (61)(62)(63). In Neuropilin-2 null mice increased numbers of tangentially migrating cells enter the striatum. Similarly, interneurons that express a dominant negative form of Neuropilin-1 also aberrantly enter the striatum (15).
Previous studies demonstrated that Dlx1 and Dlx2 function is necessary for migration of more than 75% of tangentially migrating interneurons to the murine cortex (5), olfactory bulb (13), and hippocampus (16). Upon loss of Dlx1 and Dlx2 function, there is ectopic accumulation of cells with molecular properties of cortical interneurons within the SVZ of the ganglionic eminences. These cells, which express high levels of Nrp-2 (15) (Figs. 5, B and C, and 6) are presumed to be collections of interneurons that have failed to migrate to the cortex. Normally, Neuropilin-2 expression patterns minimally overlap with Dlx1 and Dlx2 expression (Fig. 4) and with the pathways followed by tangentially migrating interneurons (64). These data support our hypothesis in which DLX1-and/or DLX2expressing cells could down-regulate Nrp-2 expression after E13.5, enabling later born interneurons, most derived from the MGE and the anterior entopeduncular area, to take the deep route to the striatum toward the neocortex. Repression of NRP-2 expression may therefore allow these interneurons to migrate through semaphorin-expressing cells in the striatum and at the pallial-subpallial interface (15,25,65). Late born interneurons that aberrantly express NRP-2 in the absence of Dlx1 and Dlx2 function could fail to migrate to neocortex due to the repulsive guidance cues from semaphorin-3F-expressing cells mediated via NRP-2 and accumulate as subventricular ectopias (15,25,65). In the Dlx1/Dlx2 mutant, loss of DLX-dependent repression of Nrp-2 transcription may explain, in part, the ectopic accumulation of GABAergic interneurons in the ganglionic eminences. These studies do not exclude the likely possibility that other molecules, some directly regulated by Dlx genes and others independent of Dlx function, contribute to the tangential migrations of interneurons especially after they have passed the pallial/subpallial junction (15). Other transcriptional factors, in addition to Dlx-dependent repression, may contribute to repress Nrp-2 expression in interneurons entering the striatum, as evidenced by the presence of semaphorinexpressing domains and the lack of Neuropilin-2 expression in sorted striatal interneurons (15). However, our results suggest that Dlx1 and Dlx2 repression of Nrp-2 may contribute to the tangential migration of late born inhibitory interneurons from the subcortical telencephalon.